System Configuration 6970 base system AMD Phenom II 965BE @ stock AMD 6970 single and crossfire Mushkin 1333 Cl7 4gigs Gigabyte 790fx UD5P 5850 xfire base system AMD Phenom II 940BE @ 3.4ghz cpu 2200nb ATi 5850s in crossfire Gskill 800 CL 5 Asrock 790gx Games and Benchmarks to be run in comparison to the 5850s Crysis Warhead Devil May Cry 4 The Last Remnant Games and Apps to be testing in single and dual card 6970s Metro 2033 Crysis Warhead Devil May Cry 4 The Last Remnant Dirt 2 Fallout New Vegas Battlefield Bad Company 2 Just Cause 2 Left 4 Dead 2 Alien vs Predator Bench Heaven Bench Mecha Rampage Demo 6970 xfire vs 5850 xfire 1680x1050 Now it is obvious that at 1680x1050 the cpu limit is more evident and as other reviews have shown the higher the resolution the better the 6900s scale but since this is a direct comparison apples to apples of 5850 xfire vs 6970 xfire at release, those limitations will be ignored entirely Crysis Warhead 1680x1050 Max Settings 4xAA Devil May Cry 4< DX9 1680x1050 8xAA It would seem in DX9 the 6900 series offers almost no real improvement in this title untill you hit higher resolutions. This is a situation where an i7 is needed to break the deadlock or a heavily overclocked (4ghz+) Phenom II The Last Remnant DX9 1680x1050 Max Settings< 6970 Stock CPU + GPU Review All games are run at 1920x1200 in this portion of the review unless otherwise noted. Unigine Heaven Bench 2.1 Tessellation Moderate Crossfire Scaling 75% Tessellation Normal Crossfire Scaling 86% Tessellation Extreme Crossfire Scaling 92% Metro 2033 Max Settings AA mode = AAA DX11 Tessellation = On DX11 DoF = off Crossfire Scaling 69% Metro 2033 Max Settings AA mode = AAA DX11 Tessellation = On DX11 DoF = On Crossfire Scaling 96% Battlefield Bad Company 2 Max Settings 4xAA Single player was tested in Mission 12 Single Gpu mode Battlefield Bad Company 2 Max Settings 4xAA Single player was tested in Mission 12 Dual Gpu mode Crossfire Scaling 58% Single Player Crossfire Scaling 18% Multiplayer Dirt 2 Max Settings 4xAA Crossfire Scaling 24% Alien Vs Predator Max Settings 4xAA Crossfire Scaling 89% Fallout New Vegas Max Settings 4xAA Crossfire Scaling 0% Seems i have the bug where the Gamebryo engine is stuck with Vsync on even when disabled via CCC in game and via INI file settings Just Cause 2 Max Settings 4xAA Crossfire Scaling 33% Left 4 Dead 2 Max Settings 4xAA Crossfire Scaling 17% Mecha Rampage Demo Default settings 1920x1080 Crossfire Scaling 64% Crysis Warhead Enthusiast 4xAA Ambush bench Crossfire Scaling 8% It would seem a stock Phenom II heavily Bottlenecks the 6970s in dual gpu setup. I shall return to this bench once we reach the Overclocking portion of the review. Devil May Cry 4 DX10 8xAA Super High Crossfire Scaling Scene 1 92%</b><b>Crossfire Scaling Scene 2 84% Crossfire Scaling Scene 3 44%</b><b>Crossfire Scaling Scene 4 73% Crossfire Scaling Average of all Scenes 73% Please note the bar graph below is set up as Single vs Dual per scene Devil May Cry 4 Bench has 4 major scenes each with there own score. Grey, Red, Blue, Light Green are Single GPU Black, Brown, Dark Blue, Dark Green are Dual GPU The Last Remnant Max Settings 1920x1200 Crossfire Scaling 63% Overall the extra performance gained from Crossfire on a stock AMD rig is around 53% in terms of scaling if we remove the bugged Fallout New Vegas run its 57%. This is without a doubt better then the 5800 series scaling on this platform but the performance gain isnt what we would like to see. The next round of benches will focus on overclocking both the CPU and the GPU to see what gains can be had on an AMD gaming rig. 6970 Crossfire Stock vs Overclocked Review CPU @ 3800mhz NB @ 2600mhz GPU @ 950 core 1400 memory RAM @ 1333mhz 7-7-7-20-1T Crysis Warhead Enthusiast 4xAA Ambush bench Alien Vs Predator Max Settings 4xAA Metro 2033 Max Settings AA mode = AAA DX11 Tessellation = On DX11 DoF = On It seems even when Overclocked to 3800mhz CPU 2600NB the AMD Phenom II 965BE just isnt enough to push these gpus especially when overclocked. Crysis Warhead was particularly bad with GPU utilization on the AM3 platform at a dismal 45% per GPU or less. Where as Metro 2033 and Alien vs Predator show a GPU usage of 90-99%. Overall the 6970s in crossfire are a potent combination but there something that should be avoided on an AMD platform. Bulldozer might change things in 2011 but for now a full AMD platform just dosent make sense. I am fairly disappointed that crossfire scaling while good is severely hampered on the AMD cpu compared to almost every Core I series cpu. After comparing with reviews around the web, in many other games that ive not listed here the AMD cpus results in a near 30% reduction in performance in some cases, this i feel is unacceptable the AMD driver team needs to seriously pick up the pace and optimize there drivers for there own platforms, a 965BE overclocked with dual gpus is in comparison to the Core i5 750 @ stock slower in almost every test, and while I am impressed with the GPUs themselves GPU usage due to a cpu bottleneck is nothing short of tragic, games like crysis sitting at 30-40% usage per gpu means theres alot of untapped potential. Then theres games like Bad Company 2 that GPU usage also suffers with the average usage being around 50-55% per GPU again this greatly hinders performance. There is still much work and optimizations that need doing but so far the 6970s in crossfire come in at $260 cheaper then a GTX 580 SLI setup and perform only 5-8% slower while being faster then GTX 570 sli, thats extreme performance for your dollar, Lets just hope for those AMD diehards that they can improve there Crossfire performance on there own CPUs, as it really is a shame that right now Intel is the only game in town when it comes to being able to push a multi gpu system. And with bulldozer still months away the potential cure all it may or may not offer is still nowhere in sight. Till next time everyone happy gaming overclocking folding and crunching.