There are many GPU and CPU Benchmarks out there.
Especially in the World of building a Workstation for 3D, VFX and Animation, putting your CPU, GPU, and other components through a series of tests and comparing them to the performance of other Systems and configurations is so much fun, that sometimes you upgrade, tune or overclock your System, just to see those numbers rise.
Most importantly though, the benchmark should be able to simulate a real-world work environment, because this is what it’s all about: Figuring out what Hardware Components will get you the maximum speed for the type of work you do.
Depending on what you use your computer for, there are lots of different CPU and GPU Rendering Benchmarks, that will fit your needs in finding and comparing the optimum Hardware Components.
What are the CPU and GPU responsible for in 3D?
Before diving into what Benchmarks to use (I’ll get into that a bit later), let’s take a look at what components in your System are mainly responsible for what tasks. This way you will be able to identify what component is responsible for the type of work you do on a daily basis.
CPU, Important Component in (among others):
- Calculating Scene-States (Objects, Modifiers, Deformers and the like)
- Animation-Calculation (Rigs, Dynamics)
GPU, Important Component in (among others):
- Displaying Calculated Scene Contents (Viewport)
- Displaying Polygons / Hi-Res sculpting (after they run through CPU-Calculations)
What makes a CPU and GPU strong in a 3D specific task?
Now that we know, what the CPUs and GPUs main tasks are, let’s take a look at what features these Hardware components must have to optimize these tasks.
- CPU-Rendering: The more cores and higher the clocks, the better
- Calculating Scene States: The Higher a single Core-Clock, the better (Think Turbo-Boost)
- Simulation: The Higher a single Core-Clock, the better (Think Turbo-Boost) (unless you have lots of independent sims that can be calculated simultaneously)
- Animation Calculations: The higher a single Core-Clock, the Better (again think Turbo-Boost)
- GPU-Rendering: The more CUDA-Cores or OpenCL Cores, the better, the more VRAM the better, the higher the GPU and VRAM clock speed the better
- Displaying Scene Contents / Displaying Polygons / Hi-Res sculpting: Usually a variety of OpenGL features, the higher the VRAM the better, the higher the Triangles/s the better
A common misconception, for example, is, that you need a faster GPU if the Viewport in your 3D-Application gets slow.
Even though the GPU is responsible for displaying Scene-Contents, most of the Time, the CPU, that first has to calculate deformers, modifiers, rigs and the like, before the GPU can display the resulting object/mesh, is responsible for slow viewport Speed.
Unless you display RAW Meshes without any type of Mesh-modifications, your CPU usually is the culprit in slow viewport speed.
Reasons for benchmarking your Hardware
For many of you, it will be fairly obvious, why you benchmark your System and specifically the Processor and Graphics Card since these are so important in a Digital Artist’s Day. But there are quite a few more reasons for benchmarking your hardware, that not everyone knows about:
Are your components running as fast as they should?
Probably the most obvious reason for benchmarking your CPU and GPU is, to compare your results with online Databases and see if your Hardware components reach the Speeds they are supposed to.
Possible problems when not reaching common speeds
If your hardware component does not reach the Speed other User’s are reaching, be sure you don’t have any background programs running while benchmarking, that will take away some performance. It is best to benchmark your System right after a clean install.
When benching your CPU and reaching scores that are too low, be sure your BIOS settings, such as Turbo-Boost, are correct and the CPU is not running too hot.
When benching your GPU and not reaching common scores, again be sure no other software is running in the background and your GPU driver is up-to-date. That said, though, sometimes reverting to an older Driver does the trick.
GPU Problems can also be:
- Too hot
- not enough PCIe-Lanes (or wrongly configured)
- the PCIe wrong slot
- not enough power from the PSU.
Also be aware, that GPU performance doesn’t always scale linearly when using Multiple GPUs. Using 2 GPUs might give you 1.9 times the performance, 4 GPUs might only give you 3.5 times the performance, depending on the benchmark you are using. Check out this article about GPU-Hardware here for more in-depth information.
Octane does a fairly good job in scaling linearly, but redshift, for example, will give you less performance per card, the more cards you add.
Planning on upgrading your PC? Bench it first!
Another great time to benchmark your Computer is when you are planning on buying new components or an entirely new PC.
You will only know if spending all that money is worth it if you know what performance improvement over your current Hardware you can expect.
In most cases, the higher the price does not always mean a better Hardware component for your specific use case. Maybe you only need to upgrade a specific bottleneck that slows down your overall system performance and not get an entirely new PC.
If you have a certain amount of budget you are willing to spend, benchmarking your current system first, will give you the possibility to roughly calculate what you can expect when getting new components within your budget.
Want to tune your System to have it run the best it can?
Overclocking and tuning/optimizing your System can greatly improve performance. There is no way around knowing if your optimizations have any effect if you don’t have a benchmarking baseline of your System’s default performance.
Benchmarks are also great for testing your System’s stability after overclocking.
Utilizing online Renderfarms
With Internet Connections getting faster and faster, online Renderfarms are more popular than ever. Calculating the cost of a Scene you would like to render on a Farm usually requires knowing your own System’s speed to be able to extrapolate.
Here is a screenshot of the Cost Estimator on the Ranch Computing Renderfarm. Notice how you have to know your Cinebench Score to be able to calculate the cost of rendering your scene on the Renderfarm:
It is also great to know the scores of your own PCs, If you have many of them, to best distribute specific tasks that run optimally on the different systems. Benchmarking your PCs regularly also notifies you of any performance issues that might arise after some time, that otherwise might go unnoticed.
Best Benchmark for testing your CPU Performance
Ok, so here they are: The actual Benchmarks, that let you test your System and Hardware Components’ performance.
The most popular Benchmark for testing Multi-Core and Single-Core performance of your CPU, especially in the 3D-Rendering world, is Cinebench R15.
The Cinebench CPU Render Benchmark in itself is quite simple. It renders a pre-defined Scene on your CPU.
Since rendering Scenes on your CPU is something you (in the CG Industry) probably do quite often, this CPU benchmark comes very close to real-world applications and is based on the 3D-Software Cinema 4D.
Cinebench can both render the Scene on all Cores, for multi-Core performance, and on a single Core, to obtain single-Core performance results.
Be sure to check “Advanced Benchmark” in the Menu to be able to test your single-core CPU performance too.
It also has an inbuilt GPU OpenGL performance benchmark, that is a fairly good indicator of your possible viewport performance. Note though, that this OpenGL benchmark can produce very inconsistent results in different CPU/GPU configurations. Cinebench runs on all major CPUs.
Just hit “Run” on any of the Benchmarking options, and the Benchmark will fire away!
Download the Cinebench R15 CPU Render Benchmark here.
Another great Benchmark for testing your CPU Render performance is the VRAY Benchmark. It is quite similar to Cinebench, as it renders a predefined Scene on your CPU (or GPU see below) and has an extensive online database to compare results in various configurations.
Download the Vray CPU Render Benchmark here.
Best Benchmark for benching your GPU
There are 3 very popular GPU render engines: Octane, Redshift, and VRAY-RT. All of these have GPU Render Benchmarks based on their engines. Octane and Redshift though only run on Nvidia GPU with CUDA-Cores.
Octane and VRAY GPU Benchmarks are easy to download, the Redshift Benchmark is currently only available for Customers.
The procedure is the same as in the Cinebench Benchmark. Hit “Run” to get your results.
Here are some CPU Benchmark results for Cinebench Single and Multi with performance/price ratio:
|CPU Name||Cores||GHz||Cinebench R15||Price($)||Performance/Dollar|
|AMD Ryzen 5 2600||6||3.4||1307||115||11.26|
|AMD Ryzen 5 2600X||6||3.6||1373||126||10.89|
|AMD Ryzen 7 2700X||8||3.7||1783||165||10.80|
|AMD Ryzen 7 2700||8||3.2||1526||145||10.52|
|AMD Threadripper 1920X||12||3.5||2431||240||10.12|
|AMD Threadripper 1900X||8||3.8||1711||200||8.55|
|AMD Ryzen 5 3600||6||3.6||1581||199||7.94|
|AMD Ryzen 7 1700X||8||3.4||1540||215||7.16|
|AMD Ryzen 7 1700||8||3.0||1426||209||6.80|
|AMD Ryzen 7 1800X||8||3.6||1613||239||6.74|
|AMD Ryzen 5 1600||6||3.2||1147||169||6.71|
|AMD Ryzen 5 1400||4||3.2||787||119||6.61|
|AMD Threadripper 2920X||12||3.5||2604||400||6.51|
|AMD Ryzen 7 3700X||8||3.6||2116||329||6.43|
|AMD Ryzen 9 3900X||12||3.8||3168||499||6.34|
|AMD Ryzen 5 1500X||4||3.5||803||139||5.77|
|AMD Ryzen 5 1600X||6||3.3||1250||219||5.71|
|AMD Threadripper 1950X||16||3.4||3062||550||5.56|
|AMD Ryzen 3950X||16||3.5||4070||750||5.42|
|AMD Ryzen 7 3800X||8||3.9||2166||399||5.42|
|Intel i5 8400||6||2.8||966||179||5.39|
|Intel i7 8700||6||3.2||1389||289||4.80|
|Intel i7 9700K||8||3.6||1542||360||4.28|
|AMD Threadripper 2970WX||24||3.0||4323||1025||4.21|
|Intel i9 9900K||8||3.6||2077||500||4.15|
|AMD Threadripper 2950X||16||3.5||3210||795||4.04|
|Intel i9 9940X||14||3.3||3173||829||3.82|
|Intel i7 8700K||6||3.7||1428||359||3.79|
|Intel i9 9980XE||18||3.0||3799||1049||3.62|
|Intel i7 7800X||6||3.5||1333||369||3.62|
|Intel i9 9960X||16||3.1||3211||929||3.45|
|Intel i9 9920X||12||3.5||2438||719||3.45|
|Intel i7 8086K||6||3.7||1386||425||3.26|
|Intel i7 7700K||4||4.2||996||310||3.21|
|Intel i7 7740X||4||4.3||986||329||2.99|
|Intel i5 9600K||6||3.7||1068||360||2.96|
|Intel i5 7600K||4||3.8||701||239||2.93|
|AMD Threadripper 2990WX||32||3.0||5224||1799||2.90|
|Intel i7 7820X||8||3.6||1734||599||2.89|
|Intel i7 7820X||8||3.6||1734||599||2.89|
|Intel i7 6800K||6||3.4||1096||419||2.61|
|Intel XEON E5-2620 v4||8||2.1||1096||420||2.60|
|Intel i9 7900X||10||3.3||2169||999||2.17|
|Intel i7 6850K||6||3.6||1235||570||2.16|
|Intel i9 7920X||12||2.9||2438||1200||2.03|
|Intel i9 7940X||14||3.1||2849||1450||1.96|
|Intel i9 7960X||16||2.8||3161||1700||1.89|
|Intel i9 7980XE||18||2.6||3455||1900||1.81|
|Intel i7 6900K||8||3.2||1562||1049||1.48|
|Intel XEON E5-2650 v4||12||2.2||1589||1200||1.32|
|Intel i7 5960X||8||3.0||1324||1069||1.23|
|Intel i7 6950X||10||3.0||1788||1649||1.08|
|Intel XEON E5-2687W v4||12||3.0||1860||2444||0.76|
|Intel XEON E5-2699 v4||22||2.2||2460||4500||0.54|
|AMD Threadripper 3960X||24||3.8||5933||1399||4.24|
|AMD Threadripper 3970X||32||3.7||7398||1999||3.70|
|Intel i9 10980XE||18||3.0||3799||999||3.80|
|AMD Threadripper 3990X||64||2.9||10449||3990||2.61|
|CPU Name||Cores||GHz||Cinebench R15||Price($)||Performance/Dollar|
Octanebench GPU results with performance/price ratio:
|GPU Name||VRAM||OctaneBench||Price $ MSRP||Performance/Dollar|
|GTX 1070 Ti||8||153||450||0.340|
|GTX 1080 Ti||11||222||700||0.317|
|RTX 2080 Ti||11||304||1199||0.253|
|GTX TITAN Z||12||189||2999||0.063|
Extensive VRAY Benchmark List can be found here.
Need help benchmarking your CPU / GPU? What scores did you reach? Let us know in the comments!