NVIDIA A100 PCIe vs NVIDIA Tesla P40
Processeur graphique
| Nom du GPU | GA100 | GP102 |
|---|---|---|
| Architecture | Ampere | Pascal |
| Fonderie | TSMC | TSMC |
| Taille du processus | 7 nm | 16 nm |
| Transistors | 54,200 million | 11,800 million |
| Taille des matrices | 826 mm² | 471 mm² |
Carte graphique
| Date de publication | Jun 22nd, 2020 | Sep 13th, 2016 |
|---|---|---|
| Génération | Tesla | Tesla |
| Production | Active | End-of-life |
| Bus interface | PCIe 4.0 x16 | PCIe 3.0 x16 |
| Prix de lancement | — | 5,699 USD |
Vitesse clock
| GPU clock | 1410 MHz | — |
|---|---|---|
| Mémoire clock | 1215 MHz 2.4 Gbps effective | 1808 MHz 14.5 Gbps effective |
| Base clock | 765 MHz | 1303 MHz |
| Boost clock | 1410 MHz | 1531 MHz |
Mémoire
| Taille de la mémoire | 40 GB | 24 GB |
|---|---|---|
| Type de mémoire | HBM2E | GDDR5X |
| Bus mémoire | 5120 bit | 384 bit |
| Bande passante | 1,555 GB/s | 694.3 GB/s |
Render Config
| Shading units | 6912 | 3840 |
|---|---|---|
| TMUs | 432 | 240 |
| ROPs | 160 | 96 |
| Comptage SM | 108 | 30 |
| Tensor cores | 432 | — |
| Cache L1 | 192 KB (per SM) | 48 KB (per SM) |
| Cache L2 | 40 MB | 3 MB |
Performance théorique
| Taux de pixel | 225.6 GPixel/s | 147.0 GPixel/s |
|---|---|---|
| Taux de texture | 609.1 GTexel/s | 367.4 GTexel/s |
| FP16 (half) performance | 77.97 TFLOPS (4:1) | 183.7 GFLOPS (1:64) |
| FP32 (float) performance | 19.49 TFLOPS | 11.76 TFLOPS |
| FP64 (double) performance | 9.746 TFLOPS (1:2) | 367.4 GFLOPS (1:32) |
Board Design
| Taille Slot | Dual-slot | Dual-slot |
|---|---|---|
| Longueur | 267 mm 10.5 inches | 267 mm 10.5 inches |
| TDP | 250 W | 250 W |
| Suggestions pour le PSU | 600 W | 600 W |
| Sorties | No outputs | No outputs |
| Connecteurs de puissance | 8-pin EPS | 1x 6-pin + 1x 8-pin |
| Board number | P1001 SKU 200 | PG610 SKU 200 |
| Taille | — | 111 mm 4.4 inches |
Caractéristiques graphiques
| DirectX | — | 12 (12_1) |
|---|---|---|
| OpenGL | — | 4.6 |
| OpenCL | 3.0 | 3.0 |
| Vulkan | — | 1.2 |
| CUDA | 8.0 | 6.1 |
| Shader model | — | 6.4 |
