backtop


Print 61 comment(s) - last by johnsonx.. on Nov 30 at 1:43 PM


Image courtesy PC Watch
Four cores, four graphics cards, four hard drives, four everything

PC Watch has the scoop on AMD’s upcoming 4x4 enthusiasts platform. The article claims AMD dubbed its 4x4 platform Quad FX. The upcoming Quad FX platform is based around NVIDIA’s unannounced nForce 680a chipset with SLI compatibility. DailyTech previously revealed images of ASUS’ nForce 680a offering—L1N64-SLI WS. Initial Quad FX systems will be powered by AMD dual-core processors, though the platform should be compatible with AMD’s upcoming Stars processors.

AMD is expected to launch Quad FX with three processors initially—the FX-74, FX-72 and FX-70. AMD is shipping the FX-74, FX-72 and FX-70 in pairs at $999, $799 and $599 respectively. This undercuts Intel’s recently released quad-core Kentsfield Core 2 Extreme QX6700 processor priced at $999.

The processors will utilize AMD’s socket F and feature a 125 watt TDP and manufactured using a 90nm SOI process. Unlike AMD’s workstation Opteron 2200 series processors, AMD Quad FX systems will not require registered DIMMs and function with regular unbuffered DDR2.

PC Watch has also posted benchmarks of the upcoming Quad FX platform as well. The early numbers do not favor too well against Intel’s Core 2 Extreme QX6700 though.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Yah well...
By johnsonx on 11/29/2006 6:04:41 PM , Rating: 3
quote:
Intel tried t ocalled their initial dual cores that when it was just 2 dice on the same package


That was not dual core because.....?

Seems to me if you plug one chip package into one socket, and you get 2 full CPU cores, then that's dual core.

By your logic then, the Core2Extreme QX6700 isn't Quad Core, because Intel put two dual-core dies on the same package?

What if AMD implements Quad Core the same way? Then there won't be any quad cores at all? Or if AMD does it, will that make it a wise choice chosen for manufacturing flexibility and yields?


RE: Yah well...
By KristopherKubicki (blog) on 11/29/2006 6:31:50 PM , Rating: 5
AMD's original roadmaps last year showed the company was working on 65nm processors with two dice on the same package. This project seemed to have gone to the wayside due to K8L -- the native quad-core design.

Of course the argument on whether or not a dual/quad chip is glueless is ridiculous. I have never seen a single solid argument as to why a multi-die package should not be considered multi-core.


RE: Yah well...
By johnsonx on 11/29/2006 10:32:39 PM , Rating: 2
Kristopher,

The solid argument is that Intel did it with two dies, which is therefore wrong, while AMD did it right with a single dual-core die. Doing multi-core with multiple dies will not become correct until AMD does it, at which point it will be correct because of the *special way* AMD does it. Indeed, at that point, if Intel is using a single die that will become wrong.


RE: Yah well...
By Marcus Pollice on 11/30/2006 12:53:07 AM , Rating: 2
What people seem to forget is that Smithfield was one single die. Only with Presler they simply combined 2 Cedar Mills as they can achieve better yields that way.

The only argument one can bring up is the way the cores communicate with each other. Since Intels "FSB-hampered" MCMs don't scale considerably worse than a "native" dual-core (we have yet to see if that applies to quad-core as well), you cant argue its not multi-core. As long as performance scaling is there it doesn't matter how it is done.


RE: Yah well...
By Chillin1248 on 11/30/2006 12:47:10 AM , Rating: 3
Anyone notice that the Intel system in the test had 2X1GB RAM modules and the AMD system had 4X1GB RAM modules (2X1GB per CPU) total? And they were still defeated in benchmarks where memory bandwith counts, interesting.

-------
Chillin


RE: Yah well...
By edpsx on 11/29/2006 6:38:33 PM , Rating: 2
Just because you put two cores on one die doesnt make it "true" dual core. The original design only allowed for one core to talk to anything at any one time, essentially leaving one core sitting idle till either the system bus was done or the other cpu was done doing whatever. I dont know if they changed it since then or if the die shrink alone added this performance. AMD's dual core allows for both cores to work simultaneously together or seperate allowing for a much more effecient working process. Im not a fanboy by any means but in the long run it just seems to me that AMD's design will eventually win out. Intel is just a powerhouse due to the amount of capitol they have.


RE: Yah well...
By cochy on 11/29/2006 9:13:39 PM , Rating: 2
Actually if you do research there are pro's and con's to both AMD's method and Intel's method. Also if you do research you will see that both methods do not any raw performance benefits over the other. Maybe there are theoretical ones, but we all live in the real world for the time being at least.


RE: Yah well...
By zsdersw on 11/29/2006 9:56:33 PM , Rating: 1
Oh please.. not this "true" multi-core crap again. Two cores in one processor package is "dual-core".. period. Four cores in one processor package is "quad-core".. period.

Aside from that, your comments about the "long run" are also totally ridiculous. Apparently you've never heard of Intel's Nehalem; four cores on one piece of silicon. Where did you get the idea that two pieces of silicon on one chip package is going to be the be-all-end-all of how Intel does multi-core chips?


RE: Yah well...
By DallasTexas on 11/29/2006 7:12:57 PM , Rating: 4
Who cares how the sausage is made? You plug in one chip in one socket and the OS sees 4 cores. Seems like a 4 core solution to me.

This "true 4 core" argument doesn't hunt. It's you and AMD fanboys feeble attempt to deflect the fact that Intel's microarchitecture at this point in time is better.

Instead of just getting accepting it and moving on, you dwell on stupid sausage making recipes instead of enjoying the final product deliverables - performance.


RE: Yah well...
By Targon on 11/29/2006 9:10:24 PM , Rating: 2
While you have a right to insult those who don't have a solid argument, there is a solid argument about why AMD has the better system design that you may not have thought about in your "anti-fanboy" rants.

If a processor is designed to be either dual-core or quad-core, there is no need to go through the main system bus for the cores to talk together. As a result, performance will be better if the processor is a "true dual or true quad core processor". I hope you can understand that concept about why the initial dual-core Intel chips(Pentium D), and initial quad-core chips are inferior by design to a true dual/quad core design.

Now, Intel has the advantage on a core to core comparison at this point in time, and most AMD fans do not deny this fact. So, with that fact not being a part of current discussions, the discussion moves on to the overall system architecture. Remember that when AMD releases their K8L based chips, even if performance matches that of Intel's best at that point, you can then look at the system architecture as a reason for one being faster than the other.

The Quad FX concept then is a good idea, even if the Core 2 Duo holds the lead in performance because of that core to core advantage. That's what the whole thing comes down to, what makes for a good system architecture, not just CPU. K8L will be able to be dropped into a Quad FX system, and you will hopefully see at that point that what you were bashing in November/December of 2006 will end up being the basis for a system you envy in Aug/Sept of 2007.


RE: Yah well...
By Lakku on 11/29/2006 11:52:34 PM , Rating: 2
Well, that's all fine and dandy, but I'll stick to what Valve had to say on a single die solution of dual/quad core compared to two packages 'glued' together to make multi-core, and they said the performance difference was "not enough to matter". Perhaps it will make a difference in server applications, but in terms of home use, I doubt there will be much of a difference. Why is it assumed K8L will be a chip of envy? What if Nehelem beats it like Core 2's beat Athlons now? Either way, I don't care. I would MUCH rather have a single socket and single chip (and thus a single heatsink and a single intense thermal area) then having two hot runnig chips side by side. That looks like a total of 5 to 6 fans (2 heatsinks, 3-4 aux.) for this QuadFX compared to 1 fan for the Kentsfield.


RE: Yah well...
By DallasTexas on 11/30/2006 11:02:24 AM , Rating: 2
"..If a processor is designed to be either dual-core or quad-core, there is no need to go through the main system bus for the cores to talk together. ..."

Yes, I agree that's an interesting sausage recipe. I'll write that down.

"..Now, Intel has the advantage on a core to core comparison at this point in time, and most AMD fans do not deny this fact..."

Acually, most AMD fans counter with the hypertransport red herring. The fact that HT is not a limitation in real world applications vs FSB is tally dismissed. What do you expect them to say? yeah, Intel is faster? Doubtful, it's all about religion here.

".. the discussion moves on to the overall system architecture. Remember that when AMD releases their K8L based chips, even if performance matches that of Intel's best at that point, you can then look at the system architecture as a reason for one being faster than the other..."

It's great to hypothesize the what if and when if some future processor compares to another. I like to point out that for 3 years, integrated mem controller and hypertransport was some "overall system architecture" panacea and here we have Intel kicking AMD's butt every which way and it has NEITHER technology! Explain that one. Now the ball shifts to some future conjecture of who's better because you know when having an argument on future speculation, you can always place AMD ahead or on equal footing with Intel. The future is AMD's safe haven to keep their followers in church.


"..The Quad FX concept then is a good idea, even if the Core 2 Duo holds the lead in performance because of that core to core advantage.>."

It's just an amazing phrase above. Read it again. I'll dig deep and give you the benefit of the doubt. You're saying buy this AMD 4x4 Hummer because it will scale for a long long time. Never mind it is an energy hog, costs a fortune and will be obsolete in 6 months. Let's invest a few $1000 in DDR2 memory in it although DDR3 is now sampling. So basicaly, you have a pretty nice system for the winter. Is that a good deal? Ask yourself that.


RE: Yah well...
By Phenick on 11/29/2006 8:11:00 PM , Rating: 1
Core 2 is a better product for sure right now... Which I own and use and enjoy... But the Core 2 Quad is a joke and a waste in my opinion... I am not an AMD fanboy in the least I just do not understand how people cannot see the benefit of a dual cpu setup... If not for any reason other then... It's fun... I used to run dual Xeon systems and it was a blast... However the lack of enthusiast level hardware support made it a less powerful system for games and so I never went dual cpu again... If roadmaps don't change for Intel or AMD for a year.... This time next year I will own an 8 core Quad FX system... I will bet money when that happens, Intel won't have similar performance numbers.... Right now Core 2 Extreme is fine for me... 2 Core2 dice on one package doesn't interest me for the price, the performance just does not impress me at all... and it isn't any more fun to play with.... Freaking power hog...


RE: Yah well...
By ShapeGSX on 11/29/2006 8:34:59 PM , Rating: 2
So 2 AMD chips on a motherboard is fun, but 4 Intel cores in one package is a "power hog?" Huh?

Did you miss this graph?
http://pc.watch.impress.co.jp/docs/2006/1129/graph...

Which one is the power hog, again?


RE: Yah well...
By Phenick on 11/29/06, Rating: -1
RE: Yah well...
By ShapeGSX on 11/29/2006 9:31:15 PM , Rating: 2
This isn't innovation. This is a server platform. It has been done for years.

Have you seen the threads on xtremesystems that show the overclockability and performance of the Intel Core 2 Quad parts? They still blow the AMD stuff right out of the water. They don't seem to be having problems with having the two dies sitting in the same package. Intel doesn't have a problem doing this.

Of course, AMD is stuck on 90nm at the moment, so they don't have that option.

Funny that the company that trumpeted their power savings over the competition just 9 months ago is now releasing a power hungry platform like this.


RE: Yah well...
By Steve Guilliot on 11/29/2006 9:54:38 PM , Rating: 3
Have some more koolaid to wash down that helping of hyper-bole.

The superior performance and overclockability of C2Q vs. 4x4 belie all that blustering you just subjected us to. The situation may change next year after AMD launches K8L, but then you are just replacing inaccuracies with speculation; equally useless.


RE: Yah well...
By cochy on 11/29/2006 9:20:29 PM , Rating: 2
lol well if it's fun for you then do it. But you are confused in that the current trend in the cpu industry is to more cores in a single package. Guess you won't be having fun to much longer ;)


"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson

Related Articles
AMD 4x4 Motherboard Details Unveiled
November 20, 2006, 3:31 PM
AMD Beyond "Brisbane"
November 14, 2006, 6:31 PM
AMD Opteron 10/30/2006 Price Cuts
October 10, 2006, 8:03 PM
AMD Q4'06 Dual-Core Roadmap
October 3, 2006, 8:23 AM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki