Anand doesn't like 4x4

[Viditor]

Originally posted by: DrMrLordX

Originally posted by: Idontcare
Furen, exactly, I'm with you on this. Sad how folks are focusing on how minimal the waste heat could be if they had bought 4 processors and proceeded to do nothing with them.

Only one number should matter when one is talking about an extreme performance product, and that is its extreme performance.

Reading all this talk over whether these quad-core platforms are the equivalent of 3 100W light bulbs or 6 100W lightbulbs makes me feel silly. What are folks buying these systems for if 6 100W lightbulbs is their metric of success?

uh? The problem isn't the power consumption alone. If the QuadFX platform could actually do something worth mentioning that put it head-and-shoulders above the QX6700, then the extra power would be "worth it" to many people. Fact is that it doesn't. Under a 32-bit OS like XP, an environment in which most QuadFX systems will be run (or 32-bit Vista), QuadFX loses to the QX6700 all the while costing more money and chewing up more power. And let's be clear here that it chews up significantly more power AT LOAD. The idle power figures aren't the issue here, at least not for me.

Furthermore, the high stock vcore on the FX-74 indicates that these chips have little, if any headroom for overclocking. OC a Kentsfield to 3 ghz (which can and has been done repeatedly) and you'll whip the QuadFX system soundly, presumably even in a 64-bit OS (that few people will be using with either the QX6700 or QuadFX). The power consumption figures for the Kentsfield would look less impressive after an overclock, but the performance gain would be considerable.

Bottom line is this: if it's going to suck up more power, it'd better give me something in return. The QuadFX platform currently does not produce performance that can justify its power consumption.

For the record, the Prescott sucked because it chewed up more power/produced more heat than the Northwood and did not yield appreciable gains in performance.

I've been looking through those numbers, and I really don't think that the 64bit accounts for much. Being NUMA aware (I believe that even 32bit Vista is NUMA aware, and Linux certainly is...) looks to give Quad FX a 20-25% advantage over WinXP...

 

[biostud]

Once the need for 8 cores arise the motherboard will probably be outdated, so opting for Kentsfield or K8L single socket is the only sensible way to go.

 

[A554SS1N]

Amazing how AMD has actually gone worse than Intel ever was with it's P4 series.... truly abysmal. That kind of power consumption could only be justified if it was twice as fast as Intels best, but it's miles slower, so this is a complete failure. I think this'll be bad for AMD's image...

 

[Accord99]

Originally posted by: Viditor
I've been looking through those numbers, and I really don't think that the 64bit accounts for much. Being NUMA aware (I believe that even 32bit Vista is NUMA aware, and Linux certainly is...) looks to give Quad FX a 20-25% advantage over WinXP...

Why then, other a few 3D rendering tests and DVD Shrink wasn' t this performance gain seen? How come the megatasking test, the reason for existence for 4X4 actually runs slower in Vista, and gets beaten by Kentsfield.

 

[Compellor]

Originally posted by: peternelson

Can anyone point me to a benchmark review of QuadFx 4x4 using anything other than 32 bit XP?

Didn't think so.

These reviewers need to run some XP Pro x64, Server 2003 SP1, Vista, Linux before writing the final words.

Also given that AMD is already testing true quadcore Barcelona (also in today's news), the 8 core potential of this platform comes closer.

TR used XP Pro x64:

http://techreport.com/reviews/2006q4/quad-fx/index.x?pg=1

 

[Pugnate]

That was liking watching someone take on Uwe Boll in the boxing ring. You want them to kick Boll's ass, but you know there is no chance in hell.

 

[biostud]

Originally posted by: A554SS1N
Amazing how AMD has actually gone worse than Intel ever was with it's P4 series.... truly abysmal. That kind of power consumption could only be justified if it was twice as fast as Intels best, but it's miles slower, so this is a complete failure. I think this'll be bad for AMD's image...

maybe, but the good thing is that this does not affect AMD's entire line of CPU's, as was the problem with Intels line of dualcore CPU's. But launching this as a viable competitor to kentsfield is stupid. As a proof of concept it's fine, but until 65nm CPU's from AMD can be used for 4x4 it's not useful.

 

[RichUK]

Shock Horror

 

[MDme]

it's quite unfortunate that AMD had to resort to this system. Honestly, I think they did not anticipate intel moving to cough...quad core...cough CPUs this soon. they probably meant this to compete with Core 2 Extreme CPUs. might also be that intel saw 4x4 and decided to do release a "quad-core" cpu (using their glue technique).

Anyway, the reviews have definitely proven 1 thing: the 4x4 consumes a lot of power. One thing you have to notice though is that most of the reviews were not that well done. The 4x4 systems were not configured that well. Only TR's review properly had CnQ enabled (though obviously no amount of CnQ will tame the load draws of 4 3.0Ghz K8 cores). It does show however that idle power is not that different albeit still higher than QX6700. Also the performance of the 4x4 widely varied with the proper OS support. 64-bit and NUMA can contribute much to the performance that actually makes it competitive from a purely performance standpoint with the Core 2 Quad. Performance per watt is another story though.

So in summary what we have with 4x4 is an extremely niche product that is also very expensive. it will probably cater to people looking to eventually run an 8x4 with K8L (aka as K10). With the proper configuration, optimization, (and a lot of cash for the electric bill), the 4x4 offers a powerful (no pun intended) solution with a decent upgrade path.

 

[Atheus]

What have AMD actually been doing for the last few years? They must have known Intel would come out with a new architecture sooner or later, and it was always going to be good... they seem incredibly badly prepared for this.

 

[Idontcare]

AMD sucker-punch for Intel?

Why would AMD want to release a low-volume part/platform now that succeeds in motivating Intel to step-up their 45nm or quad-core frequencies on any faster of a pace than Intel already are on?

If 4x4 were a high volume part, or anything that really materially impacted the financial bottom line in the next 6 months, then I could see cause for concern.

This reads strategy to me. No way AMD did not fully know what they were putting into the market with 4x4's now publicly observed performance and wattage.

Time to convince Intel they are the dealing with the same old AMD of K5/K6 days until 65nm is matured past first wave of production silicon before native quads are tossed onto the market.

 

[herbiehancock]

Originally posted by: MDme
One thing you have to notice though is that most of the reviews were not that well done. The 4x4 systems were not configured that well. Only TR's review properly had CnQ enabled (though obviously no amount of CnQ will tame the load draws of 4 3.0Ghz K8 cores). It does show however that idle power is not that different albeit still higher than QX6700.

While TR's review showed CnQ enabled, it is NOT the only one that got it functioning....the aforementioned review this posting was about, Anandtech's review, was updated to remark that they, too, got the CnQ function operating properly and showed the idle power draw was only a few watts above Kentsfield. Full power draw, on the other hand, and we all know how it went........like >400 watts. Hmmmmmm....do I need a new heater for this winter?

 

[MDme]

must've read AT's article before they got CnQ running. (they were quoting 300+ watts at idle before). anyway, I guess AMD released this for us folks with COOOOOLLLLLD winters.

IF AMD was smart.....IF that is....they'd just make all their UPCOMING processors capable of using the 4x4. that way all their future offerings would have the option of going x2 # of cores. i.e. they'd have an 8-core system when they come out with a quad core chip. this will partially offset intel's manufacturing advantage.

 

[hectorsm]

Originally posted by: Atheus
What have AMD actually been doing for the last few years? They must have known Intel would come out with a new architecture sooner or later, and it was always going to be good... they seem incredibly badly prepared for this.

That's exactly what I thought. AMD has no excuse. They deserve to be in the position they are now.

AMD knew about the Pentium M technology moving to desktop. Core 2 and Kentsfield are somewhat based on the Pentium M technology. All they had to do is look at the potential of the Pentium M to see what was coming.

Under estimating Intel does not even come close to what they did.

 

[crazylegs]

this 4x4 concept from AMD almost makes me cry - it smacks of desperation imo

i just pray they have somthing better coming out later in 2007...

 

[jazzboy]

I remember when AMD was trying to claim it still had better performance-per-watt compared to Intel's Core 2 Duos (wasn't true, unless you counted the Energy Efficient AMD X2s).

And now they just look like a laughing stock.

I think I've said this before here, but AMD have just got cocky ever since they launched the original athlon 64. Surely just looking at the original Banias Pentium M should've made them realise that Intel had something good for the future. They obviously didn't take that much notice and are now paying the price.

Barcelona/K8L had better be good or there's going to be trouble for AMD.