AMD Instinct MI250 128GB vs AMD Instinct MI210 PCIe

The Instinct MI250 delivers 2.0x the FP16 throughput of the Instinct MI210 (362 vs 181 TFLOPS dense). The Instinct MI250 also carries 64 GB more memory (128 GB vs 64 GB).

Instinct MI250: CDNA 2, 2021 Instinct MI210: CDNA 2, 2022
FP16 dense lead
2.0x
Instinct MI250: 362 vs 181 TFLOPS
Memory
128 vs 64 GB
64 GB more for the Instinct MI250
Bandwidth
3.3 TB/s vs 1.6 TB/s
Instinct MI250 moves data faster
TDP
500 W vs 300 W
Instinct MI210 draws 200 W less

Specifications

Instinct MI250
Instinct MI210
Architecture
CDNA 2CDNA 2
Launch Year
20212022
Form Factor
OAMPCIe
Memory
128 GB64 GB
Memory Bandwidth
3.3 TB/s1.6 TB/s
TDP
500 W300 W
Process Node
6nm6nm

Performance (TFLOPS)

Instinct MI250
Instinct MI210
FP64
45 TFLOPS 23 TFLOPS
FP32
45 TFLOPS 23 TFLOPS
TF32
No verified data available No verified data available
BF16
362 TFLOPS 181 TFLOPS
FP16
362 TFLOPS 181 TFLOPS
FP8
No verified data available No verified data available
FP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
362 TOPS 181 TOPS

FLOPS by Precision

What actually differs

The Instinct MI210 is the newer part: CDNA 2, launched in 2022, against the Instinct MI250's CDNA 2 from 2021. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the Instinct MI250 against the Instinct MI210: FP64 45 vs 23 TFLOPS, FP32 45 vs 23 TFLOPS, FP16 362 vs 181 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 128 GB against 64 GB, fed at 3.3 TB/s versus 1.6 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

Power budgets are 500 W for the Instinct MI250 and 300 W for the Instinct MI210. At FP32 that works out to 0.09 against 0.08 TFLOPS per watt.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.

Frequently asked questions

Is the Instinct MI250 faster than the Instinct MI210?

At FP16 precision the Instinct MI250 reaches 362 TFLOPS dense against 181 TFLOPS for the Instinct MI210. The performance table on this page lists every published precision for both GPUs.

Which has more memory, the Instinct MI250 or the Instinct MI210?

The Instinct MI250 carries 128 GB of memory versus 64 GB for the Instinct MI210. Memory bandwidth is 3.3 TB/s for the Instinct MI250 and 1.6 TB/s for the Instinct MI210.

How much power do the Instinct MI250 and the Instinct MI210 draw?

The Instinct MI250 is rated at 500 W TDP and the Instinct MI210 at 300 W. On FP32 throughput per watt, the Instinct MI250 is the more efficient part.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.