Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Apoppin is correct in the fact that the cards are able to be clocked independently and still work in crossfire and none of the cards will change their clocks to match the others.
However... it will still load balance to the extent that each card will be doing an equivalent amount of work. If one card is more powerful you will actually see lower GPU utilization on that card compared to the weaker card.
From what I remember about these discussions, it turned out that all the cards *will* downclock to match the lowest clocked card, but that this downclock won't actually be visible to the user.
From what I remember about these discussions, it turned out that all the cards *will* downclock to match the lowest clocked card, but that this downclock won't actually be visible to the user.
only SLI will do that I believe, not CF. since you can do multiple different card together for CF, but not SLI.
Ah yeah, could be it was about SLI. Am I correct in thinking that the memory clocks of the two cards in CF must match though?
Yep, I can confirm I had no issues running CF 5850's at different speeds. People seem to think the rules that apply to SLI apply to CF.
Let me try againI don't think that is really the main source of confusion here. It's pretty clear at this point that clocks aren't changing.
The remaining point of contention seems to be whether mixing uneven cards results in a performance boost over two of the slower cards, or to what degree Crossfire is able to balance the load between cards with different capabilities.
Now there are quite a few benchmarks on the net that do show crossfire slowing down to the "lowest common denominator" when using mixed cards, though there are also some user results such as those above that show Crossfire being able to take advantage of the faster card more than the slower card.
If anything this is the point that needs clarification, and preferably some idea of what is going on under the hood rather than conjecture based on user experiences.
Let me try again
First of all, i don't think you should benchmark with two CrossFire interconnects.I read your article.
Here is an article using newer cards and newer drivers and out of all the tests they did 5870+5850 was only faster than 5850 crossfire in one benchmark and that was well within the margin of error (I think it was like 300 vs 304 fps). In most cases, 5870+5850 crossfire was actually slower than 5850 crossfire.
http://www.legitreviews.com/article/1107/1/
Again, I value the work you've done to investigate this issue but the question is clearly not answered yet with any degree of clarity or certainty.
Could it be due to the different stream processor count in the 5850 vs. 5870 whereas 4870, 4890, and 4870x2 all have the same number of stream processors per GPU? That's just one question that is still out there.
Very much looking forward to this....i am working on it as part of a series dealing with finding the limits of HD 5870 architecture
Thirdly, they only tested at 1920x1200 which does not stress 5850/5870 CF at all
What conclusions can you draw from 3 games that may be also CPU-dependent ?
Well they tested 5850 crossfire and 5870 crossfire. 5870 crossfire was faster than 5850 crossfire, as expected, so clearly there is room left for improvement with more GPU power.
If the resolution or lack of CPU power were the reason 5850+5870 is not faster than 5850 crossfire, then it would stand to reason that there would be little to no benefit from going to 5870 crossfire either.
Looking forward to reading your results.