AMD Instinct MI250
The MI250 delivers dual-GCD CDNA 2 performance with 128GB HBM2e for demanding workloads. Optimized for high-performance computing and large-scale AI training.
VRAM
128 GB
Memory
HBM2e
Bandwidth
3,277 GB/s
FP16
362 TF
TDP
500 W
Memory and compute
Architecture and power
Notes
Dual-GCD design for HPC and AI training workloads.
Models that may fit on AMD Instinct MI250
Estimates based on INT8 quantization. Actual fit depends on framework and batch size.
More GPUs by AMD
The manufacturer's other cards in the catalog, largest memory first, with bandwidth and power where published.
Frequently asked questions about AMD Instinct MI250
Answered from the entry's own figures: memory, which models fit, bandwidth, power and the card's class.
AMD Instinct MI250 has 128 GB of HBM2e at 3,277 GB/s. At INT8 a model's weights take about one byte per parameter, so the card holds roughly a 102B-parameter model with room for context; larger models need several cards or a lower precision.
From the catalog's estimates at INT8, Gemma 4 31B IT, Llama Nemotron Embed VL 1B V2, GLM Z1 9B 0414 and NVIDIA Nemotron Nano 12B V2 VL BF16 fit on a single AMD Instinct MI250 with their default context. The list on this page shows the estimated memory each takes; the capacity planner sizes any model against this card for your own context length, batch size and request rate.
3,277 GB/s. Bandwidth bounds how fast a model generates tokens, because every token reads the whole set of weights and the KV cache from memory; at this rate a 102B model at INT8 could be read on the order of 32 times a second on one card, before batching and framework efficiency.
AMD Instinct MI250 has a TDP of 500 W, with a maximum of 500 W, in the OAM form factor. Budget the TDP per card plus host overhead when sizing a server's power and cooling.
AMD Instinct MI250 is a datacenter accelerator, meant for servers and multi-GPU serving, built on the CDNA 2 architecture and released in 2021. It supports a multi-GPU interconnect at 800 GB/s, so a model larger than one card can be split across several.
Specifications from the manufacturer's published material; a figure that is not published is shown as such, and a derived one is explained in the notes.
Added Jan 27, 2026 · Last updated Apr 30, 2026
Explore more
Companion tools that draw on the same catalog and routing engine.
Put intelligence in the middle.
One gateway in front of every model, with your policies applied and every decision on record. Start on the free plan with 5,000 requests a month, no card required.