Can I deploy NVIDIA GH200 or NVIDIA GB200 on Runcrate today?

NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest. NVIDIA GB200 is not in stock on Runcrate right now; reserve a cluster, and once we've verified your company and your need we take your exact spec to GPU owners that fit.

How much VRAM does each GPU have?

NVIDIA GH200 has 96GB HBM3. NVIDIA GB200 has 186GB HBM3e.

GH200 or GB200: which holds a 70B model in 8-bit weights?

It needs 70 GB of 8-bit weights. On the GH200: 1 GPU, 26 GB spare. On the GB200: 1 GPU, 116 GB spare. Pick the GB200: same GPU count, at least 2.0x the memory bandwidth.

GPU comparison

GH200 vs GB200.

The GB200 is ahead on memory (186 GB vs 96 GB) and speed (2.5x 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.

Rent either on Runcrate

GH200Not in stock right now
Register interest GH200 pricing
GB200Not in stock right now
Reserve a cluster GB200 pricing

Which is better for

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

JobGH200GB200Pick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
GB200: 1 GPU, 116 GB spare
GB200. Same GPU count, at least 2.0x the memory bandwidth.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
GH200: 1 GPU, 80 GB spare
GB200: 1 GPU, 170 GB spare
GB200. Same GPU count, at least 2.0x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
GB200: 1 GPU, 140 GB spare
GB200. Same GPU count, 2.5x the FP16 throughput.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
GB200: 8x, 20,000 TFLOPS, NVLink
GB200. 8 GPUs per node, not 1.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
GB200: 250 TFLOPS per 100 W
Either. Not comparable.

Compare another pair

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

Blackwell

NVIDIA GB200

Not in stock right now

Memory: 186 GB HBM3e
FP16 tensor: 2500 TFLOPS
Bandwidth: 8.0 TB/s

Reserve a cluster Deploy B200 instead

Full pricing details

At a glance

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

Memory

NVIDIA GH20096 GB

NVIDIA GB200186 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

NVIDIA GB2008.0 TB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

NVIDIA GB2002500 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

NVIDIA GB20090 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	NVIDIA GB200
Architecture	Hopper	Blackwell
Memory	96 GB	186 GB
Memory type	HBM3	HBM3e
Memory bandwidth	up to 4.0 TB/s	8.0 TB/s
FP32	67 TFLOPS	90 TFLOPS
FP16 tensor	990 TFLOPS	2500 TFLOPS
INT8	1979 TOPS	3600 TOPS
Power	up to 1000 W per superchip (CPU, GPU and memory)	1000 W per GPU (TDP)
Form factor	Superchip	Superchip
NVLink	One GPU	Yes
GPUs per node	1	8
Hourly price (live)	Not in stock right now	Not in stock right now

Questions

GH200 or GB200. What buyers ask first.

Other comparisons