Can I deploy NVIDIA GH200 or NVIDIA A100 40GB 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 40GB is on demand in 3 regions from $1.99/hr, billed by the second on most GPUs.

How much VRAM does each GPU have?

NVIDIA GH200 has 96GB HBM3. NVIDIA A100 40GB has 40GB HBM2.

GH200 or A100 40GB: 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 40GB: 2 GPUs (80 GB). Pick the GH200: fits on one GPU.

GPU comparison

GH200 vs A100 40GB.

The GH200 is ahead on memory (96 GB vs 40 GB) and speed (3.2x the FP16 throughput and up to 2.6x the memory bandwidth). The A100 40GB makes sense when your model fits in 40 GB and its hourly price is lower.

Launch A100 40GB Reserve a cluster

Rent either on Runcrate

GH200Not in stock right now
Register interest GH200 pricing
A100 40GBfrom $1.99/hr
$0.050 per GB/hr$0.64 per 100 TFLOPS/hr
Launch A100 40GB A100 40GB 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 40GBPick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
A100 40GB: 2 GPUs (80 GB)
GH200. Fits on one GPU.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
GH200: 1 GPU, 80 GB spare
A100 40GB: 1 GPU, 24 GB spare
GH200. Same GPU count, up to 2.6x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
A100 40GB: 2 GPUs (80 GB)
GH200. Fits on one GPU.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
A100 40GB: 8x, 2,496 TFLOPS
A100 40GB. 8 GPUs per node, not 1.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
A100 40GB: 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 40GB

from$1.99/hr

Available now · 3 regions

Memory: 40 GB HBM2
FP16 tensor: 312 TFLOPS
Bandwidth: 1.6 TB/s

Launch A100 40GB 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 40GB40 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

NVIDIA A100 40GB1.6 TB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

NVIDIA A100 40GB312 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

NVIDIA A100 40GB19.5 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	NVIDIA A100 40GB
Architecture	Hopper	Ampere
Memory	96 GB	40 GB
Memory type	HBM3	HBM2
Memory bandwidth	up to 4.0 TB/s	1.6 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 $1.99/hr

Questions

GH200 or A100 40GB. What buyers ask first.

Other comparisons