ATI FireGL V7700 vs NVIDIA GRID A100B
Processeur graphique
| Nom du GPU | RV670 | GA100 |
|---|---|---|
| GPU variant | RV670 GL | — |
| Architecture | TeraScale | Ampere |
| Fonderie | TSMC | TSMC |
| Taille du processus | 55 nm | 7 nm |
| Transistors | 666 million | 54,200 million |
| Taille des matrices | 192 mm² | 826 mm² |
Carte graphique
| Date de publication | Nov 19th, 2007 | May 14th, 2020 |
|---|---|---|
| Génération | FireGL | GRID |
| Production | End-of-life | Active |
| Prix de lancement | 899 USD | — |
| Bus interface | PCIe 2.0 x16 | PCIe 4.0 x16 |
| Critiques | 5 in our database | — |
Vitesse clock
| GPU clock | 777 MHz | — |
|---|---|---|
| Mémoire clock | 1126 MHz 2.3 Gbps effective | 1215 MHz 2.4 Gbps effective |
| Base clock | — | 900 MHz |
| Boost clock | — | 1005 MHz |
Mémoire
| Taille de la mémoire | 512 MB | 48 GB |
|---|---|---|
| Type de mémoire | GDDR4 | HBM2E |
| Bus mémoire | 256 bit | 6144 bit |
| Bande passante | 72.06 GB/s | 1,866 GB/s |
Render Config
| Shading units | 320 | 6912 |
|---|---|---|
| TMUs | 16 | 432 |
| ROPs | 16 | 192 |
| Compute units | 4 | — |
| Cache L2 | 256 KB | 48 MB |
| Comptage SM | — | 108 |
| Tensor cores | — | 432 |
| Cache L1 | — | 192 KB (per SM) |
Performance théorique
| Taux de pixel | 12.43 GPixel/s | 193.0 GPixel/s |
|---|---|---|
| Taux de texture | 12.43 GTexel/s | 434.2 GTexel/s |
| FP32 (float) performance | 497.3 GFLOPS | 13.89 TFLOPS |
| FP64 (double) performance | 99.46 GFLOPS (1:5) | 6.947 TFLOPS (1:2) |
| FP16 (half) performance | — | 55.57 TFLOPS (4:1) |
Board Design
| Taille Slot | Dual-slot | IGP |
|---|---|---|
| Longueur | 254 mm 10 inches | — |
| Taille | 111 mm 4.4 inches | — |
| TDP | unknown | 400 W |
| Suggestions pour le PSU | 200 W | 800 W |
| Sorties | 1x DVI1x DisplayPort1x S-Video | No outputs |
| Connecteurs de puissance | 1x 6-pin | None |
| Board number | B439 | — |
Caractéristiques graphiques
| DirectX | 10.1 (10_1) | — |
|---|---|---|
| OpenGL | 3.3 (full) 4.0 (partial) | — |
| OpenCL | — | 3.0 |
| Vulkan | — | — |
| Shader model | 4.1 | — |
| CUDA | — | 8.0 |
