Processor benchmarks with Resident Evil 5: Core i7 reigns, Phenom strong - Update: Lynnfield results
The official benchmark demo of the Resident Evil 5 PC version has been released. PC Games Hardware ran tests with 20 CPUs.
Resident Evil 5: Engine
Capcom's new game is, like for example Devil May Cry 4, based on the MT Framework Engine. This architecture utilizes almost any modern rendering technology from HDR, Hemisphere Lighting as well as Soft Shadows and Soft Particles to Depth of Field and Motion Blur or Alpha To Coverage and static Ambient Occlusion. Besides the massive multi-core optimizations offered by the MT Framework (2.0), it is also DirectX 11 ready, but the version used in Resident Evil 5 is "only” a DirectX 9 respectively DirectX 10 renderer. In comparison to the console version the PC version of Resident Evil 5 delivers sharper textures and offers higher resolutions.
Resident Evil 5: Benchmark Tool and Results
Capcom has released the benchmark tool of Resident Evil 5. While we used the "Variable Benchmark” for our Resident Evil 5 graphics cards benchmarks, we switched to the "Fixed Benchmark” for the processor tests. This mode repeatedly shows the same sequence (real-time demo). Conclusions on the performance of the final Resident Evil 5 can only be drawn to a certain point, but the "Fixed Benchmark” nevertheless gives a suitable indication at what you will have to expect in CPU limited scenes - for example when hordes of infected are attacking you. For more specific results we once again use the tool Fraps and record the minimal framerate. Attention: The "JobThread” in the Resident Evil 5 config.ini has to be set to at least the number of thread minus one!
As its name indicates the MT Framework Engine scaling very good with multi-core processors: A C2Q Q6600 (4x 2,4 GHz) is 41 percent faster than C2D E6600 (2x 2.4 GHz), while the C2D E8400 is regardless of its 2x 3.0 GHz and 6Mibyte L2 cache only as fast as the Q6600. In contrast to many other engines Capcom's development does benefit when running on four instead of three cores - depending on the frequency the framerate is increased by about 13 percent. The MT Framework also likes more L2 cache - the C2Q Q9650 (12 MiByte) beats the C2E QX6850 (8 MiByte) which is running at the same frequency by about 9 percent.
AMD's Phenoms are gleaming in Resident Evil 5: the older Phenom X4 9950 beats the Q6600 and the Phenom II X4 945 (4x 3.0 GHz), if running at the same frequency, beats the more expensive C2Q Q9650 by about 6 percent - normally the Intel CPU is the faster one if those two are compared. Neverthless even a (simulated) Phenom II X4 965 (4x 3.4 GHz) doesn't stand a chance against Intel's Core i7 array - even the smallest model, the Core i7-920 with 2.67 GHz is about 17 percent faster. If running at the same clock speed the Nehalem beats the Deneb by almost 25 percent and the Yorkfield even by about 32 percent - even though the Core i7s are slowed down by the Geforce GTX 285. Smaller resolutions like 800 x 600 place the Core i7 more than 50 percent in front - so the "Runs great on Intel Core i7” logo hasn't been placed at the beginning of the benchmark without a reason.
SMT doesn't deliver more fps, but doesn't slow down the system either. The Turbo mode vanishes in 1,680 x 1,050 since at around 108 fps the GPU becomes the limiting factor.
Update: September 15, 2009
Intel's new Lynnfield processors, the Core i5-750 (4x 2.67 GHz) and the Core i7-860 (4x 2.8 GHz) to be exact, place themselves right where they were expected to: The Core i5 gets close in on the Core i7-920 and beats the Phenom II X4 965 BE. The i7-860 is able to beat the i7-920 because of its higher core frequency.