Some AMD K10 Updates-3DMark06 Score of 30000

I'm not sure how relaible this story is but from the way it sounds it looks serious.

http://www.theinquirer.net/?article=41970

What do you guys think?

 

The INQ is so unreliable I stopped taking anything it said to heart a LONG time ago. Story after story of 'oops our last article was wrong.'

Plus, why would a 3DMark benchmark say much about a CPU?

 

yeah processor makes hardly any diffrene at all to 3Dmark06. going from dual to quad core yeilds about a 1000mark improvment on a high end system. So you could think of it like going from 6Ghz (2 x 3Ghz) to 12Ghz (3 x 3Ghz) on an intel system makes about 1000 marks diffrnce.

Also I have no idea what they did to teh setting if they wanted to achive a score like that. I can get to 13K on my system (Quad 3.2Ghz with 8800GTX (641Mhz core 2100mhz memory)) Now thats faster than what stock 8800 Ultras score by having an overclocked Ultra you can get upto 14K-15K marks. And by having them in SLI people are getting into the 25K - 26K range. But 2 overclocked 8800Ultra's. But that space is mostly dominated by nvidia cards.

The best score I have found for a single 2900Xt is this
14253Marks running on Q6600 @ 3.5GHz | 2900XT @ 874/999
So teh graphics card is clocked way way higher than what they mentioned in that artical. And even if you then put that in crossfire you would not get double the perfomance so so you would get 27K marks which is being friendly to it. Then we consider the fact that with Intel CPU's 6Ghz is equating to about 1000 marks. So for this thing to have gotten 30K iit would have to be equivelant to a 8ghz Core 2 Quad. Considering each core would need to be about 4.5ghz faster than it currently is to see that amount of gain.

So does anyone think AMD's new processor is really that fast?

 

A Quote from a guy on teh [H]ard forums who knows teh 3Dmark06 maths a little better than I do but still says the same thing. to get that score you need more that twice the speed of any current Intel CPU which i doubt will be happening for awhile.

 

No, that's not true.
They never say "oops our last article was wrong". They just pretend that they never said otherwise, and that they're really correcting *other people*.

But going from dual to quad core is not the same as having a faster CPU with the same number of cores.

Um, yes. Or you could think of it as phoning up Santa Claus and ask him for his 3dmark score. It has about the same bearing on reality.

But according to the article, they ran Crossfire, not a single GPU.

No, but I think your maths are very flawed.

(Which is not to say that the story is true, of course. Just that your maths aren't right. First, you can't extrapolate how 3dMark will scale with a faster CPU or GPU. It always depends on what's the current bottleneck. And of course, 2 cores at 3GHz are nowhere near the same as 1 core at 6GHz as you seem to think. And 3dMark will certainly not scale the same way between the two)

Is the story true?
Well, In my experience, The Inq aren't exactly liars. They're just fools who tend to get things wrong. (and won't admit it when they do)
There's a big difference.

 

OK so why do I see double the perfomance from my Quad core processor when compared to a dual care at the same clock speed in movie encoding and the CPU score in 3D-mark. In proper multitheaded applications thescaling is pretty close to linear in that a dual care is twice as fast as a single core etc. run some benchmarks you might be supprised. As it happens the CPU benchmark in 3dmark06 is multitheaded.

I know it doesn't work exactly how my math states but its easier to do it that way or its a mess.

Also you can extrapolate how 3dmark scores will scale with a faster CPU. Beacue 3dmark just uses a mthematical algorithm to calculate the scores. And once you break teh scores into their categories (CPU, and the diffrent GFX card tests) you see that once your running on a 3ghz quad even in crossfire or SLI the graphics cards are still the bottleneck becasue overclocking to 3.5Ghz only changes your CPU score the graphics card scores don't move.

Also I was being friendly when I took the single 2900XT score and basically doubled it to give me the score the same system would get in crossfire. Beacuse in reality it would only get about 25K not 27K like I calculated it to be. Tehn take into account that teh 2900Xt's in the enquirers artical were clocked much lower and you would probably only get 23K or so from those cards on a 3.5Ghz C2Q. The scary thing is that you would get 22K on a 3.5ghz C2D which shows you how much teh graphics cards are bottlenecking the system.

Also I know my maths was flawed (thats what happens when you throw it together in 2 min) otherwise I wouldn't of bothered fining someone to quote who did the math properly.

 

hahahahha they called vista MeII

 

Why have global temperatures risen over the last couple of hundred years, at the same time as the number of pirates worldwide has decreased?
Clearly, the lack of pirates is what causes global warming!

I'm sorry, but it doesn't work like that. You see double the CPU score in 3dmark with your CPU at this clock speed, but that doesn't mean it'll be the case for any CPU at any clock speed.

Close to linear is still not linear. And it depends on the number of cores still.
And 3dmark is not "properly multithreaded". They just pretend.

And what does it use as input to that algorithm, do you think?

Yes, the inq thinks people will take them more seriously if they never call anything by its name. They think they're cool in a kindergarten kind of way because thye come up with nicknames for everything.