Cinebench R10 is an OpenGL 3D rendering performance test based on Cinema 4D from Maxon. Cinema 4D is a 3D rendering and animation tool suite used by 3D animation houses and producers like Sony Animation and many others. As such, it's an excellent (and free) rendering benchmark. We've historically used Cinebench R10, but are in the process of moving to R11.5 and have included both tests here for reference.
 |
|
Cinebench R10 64-bit
|
|
3D Rendering
|
|
Cinebench is a multi-threaded, multi-processor aware benchmark that renders a single 3D scene and tracks the length of the entire process. The rate at which each test system could render the entire scene is represented in the graph below.
Origin Genesis: Core i7 920 @ 3.8GHz, Maingear Shift: Core i7 980X @ 4.2GHz
The Shift is one of the fastest systems we've ever seen, smoking past the Origin Genesis that was setting HH records in this test a few months ago. It would be a mistake to present either system as slow, but Intel's Gulftown has wings (and cores) that 45nm Nehalem processors can't match, and a 10.5 percent clockspeed advantage to boot. Now let's check Maxon's latest Cinebench 11.5. One of the reasons Maxon released the update was to address a problem in R10's scaling. One of the challenges of parallel programming is that it takes a non-zero amount of time to spin off and retire additional threads. In Cinebench R10's case, this purportedly had a negative impact on the program's scaling in multicore configurations. Cinebench 11.5 fixes the problem—hopefully we'll see a difference.
|
|
|
Cinebench R11.5 64bit
|
|
Rendering Performance
|
|
|
Cinebench R11.5
|
Cinebench 11.5 is the latest update to Maxon's 3D rendering benchmark suite and the third major iteration of the Cinebench series. As with R10, CB11.5 includes a single-threaded, multi-threaded, and OpenGL test. We've focused on the first two tests as part of our processor comparison; the OpenGL test is a GPU-specific benchmark and is meant to represent professional graphics performance. Scores between the two benchmarks are not directly comparable, although it is possible to render R10's workload using 11.5, should you feel inclined.
|
Crunch through the math (we'll spare you), and Cinebench 11.5 scales quite a bit better than Cinebench R10, even at the quad-core level. In R10, the quad-core Origin's multi-threaded performance is exactly 4x faster than its single-threaded performance. That's actually quite good for a real-world application, but CB 11.5 does it one better. There, the Origin's multi-threaded score is 4.94x higher (up 23.5 percent) than its single-threaded result.
The Shift jumps even more. In R10, its MT score was 5.53x higher than its single-thread. In CB11.5, it was 7.35x higher, for a 33 percent increase in scaling rate.
The different scaling rates in R10 vs. CB11.5 is an excellent illustration of how difficult it is to take full advantage of theoretical performance boosts in the multi-core era. It wasn't until Maxon revised the benchmark that anyone released the previous version had left 1-1.5 cores worth of performance lying on the table. The fact that the previous version of the benchmark was fully multiprocessor-aware is that much more evidence of the complexity of the problem.
