[citation][nom]clicky clicky[/nom]Here ya go Cleeve:http://www.overclockersclub.com/re [...] _955/8.htmhttp://benchmarkreviews.com/index. [...] mitstart=6[/citation]
Thank you! Some stuff to dig into.
First, the overclockersclub article: single GTX 260 used, the Phenom II benched at both stock clocks as well as 3.8 GHz.
far Cry 2 benched at the same Dx10 'Very high' setting which is nice, although they added 2xAA to the mix which will confuse things a bit. but for the most part the AA impact should be minimal as 2xAA isn't overly demanding for a GTX 260, although at 1920x1200 it might start to be a pretty big burden.
Let's look at the numbers:
Cyberpower Phenom 2, 3.6 GHz, dual Radeon 4890s:
1280x1024: 67.1
1680x1050: 62.7
1920x1200: 62.0
Overclockersclub Phenom 2, 3.79 GHz, single GTX 260:
1280x1024: 73
1680x1050: 70
1920x1200: 62
Delta:
1280x1024: 6.1 fps
1680x1050: 7.3 fps
1920x1200: 0 fps
There is an anomaly here, but if this game is CPU limited as I suggest, is it that big of an anomaly? The overclockersclub system is running 200 MHz faster, so if this game is indeed CPU bottlenecked, isn't this *exactly* what we'd expect to see... the dual-card setup having little effect, but the CPU overclock having more of an impact?
Digging a little deeper though, the overclockersclub Phenom II had very little benefit from the 3.8 GHz overclock, only gaining a single fps at each resolution, suggesting the system speed is having less of an effect than the graphics system.
Having looked at this however, it's hard to argue the Cyberpower system is grossly underperforming untile we have difinitive evidence that the game can perform much better with a better graphics subsystem. However, when we look at the Core i7 results...
SBM Core i7, stock, dual GTX 260s:
1280x1024: 89.5
1680x1050: 88.1
1920x1200: 86.8
Overclockersclub i7 920, stock, single GTX 260:
1280x1024: 64
1680x1050: 55
1920x1200: 49
Delta:
1280x1024: 25.5 fps
1680x1050: 33.1 fps
1920x1200: 37.8 fps
Wow. We're seeing a massive discrepancy here, when the i7 920 is paired with the single GTX 260 it's getting killed by the Phenom II, and when it's paired with two GTX 260's in SLI, it's soaring.
Conflicting evidence, to be sure. But we've experienced this type of thing before... Chris Angelini did a bit of research into it, as the i7 920 has performed poorly in conjunction with a single GTX card in the past, specifically with Far Cry 2 no less:
http://www.tomshardware.com/reviews/geforce-gtx-280,2156.html
This looks as wonky to me as it does to you guys, so I'm certainly going to dig a little deeper into this puzzle and spend a good deal of time putting Far Cry 2 through some different testing scenarios. But honestly, based on all of this information, it seems a little more likely to me that Far Cry 2 is demonstrating some anomalously low numbers when a single GTX card is paired with the i7, and that problem looks to be dissapearing when dual cards are used.
We'll see what happens after a thorough examination though.
As for the benchmarkreview.com numbers, it mentions they're using the Crysis benchmark too, but I can't find any information about whether they're using the default GPU test, or the optional CPU test.
At Tom's, we use the CPU test because we find it better simulates gameplay. In the CPU test, the camera follows the character actually shooting buildings and destroying the environment from a first person view, just like the game is played. We find this preferable to the GPU test, which simply renders a flyby of the island - not something a player will really experience during the game. And since we prefer to generate numbers closer to real-life game expectations, we use the CPU, not GPU test (which incidentally also relies on GPU power quite a bit, hey, it's Crysis). We've foun d Crysis to be quite reliant on both CPU and GPU speed to get decent frame rates.
in any case, thanks for the linkys, gives me some direction into what I can be looking for.