ATI Radeon X1900XTX CrossFire
Date: 24.01.2006
|
|
Overclocking
In view of the fact that overclocking of X1900XTX hasn't brought any noticeable results, we were not after making the most of X1900 ÑrossFire. A slight overclocking to 650/1550 was done only with the purpose too see how two X1900XTX cards are faster than one.
Test setup
X1900 CrossFire test-bench
|
Bus
|
PCI-E
|
|
CPU
|
|
|
MB
|
|
|
Memory
|
|
|
OS
|
WinXP + SP2 + DirectX 9.0c
|
|
PSU
|
|
To test X1900 CrossFire, we used engineering drivers.
And this is how the X1900 CrossFire test-bench looks alive.
7800GTX-SLI test-bench
|
Bus
|
PCI-E
|
|
CPU
|
|
|
MB
|
|
|
Memory
|
|
|
OS
|
WinXP + SP2 + DirectX 9.0c
|
|
PSU
|
|
Test Results
The values in brackets stand for the default frequencies of GPU/memory of the video cards in question, and those with the character @ - for frequencies when overclocked. The value in the middle of the bar is just the test result (in score marks or FPS). The value to the right of the bar shows the percentage the performance of video cards has changed in the mode of operating two video cards as compared to a single video card. That is, for 7800GTX-SLI it shows a variation of performance versus the standard 7800GTX running at default frequencies. Similarly - for X1900 CrossFire and single X1900CF.
 Tests of 7800GTX-SLI with the recently released 3DMark'06 came up against certain issues. The system wouldn't pass the test to the end and simply hung. We tried using varied drivers, including the most recent 81.85 and 81.95 WHQL. It was once when the test did finally ran to the end and issued a result which, I must admit, was discouraging. The dual 7800GTX-SLI score less marks than the single 7800GTX. Perhaps it was due to the "raw" state of 3DMark'06. As I already said, we'll go into the causes of that later. Perhaps we should wait for one or two more patches from the developers prior to "commissioning" the new tool.
As regards the results for X1900 CrossFire, the performance boost of over 60% can be seen as an excellent result! OK. Going on.
The performance boost attained through merging X1900 video cards in 3Dmark'05 is not impressive indeed, while that for 7800GTX-SLI demonstrates a substantial rise. Anyway, fanciers of ATI should not worry at all because the performance seems to be CPU-bounded. To all appearances, for Athlon 64 4000+ the result 12000 "marks" in 3DMark'05 seems to be the best what it can do.
3DMark'03 is not the most "comfortable" application for ATI video cards of the new architecture. Surely, we can say that the victory is with X1900 CrossFire, but it is highly probable that at this test a couple of video cards based on 7800GTX 512 in the SLI-mode could have shown much more impressive results. Again, unless the "marks" are CPU-bounded.
Judging by the synthetic benchmarks, the ATI CrossFire is quite capable of demonstrating its "muscles". Let's see what it is worth at games.
Traditionally, it is hard to expect miracles in a game like Half-Life 2. Again, all is CPU-bounded.
Increase of resolution to 1280õ1024 doesn't change anything. All the results seem to by stuck to the 140 FPS mark. Needless to say, in such a situation the greater boost will be with the couple of video cards whose single partner is "weaker" than the opponent. Which we actually see in 7800GTX-SLI.
The picture is absolutely similar. To all appearances, 133 FPS at 1600x1200 is the limit for Athlon 64 4000+ in Half-Life2.
And here as well the picture is almost the same. To the joy of "red-white" fans, X1900 CrossFire takes a slight lead.
But at 1600õ1200 the load upon the video subsystem rises so high that the effect from merging two video cards on the base of two X1900-based video cards becomes substantial enough and amounts to over 50%. At the same time, for X1900 CrossFire the FPS even at this resolution didn't drop below 100. Isn't that the figure of the day? :)
Finally, the F.E.A.R of course. On the one hand the game is abundant with shaders, on the other - there is a couple of X1900 abundant with pipelines. Let's see who beats whom.
In fact, nobody doubted that X1900 CrossFire just at that game would show its worth. But how comes the difference is so immense!
Increase of resolution in F.E.A.R. produces a substantial effect upon even the video subsystem which totally uses as many as 96 (!) pixel pipelines. The more strange it is that the performance boost turns out to be less than that in 1024x768. Normally, as the resolution goes up the result is reverse.
It is hard to believe, but the boost gained from merging two X1900-based video cards becomes even smaller and hardly reaches even 20%. That may be a bit disappointing for enthusiasts who are willing to pay extra for two latest and very expensive video cards in order to play at high resolutions on large-size monitors. Nevertheless, it's not yet the time to get upset. The drivers will definitely be improved, and the R580 architecture is flexible enough to adapt itself to any tasks. So, we'll come back to testing ATI CrossFire some time later once certified drivers have been released.
Finally, let's bring in a graph showing the results for X1900 CrossFire in the most demanding mode which can even be set in Catalyst Control Center.
As you see, even under these settings you can play quite comfortably at high resolutions. As regards F.E.A.R, let's wait for the final version of the drivers.
Conclusion
X1900XTX is undoubtedly a successful product. I believe the serious breakthrough in performance and functionality of video accelerators which is demonstrated by X1900XTX and X1900 CrossFire will make the dream of "cinematic" games of the future come true much earlier. As to the availability of video boards - ATI has learned the lessons of the past, and video cards are already on sale worldwide. They are still missing in Russia, but that is a matter of a few days.
 |
Top Stories: |
 |
 |
|
MoBo:
| |
|
|
VGA Card:
|
|
|
|
CPU & Memory:
|
|
