Can I deploy NVIDIA B200 or NVIDIA GH200 on Runcrate today?

NVIDIA B200 is on demand in 2 regions from $7.67/hr, billed by the second on most GPUs. NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest.

How much VRAM does each GPU have?

NVIDIA B200 has 180GB HBM3e. NVIDIA GH200 has 96GB HBM3.

B200 or GH200: which holds a 70B model in 8-bit weights?

It needs 70 GB of 8-bit weights. On the B200: 1 GPU, 110 GB spare. On the GH200: 1 GPU, 26 GB spare. Pick the B200: same GPU count, at least 2.0x the memory bandwidth.

GPU comparison

B200 vs GH200.

The B200 is ahead on memory (180 GB vs 96 GB) and speed (2.3x the FP16 throughput and at least 2.0x the memory bandwidth). The GH200 makes sense when your model fits in 96 GB and its hourly price is lower.

Launch B200 Reserve a cluster

Rent either on Runcrate

B200from $7.67/hr
$0.043 per GB/hr$0.34 per 100 TFLOPS/hr
Launch B200 B200 pricing
GH200Not in stock right now
Register interest GH200 pricing

Live hourly price, then per GB of memory and per 100 FP16 TFLOPS.

Which is better for

Which to pick for each job. Memory fit and throughput, from the datasheets.

JobB200GH200Pick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
B200: 1 GPU, 110 GB spare
GH200: 1 GPU, 26 GB spare
B200. Same GPU count, at least 2.0x the memory bandwidth.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
B200: 1 GPU, 164 GB spare
GH200: 1 GPU, 80 GB spare
B200. Same GPU count, at least 2.0x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
B200: 1 GPU, 134 GB spare
GH200: 1 GPU, 50 GB spare
B200. Same GPU count, 2.3x the FP16 throughput.
Train on a full node
FP16 tensor throughput per node
B200: 8x, 18,000 TFLOPS, NVLink
GH200: 1x, 990 TFLOPS
B200. 8 GPUs per node, not 1.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
B200: 225 TFLOPS per 100 W
GH200: Superchip power, not a GPU TDP
Either. Not comparable.

Compare another pair

Blackwell

NVIDIA B200

from$7.67/hr

Available now · 2 regions

Memory: 180 GB HBM3e
FP16 tensor: 2250 TFLOPS
Bandwidth: 8.0 TB/s

Launch B200 Reserve a cluster

Full pricing details

Hopper

NVIDIA GH200

Not in stock right now

Memory: 96 GB HBM3
FP16 tensor: 990 TFLOPS
Bandwidth: up to 4.0 TB/s

Full pricing details

At a glance

Spec by spec. The longer bar is the better number.

Memory

NVIDIA B200180 GB

NVIDIA GH20096 GB

Memory bandwidth

NVIDIA B2008.0 TB/s

NVIDIA GH200up to 4.0 TB/s

FP16 tensor

NVIDIA B2002250 TFLOPS

NVIDIA GH200990 TFLOPS

FP32

NVIDIA B20075 TFLOPS

NVIDIA GH20067 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA B200	NVIDIA GH200
Architecture	Blackwell	Hopper
Memory	180 GB	96 GB
Memory type	HBM3e	HBM3
Memory bandwidth	8.0 TB/s	up to 4.0 TB/s
FP32	75 TFLOPS	67 TFLOPS
FP16 tensor	2250 TFLOPS	990 TFLOPS
INT8	4500 TOPS	1979 TOPS
Power	1000 W per GPU (TDP)	up to 1000 W per superchip (CPU, GPU and memory)
Form factor	SXM	Superchip
NVLink	Yes	One GPU
GPUs per node	8	1
Hourly price (live)	from $7.67/hr	Not in stock right now

Questions

B200 or GH200. What buyers ask first.

Other comparisons