Specifications Compared
| Spec | B200 | GH200 |
|---|---|---|
| TDP | 1000W | 900W |
| VRAM | 180-192 GB | 96 GB |
| CUDA Cores | 18,432 | 16,896 |
| FP4 (dense) | 9,000 TFLOPS | Not published |
| FP8 (dense) | 4,500 TFLOPS | 1,979 TFLOPS |
| Memory Type | HBM3e | HBM3 |
| Architecture | Blackwell | Hopper |
| FP16 (dense) | 2,250 TFLOPS | 989 TFLOPS |
| Form Factors | SXM, NVL | SXM |
| INT8 (dense) | 4,500 TOPS | 1,979 TOPS |
| Interconnect | NVLink, PCIe 6.0, InfiniBand | NVLink-C2C, PCIe 5.0 |
| Tensor Cores | 576 | 528 |
| FP32 Performance | 75 TFLOPS | 67 TFLOPS |
| FP64 Performance | 37 TFLOPS | 34 TFLOPS |
| Memory Bandwidth | 8,000 GB/s | 4,000 GB/s |
| FP4 (with sparsity) | 18,000 TFLOPS | Not published |
| FP8 (with sparsity) | 9,000 TFLOPS | 3,958 TFLOPS |
| FP16 (with sparsity) | 4,500 TFLOPS | 1,979 TFLOPS |
| INT8 (with sparsity) | 9,000 TOPS | 3,958 TOPS |
Performance Analysis
The FP16 dense performance of 2250 TFLOPS on the B200 exceeds the 989 TFLOPS dense figure on the GH200 by more than double which accelerates matrix operations during both training and inference phases. FP32 performance remains close with 75 TFLOPS on the B200 and 67 TFLOPS on the GH200 so workloads heavy in single precision arithmetic see limited gains. Memory bandwidth doubled to 8000 GB/s allows larger batch sizes in memory intensive tasks without swapping data as frequently as the 4000 GB/s limit on the GH200. Sparse FP16 reaches 4500 TFLOPS on the B200 compared to 1979 TFLOPS with sparsity on the GH200 which benefits pruned models in inference scenarios. The FP8 dense rating of 4500 TFLOPS on the B200 versus 1979 TFLOPS dense on the GH200 further widens the gap for quantized inference pipelines. Overall the higher TDP of 1000W on the B200 supports these elevated compute rates while the GH200 operates within 900W.
Current On-Demand Offers
Cheapest secure on-demand offer per provider that is reported in stock right now, per GPU per hour; multi-GPU instances show the whole-instance price alongside. Deploy opens the DeployGPU console for providers on the platform, otherwise the provider's own site. Stock is rechecked every minute.
B200
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| RunPod | global | 1 | $7.99 | — | Deploy |
| VERDA | FIN-HEL | 8 | $8.03 | $64.23 | Deploy |
| Vast.ai | , US | 4 | $8.38 | $33.50 | Deploy |
3 providers in stock, 3 offers. All B200 offers, price history and alerts
GH200
1 provider in stock, 2 offers (cheapest per provider shown). All GH200 offers, price history and alerts
Notify me when B200 drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold. Cheapest right now: $7.99/GPU-hr.
QuantaCloud
Comparing H-series providers? We broker across all of them.
Hopper stock changes by the hour and prices differ by provider. If you need 16+ GPUs reserved or a cluster in the next 90 days, we quote H-series or B300 inventory at partner rates: one quote, 24h turnaround.
When to Choose the B200
The B200 suits environments that require maximum memory capacity of 180 to 192 GB HBM3e for handling very large models without partitioning across multiple devices. Its 8000 GB/s bandwidth and 2250 TFLOPS FP16 dense performance make it preferable for training sessions that demand high throughput in dense tensor operations.
When to Choose the GH200
The GH200 fits deployments where the 900W TDP allows higher density packing in racks compared to the 1000W requirement of newer options. Its 96 GB HBM3 and 4000 GB/s bandwidth remain adequate for workloads that fit within those limits such as certain scientific computing tasks using the 67 TFLOPS FP32 rating.
Use Cases
The B200 provides 2250 TFLOPS FP16 dense and 180 to 192 GB HBM3e which exceed the 989 TFLOPS FP16 dense and 96 GB on the GH200.
The B200 delivers 4500 TFLOPS FP8 dense along with 8000 GB/s bandwidth that support larger batch processing than the 1979 TFLOPS FP8 dense and 4000 GB/s on the GH200.
The B200 offers 4500 TFLOPS FP16 with sparsity and 180 to 192 GB memory capacity which accommodate adapter updates more efficiently than the GH200 specifications.
Both GPUs handle image generation workloads adequately with the B200 at 9000 TFLOPS FP4 dense and the GH200 at 1979 TFLOPS FP8 dense providing sufficient throughput.
The GH200 operates at 900W TDP with 67 TFLOPS FP32 which suits sustained simulations where the B200 1000W limit and 75 TFLOPS FP32 offer marginal benefit.
Frequently Asked Questions
How much more memory does the B200 have than the GH200?▾
The B200 provides 180 to 192 GB HBM3e while the GH200 provides 96 GB HBM3. This difference directly affects the size of models that fit on a single device during training or inference.
What is the memory bandwidth comparison between the two GPUs?▾
The B200 reaches 8000 GB/s while the GH200 reaches 4000 GB/s. The doubled bandwidth on the B200 supports larger batch sizes without increased data movement overhead.
Which GPU has higher FP16 dense performance?▾
The B200 achieves 2250 TFLOPS FP16 dense compared to 989 TFLOPS FP16 dense on the GH200. This gap applies to both training and inference workloads that rely on dense matrix multiplications.
What are the TDP values for the B200 and GH200?▾
The B200 lists a TDP of 1000W while the GH200 lists a TDP of 900W. The 100W difference influences rack level power density and cooling requirements in large clusters.
Do the GPUs share the same interconnect options?▾
The B200 supports NVLink, PCIe 6.0 and InfiniBand while the GH200 supports NVLink-C2C and PCIe 5.0. These specifications determine multi GPU scaling behavior in distributed setups.
Which is cheaper to rent, the B200 or the GH200?▾
Cloud rental prices for both the B200 and GH200 vary by provider, configuration, and availability. This page shows live pricing from 25+ providers updated every 60 seconds. Scroll to the Live Cloud Pricing section to compare current rates.
How much VRAM does the B200 have compared to the GH200?▾
The B200 has 180 to 192 GB of HBM3e memory. The GH200 has 96 GB of HBM3 memory.
Can I find B200 and GH200 GPUs available to rent right now?▾
Yes. This page shows real-time availability across 25+ cloud GPU providers. The Live Cloud Pricing section displays only in-stock offers with current pricing.
What is the main difference between the B200 and the GH200?▾
The B200 uses the Blackwell architecture (2024) while the GH200 uses Hopper (2023). The B200 delivers 2.3x the dense FP16 throughput (2,250 vs 989 TFLOPS, both without sparsity) and 2.0x the memory bandwidth of the GH200.
Rent these GPUs
Each GPU page lists every current on-demand offer by provider, per GPU-hour, updated every minute.
Related comparisons
How this page is made
- Specifications come from the NVIDIA, AMD and Intel datasheets for the B200 and the GH200. Dense and sparse throughput are listed separately.
- Prices and availability are live: every offer shown is an in-stock, on-demand listing from a provider we track, rechecked every minute.
- The written comparison (overview, performance notes, when to choose each card, verdict, use cases and FAQ) was drafted with an AI model from the specification table above and passed an automated check that rejects any figure not in that table. It was last generated on . It contains no prices; those are always read live.
- Read how we collect the data, or report an error on this page.
Next steps
- Rent B200Every current offer by provider, daily price history and a price alert.
- Rent GH200 Grace HopperEvery current offer by provider, daily price history and a price alert.
- GPU price indexHow on-demand prices for the major GPUs have moved, updated daily.
- All GPUsEvery GPU we track, with current lows and provider counts.