Can I deploy NVIDIA GH200 or NVIDIA A100 80GB on Runcrate today?

NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest. NVIDIA A100 80GB is on demand in 6 regions from $1.73/hr, billed by the second on most GPUs.

How much VRAM does each GPU have?

NVIDIA GH200 has 96GB HBM3. NVIDIA A100 80GB has 80GB HBM2e.

GH200 or A100 80GB: 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 A100 80GB: 1 GPU, 10 GB spare. Pick the GH200: same GPU count, up to 96% more memory bandwidth.

GPU comparison

GH200 vs A100 80GB.

The GH200 is ahead on memory (96 GB vs 80 GB) and speed (3.2x the FP16 throughput and up to 96% more memory bandwidth). The A100 80GB makes sense when your model fits in 80 GB and its hourly price is lower.

Launch A100 80GB Reserve a cluster

Rent either on Runcrate

GH200Not in stock right now
Register interest GH200 pricing
A100 80GBfrom $1.73/hr
$0.022 per GB/hr$0.55 per 100 TFLOPS/hr
Launch A100 80GB A100 80GB 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.

JobGH200A100 80GBPick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
A100 80GB: 1 GPU, 10 GB spare
GH200. Same GPU count, up to 96% more memory bandwidth.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
GH200: 1 GPU, 80 GB spare
A100 80GB: 1 GPU, 64 GB spare
GH200. Same GPU count, up to 96% more memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
A100 80GB: 1 GPU, 34 GB spare
GH200. Same GPU count, 3.2x the FP16 throughput.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
A100 80GB: 8x, 2,496 TFLOPS
A100 80GB. 8 GPUs per node, not 1.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
A100 80GB: Varies by card
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

Ampere

NVIDIA A100 80GB

from$1.73/hr

Available now · 6 regions

Memory: 80 GB HBM2e
FP16 tensor: 312 TFLOPS
Bandwidth: up to 2.0 TB/s

Launch A100 80GB Reserve a cluster

Full pricing details

At a glance

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

Memory

NVIDIA GH20096 GB

NVIDIA A100 80GB80 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

NVIDIA A100 80GBup to 2.0 TB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

NVIDIA A100 80GB312 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

NVIDIA A100 80GB19.5 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	NVIDIA A100 80GB
Architecture	Hopper	Ampere
Memory	96 GB	80 GB
Memory type	HBM3	HBM2e
Memory bandwidth	up to 4.0 TB/s	up to 2.0 TB/s
FP32	67 TFLOPS	19.5 TFLOPS
FP16 tensor	990 TFLOPS	312 TFLOPS
INT8	1979 TOPS	624 TOPS
Power	up to 1000 W per superchip (CPU, GPU and memory)	up to 400 W per GPU (TDP)
GPUs per node	1	8
Hourly price (live)	Not in stock right now	from $1.73/hr

Questions

GH200 or A100 80GB. What buyers ask first.

Other comparisons