Can I deploy NVIDIA GH200 or NVIDIA A40 on Runcrate today?

NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest. NVIDIA A40 is on demand in 1 region from $2.42/hr, billed by the second on most GPUs.

How much VRAM does each GPU have?

NVIDIA GH200 has 96GB HBM3. NVIDIA A40 has 48GB GDDR6.

GH200 or A40: 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 A40: 2 GPUs (96 GB). Pick the GH200: fits on one GPU.

GPU comparison

GH200 vs A40.

The GH200 is ahead on memory (96 GB vs 48 GB) and speed (6.6x the FP16 throughput and up to 5.7x the memory bandwidth). The A40 makes sense when your model fits in 48 GB and its hourly price is lower.

Launch A40 Reserve a cluster

Rent either on Runcrate

GH200Not in stock right now
Register interest GH200 pricing
A40from $2.42/hr
$0.050 per GB/hr$1.62 per 100 TFLOPS/hr
Launch A40 A40 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.

JobGH200A40Pick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
A40: 2 GPUs (96 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
A40: 1 GPU, 32 GB spare
GH200. Same GPU count, up to 5.7x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
A40: 1 GPU, 2 GB spare
GH200. Same GPU count, 6.6x the FP16 throughput.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
A40: 4x, 599 TFLOPS
GH200. 6.6x the FP16 throughput per GPU.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
A40: 50 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

Ampere

NVIDIA A40

from$2.42/hr

Available now · 1 region

Memory: 48 GB GDDR6
FP16 tensor: 149.7 TFLOPS
Bandwidth: 696 GB/s

Launch A40 Reserve a cluster

Full pricing details

At a glance

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

Memory

NVIDIA GH20096 GB

NVIDIA A4048 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

NVIDIA A40696 GB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

NVIDIA A40149.7 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

NVIDIA A4037.4 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	NVIDIA A40
Architecture	Hopper	Ampere
Memory	96 GB	48 GB
Memory type	HBM3	GDDR6
Memory bandwidth	up to 4.0 TB/s	696 GB/s
FP32	67 TFLOPS	37.4 TFLOPS
FP16 tensor	990 TFLOPS	149.7 TFLOPS
INT8	1979 TOPS	299 TOPS
Power	up to 1000 W per superchip (CPU, GPU and memory)	300 W per GPU (TDP)
Form factor	Superchip	PCIe
NVLink	One GPU	No
GPUs per node	1	4
Hourly price (live)	Not in stock right now	from $2.42/hr

Questions

GH200 or A40. What buyers ask first.

Other comparisons