GPU performance in GFLOPS
This is from this project https://github.com/ProjectPhysX/FluidX3D which got some of my results:
Colors: 🔴 AMD, 🔵 Intel, 🟢 Nvidia, ⚪ Apple, 🟡 ARM
| Device | FP32 [GFlops/s] | Mem [GB] | BW [GB/s] | FP32/FP32 [MLUPs/s] |
|---|---|---|---|---|
| 🟡 Raspberry Pi 3 | ||||
| 🔵 UHD Graphics 620 (i5 7300U) | 422 | 6 | 10 | 72 |
| 🔵 UHD Graphics 630 (i5 8500T) | 422 | 13 | 23 | 150 |
| 🔵 UHD Graphics 770 (i7 13700T) | 819 | 4 | 20 | 135 |
| ⚪ Apple M1 | 2,048 | 11 | 56 | 727 |
| 🟢 GTX 960 | 2,593 | 2 | 79 | 513 |
| 🔴 RX 470 | 5,022 | 4 | 152 | 1006 |
| 🟢 P106-100 | 4,372 | 6 | 145 | 2045 |
| 🟢 GTX 1060 | 4,372 | 6 | 149 | 978 |
| 🟢 GTX 1070 | 6,463 | 8 | 201 | 1312 |
| 🔴 RX 6600 | 7,326 | 8 | 141 | 922 |
| 🟢 RTX 3060 Ti | 16,197 | 8 | 398 | 2604 |
| 🟢 RTX 3070 Ti | 21,750 | 8 | 529 | 3465 |
A run looks like this
.-----------------------------------------------------------------------------.
| ______________ ______________ |
| \ ________ | | ________ / |
| \ \ | | | | / / |
| \ \ | | | | / / |
| \ \ | | | | / / |
| \ \_.-" | | "-._/ / |
| \ _.-" _ "-._ / |
| \.-" _.-" "-._ "-./ |
| .-" .-"-. "-. |
| \ v" "v / |
| \ \ / / |
| \ \ / / |
| \ \ / / |
| \ ' / |
| \ / |
| \ / FluidX3D Version 2.13 |
| ' Copyright (c) Dr. Moritz Lehmann |
|-----------------------------------------------------------------------------|
|----------------.------------------------------------------------------------|
| Device ID 0 | NVIDIA GeForce RTX 3070 Ti |
| Device ID 1 | NVIDIA GeForce GTX 1060 6GB |
| Device ID 2 | Intel(R) HD Graphics 530 |
|----------------'------------------------------------------------------------|
|----------------.------------------------------------------------------------|
| Device ID | 0 |
| Device Name | NVIDIA GeForce RTX 3070 Ti |
| Device Vendor | NVIDIA Corporation |
| Device Driver | 551.23 (Windows) |
| OpenCL Version | OpenCL C 1.2 |
| Compute Units | 48 at 1770 MHz (6144 cores, 21.750 TFLOPs/s) |
| Memory, Cache | 8191 MB, 1344 KB global / 48 KB local |
| Buffer Limits | 2047 MB global, 64 KB constant |
|----------------'------------------------------------------------------------|
| Info: OpenCL C code successfully compiled. |
| Info: Allocating memory. This may take a few seconds. |
|-----------------.-----------------------------------------------------------|
| Grid Resolution | 256 x 256 x 256 = 16777216 |
| Grid Domains | 1 x 1 x 1 = 1 |
| LBM Type | D3Q19 SRT (FP32/FP32) |
| Memory Usage | CPU 272 MB, GPU 1x 1488 MB |
| Max Alloc Size | 1216 MB |
| Time Steps | 10 |
| Kin. Viscosity | 1.00000000 |
| Relaxation Time | 3.50000000 |
| Reynolds Number | Re < 148 |
|---------.-------'-----.-----------.-------------------.---------------------|
| MLUPs | Bandwidth | Steps/s | Current Step | Time Remaining |
| 3456 | 529 GB/s | 206 | 9997 70% | 0s |
|---------'-------------'-----------'-------------------'---------------------|
| Info: Peak MLUPs/s = 3465 |
-------------------------------------------------------------------------------
I need more time to find software and do the measurements, but I was inpired by the comparison of the graphics performance of my PS4 Pro to other consoles.
FP32 single
Let’s assume this is possible max raw performance in long (32 bit or single) FP32
- PS2 GFLOPS 16 Pixel shaders
- PS3
- PS4 1840 GFLOPS 18 CU, 8 GB GDDR5 memory 5500 MT/s
- PS4 Pro 4198 GFLOPS 36 CU, 32 ROPs, 144 TMUs, 2304 Cores, 256 bit bus, 217.6 GB/s
- PS5
- RX 470 3793 GFLOPS
- Apple M1 2600 GFLOPS (8-core, 128 CU or execution units, handle nearly 25,000 threads
- XBOX 360
- Xbos one S
- XBox Series S
- XBox Series X
In many cases it can be simple calculated by the CPU architecture and the frequency. For example my dual https://ark.intel.com/content/www/us/en/ark/products/37106/intel-xeon-processor-x5550-8m-cache-2-66-ghz-6-40-gt-s-intel-qpi.html with 2.67 GHz has a https://en.wikipedia.org/wiki/FLOPS (Nehalem EP) which results in 2.67 x 8 = 21.36 gflops.
FP64 double
- RX 470 237 GFLOPS