NVIDIA GB200 NVL72 GPU 186GB vs NVIDIA GeForce RTX 5090 32GB

The GB200 delivers 6.0x the FP8 throughput of the GeForce RTX 5090 (5,000 vs 838 TFLOPS dense). The GB200 also carries 154 GB more memory (186 GB vs 32 GB).

GB200: Blackwell, 2024 GeForce RTX 5090: Blackwell, 2025
FP8 dense lead
6.0x
GB200: 5,000 vs 838 TFLOPS
VRAM
186 vs 32 GB
154 GB more for the GB200
Bandwidth
8.0 TB/s vs 1.8 TB/s
GB200 moves data faster

Specifications

GB200
GeForce RTX 5090
Architecture
BlackwellBlackwell
Launch Year
20242025
Form Factor
SXMPCIe
VRAM
186 GB32 GB
Memory Bandwidth
8.0 TB/s1.8 TB/s
TDP
575 W
Process Node
4nm4nm

Performance (TFLOPS)

GB200
GeForce RTX 5090
FP64
40.0 TFlops 1.6 TFlops
FP32
80.0 TFlops 104.8 TFlops
TF32
1250.0 TFlops
2500.0 TFLOPS sparse
209.5 TFlops
sparse not published
BF16
2500.0 TFlops
5000.0 TFLOPS sparse
419.1 TFlops
sparse not published
FP16
2500.0 TFlops
5000.0 TFLOPS sparse
419.1 TFlops
sparse not published
FP8
5000.0 TFlops
10000.0 TFLOPS sparse
838.2 TFlops
sparse not published
FP6
5000.0 TFlops
10000.0 TFLOPS sparse
No verified data available
FP4
10000.0 TFlops
20000.0 TFLOPS sparse
1676.0 TFlops
sparse not published
INT8
5000.0 TFlops
10000.0 TOPS sparse
838.2 TFlops
sparse not published

FLOPS by Precision

What actually differs

The GeForce RTX 5090 is the newer part: Blackwell, launched in 2025, against the GB200's Blackwell from 2024. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the GB200 against the GeForce RTX 5090: FP64 40 vs 1.6 TFLOPS, FP32 80 vs 105 TFLOPS, FP16 2,500 vs 419 TFLOPS, FP8 5,000 vs 838 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 186 GB against 32 GB, fed at 8.0 TB/s versus 1.8 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.

Frequently asked questions

Is the GB200 faster than the GeForce RTX 5090?

At FP8 precision the GB200 reaches 5,000 TFLOPS dense against 838 TFLOPS for the GeForce RTX 5090. The performance table on this page lists every published precision for both GPUs.

Which has more memory, the GB200 or the GeForce RTX 5090?

The GB200 carries 186 GB of VRAM versus 32 GB for the GeForce RTX 5090. Memory bandwidth is 8.0 TB/s for the GB200 and 1.8 TB/s for the GeForce RTX 5090.

Can I rent the GB200 or the GeForce RTX 5090 in the cloud?

Yes. Live cloud listings tracked by Flopper start at $0.32 per GPU hour. The rental pricing section on this page lists current providers and rates for both GPUs.

Where to Rent

NVIDIA GeForce RTX 5090 32GB

ProviderConfigurationPrice/GPU-hrChecked
Vast.ai
1× GeForce RTX 5090 32GB Community $0.32 2h ago View →
Excess Supply
1× GeForce RTX 5090 32GB $0.65 2h ago View →
RunPod
1× GeForce RTX 5090 32GB Community $0.69 2h ago View →
RunPod
1× GeForce RTX 5090 32GB $0.99 2h ago View →
All GeForce RTX 5090 listings and price history →

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.