Slashdot Mirror
45nm Phenom II Matches Core 2 Quad, Trails Core i7
An anonymous reader writes "AMD recently debuted its 45nm Phenom II processors, and The Tech Report has already run them through a complete suite of benchmarks to see how they perform compared to Intel's latest and greatest. The new 2.8GHz and 3GHz Phenom IIs are in a dead heat with like-priced Core 2 Quads, but they generally fall well behind Intel's new Core i7 chips. TR concludes that AMD's future doesn't look as bleak as some say, and future Phenom IIs could compete favorably with more affordable Core i7 derivatives."
But it could have been if I had a Core i7!
is that its not the code you actually will run.
is that the CPU price is only one component of a significantly higher overall platform cost. Both memory and mainboard cost significantly more if you want to build an i7 rig.
If libertarians are so opposed to effective government, why don't they all move to Somalia?
The main problem is that AMD is doing the exact same thing Intel did when the P4 was out: they went to a smaller litho process, slapped on cache, and cranked up the clockspeed. If you read the review carefully you'll note that while the new Phenoms are faster than some Core 2 quads, they are not faster on a clock-for-clock basis. Remember back when AMD was leading in that category and it was such a big deal?
As of right now the Phenoms are a good deal IF you already own an AM2+ mobo... otherwise they are not a good deal for 2 reasons: 1. AMD is coming out with the incompatible AM3 socket that will use DDR3 memory in the next few months, so these current chips will have a very short shelf life; 2. Intel doesn't have to do any innovation at all to beat these chips, all it has to do is drop the prices on current Core 2 quads like the 9550 that outperform the Phenoms but are currently priced higher... dropping prices ain't rocket science and there are rumors these cuts could be coming by the end of the month.
As for the Core i7, sure it is more expensive, but even the 920 model appears to wail on these chips, and there is a whole lot more future-proofness in buying a low-end i7 right now. Interestingly, the review mentions the new Phenoms have 758 million transistors which means they have about 27 million more transistors than Nehalem... but Nehalem at 2.66 Ghz is easily beating a Phenom at 3.0Ghz. It looks like what AMD really needs is a new architecture, but that does not appear to be coming any time soon.
AntiFA: An abbreviation for Anti First Amendment.
Since I'd hate to go back to the bad old days, before the megahertz wars, when processors were expensive. For desktop/low-end workstation use, the phenom IIs seem to be merely ok, price competitive with the chips intel currently has down there; but not in a position to beat the i7s. Where these new cores will be quite interesting, though, is in 4 socket and higher configurations. Even with substantially inferior core designs, AMD has been stomping Intel in the 4+ socket area, since hypertransport is markedly superior. With actually competitive cores, AMD should find 4+ sockets to be a party, at least until Intel gets quickpath stuff ramped up.
Does anybody know if there are plans for 95 watt Phenom IIs? It looks like the first models are 125 watt only (and I just got a AM2+ motherboard that only supports the 95W Phenoms).
Uhm, Athlon 64 vs Pentium 4 didn't made it into your comparison? I'd say I'd prefer Athlon-XP over Pentium 4 to.
I don't know which I'd prefer of Core Duo and Athlon64 X2.
So you didn't notice how you can't spell "Islamo-Communist mental case" without I n t e l then?
which is totally what she said
If AMD wants to improve sales they should do what they did for the X2 line: lower power consumption. I don't care what any of the "experts" say - the moment I saw the whopping 130w listed next to the i7 920 I immediately decided I didn't want one. The Athlon 4850e already has the crown on the dual core front, so if they can manage a respectable 95w quad core, AMD could corner the efficiency market. I know they had to release this chip to generate revenue, but I'd hoping that low power quad is in their future plans.
On another note, it's quite funny to see such high power requirements for the new intels. Am I the only person who remembers AMD getting ridiculed about the Phenoms power consumption? Now that intel has finally released a true quad core chip, their power consumption is the same or more than AMDs. Granted that does nothing about the performance gap, but at least it quiets the power critics.
So you didn't notice how you can't spell "Islamo-Communist mental case" without I n t e l then?
Rock on!
What is "port one"?
Again it seems that the benchmarks are running Intel Optimized code on AMD...
[...]Adding to that fact is amd's discreet graphic offerings which are far superior to intel's offerings.[...]
Unless you're using GNU/Linux. The drivers just aren't quite up to snuff yet.
most of MSI's mobo aren't getting an update that will allow the use of am2+, which pisses me off because I bought that board with plans to upgrade to a AM2+ chip
Interestingly, the review mentions the new Phenoms have 758 million transistors which means they have about 27 million more transistors than Nehalem... but Nehalem at 2.66 Ghz is easily beating a Phenom at 3.0Ghz.
The other obvious problem is that pretty much every compiler on the market [especially the Intel C/C++ compiler] is optimized for the Intel circuitry as opposed to the AMD circuitry - i.e. most compilers probably aren't even aware of the functionality of those excess 27 million transistors.
I can't for the life of me understand why AMD couldn't come up with $10 or $15 or $20 million to fund their own in-house compiler - for instance, it seems like there was a time when they could have scooped up Metrowerks from Motorola for pennies on the dollar.
Until there's a compiler which truly understands the AMD circuitry, I don't see how anyone can know [definitively] the upper bounds of the capabilities of the AMD CPUs.
PS: And as AMD tightens up the integration of their Opteron CPU circuitry with their ATI GPU circuitry [it can't be all that long now until they're both just isolated, disparate cores on the same multi-core matrix, and maybe even completely integrated into the very same core], it seems to me that having an in-house compiler - which is very tightly integrated with the circuitry - will be of paramount importance.
In all honesty, I often wonder if maybe the Intel C/C++ compiler team is really the secret ace up Intel's sleeve which does the most important work in distinguishing them from AMD.
It only works in class where the problem kid isn't more powerful then the rest. Collective punishment fails if the annoying kid can kick you back down for blaming him.
The Opteron 838x (shanghai) stomps all over anything available from Intel, they won't have a part that competes until 2H'09 at the earliest when they get the Xeon Corei7 working.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
How many businesses do video and graphics work? Not many. I'd wager that not 1 in 10,000 business machines is used regularly for any CPU intensive task. Developers would definitely be somewhat of an exception, but there really aren't that many development groups out there doing compiles of 10+ million line code bases all day every day. I mean I can compile a Linux Kernel on my 2.4 ghz P4 machine (vintage 2004) in a few minutes as long as I have 2 gigs of RAM in it.
No, 99.9999% of all business machines don't need to be even as powerful as your average home PC, where someone might actually WANT to edit some video, etc. now and then. Those business machines run Office, Web Browser, Outlook, and maybe one or two other business apps.
In fact I would venture that business is where the desktop performance requirements are falling and will fall the fastest and furthest, and where energy efficiency makes the biggest difference. Any heavy lifting that needs to be done can always be moved to server side applications. There will always be a few high end workstations, but the average business PC is pretty close to being a diskless thin client now in any good shop, and that will be the norm within 3 years.
"Malo periculosam, libertatem quam quietam servitutem." -- Jefferson
Well, I just built a new system and looked all this crap up the other day, so I feel qualified to comment.
Yes, the new chips finally just about match equally priced Intel chips - at stock.
The problem is they only go down to $250 or so. Not down to $200 or under, where the lowest-priced Core 2 Quad lives: the Q6600. And the thing about the Q6600 is, it can be trivially overclocked well past 3GHz on stock cooling with no enthusiast messing about. Just bump up the FSB speed and you're good. So for $200 or less you can get a year-old chip that will comfortably outperform AMD's latest and greatest. That's not such great news for AMD, is it?
And, don't forget, i7 is just Intel bleeding the 'enthusiast' market dry. Essentially they're getting an overpriced version of the new platform out first to capitalize on the market that'll spend any kind of money to have the fastest system on the block. Once they're done screwing those guys, they'll be making essentially the same chips available at much lower prices. Which will in turn drive down the price of the Core 2 generation, making them even more of a no-brainer purchase over the new AMD chips. AMD is basically a whole generation behind, and not looking like it'll catch up any time soon. I just don't see who - except someone who happens to have an existing AM2 motherboard that'll take one of these new chips and is looking for a modest upgrade - is going to buy one of these.
... "AMD finally on par with Intel tech from two years ago."
Seriously. I love AMD. I've been using AMD chips since my very first system of my own, which had a mighty K6-166.
This past Novemeber, I finally bit the bullet, sold my Athlon X2 system, and upgraded... To a Q6600-based rig. Some may scoff that it's only a 2.4ghz chip, but I'm running it at 3.4ghz right now, and I'm fairly sure I'll reach 3.6ghz with a bit more work.
To be sure, those who don't overclock, but want a powerful AMD-based system will find this chip worthwhile. I would expect those to be people who already have AM2+ systems looking for an upgrade (I have a good friend that wanted an ultra cheap upgrade for her desktop a few months back... I built her an AM2+ X2 setup. Guess what? Now she can upgrade to something worthwhile. Yay!).
However, for the enthusiast, the hardware tweaker, the overclocker... The Phenom II is a disappointment. It has been said it can overclock to as high as 3.8ghz with good air cooling, maybe even 4ghz if you go with water. Awesome. C2Q can clock every bit as high, and you'll get more performance per clock out of it as well. I paid $180 for my Q6600. At 3.4ghz, it'll out-pace the top-end Phenom II. That Phenom II cost $275. Which is about the same price as... A low-end Core i7 chip that will overclock to 4ghz and beyond. With the i5 chips coming down the line, what's stopping Intel from slashing Core 2 prices?
All-in-all, the Phenom II is a powerful chip, and would serve well as the heart of an AMD rig, delivering worthy performance for almost anything. The problem is, a Core 2 Quad will do ya one better every time, for the same cost or less, and has been doing so for the past two years.
I'm anxiously waiting for AMD to bring back the glory days when they actually tried to compete with Intel, not just chase their taillights.
[Trojan.]
The first thing that came to my mind when I looked at the pic of their test system was, "can I have some motherboard to go with my heat sink please?" How much longer before people are trying to squeeze components into some kind of massive rectangular radiator instead of a case?
Anyway. I'm glad AMD was able to get to .45 I knew that was the reason they weren't performing on par with the Intel Duo chips.
Having to work for a living is the root of all evil.
Man, have you tried to use a recent build of Firefox on a 5-year-old machine lately? It's barely usable unless you have at least 512mb of RAM and a fairly fast CPU.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
If AMD wants to improve sales they should do what they did for the X2 line: lower power consumption. I don't care what any of the "experts" say - the moment I saw the whopping 130w listed next to the i7 920 I immediately decided I didn't want one. The Athlon 4850e already has the crown on the dual core front, so if they can manage a respectable 95w quad core, AMD could corner the efficiency market. I know they had to release this chip to generate revenue, but I'd hoping that low power quad is in their future plans.
Don't worry, AMD will keep the same power bands they kept in the past. They'll have their 130W, 95W, and 65W desktop parts. I've got a 65W Athlon X2 and it performs well and my CPU fan barely has to spin. When the Phenom II comes out I'll be grabbing a new one. Sadly I'd have to get a new mainboard to take advantage of its additional power features, but oh well.
The enemies of Democracy are
Honestly in this day and age, no one should buy a extremely powerful processor. That is unless you are a data center or need to crunch numbers at a fast pace. Today I would rather see I/O bottlenecks get smaller and a smaller thermal footprint than pay for a faster, higher performance processor that has a TDP of 100W+ and costs $300 or more.
last I checked, OCZ didn't qualify as "value ram".
Then please check again. Last summer I finally got around to buiding a new computer, and I had been reading specs and reviews on parts for at least 6 months straight. Trust me when I tell you that OCZ isn't the high quaility ram it was back in 2003. They have quaility control complaints all over the place, to the the point where they have customer service people patrol the online feedback just to quiet the fires. That's the problem with reputations: they don't always get updated as quickly as they should.
Having said that, value ram has nothing to do with brand name or brand quality. Value ram is merely a manufacturer's entry level ram that usually comes without a heat spreaders and thus has highter (factory) latency timings. I think this is what you said didn't exsist.
Most gamers are better off spending under $200 on a CPU, and most consumers won't tell the difference.
People who buy quad core processors nowadays either want extremely performance for multithreaded tasks and are willing to pay (a lot!) or they're total dumbasses, in either case they'll buy an i7.
If AMD wants to catch up they need to cut these things down to duals like they did the original Phenom to 7750+.
Uh, isn't that backwards? Unless something's changed recently I believe the Intel TDP _is_ 'worst case', whereas AMDs figures are for 'typical use'.
No, AMDs TDP is a fairly true representation of maximum power. It has to be, since it is what OEMs use to design cooling solutions, and those cooling solutions must be designed around the worst-case scenario (at least in terms of processor power output, not necessarily other environmental factors).
Unlike Intel, AMD does not have a thermal clock gating circuit that will slow the processor down should it ever rise above the stated TDP number, so AMD's TDP must truly be a maximum. Intel on the other hand can pick a TDP that chops off most of the 'long tail' of high power but unlikely circumstance, and count on their clock gating to enforce the TDP value should one of those circumstances arise.
This is part of why AMD started using "average power" in their advertising, because Intel had an advantage in perceived power consumption based on TDP, since Intel's was much closer to average power. Starting with Phenoms they call this number ACP (Average CPU Power).
So you're not completely out in left field, just mistaken about the relationship between AMD's TDPs and "typical use" numbers.
The enemies of Democracy are
I had an Athlon 500Mhz (the oldest, slowest, first one they ever made -- it was back when AMD was just jumping on that fad of putting the chip on a slot instead of a socket). First of all, this was before NetBurst, so it was competing against P3s, not P4s. Second, it was better than the P3. I know this because if the P3 had been better, I would have gotten it instead. The K6-2s were the last "not decent" chips AMD made.
And it also did not run hot. A few years later, when I was finally upgrading the graphics card (from a TNT2 to a GeForce 3), I noticed that I had never even plugged in the CPU fan! It had run perfectly well passively-cooled that whole time!
"[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz
Hard[OCP] has had one of the best writeups so far imo. Admittedly they compare a Phenom II to a top of the line Core2 and i7 in that article, but that's only so they can compare the cores clock for clock which isn't feasable using the cheaper intel parts as an unlocked multiplier is required.
Anyhoo, the point is that Phenom II is, clock for clock, slower than Core2. Given how cheap and overclockable a Q6600 w/DDR2 is then it's hard to justify Phenom II for enthusiast purposes - tho I suppose when that reality hits the market retailers will start doing deep discounts and I expect good deals will appear.
What I'd love is, given the above facts, a proper overclocking competition article between a Q6600 or similar Core2 and the new Phenom IIs. The unlocked Phenom II can be had for about £140 here in the UK so if it generally overclocks well then it may be able to whoop Core2s' arse!
Nick
Is that still true? This is one of those areas where I've read conflicting word-of-post, some people say they're still not quite as good as the Nvidia drivers and others say they're OK for games.
Nick
Howso? At least link to the page in TFA that says the code they're using is Intel optimised. I've read two links out of this discussion but I have no intention of reading TFA unless I really have to. :)
Nick
I had one of those old slot-A systems that had had its fan seize up. It overheated and eventually failed and when I went in to check it out the can had a coating of dust on it. It had evidently been running for some months with a seized fan and only the heat of oncoming summer had revealed the problem.
Not to take away from some impressive stuff AMD has done, but AMD's glory days were also helped out by Intel shooting themselves in the foot. Back when AMD had the top-end x86s, invented AMD64, etc., Intel's 900-pound-gorilla R&D machine was off working on Itanium, running their x86 line mostly on autopilot. Once they mostly gave up on Itanium and swung their resources back to x86, AMD, as you might expect, has had a much harder time.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
I think the biggest problem with the AthlonXP/MP was the lack of a heat spreader, which caused lots of people (you paying attention Patric Norton?!) lots of headaches as they attempted to put the heatsink on without chipping the processor.
Another issue was not getting the heatsink quite flat on the chip, causing it to be rather ineffective in removing heat from the processor, leading to temperature shutdowns (either due to blown chips, or from the newer mainboards that had temperature sensors built in.
That said, I never managed to break one in that way, and I built quite a few machines based on those chips. I think I fared better than Patric because I put the heatsink on the chip while the board was sitting on my bench instead of trying to do so while the power supply was in my way...
bork bork bork!
AMD has been headquartered in Sunnyvale, California for almost 40 years.
Contribute to civilization: ari.aynrand.org/donate
I disagree. AMD is in an ok position, not great obviously, but their platform is competitive. The Core i7 is a better processor than the Phenom II, but I don't expect its price to come down any time soon. Intel will milk the high margins for as long as they can and sell old Core 2 quads to consumers for at least this year.
Given the choice between a Core 2 Quad and a Phenom II, you should pick the Phenom. No question about it. The Core 2 quad has a split cache so multithreaded performance is crap. The cores have to transfer data through the slow memory interface, which limits parallel speedup in a lot of cases. This wasn't really an issue when Intel released the processor, but in the near future it will be a serious issue because the parallel software is coming.
In Soviet America the banks rob you!
responses to that.
The trend towards reconsolidation of functionality in the back room and away from the client desktop only grows stronger with time.
The conditions under which fat 'client' PC computing originally appeared no longer pertain. I'll illustrate with an example. In the mid 80's I was working for an engineering firm. They had a few 1000 users, a LAN, a cluster of computers in the basement, and a dumb terminal on everyone's desk. Over the next 5 years disconnected PCs rapidly replaced a lot of these terminals for several reasons.
1) Networking PCs (or anything) was expensive, so if you NEEDED a PC, it was almost guaranteed not to be on the network.
2) People needed graphics capabilities and the ability to read data from floppies etc which were not possible with a dumb terminal in that day.
3) Department managers loved to control stuff, so of course they'd rather have a PC that was THEIRS and not a terminal, which 'belonged' to IT.
In every other respect the switch to PCs was negative overall. They were MUCH more expensive, required management, couldn't be backed up easily, and usually couldn't get access to resources on the network except via sneakernet.
Nowadays the 'positives' for PC/workstation fat client type machines are virtually gone. They WILL be networked and since x86 PC architecture is virtually ubiquitous there is no longer a distinction in capabilities between a PC and a 'terminal' (thin client). Furthermore the costs of unmanaged systems have finally begun to become fully appreciated. IT can't tolerate machines on the network they don't control and departments can't afford to or be bothered with dealing with the associated costs/risks themselves, so both IT and the client department would rather let IT manage things, and IT has no incentive not to then make everything a thin client.
Now, maybe the environment you work in is really backward and has not figured all this out yet. That's true with a lot of small businesses, but they are learning, and it is getting quite easy to provision network centric applications which often only require a web browser on the client side. So even the small businesses are less and less needing fat clients. I'll predict that no such thing will exist as a full up PC in business at all in 10 years. There will still be workstations for very specific uses, CAD or high end video/graphics work, but even those will mostly or even entirely utilize networked resources to boot and for storage.
In my business this sort of transition has already pretty much completed. The existing PCs are just old legacy hardware that will be replaced with diskless machines when they go end of life.
"Malo periculosam, libertatem quam quietam servitutem." -- Jefferson