Good morning, fellow 6990 users. I've accumulated the last week or so a variety of bench results, having swapped out my dual set of 6970s for a 2nd 6990. Here's a fast-access chart of those results, providing a bit of insight into the capability of these great cards in tandem. Without further aplomb, here's a summary of the testbed used to fashion the results:
i7-2600K @ 5.25 GHz (53 x 99.3)
XFX Radeon 6990 (air)
Sapphire Radeon 6990 (Koolance VID6990)
Patriot Sector 5 2GB x 4 @ 2154 MHz
MCP-35X + MCP-600
4x120 Swiftech Radiator
3x120 Swiftech #2
Kingwin Mach 1 1220 Watts #2
This set of tests derive from a relatively standard 3x1 Eyefinity portrait resolution. The GPUs were running a mere 950/1435 (exception: Crysis Warhead), variables since pushed further to 975/1500 (&, God willing, soon exceeding 1 GHz subsequent implementation of my 2nd Koolance WB). Resolution @ 3468x1920 pixels, equal to 30% more real-estate rendering than a 30" 2560x1600 surface.Alien vs. Predator:
3468x1920 | Maximum / Tessellation / SSAO / Very High | 0X AA | 16x AF | Ambush | Minimum: NA, Average: 100.8
, Maximum: NACrysis Warhead:
3468x1920 | Enthusiast | 2X EQAA | 16x AF | Ambush | Minimum: 43.88 , Average: 68.66
, Maximum: 88.42
3468x1920 | Enthusiast | 0XAA | 16x AF | Ambush | Minimum: 55.96 , Average: 75.07
, Maximum: 96.78Battlefield Bad Company 2:
3468x1920 | All Maximum | 16X EQAA | 5 minutes (first level) | Minimum: 15.919 , Average: 113.89
, Maximum: 165.40H.A.W.X. 2
3468x1920 | Maximum (minus Depth of Field) | 8XEQAA | 16x AF | Benchmark | Minimum: NA , Average: 154
, Maximum: 213Heaven 2.1
3468x1920 | Default | Tessellation Normal | 4xEQAA AF | Benchmark | Minimum: 25, Average: 100.2
, Maximum: 195.9Stalker Call of Pripyat:
I've posted the relevant images in full as to ease formatting the various results via text. The first graphic is with full MSAA 4xEQAA; the second, without AA.
Street Fighter IV
3468x1920 | Maximum | 16X EQAA | Benchmark | Minimum: NA, Average: 237.71
, Maximum: NA
As for now, that's the sum of 3x1 portrait results I'm able to provide. For the interested, I'm including separately a slew of 5x1 screenshots (which include a few benchmark results, although too few in number to illicit a second category) with relevant FPS recordings. For those who haven't tried 5x1, here's a very slight insight for what kind of frames & IQ you can expect. This is an 11+ megapixel 6990/Eyefinity 6 exclusive model, and extremely demanding for any team of GPUs to attempt! Just note that, despite what you may have heard, AMD gpus can easily handle these resolutions without encountering a phantom crossfire-bridge limitation (for transmitting screen-data). For the images below, and similar to the preceding preceding results, all in-game quality settings are maximized with 16xAF. Batman, Mafia 2, Metro 2033, & Battlefied: Bad Company 2 are in-operation with @ least 2xEQAA enabled. Crysis & Heaven 2.1 stand as exceptions, featuring no anti-aliasing.
No microstutter, stuttering, or ANY frame synchronization anomaly is observed, nor detected, in any of the provided examples / benchmarks. This encompasses the tested suite in entirety, from standard 1080P to 5x1 Eyefinity. AMD has made something resembling order of magnitude leaps in Eyefinity performance in surpassing it's predecessor generation. Having experienced the 5970/5870 in all possible renditions (single, crossfire, triple & quad), I'm confident to report all of the irritating and/or performance crippling issues once plaguing Eyefinity & multi-gpu builds has been vanquished. The only issues I encounter are born of driver immaturity related to individual titles. The most severe of examples this generation, Crysis 2, has been 100% fixed & stands as one of the most gorgeous & fluid running Eyefinity showcases.
As I'm going to be installing my 2nd 6990 waterblock this week, I'm looking forward to pushing the cards further & formulating a re-evaluation. If ya'll have benchs to submit for comparison, please post them up!