Can I deploy NVIDIA GH200 or AMD MI300X on Runcrate today?

NVIDIA GH200 is not in stock on Runcrate right now; leave your email on its pricing page to register interest. AMD MI300X 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. AMD MI300X has 192GB HBM3.

GH200 or MI300X: 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 MI300X: 1 GPU, 122 GB spare. Pick the MI300X: same GPU count, at least 32% more memory bandwidth.

GPU comparison

GH200 vs MI300X.

The MI300X is ahead on memory (192 GB vs 96 GB) and speed (32% more FP16 throughput and at least 32% more 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
MI300XNot in stock right now
Reserve a cluster MI300X pricing

Which is better for

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

JobGH200MI300XPick

Hold a 70B model in 8-bit weights
70 GB of 8-bit weights
GH200: 1 GPU, 26 GB spare
MI300X: 1 GPU, 122 GB spare
MI300X. Same GPU count, at least 32% more memory bandwidth.
Serve an 8B model in 16-bit weights
16 GB of 16-bit weights
GH200: 1 GPU, 80 GB spare
MI300X: 1 GPU, 176 GB spare
MI300X. Same GPU count, at least 32% more memory bandwidth.
Fine-tune a 70B model with QLoRA
About 46 GB with 4-bit weights
GH200: 1 GPU, 50 GB spare
MI300X: 1 GPU, 146 GB spare
MI300X. Same GPU count, 32% more FP16 throughput.
Train on a full node
FP16 tensor throughput per node
GH200: 1x, 990 TFLOPS
MI300X: 8x, 10,459 TFLOPS
MI300X. 8 GPUs per node, not 1.
Throughput per watt
FP16 TFLOPS per 100 W of TDP
GH200: Superchip power, not a GPU TDP
MI300X: 174 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

CDNA 3

AMD MI300X

Not in stock right now

Memory: 192 GB HBM3
FP16 tensor: 1307.4 TFLOPS
Bandwidth: 5.3 TB/s

Reserve a cluster Deploy H200 instead

Full pricing details

At a glance

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

Memory

NVIDIA GH20096 GB

AMD MI300X192 GB

Memory bandwidth

NVIDIA GH200up to 4.0 TB/s

AMD MI300X5.3 TB/s

FP16 tensor

NVIDIA GH200990 TFLOPS

AMD MI300X1307.4 TFLOPS

FP32

NVIDIA GH20067 TFLOPS

AMD MI300X163 TFLOPS

Full specs

The complete spec sheet.

Spec	NVIDIA GH200	AMD MI300X
Architecture	Hopper	CDNA 3
Memory	96 GB	192 GB
Memory type	HBM3	HBM3
Memory bandwidth	up to 4.0 TB/s	5.3 TB/s
FP32	67 TFLOPS	163 TFLOPS
FP16 tensor	990 TFLOPS	1307.4 TFLOPS
INT8	1979 TOPS	2610 TOPS
Power	up to 1000 W per superchip (CPU, GPU and memory)	750 W per GPU (TDP)
Form factor	Superchip	OAM
NVLink	One GPU	No
GPUs per node	1	8
Hourly price (live)	Not in stock right now	Not in stock right now

Questions

GH200 or MI300X. What buyers ask first.

Other comparisons