I'm curious to know why these results are in some cases so different from what I've seen on other sites as little as 6 weeks ago. Could nVidia have updated their drivers in the past 6 weeks to create these discrepancies, or is that the settings in the games are so different between the two tests.
Take for example FarCry2 perfomance from Anandtech benchmacks dated April '09 and compare them to the benchmarks here:
http://www.anandtech.com/video/showdoc.aspx?i=3539&p=19
http://www.tomshardware.com/charts/gaming-graphics-charts-2009/Far-Cry-2,1046.html
In the Tom's review the 4890 is beaten out by most of the new GTX2-- cards, including the GTX275, while in the Anandtech charts, it's barely beaten by the GTX285, and beats every other nV card in the list. Most of the other charts shake out in a similar fashion. Left4Dead at 4xAA, e.g....
http://www.tomshardware.com/charts/gaming-graphics-charts-2009/Left4Dead,1053.html
http://www.anandtech.com/video/showdoc.aspx?i=3539&p=20
Then we go over here to XBitlabs, and we are back to seeing numbers more in line with the Tom's results for Left4Dead;
http://www.xbitlabs.com/articles/video/display/bfg-gf-gtx275oc_9.html#sect1
...but when we look at numbers for CoD5 and Crysis on XbitLabs, the 4890 is once again in the lead, and these two games are conspicuous by their absence from the Tom's list;
http://www.xbitlabs.com/articles/video/display/bfg-gf-gtx275oc_7.html#sect1
Anyone have any idea what gives here? Am I missing something? Seems like a dramatic margin of error.
Perhaps the guys at each site could get together and reconcile these differences so that we humble readers could feel a little more confident in the information being presented.
Lastly, while I've always been a fan of EVGA (in fact, I'm seriously tempted to buy a pair of GTX285s from them today...), their GTX295 HydroCopper suffers from two BIG flaws;
1. It's nearly impossible to buy (Evga doesn't even offer it on their site)
2. The half aluminum / half copper heat sink design introduces mixed metals (read: 'galvanic corrosion') to the coolant in your WC loop - making it Most Foolish Choice Ever(tm) for responsible system builders
I vote for removing the GTX295 HydroCopper from the charts until these short comings are fixed. It's like a polished turd shining proudly at the top of each benchmark.
Thanks,
CList
p.s.
For those looking for crossfire and SLI numbers, I would direct you to this article at XBit which compares 285s, 275s, and 4890's in various single and x-fire combinations...
http://www.xbitlabs.com/articles/video/display/gtx285sli-hd4890cf.html