Preliminary Benchmarks Show RV670 Superiority Over G92

[[I.R.A]_FBi]

when they come out ima buy one a these and a quad ... damn former employer a hold on pon me money ...

 

[KennyT772]

WRONG.

The only applications where R600 manages to hold its own are synthetic benchmarks (3DMark 2006) - in real-world performance nVidia wins 90% of the time. Unless RV670 is something special, my next upgrade will be a G92.

Shows theres potential there, just no drivers. Its really sad. On hardware terms the 2900 is on top but the drivers are cutting it at the knees.

 

[Makaveli]

Assimilator is %100 correct. Enough with the Bungholio marks, the X2900 series has broken AA. And I seriously doubt the drivers will fix that seems to be a hardware problem. I'm hoping the refresh of these cards will fix this.

 

[AddSub]

the RV670 earns 11,400 3DMarks, and the G92 earns 10,800 3DMarks.

I'm glad I didn't go out and buy a 8800GTS for $300(US) earlier this summer, or a 2900PRO the other day(for about the same). Especially if these RV670 and G92 cards prove to be decent overclockers.

 

[Wile E]

WRONG.

The only applications where R600 manages to hold its own are synthetic benchmarks (3DMark 2006) - in real-world performance nVidia wins 90% of the time. Unless RV670 is something special, my next upgrade will be a G92.

That's one review, and to be honest, I don't trust it. Look at this article, and you see a much more even match-up. Keep in mind that the 8800GTS is the direct competitor for the 2900, yet the 2900 reaches or excedes the GTX level in some games.

I've used both cards, and I own the 2900, and I can say with confidence that they are on equal grounds. Some games go to the 8800, some go to the 2900.

 

[mandelore]

WRONG.

The only applications where R600 manages to hold its own are synthetic benchmarks (3DMark 2006) - in real-world performance nVidia wins 90% of the time. Unless RV670 is something special, my next upgrade will be a G92.

lol... so says that review, there are plenty more reviews which show something quite different.

If you add up the totality of the boosts in performance since the initial R600 release due to driver upgrades, you see it has alot of head room for performance. and if you also look at the opengl benchies, it appears that ATI is wiping the floor with NV.

 

[mandelore]

Keep in mind that the 8800GTS is the direct competitor for the 2900, yet the 2900 reaches or excedes the GTX level in some games.

You know.. alot of people keep forgetting that and thusly the 2900 is taken with a slanted oppinion. In fact, the 2900 rocks some ass against its direct competator, Id like to see an overclocked GTS beat an overclocked 1gb 2900xt.

 

[mandelore]

Assimilator is %100 correct. Enough with the Bungholio marks, the X2900 series has broken AA. And I seriously doubt the drivers will fix that seems to be a hardware problem. I'm hoping the refresh of these cards will fix this.

broken AA? are you on crack?

AA works great, and the fact I can set it far higher than any single NV card is also nice

 

[zekrahminator]

Here's my idea of the argument. Both sides are right. The NVIDIA fans will always find benchmarks where the G80 is superior, and the AMD fans will always find benchmarks where the HD 2900 is superior.

Personally, I wouldn't worry about it. They both perform well, they both have their quirks, and they're both hideously expensive. That's all that really matters, as far as video cards go .

 

[KennyT772]

Touche Zek.

 

[mandelore]

true true. I would tho, say that the 2900XT beats the GTS, is roughly on level playing with the GTX, and the 1gb 2900 heavily overclocked is near/at the ultra's level. however, as seen by some interesting benchmarks, the 2900 leads in OGL for some reason

 

[Steevo]

broken AA? are you on crack?

AA works great, and the fact I can set it far higher than any single NV card is also nice

They are speaking in regards to the lack of hardware supported AA, however not taking into account that on larger displays and higher resolutions AA is almost a non issue, as curves can be drawn smoothly, so no jagged edges, so no need for AA.



Again, ATI plans for the future, and Nvidia is still playing catchup in hardware.


I would rather have a tesselator to smooth out games objects and faces than a bunch of AA where it isn't needed. In such instances more triangles and better definition does render AA obsolete.

 

[TheMailMan78]

WRONG.

The only applications where R600 manages to hold its own are synthetic benchmarks (3DMark 2006) - in real-world performance nVidia wins 90% of the time. Unless RV670 is something special, my next upgrade will be a G92.

Maybe you should take a look at this thread.

http://www.techpowerup.com/?41696

 

[eidairaman1]

yeah these scores can't be right, the G92 is rumored to run at 800+ mhz core, if you OC a 8800GTS to this you get much better numbers than 10.8k. Also the G92 as rumor have it will have 16 scalar shader units more than the current 8800 GTS's. The only downside to this new card is its 256bit memory bus as opposed to the current 320bit memory buss of the 8800's. Who knows maybe we never needed the extra 64bits on the memory bus. I think the real numbers for the G92 will be higher than 10.8k. (by the way I'll be swapping my 8800 GTS 320mb in evgas step up program for a G92)

Overclocking doesnt always mean a Major Jump in performance

 

[eidairaman1]

btw if yall want to see how well a R600 overclocks, and the reason behind the Core Speed Drop, Read October 2007 Edition of CPU Magazine.