AMD Making a 5 GHz 8-Core Processor At 220 Watts

Vigile writes "It looks like the rumors were true; AMD is going to be selling an FX-9590 processor this month that will hit frequencies as high as 5 GHz. Though originally thought to be an 8-module/16-core part, it turns out that the new CPU will have the same 4-module/8-core design that is found on the current lineup of FX-series processors including the FX-8350. But, with an increase of the maximum Turbo Core speed from 4.2 GHz to 5.0 GHz, the new parts will draw quite a bit more power. You can expect the the FX-9590 to need 220 watts or so to run at those speeds and a pretty hefty cooling solution as well. Performance should closely match the recently released Intel Core i7-4770K Haswell processor so AMD users that can handle the 2.5x increase in power consumption can finally claim performance parity."

Comments

Awesome

[redmid17]

I always wanted to have a computer running my freezer

Re:Awesome

[wjcofkc]

I think you mean in your freezer...

Re:Awesome

[dicobalt]

It can do both.

Re:Awesome

[redmid17]

No I meant running my freezer. Hence the reason I typed: "I always wanted to have a computer running my freezer" instead of "I always wanted to have a computer running IN my freezer"

Re:Awesome

[ArcherB]

No I meant running my freezer. Hence the reason I typed:

"I always wanted to have a computer running my freezer"

instead of

"I always wanted to have a computer running IN my freezer"

Oh. Then I don't get it.

As a side note, I've always wanted to take an old mini fridge and turn it into a computer case.

Re:Awesome

[trum4n]

Done it. It was fun and functional, but you have to do everything you can to keep the condensation out. Fast a hell tho!

Re:Awesome

[Anonymous Coward]

Well, then, thanks for sharing, I guess. *walks away slowly*

Re:Awesome

[sjames]

There is a type of freezer that can operate using a heat source for power (strange but true).

I'm, guessing that's what he meant anyway.

Re:Awesome

[wjcofkc]

If your freezer is less than fifteen years old, it is most likely already being run by a computer.

Re:Awesome

[Bill, Shooter of Bul]

i think it would work better running a heated swimming pool, or a grill. But to each their own.

Re:Awesome

[Anonymous Coward]

I just bought an RV; that's how their fridges work. Most can run for a week or two on a tank of propane.

Re:Awesome

[hairyfeet]

I would urge those that wonder WTF AMD is doing copying all the old mistakes Intel did with netburst to read this post by a former employee who lays out exactly why this is happening, the former CEO did the usual Wall street move of slash and burn, get a stock bounce, and cash out.

They are stuck with the Netburst that is Bulldozer/Piledriver/Suckavator or whatever other names they want to give it because the former CEO FIRED everybody that knew how to make a chip over there and replaced them with computer layouts which as you can see blow through power like shit through a goose while giving worse performance on a per watt basis than the previous Stars arch.

This isn't coming from some Intel fanboy, I own and sell nothing but AMD at the shop, but when I can no longer get Stars and Liano chips I'm gonna have to seriously look at Intel because these new designs just suuuuck. There is a good reason why you don't see Thuban chips in most benchmarks against the new chips, its because if you matched clock for clock the Thubans and Denebs will win. That is pretty damned sad, when your old chips are actually better while using less power but the CEO they had closed down production of all the Stars cores (again to get a stock bounce and cash out) so there really is no plan B here.

I just hope the game console chips can give them enough operating capital to keep them afloat while hopefully the new chip designer they hired, the same one that did the Athlon64 and the Apple A6, can come up with a new design to make AMD at least kinda competitive. Until then I'll hang onto to AM3+ and Stars as long as I can and then start looking at the i3s and i5s.

Re:Awesome

[Applekid]

I always wanted to have a computer running my freezer

With that kind of power consumption, I wouldn't expect it to stay being your freezer for very long.

Grilled bratwurst anyone?

Re:Awesome

[Tough Love]

You got that one wrong. Netburst was about deepening the pipeline to ridiculous extremes in order to ramp the clock. The new AMD story is pure clock ramp via process technology and power management. Big difference there.

Re:Awesome

[Anonymous Coward]

he's well monged int 'e

Re:Awesome

[Impy the Impiuos Imp]

Whoosh! He was making a joke as to the hideous heat released.

Re:Awesome

[gman003]

No, this particular story is analogous to a 2004-era story about Intel releasing a new Pentium IV at yet-higher clocks. The current story is about a clock ramp, but the overarching narrative is the same.

The Bulldozer architecture is fundamentally broken, this time due to simple negligence (mainly in management) rather than a faulty assumption. The only way to get reasonable performance from it is to clock it to high speeds, which gives very diminishing returns. Power consumption scales with the *cube* of the clock speed, so you pretty quickly run into a power/heat wall. They clocked the early ones pretty aggressively already, but at the cost of power and heat (and thus, noise). But it's the same story as the Pentium IV - the smart people are on something else.

AMD seems to be trying to put itself back together. Hopefully the PS4/Xb1 wins will give them enough of a cash flow to keep them solvent until they can get a new architecture out, or at least hammer out the IPC problems with Bulldozer. On the bright side, Intel's been distracted by ARM - they threw away a year's lead on performance to chase idle power draw, which should give AMD a bit of time to catch up on performance on the desktop.

Re:Awesome

[Anonymous Coward]

On one hand we have Slashdot reader say to hell with hand optimizing software because _THEIR_ own time are so important and let the compiler/interpreters do their job and yet the same people complain that the chip designers should do the exact opposite.

AMD seem to be doing fine with their GPU that are developed on the same methodology. Going the SoC path (aka your "replaced them with computer layouts") is what make AMD quick in coming up with the custom APU for Sony and their new hUMA architectures. Since AMD went to outside fab house, it is important that they keep the machine do the routing so that it is more portable across FAB shops.

Re:Awesome

[Tough Love]

No it's not analogous to anything. It is about observing that AMD did not deepen the pipeline, therefore this story is not like the Netburst debacle.

Re:Awesome

[Cito]

yea I have a propane fridge and freezer, I pulled it out of an old RV at a junk yard for free :)

but it works, and I keep it for emergency purposes. Few years ago my town was without power for 5 days after a Hurricane came through southern Ga. So I connected the freezer/fridge duo to a propane tank and was able to keep most my food from spoiling.

Re:Awesome

[jones_supa]

I doubt that. The temperature measurement mechanics and the on/off pulsing for running the coolant does not need even a simple microcontroller. It's probably driven by completely analog electronics. Unless we are talking about some unordinary deluxe freezer.

Re:Awesome

[gman003]

Tell me, did Intel deepen the pipeline when they released the Pentium 4 HT 570J, as compared to the Pentium 4 HT 560 launched a few months before? No, they didn't, but they did increase the clock speed. Same story then as now - AMD didn't deepen the pipeline, but they are pushing the clock speeds to inadvisable levels because that's the quickest way to improve performance, and they *desperately* need to improve performance.

Look, I'm not an Intel fanboy. Even with all its problems, I'm actually still planning to use an A10 in a build. But pretending AMD doesn't have a problem with IPC or power efficiency only makes you look uninformed.

Re:Awesome

[Lumpy]

All of them work that way. Look at how a fridge works.

Re:Awesome

[Anonymous Coward]

AMD needs a die shrink badly. I don't think the power consumption would be quite so bad on their current designs. Fab is killing them.

Re:Awesome

[viperidaenz]

So who should spend their time optimising the compilers/interpreters? Somebody has to do it. It also needs to be done at the lowest levels to make the bigger impact.
Optimising the CPU is best place to do it.
If you go up too many levels you get the bastard child that was Itanium.

Re:Awesome

[wjcofkc]

Hmm... Makes me wonder if Samsung makes refrigerators. Smart Fridge, burrito please... SMART FRIDGE, BURRITO....! "Waves hands around in open air attempting to open freezer"

Re:Awesome

[steelfood]

On the bright side, Intel's been distracted by ARM - they threw away a year's lead on performance to chase idle power draw, which should give AMD a bit of time to catch up on performance on the desktop.

In the short term, this appears to be a good thing. In the long term, this is very bad for AMD.

The world is moving to low-powered portables. The future of consumer computing will not be on the desk or lap, but in the hand. Workstations will still use desktop chips, but Intel pretty much has that market cornered.

The low-powered, RISC space is where AMD needs to go. It doesn't necessarily have to be ARM. Instead, there's a market for low-powered x86, which is where Intel is going with Haswell. AMD needs to get ahead of the game and create something that is capable of power sipping (which obviously won't be x86), but is also capable of running legacy x86 code at reasonable speeds.

Basically, they need to create a migration path away from x86, which will never be as efficient as ARM and thus has no chance in the portable space. Yes, Intel tried that with Itanic, but they were aiming in the wrong direction (servers).

Re:Awesome

[0111 1110]

Power consumption scales with the *cube* of the clock speed, so you pretty quickly run into a power/heat wall.

Bullshit. There is no wall. I don't care all that much about heat/power/noise. I have a water cooling setup and I'm prepared to move to phase change if necessary. What I want is for Moore's Law to mean something again. Giving up on clock speed was a bad move on Intel's part. It's just sad that we still haven't made it to 5 Ghz. This whole shift from raw performance at any price to performance per watt or even wattage walls that cannot be exceeded just sucks.

I haven't bought a new CPU since my Wolfdale Core 2 Duo that I run overclocked at 4 Ghz when I need performance or at 1.2 Ghz, 0.9 volts when I don't. It was nice to get off the upgrade treadmill for a while, but it has gotten old. I want a 10 Ghz processor (without reducing performance per cycle), and I want it now.

Does anyone have one of those old Intel roadmaps that promised something like 12 Ghz by now? Staying under 4 Ghz for all of eternity is not what we were promised. I haven't owned an AMD CPU since they were king of the hill, but I applaud this performance at any price philosophy. Efficiency is great. I love efficiency. If/when VW ever imports their XL1 to America I'll be first in line to buy one, but I wouldn't want it to be the only car available. I like to save money on petrol, but I also might want to actually enjoy driving every once in a while.

I do realize that Intel's CPUs are just as fast while using a lot less power and that this new AMD processor is the equivalent of a car that drives like a Prius with the fuel economy of a Lambo, but it's a start in the right direction. If you can't beat Intel at their own game then change the game. I like the idea that someone has the balls to say, "Power be damned, we're going to make the fastest processor we know how to make. Full stop."

Re:Awesome

[redmid17]

This guy got it

Re:Awesome

[Anonymous Coward]

Or, AMD decided the x86 platform is dieing and is shifting focus towards ARM: They've repurposed their existing lines for consoles while focusing on their 2014 ARM x64 chips.
How long do you think would it take before capable ARM PCs will become available? A year? Two? Linux and Windows are already available for ARM. Office isn't too far behind either...

Re:Awesome

[Anonymous Coward]

In perpective 10ghz/2ghz and 8 core processor is bound to create problems so choose one (single core and 5ghz or mulicore and ~3ghz) but not both these days.

Re: Awesome

[DigiShaman]

IANAP. I don't think we will ever see CPUs clocked to 10Ghz unless there's some asynchronous timing voodoo used. I'm of the understanding that the speed of light and signal propagation is the real limitation here with regards to higher frequencies.

Re:Awesome

[gman003]

Perhaps "wall" isn't the best term - it's more of rapidly diminishing gains for rapidly increasing costs. People have hit 8GHz, but through liquid nitrogen cooling, which isn't exactly practical for consumer use.

Intel has actually been doing well at maintaining steady performance increases, except with Haswell. But instead of doing it with clock speed hikes, they've been working on IPC and ISA extensions. They've added AVX, to do 256-bit SIMD operations instead of 128-bit SSE. They've done a lot of work on making sure they do as much work per clock cycle as possible - I've seen real-world benchmarks where they exceed 1 instruction per clock. They've been growing the dispatch unit, to let a core have more operations in flight at once. They've optimized instructions, bringing some that took tens of clock cycles to complete into the single digits. They have not been sitting idly by. They may not have focused 100% on raw speed, but performance has been steadily rising by 5-15% per year, even with Haswell.

AMD's problem on the desktop isn't that they use too much power. Their problem is that they're no faster than Intel, often *slower*, while also using more power. You want an efficient computer, you go Intel. You want a powerful computer, you also go Intel, just with a larger checkbook - AMD has nothing that can compete with the higher-end i7s. You only go AMD if you want a cheap computer, or if you want good integrated graphics (which Intel actually is taking a shot at, now, in places).

However, if you are interested in massive clock speeds, you should look into RSFQ computers. Based on superconductors and some weird quantum effects (although it is not a quantum computer), they've had "transistors" oscillate in the 500GHz range. An old NSA study I read expected a first-generation computer based on the technology could reach 50GHz, eventually rising to 100GHz before speed-of-light problems become difficult to solve (at these speeds, a signal would travel only a few millimeters per clock cycle). And it's even power-efficient, running at under a millivolt. Shame they didn't follow up on the tech, as far as we know. I bet that stuff comes back into the spotlight the minute someone finds a room-temperature superconductor, though.

Re:Awesome

[fast turtle]

your full of lint there. An AMD FX-8320 compares nicely with the I7-3700K (8K+ on the benchmarks) - This also puts in spitting range of the E3-1245 Xeon as while the 1270/75 tend to push just over 9000. Not bad for a chip that's only $150 (that's $125 less then 1245 and over $200 less then the 1270/75 Xeon.

Re:Awesome

[ifiwereasculptor]

Let's get something out of the way first: that 5GHz chip will suck. Incredibly. However,

The Bulldozer architecture is fundamentally broken

is not exactly prudent phrasing. Bulldozer sucked. Period. It was obviously half-baked. Trinity is much better. If AMD can repeat their predicted 15% performance improvement on their next BD iteration, then they will have a truly good product. Yes, Intel is still better. An i3 is better than an FX-4300 and an i5 is better than an FX-8350 for most workloads. However, Intel jumped way ahead of AMD with Sandy Bridge and, since then, they have been mostly stagnant on the CPU front. Compare SB, IB and Haswell and you'll see mostly low single-digit increases in performance, while increasingly worsening heat dissipation. Currently, Intel has a big process node advantage (22nm vs 32nm) and they still aren't that far ahead. AMD launched a crappy product way before it was ready, but that doesn't mean the architecture sucks.

Furthermore, the reduced number of FP units is to be offset by their HSA, only to be seen in their next product, coupled with GCN iGPUs. So, while I wouldn't buy AMD either right now (except for the FX-6300, perhaps), it's not exactly smart to condemn a move before it's throughly played out.

Re:Awesome

[0111 1110]

Yes. As usual the law of diminishing returns applies, but I don't think that is the only reason we have seen such slow gains in performance. Somehow it has become accepted that going above 100 - 150 or so TDP is just unacceptable. For some applications it certainly is and I am curerntly running my CPU at the lowest speed and voltage it is capable of to save money on my electric bill. I only go above 1.2 Ghz when I have a reason, like playing a computer game or video compression etc.

Compared to what we used to expect out of a single generation Intel's slow, steady gains are just not very impressive. This is not the future that science fiction was expecting. I find it kind of depressing.

It used to be that one of the reasons I wanted to continue living was to see the amazing improvements in computer power that I assumed I might live to see. Now it seems like we cannot expect much improvement, at least not without some kind of fundamental breakthrough that may not happen at all this century.

One thing is certain. If things continue like this traditional artificial intelligence will always remain out of reach. Only wetware, connectionist, biology inspired processing has any chance.

Only embarrassingly parallel algorithms may show any significant gains although we haven't seen any major jumps in the number of processing cores either.

I suggest that it is about time programmers started getting used to coding in assembly once again. Order of magnitude improvements in processing power are over. Cue the nearly imperceptible gains where it takes decades before a difference in processing speed is actually noticable.

This is the first I've heard of RSFQ computers. I'll have to look into them. It certainly sounds interesting. It's starting to look like only a fundamental paradigm shift like that will allow us to continue into the sort of future that many of us like to read about in science fiction.

Re:Awesome

[Anonymous Coward]

No I meant running my freezer. Hence the reason I typed:

"I always wanted to have a computer running my freezer"

instead of

"I always wanted to have a computer running IN my freezer"

Does anyone understand what the hell is funny about this?

Unless you're referencing a show I've never seen, I have no idea what the heck you're talking about (and specifically why it's supposed to be funny).

Re:Awesome

[smash]

They do, I have one. It's not a smart fridge.

Re:Awesome

[smash]

Netburst only failed because the planned clock-rates proved to be impossible. It was intended to be running at 10Ghz by 2006 if I'm not mistaken. The major issues with getting beyond 3Ghz (I had a 3Ghz machine in 2004, and my following machines have been 2.4ghz and 2.2ghz... i.e., clocking SLOWER) were not foreseen by intel.

If it had scaled to the intended clockspeeds, netburst would have been fine.

But it didn't. Intel tried and it didn't work out. Thems the breaks.

Re:Awesome

[smash]

Newsflash: your core2 duo at 4ghz will get murdered by a core i5.

Re:Awesome

[smash]

Intel have been coasting on the CPU speed front and focused on GPU and power consumption reduction. Both of the latter two factors are now more important for both mobile and datacenter.

Re:Awesome

[GigaplexNZ]

No, that's different. A fridge typically uses an electrically powered compressor to move heat from inside the fridge to the outside, which is why it gets warm. It's not using heat as a power source.

Re:Awesome

[GigaplexNZ]

The low-powered, RISC space is where AMD needs to go. It doesn't necessarily have to be ARM. Instead, there's a market for low-powered x86, which is where Intel is going with Haswell. AMD needs to get ahead of the game and create something that is capable of power sipping (which obviously won't be x86), but is also capable of running legacy x86 code at reasonable speeds.

Look into Bobcat and derivative architectures from AMD.

Re:Awesome

[Anonymous Coward]

Unless you're referencing a show I've never seen, I have no idea what the heck you're talking about (and specifically why it's supposed to be funny).

I just wanna step in here, because it hits a chord with me.

Is merely referencing a show really funny? Like, if someone just blurts out "'e's pining for the fjords" is that funny? I don't think it is. You have to add something. There's nothing wrong with references, but you've gotta do more than that. Like, to be funny, you've gotta be starting out with a real life dead parrot or something, and then say he's not just pining for the fjords, because you just blew him up with TNT or something.

It takes more than a reference. I'm not saying how far you have to go, just that you go somewhere.

Re:Awesome

[hairyfeet]

Dude, now you are just being a douche, you really are. Your argument would be like saying "You can't compare any boat sinking to Titanic because that boat wasn't built before 1920!" and it really makes you sound like a pedantic dick, it really does.

As another pointed out Intel did NOT keep adding ever more pipelines to netburst, they just kept ramping up the speed because they didn't have the IPC, with a 5GHz on their roadmap before they pulled the plug. Likewise Amd is ramping up the speed because they don't have the IPC and as other have already pointed out you get THE EXACT SAME PROBLEM you had with netburst, the more you crank the clock on an already inefficient design the more waste power and heat you are gonna have. Intel ended up into 150w territory before they finally gave up on netburst, AMD went from 65w-95w with Athlon64 to now 140W+ and soon to be 220w and again they, just like netburst, are blinding trying to fix the fundamental flaws with the arch by simply cranking up the clocks.

So I'm sorry but everyone else can see the analogy, i could wallpaper this post with websites where tech editors make the exact same comparison to netburst, so if you think the ONLY way anything could be like netburst is by adding a 31 stage pipeline? Well you're just being a dick for the sake of arguing, please stop wasting everyone's time being a pedantic ass.

Re:Awesome

[kangsterizer]

Man yeah, thats sad :(

Re:Awesome

[hairyfeet]

Sadly it wouldn't help friend. read the article by the AMD engineer i linked to, they knew 2 years before Bulldozer there was a serious flaw in the design, the "half core" design requires BOTH the Operating System AND the programs to all be designed with that in mind and even then it will only give a boost in certain niche applications.

The simple fact of the matter is just as Netburst went for raw speed above all with the BD/PD design they made MUCH more primitive cores and thought they could make up for the primitive design by just adding more of them but most programs just won't scale. Even games benefit from having one really powerful core to run the main thread and weaker threads for AI,Sound, etc but the BD/PD design instead has a primitive core that is really only good at integer loads, the second you start adding floating point its gonna bog down.

At the end of the day the BD/PD design is really only good for servers and even then only in certain roles, with desktops and laptops we run just too many FP programs and single threaded programs for very simple weak integer heavy cores to work. This is why they keep cranking the clocks, hoping they can boost the speed enough to make up for that shortcoming but to use a car analogy that would be like saying if you just push this Pinto really REALLY hard it'll keep up with a Porsche, its just not gonna work. hell they have to go for a full 1Ghz higher clock just to beat the Thuban chips which were already not equal to the i7, there is no way in hell they can get the clock high enough to make a half core compete with a full core, its just not gonna work.

Re:Awesome

[lightknight]

If I remember correctly, Intel was saved by Intel Israel after Intel went down the NetBurst route with no way out.

AMD needs to get the DEC people back. And the CEO who allowed them to go missing needs to be executed. The current architecture, Bulldozer / Piledriver / whatever is a partially completed DEC design...which is where the problems are seeping in, it's only partially finished. AMD needs them back anyway, whether it keeps with the current architecture, or moves to something different, they're the only people with the expertise who can make AMD competitive again; and I mean that...they're DEC engineers...they aren't even being 'made' anymore, and it would take decades to create new ones with that kind of expertise.

As for getting away from x86...this albatross is here to stay. It's far too entrenched in the business world, and if you have ever tried to convince a business that rewriting a decades old 'critical' app in VB & Access would be a step in the right direction, then you'll know what I mean; people are cheap, and won't move until the old machines / programs are dead with zero chance of resuscitation.

Re:Awesome

[nighthawk243]

At 220W, it'd be running too hot. It'd be perfect for a dryer, high-BTU space heater, or an oven.

Re:Awesome

[lightknight]

Hmm, I think they're waiting for the next major shift in technology to happen, which is still be developed (got some of the parts, need some of the others). That's optical processing, and every once in a while, a new piece is found. I think Intel has done some work on it, and I know some others have looked into it, but right now, everyone is kind of twiddling their thumbs while the research & development people slowly figure it out.

The Electrical Engineers want it if only because it would make motherboard traces less painful than they currently are, and allow them to really start making use of space. All traces going to and from certain components need to be of the same length, yay!

Re:Awesome

[lexman098]

Sorry, gotta make a correction (why else would I troll slashdot?).

Power increases with the *square* of clock speed. See for yourself

Your general point that relying on increased clock speed is a bad strategy remains correct of course.

Re:Awesome

[gman003]

Ah, but for any overclock worth speaking of, you'll also need to overvolt. So while you are absolutely correct in theory, in practice is a different matter.

Re:Awesome

[hairyfeet]

Wow, a 4 digit...didn't know any of you guys were still around ;-)

But in this case you are wrong ancient one, and here is why: Like the BD/PD design the netburst design had a fundamental flaw, in the case of netburst it was that long ass pipeline that made stalls nasty and the fact that the chip never had great IPC to begin with. I mean you can take pretty much any of the Pentium D 8xx chips and hit 4GHz and yet a C2D clocked at less than half of that will curbstomp it and in the case of the i3s it'll curbstomp it while using less under load than the 8xx used at idle.

You see what Intel found out was that raw clocks for clocks sake meant jack and shit if that high clock didn't translate into useful work, or IPC, and the cost of power and cooling to those monsters got to the point people didn't want 'em. Not that I can blame 'em, i have a Pentium D 805 running in the shop while I decide whether its worth the trouble to try to find a socket AM2 board for this ULV Athlon X2 or to say fuck it and just get a Bobcat board and the damned thing will often hit over 143f on light loads and idles at 135f So having the Pentiums hit higher clocks really wouldn't have helped, the design was flawed to begin with. if you don't believe me starmicro has the Pentium D Extreme Editions for $75, grab one of those suckers and a hell of a cooler and you'll probably hit 5.5Ghz...and still get curbstomped by a 3GHz i series or Phenom II X2, because without the IPC to back it up the clock doesn't mean shit.

That is the problem AMD has now, the "half core" design has really shitty IPC (in fact you have to have a 1GHz overclock compared to the Phenom II to beat the Stars arch with the BD arch) and instead of accepting that and going back to the drawing board they are just cranking the shit out of the clock which again gets you into crazy temps and power draw. I mean who would even want a 220w CPU, when a chip drawing half the power in the Intel camp will beat it? Hell I build exclusively AMD but I won't even mess with the 140w because of the heat, i don't even want to know how much heat a 220w CPU would crank, even with liquid cooling the amount of waste heat that thing is gonna generate will just be unreal.

Re:Awesome

[smash]

The pipeline was so long because longer pipelines makes it easier to ratchet up the clockspeed. Intel ran into other issues which precluded the ability to ramp up past 3Ghz. So the trade-offs they willingly made with pipeline length were still there, without the clockspeed benefit.

IF intel got the raw clockspeed to 10Ghz like they planned (by willingly trading some IPC to get there), Netburst would have been a winner.

As it is, it didn't work out, and intel paid the IPC penalty without getting the payoff. Lesson learned. HENCE as you say, the core series, which kick its arse.

Re:Awesome

[dbIII]

I'm sure there's a wikipedia article you can read before trying to "correct" people with ignorant gut instinct. If not just google "refridgeration cycle".
In the 1950s my parents has a kerosene fridge - heat input from a wick to keep the gas moving and the compressor was a tank of water with the ammonia pipes going through it. Just like the modern ones it's using heat as the power source. The modern ones have an electrical resistance heating element instead of a kerosene wick.

Re:Awesome

[dbIII]

AMD has nothing that can compete with the higher-end i7s

You've ignored the top end where AMD have 16 core 4 way CPUs at the same clock speed as Intel's 10 core stuff for less than half the price.

Re:Awesome

[GigaplexNZ]

I'm sure there's a wikipedia article you can read before trying to "correct" people with ignorant gut instinct. If not just google "refridgeration cycle". In the 1950s my parents has a kerosene fridge - heat input from a wick to keep the gas moving and the compressor was a tank of water with the ammonia pipes going through it. Just like the modern ones it's using heat as the power source. The modern ones have an electrical resistance heating element instead of a kerosene wick.

You mean like this one? Modern ones use a compressor, they don't use a heating element.

Re:Awesome

[dbIII]

I only go above 1.2 Ghz when I have a reason, like playing a computer game or video compression etc.

A lot of recent server stuff clocks down when idle:
processor : 0
vendor_id : AuthenticAMD
cpu family : 16
model : 9
model name : AMD Opteron(tm) Processor 6174
stepping : 1
cpu MHz : 800.000
As required it ramps up to 2.*GHz on whatever core/s need it.

I think I've seen as low as 300MHz in /proc/cpuinfo in a more recent server.

Re:Awesome

[smash]

In other words.... intel planned to have a P4 running at 10ghz in 2006 (i think?). Even if its IPC was half that of a Core 2 running at contemporary speed of 2-3ghz, the 10 ghz P4 would still be 1.5x faster.

The Core 2 onwards only exist because even though intel stretched the pipline for 10ghz other issues cropped up (forget what they were exactly leakage or something?) and the P4 was left as a dead duck. Intel salvaged a little performance from it with hyperthreading and various instruction set upgrades, but it wasn't really enough.

What amazes me though is that the current architectures we are still using in intel land are still (admitedly distant) descendants of the P6 from 1995.

Re:Awesome

[c0lo]

They do, I have one. It's not a smart fridge

Of course it's not, it goes without saying.

Re:Awesome

[Anonymous Coward]

What's confusing is that the design flaw is totally different *and unspecified*, and the Netburst design flaw of incredibly deep pipelines really was in the service of increasing the upper bound on clock speed. I get that you're saying they went over the top and used even more clock speed to deal with the fact that pipeline stall was apparently a bigger problem than they anticipated by making each stage of the pipeline even faster.

I really don't know what the alleged Bulldozer design flaw is.

Re:Awesome

[dbIII]

Maybe the first few pages of a thermodynamics textbook may help since the refridgeration cycle is often used as an introductory example.

Re:Awesome

[slashdot_commentator]

Businesses are structured for the quarterly stock profit. For the suggestion your making, the management team would have to be in position to maintain the strategy for years, while taking a killing in profitability. Basically, AMD would have to be taken over by a Japanese company for even a *chance* of a comeback. AMD will have to hope that its pivot to gaming consoles gives it enough oxygen to sustain a radical, new CPU architecture. AMD is a dead man walking.

Re:Awesome

[TheLink]

On one hand we have Slashdot reader say to hell with hand optimizing software because _THEIR_ own time are so important and let the compiler/interpreters do their job and yet the same people complain that the chip designers should do the exact opposite.

1) There are only a few CPUs.
2) There are at least hundreds of thousands of software programs. Many of which require frequent changes as user demands change.
3) The CPUs affect the performance of all those hundreds of thousands of software.

The CPU layout isn't going to change every few weeks/months. Whereas software often changes every few months.

So it makes more sense to have a relatively few very smart and talented people optimize the CPUs. Than have many programmers "optimize" their programs and too often make them harder to maintain and not much faster.

Hand optimizing certain software like the kernel, high performance drivers, performance critical software does make sense. But for most other programs it does not make sense.

Re:Awesome

[slashdot_commentator]

I suggest that it is about time programmers started getting used to coding in assembly once again.

You can give up right there. The days of humans getting 100x more efficiency out of a CPU using assembler rather than a higher level language are over. Optimizing compilers are able to devise efficiencies at large scale/detail that a human can at this point. Enterprise level software requiring millions of lines of code are just too large to be optimizable by one human writing in assembler. Speed efficiencies with out of order execution, deep pipelines consumer CPUs will be better utilized by compilers able to make better predictive arrangement of code.

Don't get me wrong. You'll always be able to find ONE "John Henry" that will be able to outcode the "stream compiler". But you can't build a world economy on one programmer. And forget about finding COMPETENT assembler programmers. The people you need to extract these kind of efficiencies are like finding prima ballerinas. Sadly, the world's economy needs more mediocre programmers to generate more working code, and more higher-level, software engineers to implement new solutions for problems addressable by a computer.

Re:Awesome

[Anonymous Coward]

Just because it was built by Samsung doesn't necessarily mean it has the Downs.

Re:Awesome

[MagusSlurpy]

Moore's Law doesn't say anything about CPU speed, it's about the number of transistors on a chip doubling every two years. And it is still accurate (roughly).

http://upload.wikimedia.org/wikipedia/commons/0/00/Transistor_Count_and_Moore's_Law_-_2011.svg

Re:Awesome

[MagusSlurpy]

But it doesn't matter, because they both run at 3 GHz! Wait, what? Instruction set? What's that? Like for Lego sets?

Re:Awesome

[GigaplexNZ]

I've studied thermodynamics and I'm aware of how the various refrigeration cycles work. My point is that the now widespread vapor-compression cycle does not use a heating element, the old deprecated vapor absorption cycle is what used a heating element of some kind (such as a kerosene wick).

Re:Awesome

[Tough Love]

Heh, you of all people are complaining about dicks.

Re:Awesome

[Kjella]

The low-powered, RISC space is where AMD needs to go. It doesn't necessarily have to be ARM. Instead, there's a market for low-powered x86, which is where Intel is going with Haswell. AMD needs to get ahead of the game and create something that is capable of power sipping (which obviously won't be x86)

Actually Intel has already shown they can make x86 phones on par with existing ARM phones, not market leading or anything but middle of the road. You want AMD to out-do ARM and Intel, push a new instruction set, create the compiler support and the industry momentum behind it? With a single, financially troubled company who I wouldn't bet is there five years from now? Yes, Itanic was a huge failure but Intel still makes Itaniums for anyone foolish enough to bet on that horse, AMD couldn't make any such promises. To compare it to US politics x86 and ARM are the republicrats and anything AMD would come up with a third party. They get support on Slashdot but go nowhere in the real world.

but is also capable of running legacy x86 code at reasonable speeds.

If you want a past example that tried just this, see Transmeta with their VLIW design. It did not end well, the x86 code didn't perform and nobody made native VLIW binaries. Besides, AMD has already split their efforts two-way on ARM and x86 which I suspect was a mistake. If you're losing the battle against Intel picking a new one against Samsung, Qualcomm, nVidia, Apple and everyone else making ARM chips still doesn't seem like a good idea. As an x86 supplier they'd have value as an alternative to Intel, in the ARM market they're entirely expendable.

Re:Awesome

[Bert64]

Only processors are adding new instructions all the time and compilers are way behind in using them... You can't just compile any old random code and expect it to perform well, for instance look at john the ripper or bitcoin, if you compile the C cores even with an up to date optimizing compiler it cannot compete with the hard coded asm cores.

And people are using increasingly high level languages, some of which are extremely inefficient... Just look at how fashionable ruby is and how poorly it performs.

Re:Awesome

[hairyfeet]

Again sure if you took say the Pentium D EE and used a bad enough cooler you could get up to 6GHz plus, but how much useful work would you get for that heat? To use a /. car analogy it doesn't matter if you have an engine that can take a high redline if the transmission doesn't turn those revs into speed. I mean look at this comparison using the Pentium 4 661, which was one of the fastest P4 chips they made that wasn't an extreme. You see that C60 that scores 2 marks higher? That is a 1GHz netbook chip that uses a grand total of...9 watts. A 9 watt netbook chip scores higher than one of the fastest MHz P4s that was ever made.

So I'm sorry but reaching 10GHz really wouldn't help when the IPC is THAT bad, if the passmark stayed relative to speed you'd be scoring around 1300 marks at 10Ghz, which would be equal to a 4 year old Athlon X2 at a blistering...2.5Ghz. This is the same problem that AMD is now suffering from, their BD/PD design just doesn't have good IPC and cranking up the speed is just gonna bring it from bad to okay, and the amount of heat you'll have to deal with really isn't worth it.

Re:Awesome

[roscocoltran]

Same here:

processor : 0
vendor_id : AuthenticAMD
cpu family : 21
model : 2
model name : AMD Opteron(tm) Processor 6378
stepping : 0
cpu MHz : 1400.000

It's the default on rhel5.9 and rhel6.4. How do you control this beahavior ? Is there a way to go lower than that in frequency ?

Re:Awesome

[Anonymous Coward]

They could have picked up a MIPS design instead of ARM. A MIPS CPU core would have complemented their AMD's GPU design well enough.

Re:Awesome

[smash]

6ghz is still WELL SHORT of where intel was planning on running this CPU - on AIR. And yes the ghz would help with the IPC as it was. The whole point was they thought they could ramp clockspeed higher to more than compensate for the IPC penalty of horrendously long pipelines.

But reality differed from projections. Comparing to modern CPUs is pretty pointless as they all learned from the mistakes intel made with the P4 b ack in 2000? 2001? The heat was a byproduct of having to run a lot more power through the CPU to compensate for other issues than originally planned - to even get the speeds they achieved out of it.

If intel had hit 10 Ghz successfully with the P4 (i.e., the unforeseen problems with transistor leakage did not intervene), the desktop CPU market would likely look very different today.

Re:Awesome

[nateman1352]

I starting to fell like a broken record here but seriously Slashdot, I would have thought that this Myth that x86 ISA somehow makes it impossible to build a low power part would be dead by now. ARM is a rather complex instruction set now that they have added SIMD and floating point support. If you look at the number of op-codes it has versus x86 they are roughly equal. Both ISAs have variable length instructions (all recent ARM designs support THUMB) so the decode logic complexity is actually pretty comparable. Also much of the decode logic is implemented in software via microcode on both ISAs.

There is nothing magical about ARM that makes it lower power. The real reason why x86 implementations are so much hotter is because designers of x86 processors have been targeting high compute performance for decades, whereas ARM has been targeting low power for decades. A quick look at Medfield benchmarks show that it is comparable in performance to ARM processors that were current when it was released. Medfield is ~4W active TDP... same as a Exynos 5 Dual ARM CPU. From what we have seen from Merrifield/Bay Trail/Haswell the next gen x86 parts are continuing along this trend.

Re:Awesome

[c0lo]

Just because it was built by Samsung doesn't necessarily mean it has the Downs.

Whatever it has, whether you say Smart Fridge, burrito please or say nothing, it still goes.

Re:Awesome

[asola]

I wouldn't call absorption-cycle deprecated. In fact, I would definitely by an absorption-cycle fridge if it was available in similar size as my current fridge.

Absorption-cycle fridges can be made practically silent while the ones with compressors are very noisy (maybe not the top-of-the-line, very expensive ones but still). I have my fridge in the living room so it would be worth some premium for me to have a completely silent fridge.

Moreover, I have a huge solar-hot-water-collector array, which over-produces in the summer so I would run that fridge for free (if it had external input option)

Unfortunately, nowadays, only very small fridges can be purchased with absorption-cycle. These are designed for RVs and only have propane input.

Re:Awesome

[dj245]

...AMD went from 65w-95w with Athlon64 to now 140W+ and soon to be 220w and again they, just like netburst, are blinding trying to fix the fundamental flaws with the arch by simply cranking up the clocks.

It is far more likely they are selling the product they have while they try to come up with something better. My company is doing the same thing with one of our flagship products. The product isn't very good compared to the competition, so we apply various hacks and operation limits to try to get to their level. We don't do this because we think our product can be fixed this way. We do it so we can stay in business until we can build something better. You can't sell products that don't exist yet, and you can't stay in business very long without selling anything.

Re:Awesome

[Anonymous Coward]

No I meant running my freezer. Hence the reason I typed:

"I always wanted to have a computer running my freezer"

instead of

"I always wanted to have a computer running IN my freezer"

Oh. Then I don't get it.

Hey... It's not your fault Archer Bunker or whatever your name is. That guy's obviously some sorta fuckin' nerd — in order to get that that nerd's joke, you'd have to like, understand a buncha stupid nerd shit!

Any of you nerds tries another nerd joke like that again, and me 'n' Bunker's gonna start handing out free swirlies!

Re:Awesome

[Anonymous Coward]

I was under the impression that the long term idea was having the CPU handling the scalar/sequential workloads while letting the intregrated GPU taking of the vector/parallel stuff. So basically Bulldozer is a painful yet necessary transitional architecture, and the first REAL product showcasing that idea will be Kaveri with its hUMA (provided software doesn't lag too far behind).
I still think AMD is on to something really interesting here.
And if they don't, I'll just move over to ARM because there is not a snowball's chance in hell intel gets any of my money.

Re:Awesome

[Too Much Noise]

The thing is, you have it a bit backwards. With a refrigeration cycle, which is basically an engine cycle running backwards, you are moving heat from the lower temperature to the higher one. This requires you to supply work to the cycle - you're converting that work to heat, in fact, not the other way around as you seem to argue (and as an engine would do). It doesn't really matter what work source you use as long as it does what you need - in particular, you can use a separate heat engine system to generate it, going [kerosene wick] heat -> work, which is added to heat from inside the fridge -> heat outside the fridge, but modern appliances use electric pumps, as they scale better, are more efficient (your kerosene engine is quite wasteful) and power is more readily available.

Re:Awesome

[Stuarticus]

If your definition of the thing running your fridge is "computer" you are likely more than fifty years old (I Kid!).

Re:Awesome

[Wdomburg]

I suddenly want to re-purpose a classic Easy Bake oven as a computer case.

Re:Awesome

[Anonymous Coward]

The Bulldozer architecture is fundamentally broken

No, it is not. It is merely second-best. That isn't the same as "broken."

The only way to get reasonable performance from it is to...

..buy one. Any Bulldozer has reasonable performance. The shittiest Bulldozer you can get, kicks ass. The Phenom II kicks ass. The fucking Athlon II kicks ass. Nehelum and even Core2Duo kick ass. You would have fucking killed for this shit ten years ago, because no Pentium 4 or Athlon XP was nearly as good as these chips.

It's not broken. It just doesn't live up to your "I must have 20% improvement per year" entitlement.

People who modded up parent: you are spoiled little sissy assmunchers. I wrote 80286 assembly language (segments!!!), and could tell you plenty about what "broken" truly means.

Re:Awesome

[godefroi]

They're called "absorption refrigerators": http://en.wikipedia.org/wiki/Absorption_refrigerator

Re:Awesome

[dbIII]

I'm not sure since the cores just clocked down when idle ever since delivery when I installed vanilla Centos 5.* (2.6.18-308.13.1.el5). There's bound to be a kernel option to force it one way or the other but I don't have it as a boot option (just checked).

Re:Awesome

[redmid17]

I'm condescendingly responding to a stupid question. I assume that's why people thought it was funny. I certainly didn't expect it to be rated like it was.

Re:Awesome

[Anonymous Coward]

It's a shame do ruin and otherwise good post with a poor choice of adjectives. You could have said that much better and still make your point. Carry on, but slightly better next time.

Re:Awesome

[Anonymous Coward]

BS! Back then it was a clock rate race primarily with netburst along for the ride.

Now htf can you even say power management on that beast?

Performance parity my ass. Twice the cores and higher clock? Hth is that parity?

Re:Awesome

[hairyfeet]

Yeah it would look different...PCs would be the size of fridges to hold the giant AC unit needed to cool that sucker! And nothing in your argument changes the fact that if you take the most powerful P4 and then increase its marks just as if you had increased it to 10Ghz you would still be scoring LOWER than 35w chips do today.

When you are scoring a shitastic 440 marks at 3.6Ghz magically cranking that chip to 10Ghz isn't gonna give it 10,000 marks, its gonna give it 4 times what it is now which would be around 1350 marks. You are falling for the old MHz myth that raw GHz means shit when it don't, if you managed to get a Pentium 1 up to 20Ghz that wouldn't magically make it do more useful work than a C2Q even though the C2Q would only be running at 1/4th the speed because the C2Q would do more work per cycle than that hyper pumped P2 could EVER do. Again look at the links yourself, look up the Pentium EE chips, those were "golden chips" that were hand picked to be the absolute cream of the crop, the absolute best the P4 had to offer. How did they score? Worse than the chip in my netbook, and that is while belching out over 150w of heat.

Your argument basically boils down to "if you put a big enough motor in it a Pinto could outpull a pickup" while ignoring that there is a lot more to pulling than the motor. The insanely long pipeline made stalls a nightmare for the P4, it cranked out waste heat like you wouldn't believe, and the IPC on it was truly pathetic. You say it isn't fair to compare it to a modern chip? fine compare it to a chip from its time, the Athlon 3800 X2 on this chart. Notice how the P4 based chips are using damned near more power idling than the X2 did under full load? The IPC on those chips are just fricking terrible, I should know as I have one at the shop and its sooo damned sluggish and hot that I'm seriously thinking about replacing this Pentium D with a circa 2004 Sempron single core until I can get another Bobcat board in because the Sempron? actually more snappy than the Pentium D.

So I'm sorry friend but unless you think Intel could come up with some pixie dust to sprinkle on each P4 the extra GHz would have translated into a shitload of heat, useful work? Not so much. BTW just running that Pentium D running ONLY a browser for a couple of hours, even with flash blocked? Its at 143f and as i said it struggles with page loads worse than that old Sempron, its just not a good chip, the whole line just sucked.

Re:Awesome

[smash]

You have entirely missed the point. Of course chips of today are faster at the same clock - they have had an additional 12 years of development...

Re:Awesome

[smash]

TO clarify - i'm not claiming a 10ghz p4 would be competitive today. I'm claiming that a 10ghz P4 VARIANT in 2006 with tweaks to the original 2001 design, when put on a modern process, up against the other things around IN 2006, would wipe the floor. The plan was to have the p4 running at 10Ghz on the process tech available in 2005-2006. NOT what we have today in 2013.

Of COURSE modern CPUs have better IPC. They have billions of transistors more than the P4, bigger cache, much smaller process tech, etc. The playing field is NOT LEVEL. Put a 10ghz P4 up against a 2006 spec AMD chip (NOT SOMETHING FROM TODAY) and the story is entirely different.

IF that had worked, we wouldn't simply have 10Ghz pentium 4s today. We would have 10 ghz or higher clock CPUs that had additional execution units, multi cores, etc.

But clock speed did not scale any more with process reductions like it had since the early 80s. There was a wall at approximately 3-4ghz, and the current CPUs are of COURSE much better IPC because they were FORCED to be that way because ramping up clock and shrinking process simply no longer works.

And again, the heat the P4 had was a symptom of the CPU frequency scaling hitting a wall and intel having to ramp up power to get the thing stable at even 3Ghz. Recall that prior to the P4, CPUs were only hitting 1-1.3 Ghz. That the first P4 hit 1.6 quite easily (pretty sure people were o/c'ing them to 2+ ghz on release) was perhaps a misleading indicator of things to come.

Re:Awesome

[Anonymous Coward]

They are already going for ARM Opterons. See http://www.amd.com/us/aboutamd/newsroom/Pages/presspage2012Oct29.aspx.

Re:Awesome

[Nutria]

As a side note, I've always wanted to take an old mini fridge and turn it into a computer case.

How about a wet bar?
http://toyvax.glendale.ca.us/~vance/vaxbar.html

Re:Awesome

[toddestan]

The heat from even a modest computer would overwhelm the cooling capacity of a typical mini-fridge. Most likely you would end up cooking the components followed by burning out the compressor on the fridge (since it would end up running continuously in a vain attempt to cool its contents). On the upside, condensation would likely not be a problem.

Re:Awesome

[toddestan]

From what I've seen some higher-end newer fridges have replaced the knob that controls the temperature with a small panel with push buttons and a display. It's probably not even as "smart" as a $15 digital house thermostat but there you go.

Re:Awesome

[Anonymous Coward]

Worked fine 10 years ago.

Re:Awesome

[hairyfeet]

But again you are slapping ever bigger engines in a Pinto, you CAN NOT fix a design as broken as the P4, okay? I mean for fucks sake don't you think that Intel didn't want to piss away all the damned money they had spent on dies and design work for the P4? Of course they didn't but they realized that the entire design was a dead end and couldn't be saved.

The ONLY way your hypothetical chip could work would be if they threw away every defining characteristic of the P4...but it wouldn't be a P4 then, would it? The P4 was designed like a muscle car, big ass pipes, lots of cache, sheer brute force...and it didn't work! You really need to look closely at the P4 arch friend, there was A REASON why they kept adding bigger and bigger AND BIGGER caches to it, its because a cache miss was a fricking train wreck with a pipeline that long, the damned thing will stall every time you turn around, there is a reason why they went back to a pipeline that was less than half that of the P4, its because long pipeline chips don't fucking work well man!

Seriously do some research on the history of the P4, especially the last couple of years. Look at tejas, which was a reported 6Ghz P4 they had running in the lab, ready for the fab...they threw away ALL that work, why? With tejas they had a chip that was nearly 2.4Ghz faster than anything on the market! The answer was simple, with such a huge fricking pipeline that 6Ghz meant jack and shit and even with them putting a monster cache and cranking the living shit out of the speed they just couldn't fix what at its core was a fundamental flaw of the P4, the huge pipelines required to get it to scale.

Again we are seeing the exact same thing at AMD, only for them while Intel was all about speed above all for AMD its cores above all, but to squeeze that many cores per die and not have the chips be the size of a pack of smokes they had to primitive the shit out of the design,leaving pretty much only a primitive integer core with all the front end being shared between core modules...again it just doesn't work, they can crank the living fuck out of the clocks until hell freezes over, they WILL be curbstomped by the Intel and this is from somebody that builds AMD exclusively but I know a turkey when i see one, the P4 was a turkey and the BD/PD arch is a turkey.

Re:Awesome

[hairyfeet]

But they already have not one but two products that ARE good, they have Bobcat/Jaguar and they have Liano, and they also have the design for Stars and they have the K8.

The smart move would be to start cranking out Jaguar, Liano, and if its gonna take more than a year to get the new design up use the Stars design because the BD/PD design is expensive as hell and ends up with too many bad chips per run. I don't know if you know this but with that last Thuban chip they had the ENTIRE desktop market able to be covered by a single chip, giving them pretty much 100% yields which made those chips easily the most profitable they had made. If it came out very fast? it was a high end Thuban, those that didn't pass but had 6 good cores were the lower SKUs, if it had a bad core? it was a Phenom X4, bad cache? it was an Athlon X4. As a bonus it gave the value gamers a BIG reason to buy their chips because you had a damned good chance of getting a core unlock so even their lowest SKUs were selling briskly. Hell go look up the Sempron single and see how many bought it just to see if they could unlock to an X2!

This would give them Bobcat for embedded and cheap tablets, Jaguar for low end notebooks and convertibles, Liano for the midrange, and Stars for the desktop value to low midrange since they really don't have a high end chip that can compete. then just as Intel went back to the P3 mobile to make the Core design they could go back to the Athlon64 which was a great chip and with a little tweaking and the smaller processes it would probably be pretty damned competitive. But trying to throw good money after bad by sticking with a design that runs so hot and scores so poorly?

You can't stay in business if the product you are selling costs more for you to make than what people are willing to pay and the amount of bad chips they get with the BD/PD design means to make a profit they have to keep the price of the chips waaay too high. Up until the release of Bulldozer AMD pretty much owned all the "best CPU" lists in the under $200 market which is a pretty damned big market, after they phased out Stars for the BD/PD design? They haven't won a single slot, not one, even the sub $100 chip is now an Intel.

Re:Awesome

[smash]

What I'm saying is the only reason the P4 design is "broken" is because of the scaling problems encountered that were previously unexpected. It was deliberately designed that way with sound thinking behind it with the knowledge fo the day - because existign designs hit a brick wall at ~1.2-1.3ghz.. Unfortunately, the design didn't work as expected. Intel (and everyone else, but intel was on the bleeding edge) did not foresee the drastic power leakage problems approaching 3ghz.

Any CPU designer has to make some core design choices and they are all trade-offs. In the P4s case, the trade offs were IPC vs clock speed. Clock speed did not scale as expected, hence it was a dog. If the unforesee problems above DIDN'T happen, P4 would have scaled to 10Ghz, and no it would NOT have needed a refrigerator to cool it. The only reason they were hot was because of the above issue.

And yes, exact same thing with AMD today. They tried to automate their CPU designs to punch them out cheaper and more cost effectively than doing layout by hand. Different design choice, but it was still a risk. That didn't pay off as expected. Bulldozer is the result.

The thinking was sound, the execution, and reality differed from the theory.

Unfortunately CPU designs are a major investment, and like intel in the early 00s, AMD now have to live with it until they can spend the time and money to change direction. Hopefully, AMD have what it takes to weather the storm. I'm not so sure they do though.

Re:Awesome

[jones_supa]

You have a point there, although even that is possible to whip together with a chip that is not a CPU.

Re:Awesome

[hairyfeet]

But you are missing the point, which is why the BD to netburst comparison is apt, in that both chips come from a flawed premise that you could throw a trillion dollars at and it will NEVER work, it would be like saying if you spent enough money on the wings you could flap your arms and fly to Paris, its just never gonna work.

And you are 100% correct in that a designer has to make choices, but that is NOT what happened in either case, unless you consider "throw out everything but a single metric" to be a valid design choice, which as we saw with the P4 and now with the BD it just doesn't work. in the case of the P4 ALL that Intel cared about was the clocks, that's it. They ignored IPC, heat, power, all was thrown by the wayside so that they could have the absolute fastest clock out there but who cares what the clock says it is if you aren't able to use those clocks for useful work?

You are saying with enough time and money they could make it work and I'm pointing out that is impossible because it ignores reality. that is the same problem Intel had with itanic, itanic was a monster chip IF and ONLY IF the compiler was able to line up the instructions on every program perfectly so that it would feed the chip in the way that the VLIW design required, what I'm saying is look at reality. IRL they couldn't come up with a compiler that could feed those instructions perfectly no matter what the code was for, the same goes for the P4 in that a SINGLE cache miss just slaughtered its IPC yet due to the nature of how the majority of programs behave there just isn't enough cache in the world to make sure you never have a cache or branch miss and THAT is what made the P4 ultimately a dead end, there is no way to fix that fundamental flaw.

I mean why do you think that lowly primitive C60, with its weak dual cores and wimpy 1Ghz core speed is able to get a higher score than a P4 running nearly 4 times faster than the C60, even though Passmark is probably compiled with ICC which tends to give the P4 an advantage over AMD chips? Its because IRL you ARE gonna have cache and branch misses and thanks to the much shorter pipeline the C60 doesn't really lose any IPC whereas with the P4 you might as well tie a boat anchor to it if it has a miss. Hell the whole reason they came up with hyperthreading is so they could have a second line going when the first one inevitably stalled!

So there is just no way they could have ever fixed the P4 without throwing out what made it a P4 in the first place, which is those crazy long pipelines. at the end of the day programs simply don't behave in a way that would give the P4 what it needed to get useful work out of the thing, real life just doesn't work that way. You could take any P4 and get them up to 10Ghz and it would still throw on the brakes the second it had a cache miss, that is just what you get when you build pipes that long.

Re:Awesome

[smash]

The p4 designers didn't "throw everything out" for a single metric at all (look up the design features of the P4). Yes, they chased clock-rate at the cost of IPC but the heat and power were not design choices they were unexpected consequences, which is the point I have been trying to make. If it had scaled as expected, the 75% efficiency vs say the P3 in IPC would have been more than beneficial due to the 10x ramp up in clock speed. The P4 designers DID have a choice, and they MADE that choice based on (at the time, since disproven) sound reasoning. That being that if they went after Mhz, they could scale high enough to make the IPC trade off worth it. And what they lost in IPC would have been somewhat made up for with the simpler parts of the chip running much faster due to the higher clock.

You are again, either deliberately misunderstanding my post, or we're just talking about two different things - you're attempting to argue against a point I am not even trying to make.

I'm not saying the P4 wasn't a massively flawed design. It was. I'm saying that it wasn't originally designed to simply reach .8, and be hot doing it.

The heat wasn't a design intention. It was expected (prior to manufacturing) to take until 10Ghz to reach those temps

I'm not saying that if you were to give it minor tweaks it would be competitive wiith a core2 today. It wouldn't. Time moves on.

I AM saying that if it met the original design spec and scaled as expected, it WOULD have blown AMD out of the water. It wasn't THAT far behind in IPC when compared to the AMD CPUs of the day. If scaling to 10Ghz was possible (and again, prior to the P4 failing so spectacularly, it was EXPECTED), we wouldn't have had such a tremendous focus on IPC and multi-core in the last decade.

Unfortunately it was optimized for things that were not achievable in reality. Before the P4 hit the wall at ~3ghz, the CPU manufacturers didn't expect it to be quite so hard to scale further.

2013 AMD has a message for 2005 AMD

[CajunArson]

The message is: You got the Megahertz myth wrong! The only myth is that Megahertz isn't important!

Oh, and all that performance-per-watt stuff? You might want to walk that back. Oh and, pull those Youtube videos where you accuse Nvidia users of being fake-pot farmers because their cards pull so much power. Sure it was funny at the time, but we'd rather not have to live that one down now.

Re:2013 AMD has a message for 2005 AMD

[jellomizer]

The problem is they are so many ways to judge performance.
GHZ are good for comparing like processors.
MIPS are good for similar instruction sets.
FLOPS are good for similar code (That uses floating points)

You then add these per watt if you want to show it off for a mobile device.

Re:2013 AMD has a message for 2005 AMD

[bored]

You got the Megahertz myth wrong! The only myth is that Megahertz isn't important!

Tong in cheek and all that, but....

Frankly, today both AMD and Intel are at an IPC wall nearly as much as they have been at a clockrate wall. So, yes faster clock rate is pretty much the only way to get performance if your application doesn't scale with cores. Which is a shocking number of them.

The part I find interesting, is that if they can beat the haswell with this part then they probably have an IPC advantage over intel again. Remember the top end haswell turbo boosts to 4.9Ghz.

The only real difference then is how much power they take doing it. Which may not be as significant for an average system that is idle or running without the turbo boost for a significant portion of the time.

If AMD is within a few percent in the single thread area (especially on specint) and the thing is priced like a traditional AMD then they my next machine will probably be something like this.

Re:2013 AMD has a message for 2005 AMD

[CajunArson]

You sed: "The part I find interesting, is that if they can beat the haswell with this part then they probably have an IPC advantage over intel again. Remember the top end haswell turbo boosts to 4.9Ghz."

Please re-read everything you just said very very carefully. Especially the parts about how a design with a known IPC will magically get huge IPC boosts by only increasing the core clock and power draw (hint: it won't). Please also remember that Haswell only has a 4.9GHz boost speed for incredibly small values of "4" (it's 3.9GHz, not 4.9GHz).

Re:2013 AMD has a message for 2005 AMD

[TubeSteak]

The first 4 GHz are easy.

Here's Ivy Bridge chips pushing ~220 Watts to reach 4.7 GHz
http://www.legitreviews.com/images/reviews/1924/power-consumption.jpg

/AMD's FX-9xxx series uses a 32nm process.
/Intel's Ivy Bridge and Haswell use a 22nm process.

Re:2013 AMD has a message for 2005 AMD

[bored]

My point was, that if they can match or beat a 4.9 Ghz part with a 5Ghz part, then the IPC's are going to be similar.

I didn't say the IPC changes with clock rate...

Re:2013 AMD has a message for 2005 AMD

[CajunArson]

That's like saying if Intel increases it's IGP performance by a factor of 10 then AMD will have to worry... of course it would, but the whole problem with that statement is the pesky word "if"

My 4770K is overclocked to 4.6GHz without that much tuning right now, and I guarantee it beats these new parts even in the perfectly multithreaded synthetic benchmarks that are best-case scenarios for AMD. It does that without being a space heater, and if the rumors about prices are true, the 4770K is an outright bargain to boot.

Re:2013 AMD has a message for 2005 AMD

[realityimpaired]

It'd be interesting to see the numbers on my own system, I suppose... I have an i5-2500k (sandy bridge) that I bought right when ivy bridge came out, which I have clocked at 4.8GHz. That's running with a Cooler Master Hyper 212+, and doesn't exceed 65'C even when transcoding blu-ray/dvd videos to h.264... I've kept it pegged at 100% CPU for 6 days straight without exceeding that temperature....

I'd be surprised if that heatsink could provide >200W of cooling, given that it's a big radiator with a fan. To be fair the case has a *ton* of airflow (Antec Eleven Hundred), but I'm still doubtful that it's capable of providing *that* much cooling... Cooler Master's own specs say it can only handle about 180W of TDP....

Re:2013 AMD has a message for 2005 AMD

[CajunArson]

Did you bother to read that graph? Try looking at the bottom where it says "Wattage At the Wall"

You must be an enormous Intel fanboy to think that they have invented technology that allows every single component in the whole computer outside of the CPU to consume zero power in highly-overclocked systems....

Re:2013 AMD has a message for 2005 AMD

[Jah-Wren Ryel]

> You then add these per watt if you want to show it off for a mobile device.

Or any data center.

Re:2013 AMD has a message for 2005 AMD

[jones_supa]

The problem is they are so many ways to judge performance.
GHZ are good for comparing like processors.
MIPS are good for similar instruction sets.
FLOPS are good for similar code (That uses floating points)

Of those, I think GHz is used way too often, while it actually has lost much of its meaning these days. For example we've had 2GHz desktop CPUs for a decade now, but the performance difference between them can be worlds apart.

Re:2013 AMD has a message for 2005 AMD

[viperidaenz]

You missed the parent's comment that Haswell only boosts up to 3.9GHz, not 4.9GHz.
If you can match a 3.9GHz part with a 5GHz part, well, that proves nothing about IPC - unless you can't beat it by a significant margin, then it proves IPC is lower.

Re:2013 AMD has a message for 2005 AMD

[Skarecrow77]

describing cpu speed in ghz is like describing engine speed in rpm. it's technically accurate, but tells you nothing at all about what the product can actually do for you unless you're comparing two different examples of the exact same architecture.

Re:2013 AMD has a message for 2005 AMD

[smash]

That. Increased power = increased battery, UPS, cooling, power bills, carbon taxes and physical space to install it all. All of which are expensive. The most important criteria in a modern datacenter is performance per-watt.

Re:2013 AMD has a message for 2005 AMD

[Anonymous Coward]

I'll buy that GW is dangerous when ALGORE sells his beach house and carbon-neutrally composts his $100*10^6 from Qatar.

WTF? You base your acceptance of the science based on what Al fucking Gore does? Are you retarded?

Al Gore could burn a million rubber tires, or personally bicycle-power carbon sequester a billion tonnes of CO2, it doesn't matter; it's all about the science. Even those of us that think it's a serious issue don't give a shit what Al Gore does. It's not about him.

Fucking fuck, what the fuck is wrong with you?

Re:2013 AMD has a message for 2005 AMD

[Yvanhoe]

Up to a certain point. Actually data center care about performance per watt because it actually means performance per dollar. However, a small saving in watts does not necessarily justifies a tripling of the price for instance.

Also, some applications benefit hugely from being able to run on a single core or even a single CPU. Saves a lot of complexity and internal bandwidth.

Running a datacenter efficiently does not depend on a single factor.

Re:2013 AMD has a message for 2005 AMD

[smash]

To a point, sure. But purchase price would have to be massively different to have a significant impact on TCO. If we were to save say, 50% on compute costs, but it then took 2x the power, cooling and floorspace, it would be a net loss. Purchase cost is a one-off. Power, cooling and space costs are ongoing, and trending upwards.

Re:2013 AMD has a message for 2005 AMD

[NoImNotNineVolt]

Considering that engine output power is the product of the torque and the engine speed (rpm), it would seem that it's quite an informative measure after all.

This is why the Honda F20C engine is an impressive feat of engineering. The Honda S2000, powered by the F20C, "held the record for the highest horsepower per liter for any NA piston road production car for 10 years until the introduction of the Ferrari 458," according to Wikipedia. The F20C would redline as high as 9200rpm, which is part of the reason why 240hp could be squeezed from a 2.0L naturally aspirated piston engine. Compare this against the "5.0" that powered the iconic third generation Ford Mustang, a massive 4.9L naturally aspirated piston engine that only puts out between 140hp and 225hp (depending on the model year) due it its pathetic redline under 5000rpm.

It seems that CarAnalogyGuy is on vacation or something :P

Big deal

[Anonymous Coward]

Power 6 was running at 5.0ghz 5-6 years ago.

Re:Big deal

[Anonymous Coward]

And it seemed to run 4 times slower than old opterons on all the operations I saw it run. Awesome.

Re:Big deal

[cold fjord]

Power 6 is quite a bit more expensive per processor & system.

Power 6 is corporate, FX-9590 is power to the people.

Re:Big deal

[aliquis]

Yeah.. power at least ..

Global warming is coming.

Re:Big deal

[Anonymous Coward]

Guess you never saw one then, because that is complete bullshit.

Re:Big deal

[cold fjord]

House warming, certainly.

Global? Maybe.
What happened to global warming?
Curry: less warming than predicted. Models seem wrong

Re:Big deal

[Jah-Wren Ryel]

> Power 6 was running at 5.0ghz 5-6 years ago.

And Power6 sacrificed out-of-order execution to do it. It was also only 2 cores instead of 8

Re:Big deal

[Anonymous Coward]

Don't be silly. At the time it was the worlds fastest processor.

http://www-03.ibm.com/press/us/en/pressrelease/21580.wss

"Demonstrating its remarkable versatility, the new IBM System p 570, running the POWER6 processor, claims the No.1 spots in the four most widely used performance benchmarks for Unix servers – SPECint2006 (measuring integer-calculating speed common in business applications), SPECfp2006 (measuring floating point-calculating speed required for scientific applications), SPECjbb2005 (measuring Java performance in business operations per second) and TPC-C (measuring transaction processing capability)"

And when Power8 comes out it will likely smoke anything out there. In about a year and a half.

Poor AMD...

[Anonymous Coward]

That 220W figure is astonishing and makes me feel bad for AMD. They are getting kicked in the balls not because of any merits or demerits of their design but simply because they don't have access to the advanced process technology that Intel does.

Re:Poor AMD...

[Anonymous Coward]

That 220w figure is not correct, that is why it is so hard to believe. There has never been a cpu with a TDP that high, and this one won't be either. I would guess 125w or 140w. The 220w number is probably total power consumption, which won't be anywhere near 2.5x more.

Re:Poor AMD...

[kdogg73]

Then along comes Apple...

Re:Poor AMD...

[Anonymous Coward]

There is no way that cpu has a 220w TDP. The figure shown is probably total system power consumption, which won't be anywhere near 2.5x more.

Re:Poor AMD...

[Khyber]

"That 220w figure is not correct, that is why it is so hard to believe. There has never been a cpu with a TDP that high,"

Um, I think the Itanium II MX2 actually got higher than that, with a TDP of 260.

Re:Poor AMD...

[synapse7]

If I had the time and money I'd buy it just to see what happens.

Re:Poor AMD...

[elwinc]

True, no CPU chips draw 220 watts. But that doesn't mean it's impossible to cool a 220 watt chip

The Nvidia Titan GPU card, with a 7 billion transistor chip at its heart, draws an additional 236 watts when it goes from idle to full load. It's not hard to imagine 200 watts feeding into the GPU chip. Other GPU cards on that page draw even more power than the Titan. The Radeon HD 6950 CFX card drew 329 watts. It's not hard to imagine the chip at its heart drew over 220 watts.

If you want to cool a 220 watt CPU you might need water cooling, but it's by no means impossible.

Re:Poor AMD...

[robthebloke]

yeah, this article is guessing at about 125w....

Re:Poor AMD...

[sjames]

That makes a lot more sennse and is entirely in-line with some of Intel's Xeon line at 120-150W.

Re:Poor AMD...

[Anonymous Coward]

the 220W is pure speculation!
  from the article :(em is mine)

The top end product is the FX-9590 which has a top turbo speed of 5 GHz. This will be a full four module implementation with the 8 MB of L3 cache. AMD did not give any other details for this particular part. We do not know what the base clock is, we do not know what the TDP is, and we can only assume that the northbridge/L3 cache will be clocked at the standard 2.2 GHz that we have seen on previous Vishera parts.

Re:Poor AMD...

[ltwally]

According to Wikipedia, AMD is worth $4.5b. Possibly more. Perhaps Apple could convince their shareholders to take less. But we'll call it $4.5b for our purposes.

You think Apple wants to spend that much money to acquire a microprocessor company? A microprocessor company that doesn't even have its own fabrication plants? A microprocessor company that is noticeably lagging behind their main competitors: Intel and nVidia? Whatever your feelings towards AMD, you cannot refute that their market share has been on a decline the past few years, and that the Bulldozer lineup has not been able to resuscitate them.

About the only truly positive aspect for Apple would be that they would also get the ATI assets as well. But that's a double-edged sword. What if the ATI lineup slides? As things are, they can easily switch to nVidia GPU's. If they bought out AMD, they'd have little choice to be to stick with ATI gpu's no matter how good or bad things got.

And let's not forget, there are certainly some folks at Apple that were around for the joys of the G5 series -- another processor that was effectively a space-heater. They had problems with that, and took some flack for that. I imagine they'd like to avoid that unpleasant memory.

Personally, I would be shocked if Apple wanted to spend $4.5b, end a successful relationship with Intel, only to acquire a less efficient and often less powerful CPU lineup without acquiring a chip foundry as well. If there was the fabrication plant in there, then perhaps they could use it to make their own ARM chips for their phones/tablets. But they don't even get that.

Re:Poor AMD...

[bws111]

There are CPUs that draw that much. IBMs EC12 draws about 300 watts.

Re:Poor AMD...

[Anonymous Coward]

For the past decade, AMD has been releasing "Average" TDP draws while Intel has been releasing "Maximum" TDP draws. Which is why AMD can burst past it's rated TDP for a little in many real-world situations, while Intel CPUs will almost never reach peak, except when using hand-made ASM that can work every execution-unit at 100%.

One thing that AMD has been doing quite well is total system power consumption. Intel typically beats AMD in actual CPU draw, but then loses its edge once you include the chipsets/etc.

Re:Poor AMD...

[serbanp]

... because they don't have access to the advanced process technology that Intel does

That's a wrong way to put it. It's their own process, they paid for it, that's why they have access to it.

Intel's process is at least two generations ahead of everybody else because they understood long time ago that technology alone can crush the competition and decided to pour an insane amount of money into creating the said forefront technology.

What did AMD do? Become a fabless chip maker, at the mercy of the likes of TSMC or GF...

Re:Poor AMD...

[NatasRevol]

They already bought PA Semi. They don't need AMD.

They need to buy TSCM to get their own fab plants and get out from under Samsung's nose & thumb.

Re:Poor AMD...

[Kjella]

According to Wikipedia, AMD is worth $4.5b. Possibly more. Perhaps Apple could convince their shareholders to take less. But we'll call it $4.5b for our purposes.

That's the balance sheet, in practice the market cap is 2.8 billion - right before Christmas it was about 1.4 billion. At any rate, AMD's technology sucks at power efficiency which makes it a horrible match for all the mobile devices (iPhone, iPad, MacBook Air, MacBook Pro) that Apple wants to sell. Even trying to make "fashionable" non-mobile products like the Mac Mini, iMac or the new Mac Pro would be very much harder with an AMD processor. If you don't mind a big case, big heatsink and big fans AMD will get the job done but it's totally the opposite of everything Apple stands for.

Re:Poor AMD...

[Anonymous Coward]

>What did AMD do? Become a fabless chip maker, at the mercy of the likes of TSMC or GF...

Be serious. AMD didn't have enough revenue to compete on process technology research even when they still had their own foundries.

Re:Poor AMD...

[Shinobi]

One thing that AMD has been doing quite well is total system power consumption. Intel typically beats AMD in actual CPU draw, but then loses its edge once you include the chipsets/etc.

That is not true any more, and hasn't been since Sandy Bridge was released. Since then, the total system power consumption figures has been in favour of Intel except in the extreme low-end, such as against AMD's E-350. However, if you intend to for example build a small server or something for your home, you're better off with a SB/IB low-power Pentium. Only somewhat higher total power draw, especially if you slap in a passively cooled GT220 or something, and much better CPU performance(MUUUUCH better, because the E-350 is a steaming pile of crap in that regard)

Re:Poor AMD...

[serviscope_minor]

What did AMD do? Become a fabless chip maker, at the mercy of the likes of TSMC or GF...

They had little choice.

Chip making on the top end is getting increasingly expensive, which means it will be consolidated into fewer and fewer fabs. Since intel is much richer than AMD and also have pushed their process further, AMD was always going to lose their chip fab before intel.

Actually, given that intel has the leding chip fab, the chances are they'll start fabbing for others and will be one of the few that ever remain.

There's just no wy that AMD at its size could have ever continued down that path.

Re:Poor AMD...

[Kartu]

AMD has great GPUs for quite a while now. 5xxx, 6xxx, 7xxx all been good to excellent.
So saying it is lagging behind nVidia is unfair.
Their marketing department could have done better, yes, sure.

AMD has unique product, their APUs that will power upcoming consoles (heck and even SteamBox opted for it) and with games being optimized for their architecture AMD will get quite a boost. I wouldn't expect them to dissapear any time soon, neither Microsoft nor Sony would allow that to happen. (and as far as I remember, 4 billion $ was the sum Sony wasted on Cell alone)

They are also improving on IPC/power front, shortening the gap to Intel.
Raw CPU power is becoming less and less important.
Even today, with all the woes, you get more performance per buck if you go AMD CPU/GPU.

So

[maroberts]

I've reduced my power consumption by replacing all my light bulbs with LED versions, only for my computer to negate the savings I've made....

Re:So

[jedidiah]

ppppffffft.

I had to install a separate air conditioner to cool my computers. It's like I bought a 100lb CPU cooler.

Turbo Core

[Anonymous Coward]

Is the turbo button making a comeback?

Re:Turbo Core

[Mike Frett]

If you put it in an ASUS Mobo, in the UEFI there are 3 buttons: Power Saving, Normal and Performance; Performance being the 'Turbo' mode. I accidentally had mine in Turbo mode when I assembled my Computer, then dropped it back to normal once I knew what was going on; it was a noticeable difference. By the way, it auto-overclocks to the Turbo mode, requiring you to only turn on Performance mode.

This post should be deleted.

[Anonymous Coward]

There is no way this cpu has a 220w TDP. I can't believe a website as reputable as slashdot would post such utter nonsense. That figure is probably total system comsumption, which won't be anywhere near 2.5x more.

Re:This post should be deleted.

[johkir]

I agree. All the chips in the FX TDP arena are 95-125W. I'm thinking it'll be more, but not almost double. More in the 140-165W range. AMD is going for the 5 GHz media sensation buzz, not power buzz.

Re:This post should be deleted.

[Khyber]

Scroll down to Hondo MCM (The Itanium II MX2.)

See that 260W TDP?

Re:This post should be deleted.

[Hatta]

I can't believe a website as reputable as slashdot would post such utter nonsense.

Welcome to Slashdot.

Re:This post should be deleted.

[Anonymous Coward]

Tough choice. I'll just buy both for winter.

Re:This post should be deleted.

[CajunArson]

AMD Fanboy: This Post should be deleted because it doesn't agree with my pro-AMD RDF.

Can we please have another article about how Haswell is such a failure instead? Oh, and can you please not do any benchmarks between Haswell and any AMD products whatsoever except for a couple of IGP benchmarks that we will pretend represent the only types of systems gamers care about? K-thanks!

Re:This post should be deleted.

[tibman]

sooo, are you saying he's wrong or just attacking him?

Re:This post should be deleted.

[Anonymous Coward]

You're an idiot. I would still reccommend that people go with an Intel processor. I'm just saying the TDP of this new AMD cpu will not be 220w.

History repeating itself?

[Carnildo]

Why am I having flashbacks to the Pentium 4?

Re:History repeating itself?

[mdielmann]

Awesome sig - how appropriate.

Re:History repeating itself?

[intermodal]

Probably because the cores aren't really cores. They're four cores that are basically hyperthreading.

Re:History repeating itself?

[ericloewe]

It's hard not to. Intel wrote the book on "best way to screw-up a microarchitecture and let your competitor gain an advantage", which they have been taking into account since dropping Netboost. Now comes AMD and follows that very same book quite closely...

Re:History repeating itself?

[gagol]

more like 8 int cores and 4 shared fp cores...

Re:History repeating itself?

[Anonymous Coward]

Heat generation coupled with wattage consumption.
This was a killer to Intel losing its leverage in the Gigahertz wars with AMD back then.
That was the point in time AMD stomped all over Intel, when AMD was at the top of it's game.

The difference now, is not so much a multi-core cpu world, but how financially secure AMD is to weather through a similar maneuver if it were to fail.

Intel obviously made it through, however I am not so sure for AMD.

Re:History repeating itself?

[aztracker1]

Yeah.. I am not completely sure if this one is a good idea... I know that I really like my 8350 because of really parallel usage.. but that's not what most people need out of it. I also know that things will probably get pretty interesting in the future... Like others have mentioned, AMD letting their brightest engineers go a few years back was probably one of their stupidest mistakes. The trouble getting to next gen manufacturing is another big issue.

Re:History repeating itself?

[Bert64]

The problem was that despite AMD having a clearly superior product, most people still stuck with intel giving them enough money to recover and produce competitive products. The several years during which amd were clearly superior should have given them enough profit and momentum to keep going.

Re:History repeating itself?

[Petersson]

Compared to this power pig, Netburst Pentium 4 was a cute little piggy.

"Performance should closely match"

[somarilnos]

The summary suggests that the "performance should closely match the recently released Intel Core i7-4770K Haswell processor", but nothing in the article, or anything released about this chip so far, supports that. It's all just guesswork until we see some actual benchmarks from the chip.

I don't honestly expect we're going to be seeing performance parity from this chip (although I'd love it to be true). But that hasn't been AMD's selling point for me for a long time. Chances are, we're going to see a chip that breaks the 5.0 GHz barrier, under-performs relative to Intel's top end chip, but costs about half as much. That's been their game for a long time now, and I haven't seen anything that leads me to believe that this chip is changing that.

Re:"Performance should closely match"

[Animats]

The summary suggests that the "performance should closely match the recently released Intel Core i7-4770K Haswell processor", but nothing in the article, or anything released about this chip so far, supports that. It's all just guesswork until we see some actual benchmarks from the chip.

If they're just cranking up the clock speed of an existing design, the performance should be quite predictable. The difficult-to-predict thing is the lifespan of the part. Atoms migrate faster as heat and voltage go up.

The limit on clock speed today is from heat dissipation. AMD got 8GHz out of a CPU a few years ago by cooling with liquid helium, but it's not worth the trouble.

220W Peak

[Anonymous Coward]

If the thing idles low, who cares? Your high end GPU already pulls more power than your CPU. If AMD convinces motherboard manufacturers to build boards that handle 250W, they will certainly release a high performance APU that could replace a lot of discrete GPUs. This is the longer term strategy that people miss when they see that high TDP.

Re:220W Peak

[hairyfeet]

Never had one of the Intel Pentium D 9xx chips have you? Sure with speedstep they would idle nicely but every time you actually tried using the thing the temps would just shoot through the roof and if you wanted decent performance out of the system you pretty much had to OC. Having the CPU alone hitting 140F even with a cooler with a copper center? Not much fun, especially in the summer.

I had a box with the Pentium D 805 and just to see how big a difference it would make swapped it for the 915, know what I found? It wasn't even worth the whole $8 difference to swap the 805s for 915s because while idle naturally was lower, if the system is gonna be idling all the time? Frankly you might as well turn it off. When you began to actually use the thing the temps would shoot back up to the same as the 805 so it really wasn't worth the bother to switch.

Low idle temps is all well and good but most of the things you are gonna be doing won't leave the system idling and THAT is when those hotter chips really become a PITA.

Do not want

[Anonymous Coward]

Too much cooling. Too much heat. Too much power. In that order.

Whatever you might imagine using this for I will do without and stick with whatever Intel makes at 65-75w TDP, tyvm.

AMD slower / MHz

[SpaceManFlip]

you're probably right - I was slightly shocked recently when I compared the performance benchmarks of an 8-core AMD to a 4-core Intel. I saw the 8-core on sale for about $179 and thought "wow!" but then I was more like "wow...." after seeing the benches.

basically, the 8-core AMD was slower performance-wise the 4-core Intel with the AMD running a few MHz faster

Re:AMD slower / MHz

[hairyfeet]

Yeah but how much was the 4 core Intel? And you can probably buy that 8 core for $150 or less now if you watch the sales. I'm running the Thuban X6 and what did 6 cores cost me? $105 shipped, if you compare like to like the only chip I could get from Intel at $105 was the Pentium Dual core which the X6 outperforms so in that case the bang for the buck squarely landed in the AMD camp.

The problem with the X8s (well other than the arch, see my previous post with a link on why the BD/PD/EX platform is AMD's netburst) is they simply cost too much to make, for every X8 that comes out with all functioning core they probably get 2 dozen X4s or X6s thanks to bad cores so THAT is where the bang for the buck is, although if given a choice I'd take a Deneb or Thuban over Bulldozer any day of the week.

But if you are strictly wanting the most bang for your bucks and like most of us don't have unlimited budgets the best bets would probably be the Athlon X4 for $67 although for an extra $8 I'd probably go for the Phenom II X4 for $75 and for more than 4 cores the best bang is probably the FX6100 for $99 or the Phenom II 1035T X6 for $106. I think in the benches the Thuban beats the FX6100 but both are good deals. Nice thing about the 1035T is I have one and have sold several and with a low end gaming board like the Asrock boards they have a hell of a lot of OCing room, before deciding I didn't want to deal with the temps I had mine up to nearly 3GHz with a turbocore of nearly 3.5GHz. I probably could have gone higher with a better cooler but my apt gets hot enough as it is without adding a major OC to my system.

As you can see though you can still get crazy cheap deals on the AMD side if you just know where to look. These chips have more than enough power to do anything your average person is gonna want to do with a PC, heck my youngest is gaming on a 3.4Ghz Athlon X3 and is quite happy with the performance and with my 1035T I can game AND do a transcode AND burn a DVD at the same time with no slow downs so I would say I'm getting my $105 out of it.

Re:AMD slower / MHz

[Omestes]

basically, the 8-core AMD was slower performance-wise the 4-core Intel with the AMD running a few MHz faster

Take all benchmarks with a grain of salt. While Intel has been generally winning for awhile now, that doesn't really mean AMD is completely inferior. With like chips there are certain things a modern AMD will out-perform Intel on, such as single threaded tasks. Intel will generally smoke AMD on multithreaded tasks, though. There is also cost, while AMD might be 10% less benchmark happy than a like Intel chip, it generally is over 25% less expensive, and will generally run without need to buy a new costly motherboard.

My last big upgrade, several years ago now, the price difference between the AMD (Phenom II 4x 965 Black) and Intel (i7 something or another) was hugely dramatic, considering the fact that I'd need a new motherboard and new RAM on top of the CPU. It was about $300-400 difference (fully upgrading 8gb of RAM, with a new mobo). I took the 10% performance hit, happily. For enthusiast CPUs, you'd best take the hit, and use the cash on a better GPU.

Re:AMD slower / MHz

[bored]

What benchmark?

I've been running a bunch of single thread in house application benchmarks under linux using GCC and a generic x86_64 target.

The results of AMD machines (4300's) vs the Intel (E5's) machines are not reflected in the benchmarks being run at places like tom's hardware. Those benchmarks are pretty split, AMD wins some, Intel wins some but the price difference between the machines is about 3x. Oh and the AMD machine stomps the intel machine when all the cores load up. But that could have something to do with the fact it has 2x the cores.

Just as an example try running nbench...

Of course if I use ICC, I can shift the results one way or the other by 50% or more. And its not always the way one expects. I was trying to speed up a Reed-Solomon calculation, and I optimized it for the specific intel I was using, and it doubled in performance. Then I ran the same code on the AMD and it doubled too. The result was the AMD still won.

So, there are lies, damn lies, and benchmarks.

Re:AMD slower / MHz

[Anonymous Coward]

Intel are hugely inefficient at multicore performance in the x86 line.

So you'd really be better off checking off the 4-Core AMD rather than the 8-Core AMD if you want to compare performance on metrics designed for or predicated on Intel. Alternatively, get the cheapest 8-Core Intel.

Re:AMD slower / MHz

[Kjella]

The problem with the X8s (well other than the arch, see my previous post with a link on why the BD/PD/EX platform is AMD's netburst) is they simply cost too much to make, for every X8 that comes out with all functioning core they probably get 2 dozen X4s or X6s thanks to bad cores so THAT is where the bang for the buck is, although if given a choice I'd take a Deneb or Thuban over Bulldozer any day of the week.

None of them are good bang for the buck for AMD, the FX-8150/8350 is a big chip of 315 mm^2 versus 216 mm^2 for Sandy Bridge, 160 mm^2 for Ivy Bridge and 177 mm^2 for Haswell. Granted the last two are on 22nm but even the 32nm Sandy Bridge was way smaller than AMD's chip, which means more chips per wafer and lower defect rates. And Intel is planning to move to 14nm next year, so there's absolutely no chance of AMD closing any gap, at best they avoid widening it.

Re:AMD slower / MHz

[Anonymous Coward]

Apparently mods are high; I see nothing either "Troll" or "Flamebait" about this post.

Re:AMD slower / MHz

[Omestes]

Probably just either Intel or AMD fanboys being mad because I didn't heap unequivocal praise on whatever they like, or scorn on what they don't. I probably should have just posted "Meh. Buy what works for you."

I'll probably survive.

Re:AMD slower / MHz

[aztracker1]

If you're doing software dev, or other workstation tasks that can use 8 integer units the FX-8350 is pretty damned nice, and hard to beat for less than 50% more on the CPU side... Once you have a couple databases running in the background, a backend service, front end service, dev tools, watcher, etc.. the 8350 is a beast.. though, on gaming, I put a bit more on the GPU, and though I don't game much, it works pretty well.

Re:AMD slower / MHz

[Omestes]

I've been looking at the 8350 for my next upgrade, but sadly it is finally time for a new mobo and RAM. Its been awhile since CPUs were the big bottleneck in gaming. My 965 is perfectly adequate, to be honest, it might not be when things are being optimized for the next console generation though. RAM isn't terribly important anymore either, since it is dirt cheap, and I rarely utilize my full 8gb outside of some games, and and some image/video editing tasks. GPU and HDD are the biggest improvements now.

Granted I'm a rather light user, not doing big databases, servers, compiling, etc... Heavier than average, but still not as heavy as some people around here. For average, there still really isn't a reason to get more than a $50-100 dual core chip, 4gb of RAM, and integrated GPU. AMD should be slaughtering that market.

Re:AMD slower / MHz

[Xest]

Right, but the reason I moved away from AMD is that yes whilst the chip is cheaper, you end up spending way more on cooling, power supply, and power usage over time, and that's really the problem.

It's the TCO of going AMD that's the real problem.

Re:AMD slower / MHz

[hairyfeet]

How so? The highest wattage of any part I named is 95w and since they all have C&Q they are all gonna use far less on average. As far as PSU I only have a 600W and the only reason i went that big is that my board supports Crossfire so I wanted to have that option down the road but i could have easily went with a 400w for this X6, maybe less. Finally as for cooling I always use hyper N520s so that wouldn't change no matter what chip i used.

The ONLY way the Intel works out to be a better deal is if and ONLY if you are doing a task that requires so much IPC that nothing less than the fastest possible chip will do, like the guy that was on here the other week doing impact simulations for guard rails. In that case he actually does need the fastest possible chip he can get because he is using every cycle for his work, but for the other 90% of the population that aren't slamming their CPUs to the absolute limit for 8 hours a day? Then even that 65w Athlon X4 will work just fine and you could build a whole system for less than an i3 dual core system with less memory, HDD space, etc.

Re:AMD slower / MHz

[Xest]

"How so? The highest wattage of any part I named is 95w"

That's still higher than 65w on most equivalently priced Intel chips and the performance gap isn't as large, so ultimately you're still using more watts to gain additional performance than you are relative to Intel's offering and that does add up.

Simply arguing that cooling is irrelevant because you'd use the more expensive argument regardless is stupid, that doesn't change the fact it's an unnecessary additional expense that narrows the cost gap in terms of TCO.

With Intel you're paying for something that has a lower power cost, requires less cooling, and has greater reliability. For most people the increase in cost is completely worth it, which is why most corporates go Intel - they don't want to fuck around with systems that burn more power, result in more ambient heat and are more likely to fail.

You may be able to flog cheap AMDs to grandmas or whoever your customers usually are for their low usage scenarios where even a chip from 10 years ago would do everything they wanted and where they don't use it long enough to see the higher defect rate come to light but for professionals and gamers Intel makes far far more sense.

Re:AMD slower / MHz

[Kartu]

30 Watts difference in no way justifies buying new PSU's and coolers.

Re:AMD slower / MHz

[Xest]

Well that depends how much power you're drawing doesn't it? If you're using 390 watts on a 400 watt power supply then of course it does.

Requiring different coolers is about efficiency and AMD CPUs have always run hotter and yes, it's enough that you can get away with less cooling on an Intel CPU. 30 watts at any one moment isn't enough to cause any much noticeable heat if that's what you're referring to but the amount of heat it can create at any one moment is not the problem, it's about heat build up and dissipation and this has pretty much always been more of an issue with AMD CPUs.

Re:AMD slower / MHz

[hypergreatthing]

So... how much extra do you pay per year than a more efficient and harder working intel variation in terms of energy consumption?
An i5-4670K is what, like 220$-250$? compared to your 8 core beast at 150$? Here: http://cpuboss.com/cpus/Intel-Core-i5-4670K-vs-AMD-FX-8120
The i5 beats it hands down in almost every benchmark. Maybe i was a bit unfair. Lets look at the best amd has to offer, the FX-8350. http://cpuboss.com/cpus/Intel-Core-i5-4670K-vs-AMD-FX-8350 . The i5 beats the AMD's 200$ top cpu in almost all categories including costing 1/2 the energy per year. That extra 20-30$ per year in electricity adds up.

Should i compare it to intel's top cpu? That's a kind of worthless comparison if it can't even beat it's mid line cpu.
Yes the bottom end is dominated by amds. I myself run a a10-5700 as a server because it's a cheap all in one solution with a better onboard graphics component than anything intel has to offer. But this whole discussions is amd's top line of chips vs intels and how much cost saving's you'll have and i just don't see it from any point of view.

Re:AMD slower / MHz

[hairyfeet]

Uhhh...who is having to buy new PSUs and coolers? I bought a bigger PSU not because of the CPU, but because I wanted the option to run dual GPUs down the road and its pretty damned hard to find a PSU under 600w that supports dual PCI-E connectors.

As for the CPU cooler it worked fine, I sold it with an Athlon X4 OEM that I picked up to build a box with, the only reason i went with the Hyper N520 is I LIKE the N520, it is very compact, small enough to fit into a mATX case, yet its quiet enough and cools well enough that I can toss having an exhaust fan and just use the N520. again i didn't HAVE to, the default cooler that comes with the chip works great, i just prefer having heatpipes on my system.

The simple fact is for less than the cost of a gimped Pentium Dual you can get an X4 that WILL beat it in most tests, will only set you back $67 and makes a great basis for a multimedia or even gaming box which you can have for less than the cost of an i3 and board. hell spend just $7 more and you can have 6Mb of cache which isn't a great help in gaming but is nice if you are transcoding, or if you really feel like splurging spend a whole $89 and get an unlocked 6 core which is great for multitasking. Hell just $250 after MIR will get you a whole 6 core system, just slap on the OS and go to town.

I can't do the math for you since i don't know how much electricity costs there but before i bought my 1035T I did the math for mine and the difference in price between the X6 system and the lowest Intel X4 with equal specs would have taken at least 7 years for the difference in electricity to come out in favor of Intel. Over the life of the system you MIGHT save enough to go out for pizza, MAYBE. I'm sorry but unless you are one of those rare people that push a machine to the bleeding limit the math just doesn't work in Intel's favor, sorry.

Re:AMD slower / MHz

[aztracker1]

I've been using the AMD APU series for most of my "friends and family" suggestions... the integrated graphics are good enough for most games on a single 1080p monitor, and run really well... I will opt for at least 8GB of ram, and an SSD these days as primary concerns over CPU speed (but the 8350 being around $200 is a good deal)... if you're mostly gaming, would suggest a higher end i3 or a midrange i5 though.

Agreed that AMD should be killing with integrated graphics, and I think after the next gen consoles come out, they'll see a big bump in mobile.

TDP for 4770k == 84W

[Anonymous Coward]

TDP for this amd part: 220W

I'm certainly rooting for AMD, but this part looks like a failure.

Keep in mind in addition to providing up to that 220W of power you also have to provide 220W worth of cooling. If that's really how hot this part is going to run then it's gonna need a *HUGE* heatsink, or high end watercooling setup to keep it at acceptable temps (Which at least for me is 30-40C, not the 50-70C all the manufacturers seem to accept nowadays.)

Re:TDP for 4770k == 84W

[bored]

TDP for this amd part: 220W

Well its a lot, but I think you overestimate the cooling. 120-130W is pretty common for most high end CPUs today. And GPU's do 200+ in a double high PCI slot.

Re:TDP for 4770k == 84W

[bored]

Just to reply again, but my desktop PC probably runs at 100% for less than 30 hours a month. The extra 100 watts for 30 hours or so is going to be pennies. So over 3-5 years, making up even $50 in processor price difference is going to be hard to do.

Re:TDP for 4770k == 84W

[Anonymous Coward]

GPU power-draw is dropping every generation. 200watt GPUs is not "normal", even high and GPUs are now around 130watts, but where near 180watts only a few years back. Now if you're talking about $900 SLI/CrossFire Dual-GPU-per-card beasts, then yes, wattage is going up.

Re:TDP for 4770k == 84W

[Applekid]

TDP for this amd part: 220W

I'm certainly rooting for AMD, but this part looks like a failure.

Keep in mind in addition to providing up to that 220W of power you also have to provide 220W worth of cooling. If that's really how hot this part is going to run then it's gonna need a *HUGE* heatsink, or high end watercooling setup to keep it at acceptable temps (Which at least for me is 30-40C, not the 50-70C all the manufacturers seem to accept nowadays.)

Just curious, why is 50-70C not acceptable to you? If the whole system is designed to live happily at that range, what does it matter?

Re:TDP for 4770k == 84W

[mhajicek]

I don't think the cost of power will be significant, but it would suck to have something processing for an hour and then overheat the CPU and lose all your progress. When time is money that gets expensive real quick.

Re:TDP for 4770k == 84W

[Anonymous Coward]

Actually, it's physics. Heat, over time, destroys electronics,
and nothing you can do will change that reality.
Just because something can run at 50-70C doesn't mean it should.

Just sayin'

CAPTCHA = 'micros'

Re:TDP for 4770k == 84W

[Anonymous Coward]

Because GET OFF MY LAWN.

Re:TDP for 4770k == 84W

[dywolf]

i should think that would be because most high end cpus are only using 120W or so....
if its generating 220 W of heat...you're not gonna try and only cool it by 120W

Re:TDP for 4770k == 84W

[dywolf]

i cant think they're actually going to succeed with this chip. the extra cooling needed is going to kill any cost savings vs the intel.

Re:TDP for 4770k == 84W

[viperidaenz]

If something can run at 70C for 100,000 hours, what's the benefit of running it at 40C? So you get more than 11 years of 24/7 use? The oldest CPU I still have is only 9 years old (a trusty Pentium M, circa 2004). You're talking Pentium 4 era.

Heat, over time, only effectively destroys electrolytic capacitors.

Re:TDP for 4770k == 84W

[smash]

Haswell will still slaughter it on near idle power consumption also.

Re:TDP for 4770k == 84W

[aztracker1]

You can get a pretty decent copper/aluminum hsf for $30 these days.. the 4770K is $360, if this comes in at $250-280, could be worth it.. also, if you *are* doing highly parallel loads (virtual machines, background services) the extra cores are a boon... using an amd 8350 currently for the more workstation friendly workload.

Re:TDP for 4770k == 84W

[MagusSlurpy]

Because most microprocessors do not run for 100k hours at 70C. Maybe at 40C, but electromigration will kill just about any chip running that hot for an extended period of time.

The temps that manufacturers state tell you the temperature at which a chip will operate, not a temperature at which it will survive with extended use. If a processor runs at 70C but crashes constantly at 73C, do you really think it will happily run at 70C for 11 years?

Re:TDP for 4770k == 84W

[dywolf]

how confident are you that a cooler intended for a standard 85-130W on todays chips will be sufficient for a 220W chip?

Re:TDP for 4770k == 84W

[viperidaenz]

If you were dealing with normal -40/+85 degree parts sure, but CPU's are rated at 100+ degrees. The 4770k is rated at 100 degrees.

Re:TDP for 4770k == 84W

[MagusSlurpy]

The 4770k's maximum Tcase is 73C.
http://www.intel.com/content/www/us/en/processors/core/4th-gen-core-family-desktop-vol-1-datasheet.html

Re:TDP for 4770k == 84W

[viperidaenz]

Tcase is irrelevant. It's not the one measured by the CPU. die temp is the important one.
You only need to know Tcase when choosing the heatsink.
TJmax is the one to look for.
You can keep Tcase below 73 degrees, but exceed TJmax and cook the CPU.

Re:TDP for 4770k == 84W

[MagusSlurpy]

I thought Tjmax was unreliable because it was calculated using some standard offset number that in reality varies from CPU to CPU, and that's why most monitoring software didn't use it, preferring Tcase.

Re:TDP for 4770k == 84W

[viperidaenz]

Tcase is the calculated value. It is based on TJmax and the thermal resistance between the die and the outside of the case.

Problem is, if you over volt and over clock your CPU, it will draw more power. That means you need to derate Tcase to keep TJ under TJmax.

BTW: The J stand for junction - the transistors on the die.

Re:TDP for 4770k == 84W

[MagusSlurpy]

Not Joseph? Lame.

Re:TDP for 4770k == 84W

[aztracker1]

I'm pretty sure a CM212, or Zalman CNPS (copper) would work... there's also some water cooling kits that should work acceptably. It's not really more heat than you see from some GPUs today... and have a better thermal load.

Frying eggs with your CPU is now a feature

[sinij]

Frying eggs with your CPU is now a feature.

New AMD CPU, comes bundled with George Foreman grill heatsink.

Re:Frying eggs with your CPU is now a feature

[red_dragon]

Frying eggs with a CPU is old hat.

I still run software written in 1999

[flayzernax]

Why the fuck do I need new processors that I can cook my breakfast with?

Re:I still run software written in 1999

[BLToday]

Because the P4 is out of production and toasters can't play games.

Re:I still run software written in 1999

[Anonymous Coward]

Because bacon.

Re:I still run software written in 1999

[Anonymous Coward]

Mmmmm... bacon. Frying on the CPU. Now that's good eatin!

I don't get it.

[zacherynuk]

When did CPU become a bottleneck? Is there a new version of java or flash I haven't got yet ?

Re:I don't get it.

[Anonymous Coward]

Just because you don't do anything requiring CPU cycles, doesn't mean the rest of us don't need this..

I'm just glad they might have an upgrade solution for me in the pipeline. I run an opensource OS project and we regularly compile thousands of software packages resulting in days of time to rebuild everything.

Gamers, programmers, researchers, .... it will get used.

Re:I don't get it.

[zacherynuk]

I am a gamer and my 920 @ 3.5 with twin GTX690's doesn't get touched.
I am a programmer (of sorts) and my striped SSD is still the bottleneck.
I have clients who do geological research and the AMD GPU's do everything for them.

As a user, the only thing that ever fucks me is Adobe, Oracle, HP and the windows print spooler. And these are always flaws.
enlighten me as to you bottlenecks.

Re:I don't get it.

[Anonymous Coward]

I seriously doubt gaming will see any benefit from having faster processors. The big companies only really put in enough effort to have feature parity with the consoles, consoles which make use of hardware that's just barely better than low end PC gaming hardware. The smaller companies can't really afford to up the specs for their games, they would cut off far too much of their potential customer base. People making Flash games will still produce prototypes that they call a complete product. People making games on top of Java will continue to ask "why would we need enums" while writing sluggish code that simulates what an enum does.

Can't argue against programming side of things. Faster compile times gives us more time to actually test our stuff before concluding it's good enough.

Re:I don't get it.

[Warphammer]

One thing that really needs to be said, just in general: I bought an i7-920 new, going on *five* years ago and it was still going strong when I replaced it for a 3770k, not because I felt that I would get massive performance benefit, but because I ran out of slots trying to support new buses (USB3, SATA III, etc, etc). I've never overclocked one, but I hear they're great for that, too. If everything from that era this long-lived, or is this just quietly a really great CPU?

Re:I don't get it.

[Lumpy]

I do video editing. Rendering I would KILL for 16 cores. Yet I know AMD will not make this capable of having two of them on the same motherblard.

Re:I don't get it.

[aztracker1]

Here's a motherboard and cpu for 32 cores.. ;-)

AMD's architecture is awesome for very parallel loads... in a lot of ways much better than Intels...

Re:I don't get it.

[dbIII]

I think the above poster means the two way and four way opterons won't be running at that speed for a while and this desktop CPU won't be two way. The 48 core machine I can use is 2GHz and the 64 core ones (using 16 cores four ways) are not a lot faster. I don't know how long it will take for the two way opterons to get near this speed.

Re:I don't get it.

[Anonymous Coward]

This console upgrade is frozen, but under your theory, every time there's a console upgrade, we suddenly realize all the performance gains in the CPU for the past X years, so it's still important.

There also is some PC-first gaming, and it does tend to skew toward the CPU-bound rather than GPU-bound side of things.

The smaller companies don't have to "up their specs" -- except perhaps some games that look like they could have been written in the mid-90s or earlier, it tends to take effort to improve performance of a game so it runs on target hardware, not effort to usefully consume all available hardware. What happens if 90% of the target demographics' desktop hardware doubles in speed overnight, is that less time is spent optimizing the PC version, which can lead to more optimizations on the console if it's cross-platform, or more functional bugs fixed, or more features created, or lower prices, or higher profit margin, or less stressed developers working overtime, or some combination thereof. It is not without effect.

Re:I don't get it.

[Lumpy]

Opterons are usually 4 years behind the others. I want at least 4ghz, prefer 5ghz and we will never see that in the Opteron line.

AMD does slightly different TDP.

[Anonymous Coward]

Intel does TDP based on normal possible maximum loads. This will be indicative of TDP when your CPU is pegged at 100%.
AMD does TDP based on maximum possible stress. This will be higher than the TDP you get with normal loads but means that you'll never get a brownout.

220W Seems Resonable for an overclock

[Anonymous Coward]

The FX-8350 runs at 4.2GHz (turbo) with 1.35V stock and is rated at 125W TDP. If you overclock it to 5Ghz and need to increase the core voltage to 1.6V then we're looking at 125W * (5GHz/4.2Ghz) * (1.65V/1.35V)^2 ~= 222W. So 220W seems resonable if you're just overclocking their current top of the line chip. What i'm skeptical about is why would AMD sell what is bascially an overclocked chip.

Fastest thing on this board anyhoo.

[Impy the Impiuos Imp]

I like several cores to keep the UI from locking up, and for this or that idiotic java app hogging one of the cores.

But in general, I don't care about processors anymore. They just have to be fast enough to feed data to graphics cards. Any serious number crunching (SETI, e.g.) runs on the graphics GPU anyway.

It's not even close. My CUDA SETI doubled 8 years of normal processor totals in two weeks.

Even servers, isn't the bottleneck still I/O?

Sockets?

[GioMac]

That's untrue. AM3+ sockets or CPU pins can't handle that power.
CPU's are powered with less than 5 Volts. In case of 220W it's 220/5 = 44 Amps!!!!
OK, pins are short and 44 amps might be possible, but powering such device even with a multilayer M/B will be scary and still - doesn't look like real.
11 mm^2

Re:Sockets?

[zacherynuk]

I concur.
I would add something to your score.. but that's beyond my CPU's threshold.

Lets overclock a Core i7-4770K to 5GHz

[Glasswire]

and see how the new AMD chip compares. I assure you the i7 won't need to draw 220W to do this.
Or let's look at performance per watt at normal frequencies where, if the AMD processor really does match a 4770K in raw perf, that will mean the Intel processor will be about 2.5x better on perf / watt.
As some people have mentioned, IBM routinely clocks Power architecture processors into the 4-5GHz range AND they draw several hundred watts each. If you think that's progress, I suggest you'll want to reconsider when you see the net throughput of a dense array of low-wattage Haswells cranking out aggregate SPECcpu numbers far beyond an IBM Power 7+ processor with the same total number of watts the IBM socket draws.

Re:Lets overclock a Core i7-4770K to 5GHz

[Bagok]

Good luck getting your 4770k to 5Ghz, sound like you will need a super sweet, de-lidded Haswell for anything over 4.7. These Vishras are doing it stock, on air.

AMD's future is HSA, not clock. But hey, they had these really great parts. Why not sell them?

Re:Lets overclock a Core i7-4770K to 5GHz

[file_reaper]

Yes indeed. Let's take 22nm Haswell's, and put it up against 32nm POWER 7+ and claim the 22nm chip does better. Yes, I do indeed see your point there.

Ignore this rubbish- look to Kaveri

[Anonymous Coward]

AMD's current so-called high-end, CPU only parts, are terrible. Sure they can be clocked high, but then their power consumption figures explode (a far larger than linear relationship with clock speed). Clocked up FX parts also fail to gain a linear improvement with the higher clock speed. Worse again is that, no matter how high the clock, the parts are beaten by the high end Intel equivalents.

AMD's good stuff is the Richland (reworked Trinity), Temash/Kabini 'low-end' Jaguar based parts, and the extremely important Kaveri design that replaces Richland near the end of this year.

In comparison, AMD's old bulldozer/piledriver rubbish struggles to make any sense against AMD's extremely well priced 6-core phenom2, although the 6-core FX-piledriver part has made strides to replace this part from 2 generations back.

In the near future, HSA (with Huma) becomes essential for modern PC computing, and Kaveri is the first proper step in this direction. Ironically, the Sony PS4 is the first x86 part on the planet to have full HSA features, and AMD will not be releasing a PC part with such abilities until Summer 2014 at the earliest. Not only is Kaveri much weaker than the PS4 part (designed by AMD), Kaveri is only a partial HSA design.

Anyway, there is no sane reason to pay top-dollar to buy an 8-core FX part from AMD. They are obsolete, and even AMD admits this, having described their first attempts at their new CPU architecture as a disaster. Bulldozer/Piledriver are buggy slow power-hungry rubbish. Steamroller, AMD's first serious attempt to fix the architecture, is the core used in the Kaveri part.

It should be noted that for power reasons, the XBox One and PS4 use the Jaguar core from AMD, a core created from the famous 'stars architecture' that formed the heart of every AMD CPU from the first x64 part (and first proper dual-core) all the way through to the Phenom2. Bulldozer/piledriver/steamroller are all variants of a completely new architecture that so far has not show any benefits over 'stars'. Steamroller adds massively more resources to each CPU core, and hopefully fixes the broken cache system that is mostly responsible for the FX's current poor performance.

Unfortunately for some, Kaveri will only have 4 Steamroller cores (two modules) and so will only compete (on the CPU side) with Intel's i5 parts. The GPU side threatens to be very good (and obviously clobbers Intel's dire integrated GPU) but will be really more of interest as a 'compute' engine since power users will own a discrete graphics card.

The real change (from Intel and AMD) comes some time next year when both companies move to a 256-bit memory interface for the CPU/APU (currently 128 as 2x64) and solders the RAM on the motherboard so much faster RAM can be used as well (as with the PS4).

In Alaska, they will probably sell like hot chips!

[Anonymous Coward]

In Alaska, they will probably sell like hot chips!

Jiggahertz....

[Anonymous Coward]

... is clearly analogous to penis length for chip makers nowadays. Power consumption is penis girth. Efficiency and performance is analogous to how good you are at pleasuring her and how long you can last, but nobody brags about those things.

wattage is what drove me back to intel

[Osgeld]

I was sitting here looking to drop a new CPU in my quad core FX, but shit, to support the chip I needed a new power supply (650 watts now), and as it stands its not that far away from tripping a breaker (between 2 AMD computers hoggin power, lamps and a TV) AND its still a slower CPU, needs a replacement heatsink cause the coolermasters that come with the chip are loud as fuck

I like my 3770k

Re:wattage is what drove me back to intel

[Kartu]

How much did you pay for your 3770k + mainboard please?
And if you're a gamer, wouldn't you be better off spending part of that money on a better graphic card?

Re:wattage is what drove me back to intel

[Osgeld]

80 bucks for a gigabyte board
319 for the chip

and I upped the video card from a ati6870 to a ati 7950

or its was going to be 189 for the chip + 80 bucks for a power supply + 30-40 bucks for a decent fan to support it, so totaled up its not that much more than just the i7 (but yes I had to get another mobo, which I was planning on doing anyway, one for me, one for the wife)

Piledriver is underrated

[Anonymous Coward]

The performance of Piledriver at 4GHz on multithreaded workloads is good, and the value is good. It's an underrated chip.

this makes me wonder

[slashmydots]

A lot of the new Vishera chips can be overclocked to 5GHz on air cooling. Even AMD's own promotional and marketing materials say that word for word. So, I'm wondering, do chips really draw that much power when you overclock them like 20%? I would have thought they only hop the exact mathematical increase in clock speed in wattage. Like 10% more speed = 10% more watts roughly. Does it really go up exponentially-ish like this with other chips?

Re:this makes me wonder

[NoImNotNineVolt]

Power is increases as the cube of speed.

10% more speed => speed*1.10 => power*1.10^3 => power*1.331 => 33.1% more power

And yes, it really goes up exponentially like this with other chips. With all chips.

The funniest thing...

[Anonymous Coward]

People claiming this CPU is not worth buying because it consumes too much power are often the same people who claim low power consumption processors are not worth buying.

irrelevant

[smash]

220 watts? Haven't they heard that power consumption is the issue of the day? Both in mobile, AND in the datacenter. If you double the power consumption of your servers you need to double your UPS, double your battery, double your generator capacity and double your cooling. that is not cheap. This machine is a non-starter for the non-amd-fanboi demographic.

just for the record

[slashmydots]

In case any of you forgot, I'm pretty sure Intel's 1366 (2011 socket precursor) i7 extreme edition chips ran at 185 Watts. Their current ones are 130 Watts. 220 beats all that but it's not like Intel never upped the power handling for a forced stable overclock and called it a new chip without really changing much if anything in the infrastructure. It's basically like buying a factory superclocked model graphics card. You're paying for better power handling and guaranteed predone overclocking.

Re:just for the record

[Osgeld]

2 years ago? tech is like dog years, thats over 14 years in advancement and AMD is about 5 human years behind

just keep cranking that clock speed, it will eventually burn out

posted from my 4.2GHZ amd FX

I DON'T CARE

[Anonymous Coward]

How much power my CPU is pulling.

In fact it's starting to match up nicely with what my GPUs are pulling.

Fucking cold in here anyway.. Maybe i could finally turn the space heater off.. Well.. when we reach the 800 watt area.

Netburst debacle?

[ArchieBunker]

Can anyone explain what the problem was with Netburst chips?

Re:Netburst debacle?

[dbIII]

Many corners were cut so they performed like shit (although I still have a fileserver with some netburst Xeons running *bsd). For the long list of corners that were cut it's best to look at wikipedia. One odd thing is even my Uncle who graduated from electrical engineering before the transistor was on sale thought the pentium four was shit, although he wasn't fond of x86 in general.

Re:Netburst debacle?

[toddestan]

With Netburst Intel created a chip with a long pipeline that was inefficient on a per-clock basis, but the idea was that the design would allow fast clock speeds. Or in other words, while the Netburst may not do well against an AMD chip clocked at the same speed, it wouldn't matter if Intel was pushing 5+ GHz and AMD was still struggling to hit 2 GHz. Unfortunately for Intel, things didn't work that well and since high clock speeds required more power their chips had a tendency to run a but warm so the design topped out just past 3 GHz. Intel tried to take it even further and lengthen the pipeline more in an attempt to up the clock speeds with the Prescott core, but all that really did was make the chips even less efficient and while they were able to bump up the clock speeds a bit, it didn't even get them to 4 GHz. On the upside though, the long pipeline allowed for Intel to implement Hyperthreading, which admittedly is a pretty neat trick.

I looked at this

[Anonymous Coward]

I looked at this, and then compared the newest version of intel's processor versus the corei7-920 I have been using for about 4.5 years. I've seen several reviews where they compare the new chip versus the old and claim 66% more performance, and half the power consumption. Half the power is nice, but I'm not going to chuck out my old chip for a percentage improvement. The Corei7-920 was about 5x (500%) faster than my old CPU. Show me a new intel chip that's 5x as fast (not 1.66x as fast), and I will look and spend. On the other hand, if you gave me 50% off on a new motherboard, processor and memory, I might look now.

Not so simple any more

[dbIII]

If stuff is CPU bound you normally want to finish it ASAP anyway so performance per watt is not so important. If stuff isn't CPU bound the cores clock down these days anyway and use a lot less power, so performance per watt is a lot harder to determine.
So while you are correct that for a lot of stuff you want to get it done with the minimum power consumption it's not so clear that these CPUs are going to use more watts for the same tasks in the long run as others. If it finishes the job quicker and then runs at reduced power it may use less than something that has lower peak power consumption but takes longer to do the job.
I've got a pile of new processing nodes that use more power at peak load than the old ones but my power and cooling requirements are going down - simply because the old machines used more power while idling. If the cluster had 100% utilisation it would be a different story.

Re:Not so simple any more

[smash]

Depends if your workload has a beginning and an end. if it is a constant stream (i.e., it is a machine providing a service), then performance per watt is very important.

Re:Not so simple any more

[smash]

Also, peak power consumption is not the same thing as performance PER WATT. In your above example, your new machines are better performance per watt, hence the result you are seeing - even if their peak TDP is higher.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Comment removed

[account_deleted]

Comment removed based on user account deletion

IANAP?

[Anonymous Coward]

I Am Not A Processor?

Anandtech has more reliable servers/info

[Anonymous Coward]

http://www.anandtech.com/show/7066/amd-announces-fx9590-and-fx9370-return-of-the-ghz-race

Awesome... until the windows melt.

[Anonymous Coward]

Until we start seeing chips built on Carbon substrates (diamond) instead of Silicon we wont realize 10Ghz as the metals inside a Si chipset will fail due to weak nuclear forces which can radically change electrodynamic and physical properties when subjected to frequencies much above 5 to 7 Ghz. Although LENR research has been focused on the transition of Ni to Cu and has shown that substatial energy can be derived from the transformation, the theory also posits that the same can be true of N to C and that somewhere between 7 and 30 ghz there exists a possibility that many materials will undergo a LENR type event and potentially emit significant energy surplus (while not solving any computations).

So, slowly ramping through frequencies until your new supercooled super game tower converts metals in the chipsets and board traces and makes the room glow a bright cherenkovian blue will undoubtably be the coolest overclocking EVAR! i dont think i want you next door to me when everything goes just right and you melt the glass out of all your windows.

See http://hardware.slashdot.org/story/13/02/22/0219216/nasas-basement-nuclear-reactor for the slashdot link to TFA at http://www.gizmag.com/nasa-lenr-nuclear-reactor/26309/

from TFA
"The electrons in the metal lattice are made to oscillate so that the energy applied to the electrons is concentrated into only a few of them. When they become energetic enough, the electrons are forced into the hydrogen protons to form slow neutrons. These are immediately drawn into the nickel atoms, making them unstable. This sets off a reaction in which one of the neutrons in the nickel atom splits into a proton, an electron and an antineutrino. This changes the nickel into copper, and releases energy without dangerous ionizing radiation."

and

" In past years, several labs have blown up while studying LENR and windows have melted – showing that if it really works, it can produce an impressive amount of energy."

Sketchy Vendors

[DarthVain]

I recall seeing some sketchy vendors advertising "dual core" systems at twice their clock. So a Core2Duo @3Ghz was advertised at 6Ghz. There were a bunch like that, I wonder if they ever got sued. Anyway a friend was arguing with me about their existence (to which I believe I know much more of these matters). However when he showed me the actual print ad, I laughed and laughed, and then cried a bit.

Never underestimate the power of advertising. This was a retailer which isn't a big deal, however I recall seeing some pretty BS misleading ads from manufactures as well including nVidia, ATI, Apple, etc...

You might see 10Ghz sooner than you think! :)

LOL...

[Anonymous Coward]

Makes sense. The only way to get ahead in America is screw everyone else. But that is EXACTLY how Fox News Ayn Randian Right Wing Libertarian losers want the world to be.

i7 4770K vs AMD

[Anonymous Coward]

the intel's i7 4770K only has a 1mb l3 cache PATHETIC as long as they top that the benchmarks will prove who's better.