AMD's 3GHz K10 to break 30,000 3DMark06 (Inq)

[Donnie27]

You just can help yourself,can you...

Sure I can, if you don't post to me I, like most folks will wait for K10 results. I don't think is lovely, its being hidden. I don't think it's sweet when we're not given a taste of what it can do. Theo has been caught in tons of lies and so has Faud, Charlie and others at that site. What happened to taking anything they say with a grain of salt?

I honestly don't care about 3DBung-O-LiO marks because it doesn't correspond to anything on the Market=P I'd much rather seen SSE4 performance, maybe Crisis or Halo 3, maybe even DivX or XPeg results. Either of these processors could score 40,000 3DMarks and still get whacked in real-world games, so what in the hell did it prove?

[alfaunits]

Well, I'm sure having double the power would at max double the score. So if we had quad-core 6400+ and quad-fire cards it would give ~24K.
Phenom @ 3G will probably give more than a (theoretical quad-core 6400+ would, but dual-card would have less points than quad-fire just as much, so we're back at square one.
Even if we consider a lot faster dual-card, scoring say 15K OCed, that leaves 15K for CPU difference - not possible. Or am I missing something?

Why not? I think it's a combination of both. How exactly does 3DM06 calculates to the cpu score? Does the L3 give the score an extra boost? Does the wider datapath give it an extra boost? Or is the redesigned memory controler more efficient with memory bandwidth? Perhaps it's the native quad core that works more efficient. Well I've never digged into 3DM06's calculations so I don't know.

[CraptacularOne]

Well, I'm sure having double the power would at max double the score. So if we had quad-core 6400+ and quad-fire cards it would give ~24K.
Phenom @ 3G will probably give more than a (theoretical quad-core 6400+ would, but dual-card would have less points than quad-fire just as much, so we're back at square one.
Even if we consider a lot faster dual-card, scoring say 15K OCed, that leaves 15K for CPU difference - not possible. Or am I missing something?

I'm not sure I understood you. Are you thinking that just because you have 4 cores that you will somehow double your score in 06 vs a dual core system? I will be the first to tell you that, that's not how it works. A quad core will add some to your 06 score but it will not double it, not even close as a matter of fact. However adding more GPUs (assuming you had the CPU overhead to feed them) would give you a nice boost. I'm really interested in their multi GPU solution, as it would seem that they have a more efficient way of doing it that Nvidia did.

[informal]

Possible scenario with 3-fire system(and mark06):
SM2.0:~13100(calculated 20% more points for 3rd card,OC is factored in)
SM3.0:~14500(same as above^^)
CPU:~6900@3Ghz (hypothetical 50% better clock/clock than Kentsfield in this particular test)

[CraptacularOne]

Possible scenario with 3-fire system(and mark06):
SM2.0:~13100(calculated 20% more points for 3rd card,OC is factored in)
SM3.0:~14500(same as above^^)
CPU:~6900@3Ghz (hypothetical 50% better clock/clock than Kentsfield in this particular test)

That adds up, but I seriously doubt that the Barcelona is going to be 50% faster clock for clock. I think that if it was done on a 3 GPU set up that the cards made for more of a difference than the CPU did.

[Donnie27]

I'm not sure I understood you. Are you thinking that just because you have 4 cores that you will somehow double your score in 06 vs a dual core system? I will be the first to tell you that, that's not how it works. A quad core will add some to your 06 score but it will not double it, not even close as a matter of fact. However adding more GPUs (assuming you had the CPU overhead to feed them) would give you a nice boost. I'm really interested in their multi GPU solution, as it would seem that they have a more efficient way of doing it that Nvidia did.

QFT!

But even for 3DMarks, it would have been impressive this or even 15K was performed in software mode by QCores with just one GPU. Old school testers used to perform tests in software mode to check the processor's power. Quake and UT were tested this way. Come on Sept 10th

[Donnie27]

That adds up, but I seriously doubt that the Barcelona is going to be 50% faster clock for clock. I think that if it was done on a 3 GPU set up that the cards made for more of a difference than the CPU did.

Exactly!

[informal]

That adds up, but I seriously doubt that the Barcelona is going to be 50% faster clock for clock. I think that if it was done on a 3 GPU set up that the cards made for more of a difference than the CPU did.

I think it fits in some projections(techArp's for instance).20-30% on average was claimed,with 170% in some rare cases(beats me what that could be,apart from possible mem. bandwidth-which still can't be that higher)

[accord99]

I think it fits in some projections(techArp's for instance).20-30% on average was claimed,with 170% in some rare cases(beats me what that could be,apart from possible mem. bandwidth-which still can't be that higher)

Of course it's going be a memory bandwidth test, probably something like stream. But Barcelona will still be unable to match Intel in the major enterprise benchmarks.

[informal]

Of course it's going be a memory bandwidth test, probably something like stream. But Barcelona will still be unable to match Intel in the major enterprise benchmarks.

Are you joking?lol
Of course 1.9Ghz won't match uber 3.33Ghz Penryn.Doh

[accord99]

Are you joking?lol
Of course 1.9Ghz won't match uber 3.33Ghz Penryn.Doh

And I expect it even at 3GHz, it still won't match the 3.33GHz Penryn-based quad-core across a wide range of enterprise applications.

[CraptacularOne]

I think it fits in some projections(techArp's for instance).20-30% on average was claimed,with 170% in some rare cases(beats me what that could be,apart from possible mem. bandwidth-which still can't be that higher)

I think they were referring more to server tasks than anything having to do with gaming. I don't think they are going to be all that much faster than the core 2 architecture in gaming and benchmarks. I can see 15-20% on a good day. And that will be a monumental triumph for AMD if they can put those kind of numbers up. Which happens to be roughly the same performance advantage that C2D is currently enjoying over AMD clock for clock.

It's a see saw, they will continue to out do each other, that's what they are supposed to do. But I really don't see AMD coming out with a chip that can push a set (2) of HD2900XTs to 30k in 06.

[accord99]

I think they were referring more to server tasks than anything having to do with gaming.

Not even server tasks, more overly memory bandwidth, embarrassingly parallel benchmarks like SPECint_rate and SPECfp_rate that have very little correlation with most enterprise applications. I personally expect Phenom to be slower clock-for-clock than Kentsfield on desktop applications.

[k0nsl]

Not even server tasks, more overly memory bandwidth, embarrassingly parallel benchmarks like SPECint_rate and SPECfp_rate that have very little correlation with most enterprise applications. I personally expect Phenom to be slower clock-for-clock than Kentsfield on desktop applications.

How do you base such a theory?

[FghtinIrshNvrDi]

it's a shot in the dar, k0nsl.

Ryan

[accord99]

How do you base such a theory?

The unimpressive Cinebench score, the unimpressive POV-Ray score, the unimpressive non SPEC_rate scores in the leaked Russian AMD Power Point plus the lack of SPECint or SPECfp scores.

[informal]

Right,ok. lol

[Zytek_Fan]

The unimpressive Cinebench score, the unimpressive POV-Ray score, the unimpressive non SPEC_rate scores in the leaked Russian AMD Power Point plus the lack of SPECint or SPECfp scores.

Those are old, rendering them irrelevant, and they could have been made slow on purpose

[informal]

Those are old, rendering them irrelevant, and they could have been made slow on purpose

Exactly.

[freeloader]

The unimpressive Cinebench score, the unimpressive POV-Ray score, the unimpressive non SPEC_rate scores in the leaked Russian AMD Power Point plus the lack of SPECint or SPECfp scores.

All run with "broken" cores.

[k0nsl]

All run with "broken" cores.

Besides that, I personally find the tests questionable anyway, or at least suspect.

[knightwolf654]

i can believe it, but yet as well it is fishy, only time will tell but either way,
i'm getting one of these bad boys

[Shadowmage]

I thought Tri-fire was 2xGFX, 1xPHYSICS?

[Donnie27]

i can believe it, but yet as well it is fishy, only time will tell but either way,
i'm getting one of these bad boys

Me too if they rock. What I will NOT do is just buy because it is not Intel LOL! Just as I bought AMD when Hot Prescott sucked IMHO! I'm not using 3DMarks as a measurement of which is faster. I'll pick apps just as I did when I chose A64 over P4 because it was faster or C2D over X2 for the same reason. Not some silly Fanboy MISPLACED love for either company

as a startup year of development of this jewel of the new core seems so long ago.Bring it on AMD!

[Donnie27]

I thought Tri-fire was 2xGFX, 1xPHYSICS?

One of the GPUs can be use as a PPU