NVIDIA A10 PCIe vs NVIDIA A100 SXM4 40 GB
Graphics Processor
| GPU model | GA102 | GA100 |
|---|---|---|
| GPU variant | GA102-890-A1 | — |
| Architecture | Ampere | Ampere |
| Foundry | Samsung | TSMC |
| Manufacturing process | 8 nm | 7 nm |
| Number of transistors | 28,300 million | 54,200 million |
| Die size | 628 mm² | 826 mm² |
Graphics Card
| Release date | Apr 12th, 2021 | May 14th, 2020 |
|---|---|---|
| gpu.details.availability | 2021 | — |
| Generation | Tesla | Tesla |
| Production | Active | Active |
| Interface | PCIe 4.0 x16 | PCIe 4.0 x16 |
Clocks
| Base clock | 885 MHz | 1095 MHz |
|---|---|---|
| Boost clock | 1695 MHz | 1410 MHz |
| Memory clock | 1563 MHz 12.5 Gbps effective | 1215 MHz 2.4 Gbps effective |
| GPU clock | — | 1410 MHz |
Memory Configuration
| Memory size | 24 GB | 40 GB |
|---|---|---|
| Memory type | GDDR6 | HBM2E |
| Memory bus width | 384 bit | 5120 bit |
| Bandwidth | 600.2 GB/s | 1,555 GB/s |
Render Configuration
| Shading units | 9216 | 6912 |
|---|---|---|
| TMUs | 288 | 432 |
| ROPs | 96 | 160 |
| SM count | 72 | 108 |
| Tensor cores | 288 | 432 |
| RT cores | 72 | — |
| Cache L1 | 128 KB (per SM) | 192 KB (per SM) |
| Cache L2 | 6 MB | 40 MB |
Performance
| Pixel rate | 162.7 GPixel/s | 225.6 GPixel/s |
|---|---|---|
| Texture rate | 488.2 GTexel/s | 609.1 GTexel/s |
| FP16 (half) performance | 31.24 TFLOPS (1:1) | 77.97 TFLOPS (4:1) |
| FP32 (float) performance | 31.24 TFLOPS | 19.49 TFLOPS |
| FP64 (double) performance | 976.3 GFLOPS (1:32) | 9.746 TFLOPS (1:2) |
Dimensions & Outputs
| Slot width | Single-slot | IGP |
|---|---|---|
| Length | 267 mm 10.5 inches | — |
| Width | 112 mm 4.4 inches | — |
| TDP | 150 W | 400 W |
| Suggested PSU | 450 W | 800 W |
| Outputs | No outputs | No outputs |
| Power connectors | 8-pin EPS | None |
| Board number | PG133 SKU 215 | — |
API Support & Features
| DirectX | 12 Ultimate (12_2) | — |
|---|---|---|
| OpenGL | 4.6 | — |
| OpenCL | 3.0 | 3.0 |
| Vulkan | 1.2 | — |
| CUDA | 8.6 | 8.0 |
| Shader model | 6.6 | — |
