AMD confirms R680 is two chips on one board

Oops. >.<

 

HD3870X2 score 12056 in 3dmark2k6 at 2560x1600 with X9650 (3667mhz), 2gb ram, vista

 

So, about the same as Crossfire 3870's? Highest tested res I could find was 1920x1200, and that scored 14766.

Whats weird is that in that test the SM3.0 score is higher than the SM2.0 score, whilst in the R680 one, its the other way around...

 

Perhaps with the finer pixel resolution, dynamic branching is aiding performance much more.

 

Should have posted a link to the whole page. The same is true for all the resolutions.

 

Well, you should notice that as they increase the resolution, the gap between the SM2.0 and SM3.0 scores actually decreases.

 

Interesting performance.

However the issue of it being a dual card still exists. Quite honestly I would take a GTS 512 even if the GTS is 100USD more expensive. It might give less performance in some games, but it will be consistent with all games.

 

Here come, the Sapphire amotic cooler may ring the bell and do the trick for a single slot HD 3870X2

However, the heatsink of atomic could need re-design to fix 2 GPU on a single PCB through.

 

Firingsquad has a bit of video footage of R680 and some of Sapphire's 38x0 parts (including the Atomic) from CES. Also, TG Daily is claiming that AMD has moved up the R680 launch from January 28th to January 23rd.

 

I found a Crysis benchmark result using 1280×800 window mode (of some sort). This looks like CF is disabled.


source

 

via the tip @ R3D you would be correct.

3d clocks and CF don't work in window mode. As for I remember.

 

RV670 includes command queue monitoring that I thought was supposed to allow full 3D speed to activate even in windowed mode?
CF I think was supposed to as well but I think is waiting on newer drivers.

 

I wounder how R680 will perform under Crysis, since Crysis does not do well under Multi-GPU configuration.

http://www.techspot.com/article/83-crysis-patch-performance-multigpu/page2.html

 

I find that review a little odd, considering the other reviews I've seen and the benchmarks I've run on my 7950GX2.

http://www.anandtech.com/video/showdoc.aspx?i=3183&p=3
http://www.firingsquad.com/hardware/nvidia_geforce_8800_gts_512mb/page12.asp

 

You're right about the clocks and RV6xx parts, but I don't believe CF works in the same manner. It very well could be a driver issue.

 

You quoted the medium quality 1440x900 results. Go to the next page to see the multi-GPU performance gains.

 

I wouldn't put too much faith in a benchmark from a game I most definitely will never play.

Crysis is a dog in regards to hardware and I cannot seem to understand why.

 

http://forum.beyond3d.com/showthread.php?t=46374

Something to do with the graphics perhaps? Or maybe the physics? World size? Level of interactivity?

Just a few ideas to consider.

 

I know that absolutely nobody ever seems to read my posts, but why the hell do you benchmark a multi-GPU system in MEDIUM DETAILS?!??!

I wrote a post about the MultiGPU Crysis patch and my benchmark findings in this Beyond3D post: http://forum.beyond3d.com/showpost.php?p=1116612&postcount=19

Cliff notes:
If you actually play with settings that you'd expect to use in a multi-GPU system, the increase is quite good. If you play on a 15" screen at medium details, you don't freakin' need multiple GPU's in the first place.

EDIT: How did he come to these numbers? Did he run one of the built-in benchmarks? Did he just stand somewhere and look? He never says how these numbers were obtained. But they VERY certainly don't line up with my findings.

I wonder if he used the benchmark_cpu test instead of the benchmark_gpu? Because the results in that thread are so close together for ANY kind of video card, I have to wonder how he got ANY of these results.