Leaked Pictures of Socket F

Robbedoeske writes "Dutch language site Tweakers.net has the first pictures of AMD's Socket F, aka Socket 1207. This socket introduces support for DDR 2 memory and some say it will offer the ability for a integrated PCI Express controller on the cpu. Socket F is meant to be used in systems with more than one Opteron cpu."

Comments

Translation

[Anonymous Coward]

For those that can't read Dutch the Socket F looks like any normal chip.

Re:Translation

[Anonymous Coward]

WTF is with this 800MHz bus speed on Intel chips and still something like 200MHz on AMD? What is going on?

You'll be happy to know that the 800 "MHz" Intel bus actually runs at 200MHz, whilst the AMD bus runs at 1GHz now. The Intel bus is good for streaming data, because it is quad-pumped, meaning it can do 800MT/s, the AMD bus is double-pumped and can do 2000MT/s. However the Intel bus is 64-bits wide, whereas the AMD bus is 16-bits in, 16-bits out.

Help?

Re:Translation

[Anonymous Coward]

A pictures worth a thousand words that just happen to be in the same language as the looker.

Re:Translation

[SlowDancing]

Perhaps the 1207th connection is the frame itself, grounded. Extra shielding would be a benefit to system builders, or at any rate might make the FCC happy.

Re:Translation

[Anonymous Coward]

How does it look to those who can read Dutch?

Is that a ball grid array I see

[dremspider]

It looks similiar to Intel's new design with the pins, hopefully it isn't as easy to damage.

Re:Is that a ball grid array I see

[Anonymous Coward]

i never found LGA775 motherboards to be fragile. In fact, I find them to be the other way around since the pins are well-protected by the frame of the CPU clamp. i probably built 200 LGA775 boxen working for PC Club in the last year and never had a problem with damaging one.

Re:Is that a ball grid array I see

[Groo Wanderer]

http://www.theinquirer.net/?article=17617
http://www.theinquirer.net/?article=17652
http://www.theinquirer.net/?article=17681
http://www.theinquirer.net/?article=17709

More than you would care to know about LGA.

          -Charlie

Re:Is that a ball grid array I see

[rosewood]

It makes me sad that a PC Club would build anything but AMD boxen.

Re:Is that a ball grid array I see

[Lucan_UK]

I gotta say that this section of the first inquirer article had me in stitches

"First is the new retention mechanism. If you remember, the old way was to put a chip in the ZIF (zero insertion force) socket and close the little lever on the side. This clamped the chip in and held it down relatively tightly. If you tried hard enough, you could pull it out, there was no locking mechanism to keep it in place. You then put a large part of a battleship hull, known in the industry as a heatsink, on top, and clamped it down with a clip that locked it in place. On top of this was a fan that could lift the remaining parts of that battleship off the ground. It was big and heavy, had clearance problems, and was a bitch to put on. It could fall off in the shipping process. The thermal interface material could cement the CPU to the heatsink, and pull it out of the socket if the heatsink fell off. Since the heatsink was the only locking mechanism, there was nothing other than friction to keep the CPU in the socket. Not a good thing from a physical perspective, and having a hunk of copper weighing the better part of a pound rattling around lose inside the case was even worse."

Re:Is that a ball grid array I see

[MarkGriz]

no, but it *is* happy to see you.

Re:Is that a ball grid array I see

[Groo Wanderer]

Thanks, we try. We even succeed on rare occasions.

            -Charlie

Re:Is that a ball grid array I see

[klui]

I read that the problem is the pins on the motherboard (although well-protected) are not designed for repeated insertions/extractions. I can see the logic with that, each of the pins have a tiny spring that press against a pad on an LGA CPU.

Re:Is that a ball grid array I see

[Anonymous Coward]

What, and you think old ZIF sockets didn't have an insertion/removal cycle limit? They always have...

Re:Is that a ball grid array I see

[spectre_240sx]

Sweet. I was looking for that. Thanks!

Yes, he counted them

[Killjoy_NL]

I'm dutch so I could read the forumpost that started it all.

He actually said he counted all the pins, just to be sure to give enough information.

Funny stuff (being dutch rocks)

Re:Yes, he counted them

[Anonymous Coward]

Funny stuff (being dutch rocks)

Of course, other countries have nice rocks, too.

Re:Yes, he counted them

[Anonymous Coward]

Of course, other countries have nice rocks, too.

Shhh, don't let the Dutch know, the greedy bastards.

Re:Yes, he counted them

[Anonymous Coward]

*scratches neck*

did somebody say rocks?

Re:Yes, he counted them

[jared9900]

So what's it say then?

Re:Yes, he counted them

[FST777]

Allright, for the fun of it:

At our forum, Gathering of Tweakers, the first pictures of AMD's Socket F have emerged. In may we wrote in that AMD has set a new CPU-socket on their roadmap. The new socket would have 1207 connection point and would be meant for multi-Opteron servers. To prevent that a CPU with support for DDR-memory is placed on a DDR2-socket and vice-versa, a new socket was needed. The extra pins which are available with this step are rumored to be used for an integrated PCI Epress-controller on the CPU. Noticable on the pictures is the clear separation in the middle of the socket. This seems to point out that each core of the dual-core Opteron gets its own group of contact-point and is truly treated as a separate CPU.

The pictures further show that Socket F, like Intels Socket 775, is bestowed the pins that contact the CPU. The CPU will not be put inside the socket, but this is a so-called LGA-socket. Socket F is by the way also called Socket 1207, but like Socket 479 only has 478 pins this model only has 1206, as shown by punctual counting. This socket also supports registered DDR II 533-, 667- and 800-memory and by doing so AMD ventures the competition with Intels FB DIMM plans. The latter wil introduce its dual-core platform dubbed Dempsey coming april, with among other thingsthe Greencreek chipset with support for FB DIMM-memory.

Even the dutch text was badly written, so excuses for the grammar and spelling. It's always hard to translate anything other than your own thoughts ;)

EP?

[Anonymous Coward]

Eerste post???

Re:EP?

[Rinzwind]

That means 'First post' ;)

PGA

[theantipop]

If true, it is interesting to see AMD moving to pin grid array-style cpu connection. Intel has used this for a little while now with thier socket 775 Pentium 4 chips. While there was concern over broken pins resulting in unusable motherboards, it now seems to be a relatively robust mechanism. I wonder what advantages AMD saw that lead them to this design. I also wonder if their Socket M, 940 pin solution for next years Athlons will use the same socket design.

Re:PGA

[sarahemm]

I think you mean Land Grid Array (LGA). Pin Grid Array (PGA) is what they've been using since the 486 (386?) days...

Re:PGA

[Anonymous Coward]

That is easy to figure out, Intel still acounts for over 80% of the market and if AMD has any hope in hell with OEMs they had better make it easy for those OEMs to use similair hardware and installation tools on the assembly line. I would look for BTX support from AMD over the next year, they can posture all they want but even low wattage chips will benefit from BTXs superior cooling design.

Re:PGA

[theantipop]

Doh, yep thank you. Too early to be thinking about acronymns.

Re:PGA

[Hex4def6]

I wonder what advantages AMD saw that lead them to this design.

Well, I think its probably another cost-saving measure. I'm sure just having to put normal contacts on the CPU instead of pins makes the price of them perhaps significantly cheaper to manufacture, and passes the cost on the the motherboard manufacturer.

Re:PGA

[rajafarian]

It is not up to AMD but motherboard manufacturers to support BTX.

Re:PGA

[sarahemm]

...and when a pin breaks, AMD/Intel don't have to deal with it ;)

Re:PGA

[Halfbaked Plan]

I have Intel 80286 processors in the PGA package.

Re:PGA

[sarahemm]

I stand corrected! I guess 80286 was the first intel CPU in a PGA (IIRC the earlier ones were all just DIPs) :)
A 68pin PGA certainly looks a lot less dense than the current almost-or-over-1000 pin ones :)

Re:PGA

[Halfbaked Plan]

The 80286 was more common in a plastic package (PLCC) but there were also ceramic leadless packaged parts, and a ceramic PGA part. Not sure if they ever produced it in a gullwing SMD package or not.

AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

The more I learn about Apple and Intel the more worried I get.

IBM is cranking out killer PPC chips.
AMD is cranking out killer x86 chips.

And Intel looks like they are ready to compete in some sort of Special Olympics for Computer Chips.

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

Re:AMD Beating The Crap Out Of Intel?

[pivo]

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

That's easy: Marketing

Re:AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

Marketing money. Intel writes big checks to Dell as "co-operative marketing" funds. Also, Intel maintains its own sales force, including technical support for developers, on the corporate level. Guess how they decide which HW platform to recommend - who's laptops they show up with.

Finally, Dell isn't much of an engineering company - they need to keep they're offerings simple - both for their supply chain and support. Helps to keep it easy and cheap to acquire and sell.

Re:AMD Beating The Crap Out Of Intel?

[edxwelch]

If you mean by "marketing" bribing, then you are in fact correct ;)

Re:AMD Beating The Crap Out Of Intel?

[NixLuver]

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

Initially I think you have to consider exactly what Apple is trying to achieve. IBM won't play ball with Apple's laptop designs, and the powerbooks (as much as I love 'em) are being left behind, pretty badly, by X86 stuff. Intel mobile chips, as nearly as I can tell, offer the very best performance per watt of mobile solution at the high end (The G4 still kicks the crap out of 'em at comparable speeds, but since the fastest mobile G4 Mac you can get is 1.67 Ghz, it's a moot point).

And one thing that the geek community loses sight of is that when we talk about AMD 'kicking the crap out of Intel', it's on a pass/fail basis; overall, they have traded the 'speed lead' several times since the initial offering of the Athlon, and rarely has one lead the other, in dollars per MIP, by more than 3-5%; since most websites that do comparative benchmarks trim the chaff so you can see the difference, the average page scanning consumer or geek gets a warped impression. If we have a scale that's 1000 units long, and Intel's chip does 990, and the AMD chip does 995, and we only show the last 10, it looks like the AMD is twice as fast, when it's really only .5% faster. In these days, a hardware site will pronounce a significant win over a 3% overall difference in performance!

Application also matters. For instance, I do a lot of recording with pro hardware and software. The fact is that most of the software is optimized for the Intel chips much more so than the AMD, so in side-by-sides, I see about 20% better performance for the same hardware and software on my P4 over my Athlon. In some cases as much as 200-300%; I assume that those are REALLY optimized for the P4. But if I run up games on the two machines, the Athlon is 5-10% faster across the board (With the same video card).

Re:AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

Because AMD doesn't have the cool sound effects!

Intel Inside: *swooosh* *bum* *dum* *doom* *ding*

Re:AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

Apple users should be worried.

Their platform got fucked by Steve's ego.

1) Steve tried to get IBM to essentially give them a free new design for a low power version of the 970
2) IBM wouldn't give in to the demand
3) Steve takes his ball and goes home
4) AMD doesn't have the capacity to supply even Apple's paltry number of machines a year
5) Intel becomes Apple's 'first choice'
6) Steve goes out on stage at the WWDC and does a hour long damage control and history rewriting monologue.

IBM has really earned my respect by taking the high road and letting Jobs run his mouth off with lies about the whole mess. It is too bad Steve has fucked up the relationship so bad that the only place for Apple to go is Intel and their horrendously fucked up chip roadmap.

Re:AMD Beating The Crap Out Of Intel?

[Gibsnag]

Most non-computer savvy people that I talk to have no idea what AMD is. To be honest they barely know what a processor, is so expecting them to know about different makes when all the marketing and large white box companies are in favour of Intel is a bit much.

Personally I try and educate any people I talk to about it, but all too often I notice that glazed look come into people's eyes whenever something more complicated than "This is the power button" is discussed.

Re:AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

Intel gave them a much better deal on "intel inside" stickers.

Re:AMD Beating The Crap Out Of Intel?

[Anonymous Coward]

Intel hammers AMD on production capacity.

Re:AMD Beating The Crap Out Of Intel?

[MarkScott65]

Since when did technical superiority have anything to do with market dominance????

Re:AMD Beating The Crap Out Of Intel?

[Jozer99]

How is this +4 interesting? By any logical standard this is offtopic. Anyway, there are simple answers to all of your questions. 1. IBM is cranking out killer PPC chips. True, but they are not cranking them out fast enough, and they are not making good on their roadmap (PPCs were supposed to hit 3+ GHz in 2004). Also, it doesn't matter, as only Apple uses PPCs in mainstream computers, and Intel/AMD/Microsoft aren't just going to switch one day. 2. AMD is cranking out killer x86 chips. Also true, but AMD does not have the production cabilitites to supply the whole industry. Their market share has grown fast in the last few years, and they are scrambling to keep up. Dell does not use AMD because AMDs total production is smaller than Dell's desktop production, leaving out laptops and servers. In order for dell to have decent prices, they need a steady supply of chips, which only Intel can guarantee. 3. Intel is not in the special olympics. True, the Pentium D may not be the best processor ever, but look at the price: 80% performace of an Athlon X2 at 50% the price isn't bad. Also, their single core P4's are still blazing fast, and no one can touch the Pentium M. Plus, Intel's roadmap is full of great processors, while AMD's is just "Well, we thought we'd make some more Athlon 64s..."

Re:AMD Beating The Crap Out Of Intel?

[Kjella]

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

Quite easy when you realize that that majority of consumers don't actually use the full capacity of their CPU very often. If you look at games the GPU is far more important than the CPU, which leaves heavy CPU use to media encoders, compilers and scientic processing. That's not really a big share of the market.

Civ4 mins: 1.2 GHz or equivalent
SW Battlefront II mins: 1.5 GHz or equivalent
Call of Duty II mins: Pentium IV 1.4GHz or AMD Athlon 1.4 GHz or equivalent
Age of Empires III: 1.4 GHz equivalent or higher processor or equivalent
F.E.A.R. mins: Pentium(R) 4 - 1.7 GHz or equivalent
Sims 2: 800 MHz processor or equivalent
Quake 4: 2.0 Ghz or equivalent

Those are some of the latest games released. PIV 2.0GHz was shipping in june 2002, so they are over three years behind the state of the art. And games are normally the most intense apps a user has. Basicly, an Intel machine does pretty much everything a computer user wants to do, so does an AMD. The rest is simply mindshare and momentum. Intel can drop their prices at any time if the market is slipping. They are simply balancing out taking out extra profit versus the threat AMD poses. If they don't watch out, they'll take a spanking in the professional market though, where admins are much more aware of what they're buying...

Re:AMD Beating The Crap Out Of Intel?

[rpozz]

How the hell can AMD be making such better chips and companies like Dell still selling Intel powered crap?

Because most of the machines that Dell sells are to people who will be doing nothing more than running Windows and Office, so performance won't be a massive factor. Intel will also be selling both the CPU and chipset at an enormous discount.

The next-gen Intel chips for Apple should be pretty good if they're based on the Pentium M, which unlike the P4 seems to be able to compete reasonably well with its AMD counterparts. If it isn't, then in a few years Dell will be the only place selling Intel-based systems seeing as most places are losing their patience with Intel.

Either way, the inevitable angry, mis-informed arguments between Intel-using Apple fanboys and AMD fanboys on slashdot next year should be pretty amusing.

Re:AMD Beating The Crap Out Of Intel?

[Nykon]

Easy, AMD can't keep up with the demand. Dell has looked at the option of using AMD before,but at the time (a year or 2 ago) AMD could be making chips 100% of the time only for Dell and still not fill demand. Dell goes with Intell because of the volume Intel can produce.

Re:AMD Beating The Crap Out Of Intel?

[Glock27]

1) Steve tried to get IBM to essentially give them a free new design for a low power version of the 970

Yeah, I'm sure IBM has no use for a low-power 970... /smirk

2) IBM wouldn't give in to the demand
3) Steve takes his ball and goes home

"IBM didn't meet its obligations, so Apple looked elsewhere"

4) AMD doesn't have the capacity to supply even Apple's paltry number of machines a year

Absolutely not true, and will be even less of an issue as Fab 36 comes online next year.

I think Apple wanted Intel mainly because it IS the "market leader", and Apple/Jobs is tired of Apple's "niche" image.

5) Intel becomes Apple's 'first choice'

It most likely has been for some time. Intel was NeXT's primary target when Jobs was CEO there.

6) Steve goes out on stage at the WWDC and does a hour long damage control and history rewriting monologue.

"Steve goes out on stage at WWDC and gets the developers fired up about Intel Macs!"

He didn't do much IBM bashing, as a matter of fact.

IBM has really earned my respect by taking the high road and letting Jobs run his mouth off with lies about the whole mess. It is too bad Steve has fucked up the relationship so bad that the only place for Apple to go is Intel and their horrendously fucked up chip roadmap.

IBM is just a company, no more no less. It made its decisions just as Apple made its decisions - trying to maximize shareholder value. The two sides couldn't come to an agreement - c'est la vie!

Intel's chip roadmap isn't so bad, Intel clearly has the advantage in the mobile space (a very strategic area for Apple), and Intel's desktop chips are competitive if not great. By the time Apple brings out Intel PowerMacs, it will be using a next-generation processor. I'm sure Apple and Intel are both hoping that'll be a great part. Rumors of K10 problems are out there too.

I'd like to see Apple bring out AMD models also - and I think it'll happen once the transition is over, perhaps in 2008-2009.

Re:AMD Beating The Crap Out Of Intel?

[TheSunborn]

CIV IV, really does require a really fast computer to be really playable.

Right now my cpu just match the minimum requirement, and well I can play it, but it often includes more time spend waiting for the computer ai to move, then time spent moving myself.

But the gfx is much better then expected, even on my lowend ATI Radon 9500. But I think it really takes a Amd 4000+ to play that game well.

So for civ, cpu are MUCH more importent then gpu.

Re:AMD Beating The Crap Out Of Intel?

[Chad Page]

And Intel will be cranking out killer x86 desktop chips by the end of next year, too. They already have killer mobile chips and will have even better ones (and dual-core!) early next year. It's just their current desktop line that sucks, and Apple isn't buying those.

Re:AMD Beating The Crap Out Of Intel?

[Ryan Amos]

Intel is Apple's first choice because Intel's mobile processors are better than AMD's -- and don't try to argue this, you will lose. AMD may have the lead in desktop CPUs, but find me a major manufacturer who is making AMD laptops. Then consider that laptops are outselling desktops by a pretty decent margin.

Pentium M is not a bad architecture. It wouldn't surprise me to see Apple move the XServe line to Opterons, but the desktops will probably stay Intel for the forseeable future. Most "workstation" apps are highly optimized for Intel chips.

Re:AMD Beating The Crap Out Of Intel?

[ChadN]

As a G4 Powerbook owner, I'm quite looking forward to the switch. I want my Powerbook running OS X, but I need more oomph than I can get out of even the fastest G4 portable, and that is what switching to Intel will deliver.

Face it, Intel has the best chip for portables out there, and that is THE market to be selling to these days...

That said, when I get my new Linux development workstation in a few months, it'll be dual-core AMD based, not Intel Pentium 4. The P4 is total garbage, and I'll damned before I spend my own money on one.

Re:AMD Beating The Crap Out Of Intel?

[Glock27]

Intel is Apple's first choice because Intel's mobile processors are better than AMD's -- and don't try to argue this, you will lose.

I made a very similar point myself actually. :-)

The difference was that I don't think Pentium-M is "better", I just think it's better established since it's an older chip, and it's (as usual with Intel) well marketed and/or well foisted on manufacturers. ;-)

However, I'd personally argue that the Turions are as good or better than Pentium-M across the board, and 64-bit to boot. We'll see how uptake is going forward.

AMD may have the lead in desktop CPUs, but find me a major manufacturer who is making AMD laptops.

Um, HP? You know, the #1 PC manufacturer? How about Gateway/eMachines? Not to mention Acer, Asus and many other volume Taiwanese manufacturers.

Then consider that laptops are outselling desktops by a pretty decent margin.

By a razor-thin margin you mean. :^)

Pentium M is not a bad architecture.

Nope, the PIII has surprising legs eh?

It wouldn't surprise me to see Apple move the XServe line to Opterons, but the desktops will probably stay Intel for the forseeable future.

We'll see...it certainly depends on what happens in the intervening 24 months or so. Right now AMD's desktop chips are substantially better than Intels.

Most "workstation" apps are highly optimized for Intel chips.

Since the newest Opterons even support SSE3 I'm not sure what "optimizations" are possible for Intel of which AMD wouldn't take advantage. Plus there are lots of Opteron workstations shipping from PC manufacturers (notably IBM, HP and Sun), so one guesses that those workstation apps will increasingly support Opteron as a first-class target.

At any rate, we'll certainly see. :-)

Re:AMD Beating The Crap Out Of Intel?

[Glock27]

That said, when I get my new Linux development workstation in a few months, it'll be dual-core AMD based, not Intel Pentium 4. The P4 is total garbage, and I'll damned before I spend my own money on one.

My plan is to build a system around the Tyan Dual Socket 940 NForce4 SLI board. I'll start with one Opteron 246/248, then in a few months as prices drop add the second CPU. Then in another year or two I'll swap out the two 240-series chips and pop in the (by the hopefully dirt cheap) dual 275s or better. I'm pretty sure such a system will have the best longevity of any system I've owned. It'll be great for both Linux and Windows.

I'll also be buying Intel-based Macs when they come out I'm sure. I'm more than ready to completely migrate off Windows. :-)

Re:AMD Beating The Crap Out Of Intel?

[man_of_mr_e]

While that may be true today, the stage is set to completely reverse that in the next couple of years. Intel has not been idle, they've been working on a very major overhaul of their chips that will likely leapfrog above AMD, and AMD will be in the position Intel is in now (an older architecture that can't compete well).

I'm sure AMD is working on its next generation as well, but for years now Intel sat on it's laurels, content to tweak its existing designs. That changed about 2 years ago, and Intel has been in full bore R&D mode.

Intel has a lot of things AMD doesn't. Superior supply channels, for instance, and the ability to produce much more product than AMD currently can. That's important to a company like Dell. It means you can get enough chips when you need them.

Re:AMD Beating The Crap Out Of Intel?

[Lord Kestrel]

Even on my 3200+, a large map with 10 computers isn't the fastest thing. About the time Gunpowder shows up, there are enough units moving around to definitely slow things down a little. I want to try a Huge map with 16 civs, but I'm afraid it'll be too slow.

Re:AMD Beating The Crap Out Of Intel?

[evilviper]

If you look at games the GPU is far more important than the CPU,

And gaming is the killer app for Apple machines and OS X of course, so your argument isn't pointless crap at all...

which leaves heavy CPU use to media encoders, compilers and scientic processing. That's not really a big share of the market.

Media encoders is probably a rather big share of the market. In fact, it's probably a very big share of the APPLE market.

Besides, those are just the things that immediately came to your mind, not an exhaustive list. How about compression? How about image manipulation? How about numerous forms of encryption? How about HD video playback with CPU-intensive codecs like H.264 (Apple's preferred codec).

Basicly, an Intel machine does pretty much everything a computer user wants to do, so does an AMD.

Only if your definition of a computer user is someone who only checks their e-mail and surfs the web. Or if you are giving CPU-heavy applications an unlimited ammount of time to complete. It's not like that in the real world.

Re:AMD Beating The Crap Out Of Intel?

[therufus]

Intel are the worlds biggest marketing company!!!!!

Oh, and they also make computer cpu's on the side. And THAT'S where the priorities lie.

Re:AMD Beating The Crap Out Of Intel?

[drsmithy]

(The G4 still kicks the crap out of 'em at comparable speeds, but since the fastest mobile G4 Mac you can get is 1.67 Ghz, it's a moot point).

No it doesn't. The Pentium M is as fast - if not faster - than the G4 at the same clock speed, and has more than triple the memory bandwidth on the 533Mhz FSB models.

G4s are *spanked* by Pentium M's, even at the same clockspeed, except for those very carefully chosen benchmarks Apple likes to use. The G4 hasn't been a competitive performer for *years*.

Re:AMD Beating The Crap Out Of Intel?

[drsmithy]

Nope, the PIII has surprising legs eh?

It's even more impressive when you consider a P3 is basically just a tooled up Pentium Pro, which debuted back in 1995.

Re:AMD Beating The Crap Out Of Intel?

[lasindi]

That's easy: Marketing

I love AMD as much as anyone, but there is a reason other than Intel's brand name that Dell doesn't use AMD. AMD simply doesn't have the production capacity that Intel does, and since Dell produces a lot of computers, it needs a lot of CPUs. I always buy AMD processors for my computers, but I can do that because I'm not building hundreds of thousands of machines. I'm not saying that the Intel brand doesn't have an effect; I'm just saying that it isn't the only thing at play here.

Also, as far as mobile processors go, one reason that you don't see many AMD-powered laptops around is that AMD hasn't been quite as successful as Intel when it comes to keeping processors cool, which is especially true with laptops. I can now personally attest to this (though it involved my desktop's processor, not a laptop); I had to replace my AMD Athlon XP due to overheating and causing processor-intensive programs to crash.

Re:AMD Beating The Crap Out Of Intel?

[NixLuver]

Again, you're mislead by the same things I mentioned in the original post. A quick survey of benchmarking sites shows that the g4s (which peak at 1.67 Ghz on mac) are roughly comparable to the contemporary Pentium Ms (1.6 and 1.7); I presume you would not consider it a reasonable comparison to pit a 1.67 G4 (one or two generation old tech, depending on how you look at it) against the Pentium M's above 1.7 Ghz (current tech, or at worst one generation old). And even then the difference in overall performance between, say, a 1.67Ghz G4 and the Pentium M 725 (current gen without question) is somewhere between 1% and three %. The bandwidth of the current gen Pentium M's IS higher - I would certainly HOPE so, since the g4s are at least a full processor gen older. If you compare the M's to the current gen PPC chips, the story is a little different, but you're missing my whole point. You use the term "spanked" to describe differences of less than a full percentage point - particularly when a PentiumM 1.7 (Banias) is *not* as fast as a 1.67 PowerPC G4 - you have to go to the 725 or 735 for that; yet based on a very limited performance advantage, you're willing to pronounce the G4 "trounced". If these were humans racing down a track, they would be photo finishes. Which all illustrates the fact that we're straining over gnats in processor performance, particularly when most people cannot reliably discern a performance difference of less than 50%. We talk about games driving technology, but if you take the frame counters away, most people can't discern the difference between a system that will put up 200 fps at Q3A, and one that will do 230fps, but a 'tech site' or 'game site' would call it "pwned".

Re:AMD Beating The Crap Out Of Intel?

[drsmithy]

A quick survey of benchmarking sites shows that the g4s (which peak at 1.67 Ghz on mac) are roughly comparable to the contemporary Pentium Ms (1.6 and 1.7);

Ah. So they don't really "kick the crap out of 'em at comparable speeds", like you said earlier ? (Incidentally, I really wish people wouldn't say "speed" when they really mean "clock speed".)

Without knowing what particular benchmark sites you're referring to, it's difficult to really respond. But with a pretty much identical IPC, anywhere from ~2.4 to ~3.2 times the memory bandwidth and 2x or 4x as much L2, the Pentium M should be a noticably faster chip if you're actually doing something that doesn't fit in the CPU's L1 cache and/or isn't I/O limited (and assuming that CPU performance is something that actually matters to you).

Then then there's the fact that Pentium Ms are available at up to 2.26Ghz, or about 25% higher clock speed - so the 1.7Ghz Pentium M that stacks up quite well against a top of the line 1.67Ghz G4 is relatively slow, by Pentium M standard. This is not something that can just be waved away, as you Mac fans like to try and do.

The bandwidth of the current gen Pentium M's IS higher - I would certainly HOPE so, since the g4s are at least a full processor gen older.

Actually, by the measure you make a Pentium M "current gen", the G4 is as well, as it still receives updates (like that boost to a 167Mhz FSB - whoa, hold onto your hats - and a reduction in process size). Particularly since a Pentium M isn't really much more than a P3 sitting on a P4 bus.

yet based on a very limited performance advantage, you're willing to pronounce the G4 "trounced"

Much like you were, you mean ?

Re:AMD Beating The Crap Out Of Intel?

[NixLuver]

Ah. So they don't really "kick the crap out of 'em at comparable speeds", like you said earlier ? (Incidentally, I really wish people wouldn't say "speed" when they really mean "clock speed".)

Yeah, you're right here; I engaged in the same kind of hyperbole I was attacking. ROFLMAO. Doesn't make my point any less relevant. The point I was making, when we got sidetracked into chip bashing - and I'm not an intel basher at all, contrary to your apparent perception - is that we blow small performance differences all out of proportion (yes, I did so as well, touche) to actual difference in operational measurements and specifically human perception. 5% would make a big difference in the budget of a big-time high resolution feature movie rendered in 3d with raytracing and radiosity... cutting five days off of a hundred day job might save thousands of dollars - maybe hundreds of thousands of dollars. But running applications personally, if someone thinks they can see 5% difference between one processor and another, I'd like to see their perception tested with double blind studies.

LOL... speed bumps and die reduction don't count as 'new generations', usually, in the generally accepted parlance. And I bet Intel would take exception to your claim that a pentium M is nothing more than a p3 sitting on a p4 bus. And I wasn't 'waving anything away', and I'm not so specifically a fan of the power pc chip so much as I am a fan of Mac OS X. The fact is that there are at least two versions of the Pentium M in 1.7 Ghz, and one- the newer one - is MUCH faster than the earlier one. The original was the Banias(Sp?) and the new one is the 725(or 35, whichever is 1.7). The 1.67 Ghz g4 stacks up favorably against P4s, at clock speeds upwards of 2.2-2.6, depending on benchmark and type of application and optimizations. It beats the 1.7 Pentium M (Banias), and loses (as you pointed out) to the newest Pentium M 735 at a clock speed of 1.7 Ghz. By a tiny amount, in most applications. I am not religious about processors at all; I will be just as happy to run Mac OSX on intel as I am on PPC (I can't afford the dual dual core, as much as I'd like to, and I'm interested in seeing how Apple follows THAT one up with Intel hardware). I'm actually looking forward to laptops running Intel chips - that are, as you pointed out, MUCH faster then their PPC brethren, when we can buy a Pentium M at 2.2 Ghz, isn't it?

Regardless, you're certainly right to call me out for the hyperbole; but I would still like to see your sources for claims like "The Pentium M is just a P3 on a P4 bus".

next step?

[Anonymous Coward]

Personally I never imagined integrating a PCI Express controller in a CPU. If this trend of intregation continues, what would be the next logical step?

Re:next step?

[theantipop]

Graphics controller on chip. Intel has shown such a chip at IDF this past fall.

Re:next step?

[Danovitsch]

The next step would be a processor the size of a motherboard with all connections (USB / PCI) on it's back :)

Re:next step?

[drunkahol]

It's got to have a floppy drive built in. Things just DON'T work unless there's a floppy drive!

Re:next step?

[Barny]

Integrated cpu/northbridge/southbridge on a 12,156 pin socket, some say it will support 4 channels of ddr4 and be able to heat a large home :)

Re:next step?

[Anonymous Coward]

To invent a new marketing TLA buzzword of course! Can't keep calling it a CPU... maybe MPM. Many Pinned Monstrosity.

Re:next step?

[hal2814]

The next step would be that my co-workers would actually be correct when they refer to the box that houses the motherboard, video card, memeory, etc. as a "CPU."

Re:next step?

[vrioux]

AMD already let rumors go out about their possible integration of PCI-e. Next, we don't know but I suspect they'll get interrested in GPUs. With a modular multiple-cores architecture, it wouldn't be that hard to insert a GPU core in there. The only problem is heat... The advantages are insane : PCI-e on the cpu means the chipset doesn't need to use the hypertransport bus to route that traffic between cards and cpu! You get a direct connection between whatever cards you have and the main CPU (which is also the memory controller).

AMD really does have an edge with this technology and I hope they get as far as they can with integration. As one poster said earlier, one day we'll have a full x86 system-on-a-chip (SoC), which means cheaper boards, no need for cards, and maybe even no need for any hard drive. I hope AMD gets us to that level.

Re:next step?

[pla]

Personally I never imagined integrating a PCI Express controller in a CPU. If this trend of intregation continues, what would be the next logical step?

Single-chip computers - A CPU, and a totally passive backplane that does nothing but provide real-estate for connectors. And most likely, you wouldn't strictly need any extra cards, with a decent (but not high end, thus the need for a bus at all) GPU included right on-die.

Realistically, I expect two-chip computers as far more likely. Something along the lines of having your CPU plug directly into your video card, which has the standard video card parts on one side, and standard motherboard connectors on the other. And the whole thing could mount via a SECC-style connector to a power bus, right inside something just a tad bigger than current ATX power supplies.

Drives? Uhhh... I'll have to think about that one. ;-)

Re:next step?

[cyxxon]

Hm but what will then happen to my coworkers who always refer to it as "the harddrive" (and in reality store all their stuff on a mapped network drive)?

Re:next step?

[Anonymous Coward]

The next logical step is to integrate the user into the chip. Chips will become huge, and the housing market will merge with the computer industry.

Re:next step?

[theJML]

A north bridge/southbridge on chip would be cool... but let's not forget the tail of the TV/VCR Combo's... When the TV Breaks, what do you do with the VCR?

Personally, I think putting many things that intercommunicate on the chip is cool, but I'd rather not have a GPU on the same chip. For one, It't not going to be cheap when MS comes out with Dx10 and everyone has to support it hardware (as is happening with 9), now I'd need a new CPU chip and it can't be cheap being that video cards are so darn expensive. Or if I decide I want to go dual core, I have to buy the video card again because it's on the chip.
 
  Seems like there needs to be a line drawn somewhere.

Re:next step?

[Shanep]

Personally I never imagined integrating a PCI Express controller in a CPU. If this trend of intregation continues, what would be the next logical step?

The display, keyboard and mouse.

Re:next step?

[bhtooefr]

Actually, the buzzword already exists. System on Chip, or SoC.

Re:next step?

[cerelib]

2 GB of main memory on the chip, or better yet a GPU. Then a wireless chip and finally the audio subsystem. If we can put all of these things on the chip we could have a griddle and make pancakes.

Re:next step?

[NeoSkandranon]

With computer hardware of this day and age wouldn't that create really unmanageable heat density? Don't get me wrong, I love the idea, cause that would mean you could get something about the W x H dimensions of an LCD monitor just not as big on the large face (if I'm envisioning what you meant correctly) --just thinking of how to dissipate the heat from a Radeon 9800 and an Opteron if they're literally on the same board

Re:next step?

[edxwelch]

Did the artical really say that? I can't read dutch, but the summary only says that this new socket is for support of DDR2

Re:next step?

[Xepherys2]

Didn't we learn anything after Cyrix?

Re:next step?

[Jeff DeMaagd]

I'm not sure if puting the PCIe bus on a chip is a good idea for a general purpose CPU, but then, I've been wrong about stuff like this. I didn't think that puting a memory bus on a general purpose CPU was a good idea either, but apparently, it worked out pretty well.

I think the Alpha EV8 was supposed to have a built-in network interface, I don't know what sort of network interface it was supposed to be though, if it was Gigabit ethernet, 10Gig ethernet or just a generic processor network. Projects are underway to put network on other CPUs, the PASemi project is supposed to be one.

Re:next step?

[Dastardly]

I agree Memory. More specifically, large quantities of DRAM. This is more a result of the decreasing size and trying to figure out what to do with all the additional transistors available. The GPU is an interestign idea, but I think DRAM is easier solely because the heat densities on GPU and CPU are so high that the lower heat density of DRAM would make it easier to add to a CPU. In all likelyhood the DRAM will actually act more like a gigantic exclusive L3 cache, of course with the option of having no offboard memory. At 45nm I bet they could fit the equivalent of an 8Gbit DRAM chip, getting 1GB of memory before adding any external memory.

Re:next step?

[Anonymous Coward]

They will die and end up in a fiery, bug-ridden computer hell. Where they rightly belong.

Re:next step?

[ozbird]

Personally I never imagined integrating a PCI Express controller in a CPU. If this trend of intregation continues, what would be the next logical step?

More "pins" - it's the new MHz.

Re:next step?

[archen]

CPU is short for Central Processing Unit. It's not hard to see why a normal person would see the box as the Central Processor, since the mouse,monitor,etc do not do any processing. If the term were a bit less vague (like Machine Instruction Processor or something) then they would have less of an excuse.

Where I work there was this guy who referred to it as the 'modem'. He also gave me a crunched up paper bag to scan a logo from to reproduce customer artwork. He claimed that he "scanned" things all the time and that it was super simple. But then again maybe he's got a huge modem and a scanner that can digitally reconstruct anything...

Re:next step?

[pla]

cause that would mean you could get something about the W x H dimensions of an LCD monitor

Actually, I had something more like a double-high and slightly deeper DVD drive in mind (including an actual DVD drive as part of it, of course. But then, I prefer cubes to pizza-boxes, personally... The actual shape wouldn't much matter. "Much smaller", at any rate, with the cooling probably taking up more room than the active components themselves.

I suppose that would make water cooling a lot more attractive - Considering a flatpanel monitor as the form factor, how about an LCD on one side, an inch thicker than a typical current displays, and a 19" (diagonal) aluminum radiator/heatsink on the back?


just thinking of how to dissipate the heat from a Radeon 9800 and an Opteron if they're literally on the same board

Ever opened up a 1U rackmounted dual Xeon machine (like a PowerEdge 1750)? Desktop-stype CPU cooling could do a whole lot better if we didn't insist on trying to cool the CPU with already warm air. Think of a heatsink wrapped around on itself to form a rectangular tube 1U high, then take air from outside the case into the tube, and vent it outside the case on the opposite side.

Re:next step?

[timeOday]

If they keep putting more stuff on the CPU, at some point there should be a decrease in pin count, as there isn't much offboard to talk to any more! I like the other guy's idea of putting the DRAM on there, that should slash the pin count. Personally I've been stuck at 1GB DRAM for a few years now and don't need more. Let's make it 2GB onboard DRAM and call it good.

Re:next step?

[doorbender]

YEAH BABY!!!

Re:next step?

[Rod Beauvex]

It always irks me when they call it the modem.

Re:next step?

[ArbitraryConstant]

Given that more than two cores has quickly diminishing returns unless you have rediculously CPU-bound multi-threaded code, I would be surprised if we didn't see some very large caches or even a portion of main memory on die. OSes are getting NUMA support for Opterons, on-die memory would be another case of that.

Re:next step?

[doorbender]

Whats the limit to how many PCI slots a cpu can use? is it simply a matter of IRQs in the BIOS? Could a CPU that controls PCI slots independant of the motherboard use a larger number of PCI slots? This could be a huge breakthrough.

Are current windows oses limited to less than 14 irq?

Re:next step?

[pantherace]

Oh wait, Cyrix did that.

One might suggest that there is a reason it's not been popular.

Re:next step?

[Kjella]

Single-chip computers - A CPU, and a totally passive backplane that does nothing but provide real-estate for connectors. And most likely, you wouldn't strictly need any extra cards, with a decent (but not high end, thus the need for a bus at all) GPU included right on-die.

What about heat output? The next logical step isn't to shove them closer together - it's to use the real-estate you already have. Either the iMac G5 model, or back to the commodore model (your keyboard is your computer). If you detached the LCD, removed the space for the battery, that is essentially what a laptop is today. A keyboard with a computer inside.

Re:next step?

[MBGMorden]

Much better than where I'm at. Most of the users refer to the big box as the "hard drive". I used to work IT support at a college. One professor came down to the computer room all frantic one day because "somebody stole his hard drive". We were all thinking "Damn, they went through a lot of trouble to go in and get the hard drive." before we actually got up there and discovered that they'd taken the whole system.

Of course, there was another instance where someone did gut out a system in one of the computer labs and just left the case and power supply (and external peripherals) behind . . .

Re:next step?

[MBGMorden]

You know, you have brought up an interesting possibility (though indirectly):

Replaceable GPUs.

Think about it: why waste a whole PCB and connector set(and oodles more components) each time you need a card? I'd bet that it wouldn't be that hard for manufacturers to make one "platform" of card for each series of GPU that came out. Geforce 5000 series takes one card, 6000 series another card, ATI 9000's take one card, etc.

Couple that with removeable and standard video memory modules that plug into the cards and it could be a very nice upgradeable portion of the computer. Wouldn't it be a lot better to just buy a 256mb video memory module and be done with it (for a while at least) rather than buying video ram on every card you purchase?

Re:next step?

[Seehund]

AFAIK, calling "the box" CPU comes from when home computers ususally were all-in-one, i.e. integrated computer + keyboard and sometimes screen in one unit. For those computers that actually had separate keyboard et c. units, the "computer box" was referred to as the CPU.

Your co-workers aren't necessarily wrong, at least not if they're "old" experienced geeks.

Re:next step?

[fairyliquidizer]

Actually the traditional definition of CPU was indeed the system unit that contained what was then called the MPU (microprocessor unit, what is now called the CPU), primary storage, and possibly some other stuff I can no longer remember. Basically the rest of the computer consisted of peripherals (i.e. things that were peripheral to the CPU). So your colleagues are correct as far as the original definition of the term. Unfortunately modern usage says they are no longer correct. Try Googling and you will struggle to find the old definition but look in a computer architecture text book from the 1980s and you will find the definition that they are using.

Re:next step?

[Eightyford]

Replaceable GPUs.

I think the technology is moving too fast for that. It rarely makes sense to even replace a CPU these days.

Re:next step?

[MBGMorden]

Maybe not the CPU (I'll admit that lately I've been switching mobo's when I switch CPU's), but I think we could at least benefit from having the RAM be replaceable. That could let us cheapies hang on to our RAM from card to card (within reason; everything gets outdated eventually) and would actually open up a large market for "premium" video RAM to cater to the hard core gammers.

Re:next step?

[cskrat]

There's already alot of people whining because 3-4 gig isn't enough.

If AMD produced a chip with 2 gig onboard they'd also include a means of accessing external memory. Since some memory is on the CPU silicon and some is sitting on the mainboard, there will be a difference in speed and access methods to get to the two areas of RAM. This would end up creating something similar to the 640K barrier that emerged in the DOS days.

We do not need another one of these little quirks to further confuse the issue of memory management on the x86 platform again.

Re:next step?

[obeythefist]

Next step, intel will take AMD to court for illegally bundling the PCIe slot with the chip, forcing consumers to use AMD's PCIe instead of competing in the marketplace. We've all seen this before, embrace and extend.

Re:next step?

[Anonymous Coward]

would be the next logical step?

Not video, except for the super-cheap integrated market. Video cards would require too much silicon and put off too much heat to pack them in with the CPU cores. Plus, video cards typically get upgraded twice as often as CPUs.

Not RAM. That stuff also requires a lot of real estate on the silicon. The amount of real estate needed to bring the memory controller onto the CPU was minimal in comparison.

There's not really anything else on the board that is high bandwidth that isn't already being tucked into the system. Faster access to the drives is a possibility, but bringing the PCIe closer in probably gets you that already.

Usual reaction

[Anonymous Coward]

/me experiences a short spasm, followed by the urge to comment "imagine a beow..." but then thankfully everything's back to normal again.

Someone trasnslate the article

[majjj]

Dutch people take the lead... no point in looking at the pictures without anything to read.

Re:Someone trasnslate the article

[ghamerly]

No one else is posting a translation. I'm not Dutch, but I'll take a crack at it (I spent a year in Belgium (Leuven) learning Dutch... and I like to practice it every now and then, so here's my chance). I have to get back to work, but I did do the first paragraph (Dutch speakers please feel free to correct me).

The first photos of AMD's Socket F have emerged on our Gathering of Tweakers
forum. In May we wrote that AMD had a new processor socket on its roadmap. The
new footprint should have 1207 pins and is intended for multi-Opteron servers.
To make possible a processor with support for DDR memory on a DDR2-footprint
and vice-versa a new socket was needed. The extra pins that are available are
according to reports for an integrated PCI-Express controller on the
processors. Noticeable in the photos is the clear separation in the middle of
the socket. This seems to indicate that each core of the dual-core Opteron has
its own group of pins, and so works as two processors.

Re:Someone trasnslate the article

[Anonymous Coward]

Quick tranlation and not very tech-savvy, but here it is.

The first pictures of AMD' Socket F have appeared on our forum "Gathering of Tweakers". In May we already wrote that AMD had put a new processor socket on its roadmap. It was supposed to count 1207 pins and to be built for multi-Opteron-servers. In order to prevent that a processor with support for DDR-memory is inserted into a DDR2 socket and vice versa, a new socket was necessary. The extra pins that thus become available, would be used for an integrated PCI Express-controller on the processors. Remarkable on the pictures is the separation in the middle of the socket. It seems to indicate that every core of the dual-core Opteron has its own group of pins and is hence treated as two seperate processors.

On the pictures you can also see that Socket F, just as Intel Socket 775, is equipped with pins to contact the processor. The cpu will no langer have to be inserted into the socket; it is a so-called Land Grid Array socket. By the way, Socket F is also called Socket 1207, but just like Socket 479 only counts 478 pins, this model could also have only 1206 contact points, detailed counting work has shown. This socket also supports registered DDR II 553-, 667- and 800- memory and AMD seems to try to counter Intels DIMM plans. The latter will present its dual core platform Dempsy, with amongst others the Greencreek chipset with support for FB DIMM memory.

Re:Someone trasnslate the article

[Anonymous Coward]

I take it you've never picked up a skin mag?

Re:Someone trasnslate the article

[Anonymous Coward]

Did you go to K.U. Leuven? I know an American who went there.

Go AMD!

[Janek Kozicki]

I'm an AMD fanboy, oh yeah. Their chips are soooo good. Problem is that in Europe there wasn't enough press releases about AMD vs. Intel dual core duel. In fact there is nothing about that. Looks like all PR quietly took a large sum of money from Intel, and this duel is totally ignored by media. I feel bitter about that.

yes, I heard that AMD has launched a big ad campaign in US, but sadly this is not the case for Europe.

and all the universities in western europe are forced to buy upgrades from Dell. I really tried to buy AMD, it simply wasn't possible.

Re:Go AMD!

[stecoop]

I heard that AMD has launched a big ad campaign in US, but sadly this is not the case for Europe

Funny, with Adblock, Flashblock, Tivo, Netflix and Bitorrent, I can't recall the last commerical or ad I've seen.

Re:Go AMD!

[theantipop]

Excellent. The government eat wendys subliminal advertising scheme is working perfectly. World buy pepsi domination is at hand.

Re:Go AMD!

[ivan256]

How much did they pay you for this guerilla marketing?

Ads in slashdot comments... What's next? Ads in the stories?

Re:Go AMD!

[sznupi]

No point really. The ones who would wish to know how good is AMD already know this. Some poeple have very close relatives who have taken care of this. Rest will just believe what salesmen tell them no matter what. So they usually end up with something like P4 with 256MB of RAM on cheapest ECS board with Via chipset. And Nvidia Gf5200 (hey, the board is total no name, but it's Nvidia after all, and Nvidia is the best). And cheapest LCD monitor NOT connected via DVI (it's not that salesmenn don't tell about DVI, they say "it's pointless, when you turn on font antialiasing it looks identical on thois cheaper display" - yep, heard that). And Creative soundcard, even the simplest and most shitty, but - it's Creative. And did I mention cheapest case/keyboard/motherboard/HDD?

Re:Go AMD!

[Anonymous Cowpat]

"and all the universities in western europe are forced to buy upgrades from Dell."

Nonsense! We have Viglens with P4s in them...

NP

[kevin_conaway]

Ahh nerd porn. While the rest of the world is looking at leaked photos of Janet Jackson or Paris Hilton, we're looking at photos of AMDs new processor.

Re:NP

[Anonymous Coward]

>While the rest of the world is looking at leaked photos of Janet Jackson or Paris Hilton

People do that? Eeeeww..

>we're looking at photos of AMDs new processor.

Well, unlike Janet Jackson and Paris Hilton, a new CPU socket is actually sexy.

Re:NP

[BushCheney08]

You might be looking at the processor, but I'm looking at the hole the processor goes into... : p

Re:NP

[Anonymous Coward]

and considering the two skanks aforementioned, I thank $deity I'm looking at computer hardware and not those two...

(yes! I made a nerdy post! glass houses and stones!)

Re:NP

[evilviper]

While the rest of the world is looking at leaked photos of Janet Jackson or Paris Hilton, we're looking at photos of AMDs new processor.

No, no... This isn't even pictures of a new AMD processor... it's pictures of the SOCKET where the processor will go.

It's more like pr0n pictures of a bra or a bikini, without anyone wearing them... :-(

Re:NP

[T-Ranger]

Its a picture of holes.. So that makes it porn for straight male geeks, or lesbian female geeks.

Re:NP

[interiot]

Is that socket, erm... used?

Re:NP

[Vo0k]

Nope, that's closeup of the sexxy part of the motherboard, where the CPU inserts all its pins.
Since the introduction of the zero-force sockets, plugging a CPU in is not that arousing anymore...

Re:NP

[MouseR]

Wich goes to say that in this world, there's a socket for everyone.

Re:NP

[xtracto]

but I'm looking at the hole the processor goes into... :

Just imagine! One thousand, two hundred and seven holes!
woooooooah!

*head explodes*

Re:NP

[Pogue Mahone]

Here is a picture of the Paris Hilton, along with a few other Paris Hiltons. But probably not the Paris Hilton that you were thinking of.

Re:NP

[itomato]

Nope & nope - that's a picture of all the pins on the socket where the CPU will *rest*.

A tiny bed of nails for dual core Opterons. Helps them focus! Ssh!

Re:NP

[MindStalker]

I dont know the picture looks like a bed of pins rather than holes. But I could be wrong...

Re:NP

[zakezuke]

While the rest of the world is looking at leaked photos of Janet Jackson or Paris Hilton, we're looking at photos of AMDs new processor.

I know I'd rather see a cpu than Paris Hilton. You can at least try to carry on a convo with a cpu.. and odds are it will have something more intelligent to say.

Re:NP

[Anonymous Coward]

who was looking for janet jacksons ugly tit?

honestly, that was pretty hideous

Re:NP

[bhtooefr]

Uh, xtracto, those aren't holes.

You're watching it finger the CPU. 1,207 fingers.

Damn, that must be one happy CPU!

Re:NP

[Surt]

Are you kidding me? The socket is clearly the female side of this connection. Any nerd caught looking at the pins on his CPU may as well be checking out his friends ... you know.

Re:NP

[be-fan]

Um, not on this CPU :)

Re:NP

[Anonymous Coward]

No, its a FEMALE connector.

The cpu is Male.. you can imagine how we came up with this classification.

This is more like... lesbian porn.

Re:NP

[ichigo 2.0]

I have only one pin, you insensitive clod!

Re:NP

[turgid]

It's more like pr0n pictures of a bra or a bikini, without anyone wearing them... :-(

If someone's wearing them then is isn't pr0n, is it? Maybe I'm missing something...

Re:NP

[darkmeridian]

I thought we agreed "NP" stands for "Natalie Portman" on this forum. Wait. I guess the two are synonymous.

Carry on!

Re:NP

[evilviper]

It's absolutely ASTONISHING the number of people who have responded to me, clearly without actually clicking on the story. All with the exact same response...

Re:NP

[Deliveranc3]

If this is wrong then I don't want to be right.

That sure is alot of pins

[Ellis D Trippman]

with that kind of pin density, i wonder what a speck of dust getting into that socket would do to a system. Although it looks like a BGA socket so there may be less of a chance of a bad contact. It's cool that they are going to be intergrating more system controller logic into the main CPU, it's the evolution of the pc design, eventully I forsee single chip systems with minimal external circuitry, just one 5000 pin chip and a small wafer with some io connectors. Cool!

Re:That sure is alot of pins

[Jozer99]

Actually, probably not too much damage would come of a bad connection, just the chip would not power up until the problem was removed. Now if you want to see smoke, put the chip in the WRONG way, like rotating it 90 degrees. I wish they would make chips that were much harder to do that with. I built an LGA 775 system this summer, and all there is keeping you from inserting it wrong is a teeny little tab on the side of the chip. Now, if you were to make an "L" shaped chip, or at least an "L" shaped socket, it couldn't be put in wrong.

I think I found a UT....

[Anonymous Coward]

http://www.worldlingo.com/S9M7KdqhsbwbuIowUkjYjN1m uyzrtd1hA/translate

(If it doesn't work just go to worldlingo.com yourself ;)

Pictures of CPU sockets?

[Broken Bottle]

If that isn't geek porn, I don't know what is...

Chris

Re:Pictures of CPU sockets?

[level_headed_midwest]

A girl with taped-together glasses, a pocket protector, and a "How to Program in Machine Code" book standing naked would be geek porn.

.nyud.net:8090 images

[TomasDK]

The images:

http://img259.imageshack.us.nyud.net:8090/my.php?i mage=0015ep.jpg http://img259.imageshack.us.nyud.net:8090/my.php?i mage=0024yp.jpg http://img259.imageshack.us.nyud.net:8090/my.php?i mage=0032cd.jpg

Just in case

Translation

[Anonymous Coward]

The first photographs of AMD's Socket F have shown up on our Gathering of Tweakers forum. We wrote about AMD having put its new processor socket on its roadmap last May. The new socket is said to have 1207 connection points and is intended for multi-Opteron servers. To prevent the insertion of a DDR-supporting processor into a DDR2-socket and vice versa, a new socket design was necessary. The extra pins that came available are said to be used for an integrated PCI Express controller. What's remarkable is that there's a clear separation in the middle of the socket. This could indicate that each core of a dual-core Opteron has its own set of contacts and thus is treated as two separate processors.

The photographs furthermore show that Socket F, as Intel's Socket 775, will feature pins that make contact witht he processor. This is a so-called LGA socket: the CPU will no longer feature pins that have to be pushed into the socket. Socket F is also called Socket 1207, but carefull counting reveals that the socket only features 1206 pins. This socket supports DDR II 533-, 667- and 800MHz memory and this allows AMD to compete with Intel's FB-DIMM plans. The latter is scheduled to introduce its dual-core Dempsey platform in April, featuring the Greencreek chipset with support for FB-DIMM memory.

How do they do it? Volume Volume VOLUME!

[caveat]

While AMD and IBM make technically superior chips, they simply don't have the mass manufacturing capability to compete with Chipzilla; a side effect of the huge capacity is the ability hae the quantity of procs available to offer deep discounts to high-volume customers (e.g. Dell and Apple) and still make money.

On a side note, the stuff due to be out of Intel by the time Apple switches the PowerMacs doesn't look too shabby at all - of course, we'll have to see what IBM/AMD are offering to compete.

Re:How do they do it? Volume Volume VOLUME!

[Anonymous Coward]

"On a side note, the stuff due to be out of Intel by the time Apple switches the PowerMacs doesn't look too shabby at all"

Actually, it does look shabby. And there is still a year's worth more of Intel cancelations and slipage.

Mac users are going to be in for a shock when they see just how bad Intel's dual core chips are going to be compared to IBM and AMD's next year.

I can't wait to see how Steve Jobs tries to spin the release of desktop Intel machines next year:

1) Pretend AMD doesn't exist
2) Sandbag their PPC systems - play games with optimization settings and most likely use lower clocked 970 chips than are available

Re:How do they do it? Volume Volume VOLUME!

[Anonymous Coward]

3) Ignore desktop pentium detractors, while selling shitloads of Powerbooks and iBooks that are WAY faster than current models, and don't scorch testicles.

As horrible a chip as the Pentium 4 is, the Pentium M really is a gem. Intel will eventually (I assume) stop having their asses handed quite so badly to them with the Desktop and Server chips, but no one can laugh at their portable offerings anymore. And for Apple, I'm guessing that is what is most important.

Re:How do they do it? Volume Volume VOLUME!

[RzUpAnmsCwrds]

While AMD and IBM make technically superior chips, they simply don't have the mass manufacturing capability to compete with Chipzilla

Perhaps not for the entire industry (though, according to AMD reports, with Fab 36 and Chartered online, they will be able to supply as much as 50% of the x86 market in 2006), but AMD has more than enough capacity for Apple.

Remember, AMD was capable of 50 million dies in 2005, and it looks like they may be able to double that in 2006.

Re:How do they do it? Volume Volume VOLUME!

[hyc]

It seems to me that IBM has plenty of manufacturing capacity. So much goes unused that they hire out to other companies. The problem here wasn't lack of capacity, it was lack of importance. The number of PPC chips Apple buys is miniscule compared to all the other business IBM has going on. The biggest market for PowerPC architecture these days is in embedded systems, and the volumes there are many orders of magnitude greater than Apple will ever sell. It will be a shame when Apple drops the PPC and x86 is the only desktop architecture left, but the overall impact on PPC sales will be nil.

Re:How do they do it? Volume Volume VOLUME!

[evilviper]

While AMD and IBM make technically superior chips, they simply don't have the mass manufacturing capability to compete with Chipzilla;

WILL YOU PLEASE SHUT THE FUCK UP!!!

I'm getting so incredibly tired of hearing this same crap over and over again.

Intel has been having supply problems lately. There have been mentions of it in the press, repeatedly.

Meanwhile, AMD is opening a new fab, has a contract with 3rd parties to produce chips if AMD needs them to do so, etc.

http://slashdot.org/comments.pl?sid=158117&cid=132 47676

Yet another socket

[Cerberus7]

Yay. I'm still on the fence if all of these different sockets are a good thing or not. I've gone from Socket 7 to Super Socket 7 to Socket A over the course of the last several years. Now it seems that there are way too many different sockets to choose from, and who knows which will show the same kind of longevity that my past choices have. What's a guy to do?

Re:Yet another socket

[El_Muerte_TDS]

elmuerte is waiting for one socket to rule them all.

damn, that must be pure evil.

Re:Yet another socket

[yfkar]

But it sure can handle high temperatures.

Re:Yet another socket

[Surt]

Realize that a new motherboard is only about 10% of overall system cost, and that by the time you can afford and need a processor upgrade you can probably afford a motherboard upgrade to go with it.

As a backup, the best strategy will tend to be to buy early in a socket's lifetime. Buy a new socket with a low end chip, and figure to upgrade to the highest chip that socket will support when said chip becomes cheap.

Re:Yet another socket

[Jendi]

There really aren't that many to choose from -- your choices are basically defined by "server vs desktop" motherboard (eg socket 940 dual opteron/registered memory or socket 939 desktop) and "AMD vs Intel". After that, sure you'll want to stretch your current investment as far as possible, but at some point you have to bite the bullet and replace your motherboard and memory, how else are you going to keep getting loads more lovely memory bandwidth for your system?

IMHO, I'm going to try and wait until AMD M2 motherboards are available for a vaguely sensible price, and move to DDR2 memory with a dualcore Athlon X2 ... hopefully some time next year -- I reckon that will be the next platform with as much longevity as Socket A gave me.

Re:Yet another socket

[ninjagin]

Aye, your strategy is my own -- works great, too. I did this with Socket A years back and again with socket 478.

As a side note, I've found that I get better longevity out of a build (or "assembly", for those who are picky about terminology) if I wait for a board that has a combination of well-developed solid features on it, too.... sound, on-board VGA, SATA, SPDIF support, etc.

What it probably comes down to is letting enough time go by such that all the major moboard manufacturers have the opportunity to make competing offerings. If I buy too early in the introduction of a new socket/processor type, I find that there are still little tweaks or driver oddities to deal with. Seems like the magic time starts at about 6 mos after introduction.

Oddly enough, I've probably let the start of the window go by for Athlon 64, and with my next build (er, "assembly") slated for mid-next year, I'll have too much to choose from. Oh well, maybe next time.

Re:Yet another socket

[smoker2]

I'm going to try and wait until AMD M2 motherboards are available for a vaguely sensible price, and move to DDR2 memory with a dualcore Athlon X2 ... hopefully some time next year -- I reckon that will be the next platform with as much longevity as Socket A gave me.

Yeah right, then they'll design a new graphics card interface, or new network capabilities and the board will still be old.

I wish AMD and the m/board manufacturers would keep using / making socket A. Lots of users don't need the latest hot sh1t, but we have to at least move to 754 just to get a relatively modern processor and pci x. Why ? because all the new graphics are pci x.

Sound familiar ?

and it's not all fashion driven either... add on cards and processors do die from time to time, which is annoying when you know you have to completely replace an almost perfectly working, 4 year old machine, just because 1 removable component has been EOLed by the manufacturer. I thought we are supposed to be saving resources, not forcing the public to throw stuff away just because we got a new one to sell.

Re:Yet another socket

[NeMon'ess]

It's no good complaining about the socket, because even if that stayed the same, the rest of the board you buy today or tomorrow wouldn't be compatible with DDR2 RAM or perhaps DDR3 RAM tomorrow.

Re:Yet another socket

[William-Ely]

The short answer is none of them will. The only reason socket A lasted so long was the memory controler was still part of the north bridge. Now that the memory controler is integrated a new socket is required every time memory technology changes or a new feature is integrated.

This may put a lot of people who are looking for a long term investment in their mother board but by the time the new socket comes out it would probably be in their best interest performance wise to go ahead and upgrade. Just my $0.02

Routine Babelfish Translation:

[Anonymous Coward]

On our forum Gathering or Tweakers the first photograph of AMD's Socket f has emerged. In May we wrote all that AMD new processorsocket on its roadmap had put. The new voetstuk 1207 connection points would count and is intended for multi-Opteron-servers. To occur that a processor with support for Ddr-geheugen are pricked on Ddr2-voetje and vice-versa, therefore new socket were necessary. The extra pins which become available, could would according to reports be used for incorporated PCI Express-controller on the processors. Striking on the photograph the clear separation in the middle of the socket is. This seems indicate that each core of dual-core the Opteron his own group has contact points and this way real such as two processors it is treated.

First photograph Socket f emerged Yoeri Lauwers - Tuesday 8 November 2005 - 09.49 - sources: Redactie Tweakers.net - Submitter: Thandor - Views: 21,366 On our forum Gathering or Tweakers the first photograph of AMD's Socket f has emerged. In May we wrote all that AMD new processorsocket on its roadmap had put. The new voetstuk 1207 connection points would count and is intended for multi-Opteron-servers. To occur that a processor with support for Ddr-geheugen are pricked on Ddr2-voetje and vice-versa, therefore new socket were necessary. The extra pins which become available, could would according to reports be used for incorporated PCI Express-controller on the processors. Striking on the photograph the clear separation in the middle of the socket is. This seems indicate that each core of dual-core the Opteron his own group has contact points and this way real such as two processors it is treated. On the photograph is further also see that Socket f have equipped, just like Intels Socket 775, with pins which must make contact with the processor. The cpu will have will be therefore no longer pricked in the socket, but it concerns so-called Lga-socket. Socket f become moreover also Socket 1207 mentioned, but just like Socket 479 but 478 count pins, this model will have contact points also only 1206, this way precise telwerk expelled. This processor foot supports registered DDR II 533 -, 667 and 800-geheugen and this way venture AMD the gok to take on the competition with Intels FB Dimm-plannen. This last in April, as it happens, its dual-core will present platform Dempsey, with among other things Greencreek-chipset with support for FB Dimm-geheugen.

Re:Routine Babelfish Translation:

[AMD-lover]

Translation:
Voetstuk = Socket
Geheugen = Memory
Plannen = Plans

Back to the '80s

[cronot]

The first thing that came into my mind after reading the parent and its replies, is that this is coming closer to what microcomputers used to be back in the 80s, with the MSX, ZX-Spectrum, etc. Well, maybe the keyboard will remain detachable, as will any User Interactive peripheral, but everything else used to be much closer to the CPU back then.

Re:Back to the '80s

[Craig Davison]

Actually it seems it's going further away from what microcomputers were in the 80s. An integrated PCI controller is one step closer to a single-chip computer (imagine a board with PCI slots, DIMM slots, headers for things like power, ports on the back, and single chip in the centre).

In 80s home computers, you'd see dozens of chips on the board, but today, we see motherboards from companies like SiS with an all-in-one northbridge, video adapter, audio, modem, ethernet controller, etc., and CPUs with an integrated cache and floating-point processor.

Sexy

[FreakUnique]

Those pictures look very promising as well as getting me in trouble with the bosses for looking at porn. :p

Re:How many?

[Monkofdoom]

I only counted 1206

Online lifestyle?

[Mattygfunk1]

Funny, with Adblock, Flashblock, Tivo, Netflix and Bitorrent, I can't recall the last commerical or ad I've seen.

Try getting out of the house. Advertising is everywhere.

__
Laugh Daily adult funny video

Re:Online lifestyle?

[interiot]

Only listen to podcasts while on the road, and cover your eyes.

Re:Online lifestyle?

[stecoop]

Try getting out of the house
WTF, there is nothing outside of the basement, why would I want to leave that?

Advertising is everywhere.
Define everywhere. Yeah magazines have advertisements but I don't have to buy them. Newspapers have advertisements I don't buy them either. On the radio, it is easy enough to change channels during a commercial but NPR and the BBC don't have a bunch anyway. Billboards, I can't ignore them with pretty ladies but I can't tell you what it was advertising. Are you in a being held by some organized advertisement militia propping your eyes open with toothpicks forcing you to view advertisements? Face it, you can ignore advertisements if you want, just because it is presented dosn't mean you have to consume it.

But the AMD advertisment is still funny; I didn't know there is a media blitz untill somone from another country said so...

Re:Online lifestyle?

[xrobertcmx]

Sad thing is that there is now advertising on the back of the gasoline pumps. Your insert the handle into the car and the little plaque on the back of it tells you about NBC or soemthing.

Re:Online lifestyle?

[ckaminski]

Which is sad. Advertising and it's associated costs surely have a proportional relationship to price inflation, no?

Less ads, more profit!?!

Re:Online lifestyle?

[Anonymous Coward]

You can always live in a state that bans billboards. It's pretty hard to avoid actual business signs though, and I'm not sure why you'd want to anyway.

Re:Online lifestyle?

[Seraphim1982]

Define everywhere.
Leela: Didn't you have ads in the 20th century?
Fry: Well sure, but not in our dreams. Only on TV and radio. And in magazines. And movies. And at ball games and on buses and milk cartons and t-shirts and written on the sky. But not in dreams. No siree!

Well, all I have to say to that is....

[aicrules]

Op ons forum Gathering of Tweakers zijn de eerste foto's van AMD's Socket F opgedoken. In mei schreven we al dat AMD een nieuwe processorsocket op zijn roadmap gezet had. Het nieuwe voetstuk zou 1207 verbindingspunten tellen en bedoeld zijn voor multi-Opteron-servers.

Oh and....MULTIPASS!

Re:Well, all I have to say to that is....

[squoozer]

And the fish says...

On our forum Gathering or Tweakers the first photograph of AMD's Socket f has emerged. In May we wrote all that AMD new processorsocket on its roadmap had put. The new voetstuk 1207 connection points would count and is intended for multi-Opteron-servers.

Re:Well, all I have to say to that is....

[AvitarX]

Better than the average /. summary.

Re:Well, all I have to say to that is....

[Anonymous Coward]

does that mean it's NOT 'crazy fucking moon'?

Re:Well, all I have to say to that is....

[macdaddy357]

Firsk norp firsk norp firsky dirsky du. Firsk norp neer firsk norp neer firsk. Norp norp norp!

Ewwww

[Groo Wanderer]

No. Icky. Bad. You don't take the fastest changing part of a system and put it in a component that changes the slowest. Also, strapping a 110W GPU to a 60W CPU is not a smart thing.

          -Charlie

P.S. Bad, bad, bad. No cookie.

Re:Ewwww

[sigxcpu]

This is Intel you'r talking about surly you mean a 110W GPU on a 21.1 GigaWat CPU.

Re:Ewwww

[jandrese]

I have a feeling the integrated graphics will be, well, integrated graphics. IE, a very very low end graphics card that works just fine as long as you don't care about 3D performance.

Re:Ewwww

[ichigo 2.0]

Isn't integrating graphics on-chip a waste of transistors then? Unless Intel has given up on gamers and are aiming their processors to low-end users and workstations...

Re:Ewwww

[jandrese]

I think the target market will be office PCs and servers and the like where the graphics card is more of a necessary evil than a competitive feature. Gamers will of course buy a graphics card too, but regular folks will get their computer a bit cheaper.

Obligatory 2001 quote...

[cca93014]

My god, it's full of holes.

Wow

[springbox]

Look at all them pins!

I though having those pins on the socket was a stupid idea, but it's interesting to note that even if you did damage the pins on the motherboard, chances are it will be cheaper to replace it than the processor itself. Although only replacing the processor would be much more convenient.

Re:Wow

[Abstract_Me]

"Although only replacing the processor would be much more convenient."

Interesting that you mentioned that as its more then likely a business that will be using this socket.I can't help but wonder what would be cheaper for them. a quick swap of a processesor or completely stripping the machine down to replace the motherboard. With the price of man hours for both the tech fixing it and also everyone who may need that box to be productive the thousand dollar chip may not actually be that big of an expense

Re:Wow

[m00j]

I am waiting for the next gen where the processor AND the motherboard just have little cups on them and you have to put tiny little gold balls into each of the holes first :p

But in reality I think that this system could work out better. For starters heatsinks seem to be getting tighter and tighter each generation. On my 2 most recent computers I have found that removing the heatsink always brings the CPU with it. Such large surface areas form the heatspreaders create a lot of suction. Of course you can twist the heatsink a bit to break the suction but that always causes the heatsink to get scratched.

Plus whenever you have the cpu out and sitting on the desk you KNOW the pins are going to get bent. And then you have the fun of trying to straighten them with some tweezers! I have not seen a LGA socket personally yet but looking at the pictures it looks like they would be harder to bend unless you were very careless.

Still waiting...

[Vo0k]

...for built-in thermal guides to safely remove heat, no risk of misaligned radiator.
...for integrated liquid cooling option
...for more redundant pins so a single socket design could last 5-10 years, not replaced 5 months later with one that has two pins more
...for some good options to stack n CPUs for multi-CPU platform (socket for next CPU built into top of the previous one, thermal guides / coolant pipes running through)

Re:Still waiting...

[tomstdenis]

"not replaced 5 months later"...

You can still buy AMD64 and Sempron based 754-pin processors years after the spec was released.

In otherwords SHUT THE FUCK UP YOU IGNORANT LITTLE SHIT OMG YOU ARE TEH STUPID.

As for stacking cpus ... um you realize AMD makes *dual-core* processors right?

Tom

Re:Still waiting...

[Vo0k]

As for the socket, nevertheless, it's considered obsolete. You can still buy 486 processors too.
As for stacking the CPUs try to build 32-CPU machine without making it a distributed cluster. DUAL doesn't cut it.

Re:Still waiting...

[tomstdenis]

Um, I can go out and buy a 754-pin mobo+cpu TODAY in essentially ANY computer store.

How is it "obsolete"? Obsolete implies replaced. I'd say they have what you'd call "two product lines".

Sure they're not rolling out new designs with it, but I'd say a 2.2Ghz 3400+ is way more than enough for a desktop box. It's hardly "outdated" given the fact the top of the line 4800+ is only 200Mhz faster [but with another core].

As for your idiotic idea of stacking cpus ... they make this thing called "heat". Until you can substantially lower the power required you can't stack them, hell placing them adjacent to each other is enough of a challenge. And frankly, 32 processors is not something the bulk of their customers wants which is why they don't do it.

I say screw your bullshit SMP, I want an LCD on the top of the die so I can see the contents of the registers at all times!!!

Tom

Re:Still waiting...

[Hymer]

"...for built-in thermal guides to safely remove heat, no risk of misaligned radiator."
are "two M6 bolts integrated in the casing for mounting the radiator" an acceptable solution ?

Re:Still waiting...

[drinkypoo]

...what do you mean thermal guides? An integrated water block or something? Recent intel specifications include some pretty serious and intuitive mount points for coolers.
...integrated liquid cooling option is a piss-poor idea. What happens if it gets clogged somehow? What about mineral deposits?
...redundant anything means more cost, because they're producing hojillions of these things. It's not going to happen. Also, it means that manufacturers are tied to a design. Finally, where are those pins supposed to go? There's no reason not to just go to a new package later, rather than making a more costly socket from day one when you're not even going to take advantage of it.
...stacking CPUs? In case you didn't notice, the current packages have pins in the center. Where are you imagining we'll be removing heat from the CPU?

There's a reason we don't tie disparate types of hardware together unless they're cheap - it's the same principle followed by intelligent people who don't buy tv-dvd-vcr combos. When one things breaks, you now have a big lump of crap that's not worth repairing because it's damn near as expensive to fix as to replace.

Re:Still waiting...

[SharpFang]

Heat guides: Think standarized metal tubes that are connected with the core by metal connectors able to remove any amount of heat the CPU can produce. Then slip the tubes onto special bolts of the radiator, with minimal (0.01mm) tollerance so the heat conductivity is guaranted, no matter what. No more misaligned fans, too thin or too thick thermal paste layer etc. Alternatively instead of bolts of the radiator use water pipes. If your water cooling gets clogged, that means you're a moron. Use demineralized pure water, not Kool-aid.
Redundant 20 pins mean cost increased by $1 max. I prefer to pay $1 extra and know the hardware will be supported in 3 years.
Stacking CPUs: Removing heat through vertical bars/pipes through. Just move a few pins to sides.

"There's a reason we don't tie disparate types of hardware together unless they're cheap"
You mean, like, the industry is heading more for multi-socket motherboards than multi-core CPUs?

Re:Still waiting...

[drinkypoo]

"There's a reason we don't tie disparate types of hardware together unless they're cheap"
You mean, like, the industry is heading more for multi-socket motherboards than multi-core CPUs?

I'm not sure what your objection is but I think it comes down to a reading comprehension issue.

disparate (adj.) Fundamentally distinct or different in kind; entirely dissimilar.

The cores are [so far] identical, except in extreme cases like the Cell processor - which has one SPE and several PPEs which are all the same, unless I have that backwards.

Return to English class, please. Thank you.

Re:Still waiting...

[Lord Crc]

...for more redundant pins so a single socket design could last 5-10 years, not replaced 5 months later with one that has two pins more

Uhm, what good would that do? So you can upgrade your motherboard but keep your 10 year old CPU? Or be able to upgrade your cpu, and have your 10 year old motherboard magically have support 32 cores and whatnot that'll be down the road?

But yeah, if nothing major changes, the socket should stay the same.

Re:Still waiting...

[TheHawke]

*nukes the troll*

Ok, who's for lunch?

DDR

[Jormundgard]

cool, I love dance dance revolution

babelfish anyone?

[chocotofferts]

http://tinyurl.com/9v43q (babelfish already entered) screenshot article in jpg het gaat over de fotos geloof ik..

Egads, nerd pr0n

[Hoi Polloi]

You just made me realize that reading the article description got me as excited as looking at nekked pics of Paris Hilton. The big difference is that AMD CPUs are much more interesting than her and are more talented. They are both about as flat and prickley though.

Re:Egads, nerd pr0n

[ameoba]

Not to mention that it's hard to think of any woman, even Paris, having 1207 connectors inserted into her...

Re:Egads, nerd pr0n

[Anonymous Coward]

Well, not all at once, anyway...

Adding new functionality

[Hoi Polloi]

They've added the memory controller and now they are adding the pci-e controller. If they keep adding things to the chip soon it will be so big that they'll just put the expansion slots directly onto the cpu. It WILL be the motherboard.

Re:OUTGOING

[ScottyH]

Is this some sort of cipher? I've been seeing these around...what do they mean?

Obsolescence strikes again

[Hoi Polloi]

Great, I haven't even built my new AMD 64 system and now I have dangled in front of me the latest and greatest to come. It will not only require a new CPU, but a new motherboard and new RAM (DDR2 vs DDR).

Must...not...chase...bleeding...edge!

Re:Obsolescence strikes again

[kurokaze]

phfft.. my main desktop is still an Athlon Thunderbird 1Ghz @ 1.33 Ghz with PC133.. that's right.. PC133 SDRAM!

Re:Obsolescence strikes again

[Anonymous Coward]

Funny, I have the exact same system, except mine's a Thunderbird 1.3 GHz @ 1GHz...

Blast the gaming scene...

[lpangelrob]

I don't know what came first, Double Data Rate or Dance Dance Revolution, but I curse the second group that used the DDR acronym.

Every single time I see DDR and compatibility, I think, wait, why do you need anything else with DDR?

Re:Blast the gaming scene...

[Vo0k]

Good that Germans made Deutsche Demokratik Republik (East Germany) obsolete :)
You still can find some products "made in DDR" though.

Re:Blast the gaming scene...

[Indras]

I have it even worse, I used to work with a guy named David D. Rouse. He initialed "DDR" on all his tools and paperwork. My head gets all fuzzy trying to figure out what he's doing on Slashdot everytime I see the acronym.

Re:Blast the gaming scene...

[Anonymous Coward]

Deutsche Demokratische Republik (German Democratic Republic) was certainly first, but it gives me the shivers: are these guys actually communists?

Re:Blast the gaming scene...

[Vampo]

You can also buy DDR t-shirts in most tourist shops in the former DDR.

Re:Blast the gaming scene...

[FST777]

Now imagine memory compatible with THAT!

Re:Blast the gaming scene...

[Kjella]

Now imagine memory compatible with THAT!

Sure. Coming to a PC near you in 2007 along with Vista. Or technically, I don't think the RAM is special, it is just all encrypted on the mobo before it touches the RAM chips.

Tools?

[cbreaker]

Exactly what tools are involved in installing a CPU, with any style socket?

Re:Tools?

[sarahemm]

I assume the parent meant 'tools' as in 'robotics to insert the CPUs' or such.

Re:Tools?

[cbreaker]

I'm fairly certain that people do CPU inserts, not machines, even at places like Dell. Maybe even especially at dell, since one line of machines might change from month to month depending on who they get the board or parts from.

Either way, the point the poster was making is moot. Intel changes their sockets just as much as AMD, and the new CPU's (with the pins on the board) go in almost the same way mechanically as the ones with pins- put CPU in, pull down on a lever of some sort. I don't see how or why AMD would have to "get with Intel" on this issue.

Remember the SGI O2

[tob]

If I remember correctly, the central chip (not the CPU) that did graphics and played central IO bus in the O2 has a 1500 pin ball grid array, which seems awfully similar to this LGA stuff.

The O2 was released around 1997? A brilliant design that resulted in a very badly performing cruddy system.

For people wondering how much further this can go? Well, at least 300 pins more than AMD is cramming in now, and that with 8 years old technology.

Tob

Oh my god...

[Hymer]

...it looks just so sexy... It so big, and there are so many holes...
My GOD... YES... YES...

Sorry... It just came over me...

--

Real CPU's have the cooler mountet with two 10mm nuts...

Re:Oh my god...

[nsayer]

It just came over me...

Actually, it sounds like you came all over it.

Re:Oh my god...

[FST777]

I'm sorry, who came just all over you?

An integrated...

[eth1]

...600W power supply? :P

Now think about it....

[RJabelman]

A single chip system isn't going to need 5000 pins, as there'll be nothing else (internal) to connect to :)

Can I paint its yoohoo gold?

[SocietyoftheFist]

It's kinda my thing, you know.

Looks fake to me.

[Anonymous Coward]

Err... am I the only one who thinks that the perspective on the two "halves" of the socket is completely wrong? Take a good look at the angle at which the pins on the left hand side of the 3rd image appear to be sticking up - and then compare it to the last row of the right hand side.

The whole pin area looks too "flat" as well...

Re:FB-DIMM?

[Anonymous Coward]

Thanks for the readable translation.

I wonder why AMD isn't just skipping DDR2 and going straight to FB-DIMM?

Maybe FB-DIMM will only be for really big 4+ CPU systems? Maybe FB-DIMM negates advantages of integrated memory controller?

Re:OUTGOING

[Red Flayer]

Yeah, it's the Wilson cypher. Neal Stephenson explains this cypher pretty well in one (or two) of his novels, which is one reason why we see it so much on geek sites. I haven't bothered decrypting this post, maybe I'll get around to it this weekend.

not required.

[Anonymous Coward]

Don't worry t.net can handle it without blinking. if the editor linked to the forum it might be a different story.

Oops, gave the editor an idea... :X

Now I understand the move to pin pads

[kriston]

I can't tell from the photographs--is this socket going to be a pinless processor like Intel's Socket-775 or are we stuck with over 1000 fragile whisker-like pins? I started appreciating my new Socket-775 system after I installed my Socket-754 with all the fragile pins on it. At first I thought it was silly but after straightening out more than a couple whisker-thin pins on my Athlon 64 CPUs I'm hoping Socket-F follows the precedent of using pin pads.

Re:Now I understand the move to pin pads

[RzUpAnmsCwrds]

after straightening out more than a couple whisker-thin pins on my Athlon 64 CPUs

If you are straightening pins, you're not being nearly careful enough. Socket 754 pins are considerably beefier than, for example, socket 468, and few people complained about Pentium 4 Northwood being easy to damage.

LGA is more about better electrical connectivity than preventing bent pins. Remember that most CPUs go into OEM systems, which are aseembled by people who are much better at inserting CPUs that you are.

Re:Now I understand the move to pin pads

[kriston]

Umm, thanks, smart one, but it came out of the OEM packaging with bent pins.

I'm sure I'm better at inserting CPUs than YOU are.

Sheesh.

I/o

[LordMyren]

Sounds like you're talking about the P.A. Semi chip recently announced. =]

Well, sans GPU. But given the PCI express interface, a custom one off wouldnt be that hard to tack onto the board. Given the perposterous ram bandwidth on the PA Semi chips, a solution like nVidia's TurboCache would work great: just have one unified ram for processor and GPU.

Thats one other thing I dont expect to see integrated any time soon: RAM. As for storage in general, hopefully flash will continue growing in capacity at a vaguely exponential rate. 32gb flash would almost be sufficient to make the hd unnecessary for most systems.

Just remember though, AMD isnt doing this for integration sakes, they're in it for the I/o. Which, I should note, the P.A. Semi people have in spades. Some smart people are calling the new multi-core age "throughput" computing, which is rather apt; we are going to need a whole new architecture to support our processors.

Re:FB-DIMM?

[Wesley Felter]

Maybe FB-DIMM has licensing costs.

That explains it.

[UnknowingFool]

A new AMD Socket? No wonder the tech room was covered in drool this morning.

God, I hope that was drool.

MUULTIIPASSSS

[Broken_Ladder]

Yes! He knows it's a multipass!

Would someone please tell the Dutch...

[Anonymous Coward]

The Poles are looking for them. They want their vowels back.

Re:Would someone please tell the Dutch...

[Anonymous Coward]

The Poles are looking for them. They want their vowels back.

Naar de muggen gebruiken zij. Ken de muggen willen steken nou toe dan!

Nie dzieki. utrzymywac nasz samogloski!

grid balls

[Anonymous Coward]

Holy phallic silicon, batschmuck! For a minute, I read:

It looks similar to Intel's new design with the penis...

And then it goes on to talk about damage. Lordie!

Almost

[ichigo 2.0]

Voltage regulator on-package

given that it brings more bandwidth...

[YesIAmAScript]

SDRAM->DDR->DDR2

It adds bandwidth, so basically, your machine would be crap if the socket always stayed the same. You'd have a fast processor with only 133MB/sec of memory bandwidth (instead of 10GB/sec+).

You can't even feed the instruction pipeline on a current processor from those old busses, let alone provide the data to process.

Re:given that it brings more bandwidth...

[Cerberus7]

Yeah, but is that what all of these new pins are for? More pins for more memory bandwidth is one thing. More pins for inegrating all kinds of extra stuff into the CPU that may or may not actually show real-world benefit (integrated memory or SATA controller) seems like a waste to me. Then again, I'm still on Socket A, so what do I know.

Way back when the Pentium II moved L2 cache from the motherboard to the CPU module there were benefits gained, but that had more to do with the high-speed bus to the cache than by physically moving it to the CPU. Again it happened when everyone went back to a socket for CPUs. Is this just another repeat? Is there a real performance reason for moving system controllers onto the CPU that couldn't be gained with a little motherboard design change? Or is it just easier, or more cost effective, to get the same benefit from putting these things on the CPU?

Techno-Hater

[kicken18]

But yet without a CPU we would probably of never seen Hilton, let alone the sex tape that she kindly made for us all to laugh at, oh and also being able to steal everything from her Blue tooth phone. I think she probably doesnt like technology right now

small mistake

[elgatozorbas]

To AVOID that a processor with DDR support would be put in a DDR2-socket and vice-versa a new socket was needed. The extra pins that would thus become available could alledgedly be used for an integrated PCI-express controller on the processors. [...]

Second paragraph:
Furthermore the photo's show that the Socket F (just like the intel socket 775) is equipped with pins making contact with the processor. The CPU will no longer be pinned down into the socket, but the socket is an LGA socket. Meticulous counting showed that the socket F, also called socket 1207, only has 1206 contacts, just like the Socket 479 only had 498. This CPU socket also supports DDR II 533-, 667-, and 800 memory, being AMD's shot at competition with Intel's FB-dimm plans. The latter [Intel] will present its dual-core platform 'Dempsey' in April, with among others the Greenrcreek chipset with FB DIMM support.

Re:small mistake

[ghamerly]

Thanks for the feedback! I appreciate it. Helps me learn.

Finally, This Confirms The AMD Rumors

[danFL-NERaves]

Finally we have confirmation that all future AMD releases will be in Dutch. No English processor releases are being planned.

If the rest of the rumor is true Intel and AMD will be combining their x86 processor businesses to form the new iAMDutch conglomerate. Will any of us be safe?

Dan

"Leaked Pictures of Socket F " is like porn...

[Sr. Pato]

... for Robots. Bender would be proud... or horny; take your pick.

Somthing about sockets that arouses the geeks...

[Anonymous Coward]

I thought it was appropriate to respond to this certain Anonymous post.

/me experiences a short spasm, followed by the urge

If staring deeply into a CPU socket is any hint of gender, then does that mean geeks get laid every time they need to dismount a CPU from its socket in providing one's service? And when a geek takes out a brush or a blower... does that turn anyone on? BeoWulf clusters? Imagine an orgy of those Socket-F AMD cpus! And then there is the arousing double-team: two processing threads in the same socket, hyper-threading fscking action on a hard-drive! Oh, mount me! Oh! Oh! PANIC, core dumped.

Does that make you horny?

more bandwidth

[YesIAmAScript]

Your socket 7 has a 32-bit wide bus. This socket has a 64-bit wide bus. That's 32 new pins. And there's more stuff where that came from.

Intels slot was just a plan to get more revenue. By moving the CPU and cache onto the card, they got to sell you the cache chips, when previously the motherboard vendor sold you the cache chips. Once the cache was on the chip, there was no reason to have the slot, so they went back to the socket.

I have to ask, do you really care if the new socket has more pins? I mean, if it's incompatible anyway, why stress over the pin count? And it has to be incompatible to bring the new RAM types and thus bandwidth.

I don't think CPUs have integrated SATA controllers (yet). But they do have memory controllers and soon perhaps PCI-E bridges. Maybe SATA some day.

Re:more bandwidth

[Cerberus7]

I'm not worried about pins, I'm worried about incompatibility. A certain amount of obsolesence is expected, but here's what I used to think of as a great upgrade path (and one I was able to use at least once through Socket A's lifetime)

1. Buy new CPU, RAM, motherboard to start.
2. Upgrade CPU, keep RAM and MB
3. Upgrade motherboard, keep CPU and RAM
4. Upgrade RAM, keep CPU and MB
5. Upgrade CPU, keep RAM and MB

Lately, it seems like you need to get all three at once, and maybe you can upgrade your CPU once, but that's about it.

Strange mix of IC packages on that board

[ChumpusRex2003]

Looks funny to have a 1207 pin LGA socket in the foreground, with lots of ancillary tiny SMT discretes and TQFP or BGA ICs.

But what's that hiding in the background? A DIP packaged IC? Look at the size of it in comparison to everything else? I don't recall the last time I saw a DIP IC in a PC component.

Translation

[StikyPad]

We hope you've enjoyed looking at these pictures of a socket. Be sure to check out next week's article when we use high-speed photography to watch paint dry in slow motion.

Looks okay

[stormy78]

but I don't think I would want it. It looks like the pins might break a little too easily.

Imagine.

[Anonymous Coward]

Imagine a beowulf cluster oh...

Oh never mind, they dont look any different do they ?

it's going to be incompatible

[YesIAmAScript]

In order to feed a faster CPU you need faster RAM and a faster FSB.

So, unless you want your new CPU to work no better than your old one, that means you're going to have to have a new socket each time.

In summary:
If you don't have a new socket and new RAM, there's no point in upgrading your CPU really. So just stick with your new CPU.

If you want more performance, you won't get it without new RAM and a new socket, so you're going to have to deal with it.

This is all because faster CPUs can't go faster without more bandwidth to feed their faster instruction and data consumption rate.

It is perfectly valid to just stick with your old CPU, RAM and mobo because it is fast enough for you also.