Skip to content

GPU comparison

MI300X vs MI210.

The MI300X is ahead on memory (192 GB vs 64 GB) and speed (7.2x the FP16 throughput and 3.2x the memory bandwidth). The MI210 makes sense when your model fits in 64 GB and its hourly price is lower.

Rent either on Runcrate

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.

  • Hold a 70B model in 8-bit weights

    70 GB of 8-bit weights

    MI300X: 1 GPU, 122 GB spare

    MI210: 2 GPUs (128 GB)

    MI300X. Fits on one GPU.

  • Serve an 8B model in 16-bit weights

    16 GB of 16-bit weights

    MI300X: 1 GPU, 176 GB spare

    MI210: 1 GPU, 48 GB spare

    MI300X. Same GPU count, 3.2x the memory bandwidth.

  • Fine-tune a 70B model with QLoRA

    About 46 GB with 4-bit weights

    MI300X: 1 GPU, 146 GB spare

    MI210: 1 GPU, 18 GB spare

    MI300X. Same GPU count, 7.2x the FP16 throughput.

  • Train on a full node

    FP16 tensor throughput per node

    MI300X: 8x, 10,459 TFLOPS

    MI210: 4x, 724 TFLOPS

    MI300X. 8 GPUs per node, not 4.

  • Throughput per watt

    FP16 TFLOPS per 100 W of TDP

    MI300X: 174 TFLOPS per 100 W

    MI210: 60 TFLOPS per 100 W

    MI300X. 2.9x the FP16 per watt.

Compare another pair

vs

CDNA 3

AMD MI300X

from$4.85/hr per GPU

on an 8-GPU machine

Available now · 1 region
Memory
192 GB HBM3
FP16 tensor
1307.4 TFLOPS
Bandwidth
5.3 TB/s
Full pricing details

CDNA 2

AMD MI210

Not in stock right now
Memory
64 GB HBM2e
FP16 tensor
181 TFLOPS
Bandwidth
1.6 TB/s
Full pricing details

At a glance

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

Memory

AMD MI300X192 GB
AMD MI21064 GB

Memory bandwidth

AMD MI300X5.3 TB/s
AMD MI2101.6 TB/s

FP16 tensor

AMD MI300X1307.4 TFLOPS
AMD MI210181 TFLOPS

FP32

AMD MI300X163 TFLOPS
AMD MI21023.9 TFLOPS

Full specs

The complete spec sheet.

SpecAMD MI300XAMD MI210
ArchitectureCDNA 3CDNA 2
Memory192 GB64 GB
Memory typeHBM3HBM2e
Memory bandwidth5.3 TB/s1.6 TB/s
FP32163 TFLOPS23.9 TFLOPS
FP16 tensor1307.4 TFLOPS181 TFLOPS
INT82610 TOPS192 TOPS
TDP750 W300 W
Form factorOAMPCIe
NVLinkNoNo
GPUs per node84
Hourly price (live)from $4.85/hr per GPU on an 8-GPU machineNot in stock right now

Questions

MI300X or MI210. What buyers ask first.

Other comparisons

More pairs with the MI300X or the MI210.