Datacenter GPU Database
Compare Datacenter GPU Specs
Side-by-side FLOPS, memory, power, and architecture data for every major datacenter GPU.
50/172 · p1/4
| Compare | Rent | |||||||
|---|---|---|---|---|---|---|---|---|
NVIDIA Vera Rubin 2× GPU Superchip·Rubin | 35000.0 | 8000.0 | 576 GB | — | — | — | View → | |
AMD Instinct MI455X OAM·CDNA 5 | 20132.7 | — | 432 GB | — | — | — | View → | |
NVIDIA Rubin SXM·Rubin | 17500.0 | 4000.0 | 288 GB | — | — | — | View → | |
NVIDIA GB200 2× GPU Grace Blackwell Superchip 372GB·Blackwell | 10000.0 | 5000.0 | 372 GB | — | — | — | View → | |
M Meta MTIA 450 MTIA | 7000.0 | — | 288 GB | 1400 W | — | — | View → | |
M Meta MTIA 400 MTIA | 6000.0 | — | 288 GB | 1200 W | — | — | View → | |
Microsoft Maia 200 Maia 200 | 5050.0 | — | 216 GB | 750 W | — | — | View → | |
AMD Instinct MI355X OAM·CDNA 4 | 5033.2 | 2516.6 | 288 GB | 1400 W | — | — | View → | |
NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip 252GB·Blackwell Ultra | 5000.0 | 2500.0 | 252 GB | — | — | — | View → | |
AMD Instinct MI350X OAM·CDNA 4 | 4614.0 | 2309.6 | 288 GB | 1000 W | — | — | View → | |
Google TPU v7 192GB·TPU | 4614.0 | — | 192 GB | — | — | — | View → | |
NVIDIA B200 SXM 180GB·Blackwell | 4500.0 | 2250.0 | 180 GB | 1000 W | $6.23 | 722 | Rent → | |
NVIDIA B300 SXM 288GB·Blackwell Ultra | 4500.0 | 2250.0 | 288 GB | 1400 W | $7.65 | 588 | Rent → | |
HW Huawei Ascend 970 Da Vinci v5 | 4000.0 | 2000.0 | 288 GB | — | — | — | View → | |
NVIDIA B100 SXM 192GB·Blackwell | 3500.0 | 1750.0 | 192 GB | 700 W | — | — | View → | |
AMD Instinct MI300X 192GB·CDNA 3 | 2614.9 | 1307.4 | 192 GB | 750 W | $2.39 | 1,094 | Rent → | |
AMD Instinct MI325X OAM·CDNA 3 | 2614.9 | 1307.4 | 256 GB | 1000 W | — | — | View → | |
AWS AWS Trainium3 NeuronCore-v4 | 2517.0 | 671.0 | 144 GB | — | — | — | View → | |
RE Rebellions Rebel100 REBEL | 2048.0 | 1024.0 | 144 GB | 600 W | — | — | View → | |
HW Huawei Ascend 960 Da Vinci v4 (chiplet) | 2000.0 | 1000.0 | 288 GB | — | — | — | View → | |
NVIDIA H100 SXM5 80GB·Hopper | 1979.0 | 990.0 | 80 GB | 700 W | $1.79 | 1,106 | Rent → | |
NVIDIA H200 SXM 141GB·Hopper | 1979.0 | 990.0 | 141 GB | 700 W | $4.00 | 495 | Rent → | |
NVIDIA GH200 96GB HBM3·Hopper | 1979.0 | 990.0 | 96 GB | 1000 W | $2.29 | 864 1 quote | Rent → | |
NVIDIA GH200 144GB HBM3e·Hopper | 1979.0 | 990.0 | 144 GB | 1000 W | — | — | View → | |
AMD Instinct MI300A 128GB·CDNA 3 | 1961.2 | 980.6 | 128 GB | 760 W | — | — | View → | |
Intel Gaudi 3 128GB·Gaudi 3 | 1678.0 | 459.0 | 128 GB | 900 W | — | — | View → | |
Intel Gaudi 3 HL-338·Gaudi 3 | 1678.0 | 459.0 | 128 GB | 600 W | — | — | View → | |
NVIDIA H100 PCIe 80GB·Hopper | 1513.0 | 756.5 | 80 GB | 350 W | $2.89 | 524 | Rent → | |
NVIDIA H800 PCIe 80GB·Hopper | 1513.0 | 756.5 | 80 GB | 350 W | — | — | View → | |
AWS AWS Trainium2 NeuronCore-v3 | 1299.0 | 667.0 | 96 GB | — | — | — | View → | |
NVIDIA Groq 3 LPU Groq LPU | 1200.0 | — | — | — | — | — | View → | |
M Meta MTIA 300 MTIA | 1200.0 | — | 216 GB | 800 W | — | — | View → | |
HW Huawei Ascend 950DT Da Vinci v3 | 1034.0 | 547.0 | 144 GB | — | — | — | View → | |
HW Huawei Ascend 950PR Da Vinci v3 | 1000.0 | 500.0 | 128 GB | 900 W | — | — | View → | |
MT Moore Threads MTT S5000 MUSA 4th Gen | 1000.0 | — | 80 GB | — | — | — | View → | |
Intel Gaudi 2 96GB·Gaudi 2 | 865.0 | 432.0 | 96 GB | 600 W | — | — | View → | |
NVIDIA GeForce RTX 5090 32GB·Blackwell | 838.2 | 419.1 | 32 GB | 575 W | $0.65 | 1,290 | Rent → | |
HW Huawei Atlas 350 Da Vinci v3 | 804.0 | 425.0 | 112 GB | 600 W | — | — | View → | |
NVIDIA L40S 48GB·Ada Lovelace | 733.0 | 362.1 | 48 GB | 350 W | $0.87 | 843 | Rent → | |
NVIDIA RTX 6000 Ada 48GB·Ada Lovelace | 728.5 | — | 48 GB | 300 W | $0.84 | 867 | Rent → | |
TE Tenstorrent Blackhole p100a·Blackhole | 664.0 | — | 28 GB | 300 W | — | — | View → | |
TE Tenstorrent Blackhole p150a·Blackhole | 664.0 | — | 32 GB | 300 W | — | — | View → | |
TE Tenstorrent Blackhole p150b·Blackhole | 664.0 | — | 32 GB | 300 W | — | — | View → | |
NVIDIA RTX 5000 Ada 32GB·Ada Lovelace | 522.2 | — | 32 GB | 250 W | $0.83 | 629 1 quote | Rent → | |
FU Furiosa RNGD Tensor Contraction Processor | 512.0 | — | 48 GB | 180 W | — | — | View → | |
Google TPU v5p 95GB·TPU v5p | 459.0 | — | 95 GB | 450 W | — | — | View → | |
NVIDIA GeForce RTX 5080 16GB·Blackwell | 450.2 | 225.1 | 16 GB | 360 W | $0.59 | 763 1 quote | Rent → | |
NVIDIA L40 48GB·Ada Lovelace | 362.0 | 181.1 | 48 GB | 300 W | $0.82 | 441 | Rent → | |
NVIDIA GeForce RTX 4090 24GB·Ada Lovelace | 330.3 | 165.2 | 24 GB | 450 W | $0.40 | 826 | Rent → | |
NVIDIA H20 96GB·Hopper | 296.0 | 148.0 | 96 GB | 400 W | — | — | View → |
per page
1/4
All values are theoretical peak TFLOPS from official datasheets. Glossary