GPU comparison
B200 vs RTX 5090.
Pick the B200 for memory: 180 GB vs 32 GB, with 4.5x the memory bandwidth. A 70B model in 8-bit weights (70 GB) fits on one B200, not on one RTX 5090. For models under 32 GB pick by price.
Rent either on Runcrate
- B200from $9.59/hr
$0.053 per GB/hr
- RTX 5090Cheaperfrom $0.72/hr per GPU
on an 8-GPU machine
$0.022 per GB/hr
Live hourly price, then per GB of memory.
Which is better for
Which to pick for each job. Memory fit, from the datasheets.
Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
B200: 1 GPU, 110 GB spare
RTX 5090: 3 GPUs, more than one 2x node
B200. Fits on one GPU.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
B200: 1 GPU, 164 GB spare
RTX 5090: 1 GPU, 16 GB spare
B200. Same GPU count, 4.5x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
B200: 1 GPU, 134 GB spare
RTX 5090: 2 GPUs (64 GB)
B200. Fits on one GPU.
Compare another pair
Blackwell
NVIDIA B200
- Memory
- 180 GB HBM3e
- Bandwidth
- 8.0 TB/s
Blackwell
NVIDIA RTX 5090
on an 8-GPU machine
- Memory
- 32 GB GDDR7
- Bandwidth
- 1.8 TB/s
At a glance
Spec by spec. The longer bar is the better number.
Full specs
The complete spec sheet.
| Spec | NVIDIA B200 | NVIDIA RTX 5090 |
|---|---|---|
| Architecture | Blackwell | Blackwell |
| Memory | 180 GB | 32 GB |
| Memory type | HBM3e | GDDR7 |
| Memory bandwidth | 8.0 TB/s | 1.8 TB/s |
| FP32 | 75 TFLOPS | 104.8 TFLOPS |
| FP16 | 2250 TFLOPS, tensor | 209.5 TFLOPS, non-tensor |
| INT8 | 4500 TOPS | 838 TOPS |
| TDP | 1000 W | 575 W |
| Form factor | SXM | Consumer |
| NVLink | Yes | No |
| GPUs per node | 8 | 2 |
| Hourly price (live) | from $9.59/hr | from $0.72/hr per GPU on an 8-GPU machine |
Questions
B200 or RTX 5090. What buyers ask first.
Other comparisons