Can I deploy NVIDIA GH200 or NVIDIA V100 on Runcrate today?

NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest. NVIDIA V100 is on demand in 2 regions from $0.36/hr, billed by the second on most GPUs.

How much VRAM does each GPU have?

NVIDIA GH200 has 96GB HBM3. NVIDIA V100 has 32GB HBM2.

GH200 or V100: 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 V100: 3 GPUs (96 GB). Pick the GH200: fits on one GPU.

GPU comparison

GH200 vs V100.

The GH200 is ahead on memory (96 GB vs 32 GB) and speed (7.9x the FP16 throughput and up to 4.4x the memory bandwidth). The V100 makes sense when your model fits in 32 GB and its hourly price is lower.

Launch V100 Reserve a cluster

Rent either on Runcrate

GH200Not in stock right now
Register interest GH200 pricing
V100from $0.36/hr
$0.011 per GB/hr$0.29 per 100 TFLOPS/hr
Launch V100 V100 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.

JobGH200V100Pick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
V100: 3 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
V100: 1 GPU, 16 GB spare
GH200. Same GPU count, up to 4.4x the memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
V100: 2 GPUs (64 GB)
GH200. Fits on one GPU.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
V100: 8x, 1,000 TFLOPS, NVLink
Either. Close per node.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
V100: 42 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

Volta

NVIDIA V100

from$0.36/hr

Available now · 2 regions

Memory: 32 GB HBM2
FP16 tensor: 125 TFLOPS
Bandwidth: 900 GB/s

Launch V100 Reserve a cluster

Full pricing details

At a glance

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

Memory

NVIDIA GH20096 GB

NVIDIA V10032 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

NVIDIA V100900 GB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

NVIDIA V100125 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

NVIDIA V10015.7 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	NVIDIA V100
Architecture	Hopper	Volta
Memory	96 GB	32 GB
Memory type	HBM3	HBM2
Memory bandwidth	up to 4.0 TB/s	900 GB/s
FP32	67 TFLOPS	15.7 TFLOPS
FP16 tensor	990 TFLOPS	125 TFLOPS
INT8	1979 TOPS	125 TOPS
Power	up to 1000 W per superchip (CPU, GPU and memory)	300 W per GPU (TDP)
Form factor	Superchip	PCIe
NVLink	One GPU	Yes
GPUs per node	1	8
Hourly price (live)	Not in stock right now	from $0.36/hr

Questions

GH200 or V100. What buyers ask first.

Other comparisons