AMD Radeon Instinct MI50 vs NVIDIA Quadro K4200
Processeur graphique
| Nom du GPU | Vega 20 | GK104 |
|---|---|---|
| GPU variant | Vega 20 GLXT | GK104-850-A2 |
| Architecture | GCN 5.1 | Kepler |
| Fonderie | TSMC | TSMC |
| Taille du processus | 7 nm | 28 nm |
| Transistors | 13,230 million | 3,540 million |
| Taille des matrices | 331 mm² | 294 mm² |
Carte graphique
| Date de publication | Nov 18th, 2018 | Jul 22nd, 2014 |
|---|---|---|
| Génération | Radeon Instinct | Quadro |
| Production | Active | End-of-life |
| Bus interface | PCIe 4.0 x16 | PCIe 2.0 x16 |
| Critiques | — | 26 in our database |
Vitesse clock
| Base clock | 1200 MHz | 771 MHz |
|---|---|---|
| Boost clock | 1746 MHz | 784 MHz |
| Mémoire clock | 1000 MHz 2 Gbps effective | 1350 MHz 5.4 Gbps effective |
Mémoire
| Taille de la mémoire | 16 GB | 4 GB |
|---|---|---|
| Type de mémoire | HBM2 | GDDR5 |
| Bus mémoire | 4096 bit | 256 bit |
| Bande passante | 1,024 GB/s | 172.8 GB/s |
Render Config
| Shading units | 3840 | 1344 |
|---|---|---|
| TMUs | 240 | 112 |
| ROPs | 64 | 32 |
| Compute units | 60 | — |
| Cache L1 | 16 KB (per CU) | 16 KB (per SMX) |
| Cache L2 | 4 MB | 512 KB |
| Comptage SMX | — | 7 |
Performance théorique
| Taux de pixel | 111.7 GPixel/s | 21.95 GPixel/s |
|---|---|---|
| Taux de texture | 419.0 GTexel/s | 87.81 GTexel/s |
| FP16 (half) performance | 26.82 TFLOPS (2:1) | — |
| FP32 (float) performance | 13.41 TFLOPS | 2.107 TFLOPS |
| FP64 (double) performance | 6.705 TFLOPS (1:2) | 87.81 GFLOPS (1:24) |
Board Design
| Taille Slot | Dual-slot | Single-slot |
|---|---|---|
| Longueur | 267 mm 10.5 inches | 241 mm 9.5 inches |
| TDP | 300 W | 108 W |
| Suggestions pour le PSU | 700 W | 300 W |
| Sorties | 1x mini-DisplayPort | 1x DVI2x DisplayPort |
| Connecteurs de puissance | 1x 6-pin + 1x 8-pin | 1x 6-pin |
| Taille | — | 111 mm 4.4 inches |
| Board number | — | P2004 |
Caractéristiques graphiques
| DirectX | 12 (12_1) | 12 (11_0) |
|---|---|---|
| OpenGL | 4.6 | 4.6 |
| OpenCL | 2.1 | 3.0 |
| Vulkan | 1.2 | 1.1 |
| Shader model | 6.4 | 5.1 |
| CUDA | — | 3.0 |
