AMD Instinct MI50 32 GB
Live eBay prices, cheapest first. The best one right now is $633.68 (used).
Other sizes: 16 GB
Specs
- VRAM
- 32 GB HBM2
- Bandwidth
- 1,024 GB/s
- FP32
- 13.4 TFLOPS
- FP16
- 26.8 TFLOPS
- Power
- 300 W
- Connector
- PCIe 4.0
- Released
- 2018
- Llama 3.1 8B at Q4 (est.)
- 74–127 tokens/s
gfx906: dropped from current ROCm; runs llama.cpp (ROCm/Vulkan) and older ROCm builds.
Running it at home
- Cooling
- Passive: it has no fan. In a desktop it needs air pushed through it, front to back: a fan shroud on its end (3D-printed ones with a 40 mm blower or a 92 mm fan are sold for most of these cards) or a server chassis. Without one it overheats within minutes under load.
- Power
- Two 8-pin PCIe power inputs. Rated 300 W: allow that for each card on top of the rest of the PC.
- Size
- Full height, two slots.
- Motherboard
- Turn on "Above 4G decoding" in the BIOS: without it many boards won't start with a card with this much memory.
A server card: made for a rack server's airflow and power. With these sorted it works in a normal PC.
Price right now
Live on eBayFrom the live listings below; shipping included.
Listings
Live on eBay| Listing | Condition | Price | Checked |
|---|---|---|---|
| Used | $633.68or best offer−20%+$60.00 shipping | ||
| Used | $695.00−12%free shipping | ||
| Used | $698.25−11%free shipping | ||
| Used | $699.00−11%free shipping | ||
| Used | $699.99−11%free shipping | ||
| Used | $700.00−11%free shipping | ||
| Open box | $712.80or best offer−10%free shipping | ||
| Open box | $720.00or best offer−9%free shipping | ||
| Open box | $720.00−9%free shipping | ||
| Open box | $720.00or best offer−9%free shipping | ||
| Open box | $720.40−9%+$84.99 shipping | ||
| Open box | $720.40−9%+$84.99 shipping | ||
| Open box | $720.40−9%+$84.99 shipping | ||
| Open box | $728.20−8%+$10.00 shipping | ||
| Used | $729.00−8%free shipping | ||
| Open box | $744.00−6%free shipping | ||
| Used | $745.00−6%free shipping | ||
2PCS Amd Radeon Instinct Mi50 32GB GPU Graphics Card Server Workstation Ai DeepLot of 2, price each | Open box | $754.99or best offerfree shipping | |
2PCS Amd Radeon Instinct Mi50 32Gb Gpu Server Workstation Graphics Card Ai DeepLot of 2, price each | Used | $755.00+$60.00 shipping | |
| Used | $759.74or best offerfree shipping | ||
| Used | $779.63free shipping | ||
| Used | $780.00+$30.00 shipping | ||
| Used | $785.00free shipping | ||
| Used | $786.98+$60.00 shipping | ||
| Used | $789.98+$60.00 shipping | ||
| Used | $790.00free shipping | ||
| Used | $790.00free shipping | ||
| Used | $800.00free shipping | ||
| Used | $807.50or best offerfree shipping | ||
| Used | $835.00free shipping | ||
| Used | $841.94+$20.00 shipping | ||
| Used | $850.00free shipping | ||
| Used | $856.95+$7.00 shipping | ||
| Used | $870.98+$180.00 shipping | ||
| Open box | $887.35or best offerfree shipping | ||
10 PCs Radeon Instinct Mi50 Machine Learning HPC AI Accelerator 32GB HBM2GPULot of 10, price each | Open box | $890.00free shipping | |
| Used | $896.75free shipping | ||
| Used | $896.75free shipping | ||
| Used | $896.75free shipping | ||
| Used | $896.75free shipping | ||
| Used | $899.00free shipping | ||
| Open box | $909.98+$60.00 shipping | ||
| Open box | $929.00or best offerfree shipping | ||
| Open box | $929.00or best offerfree shipping | ||
| Open box | $929.00or best offerfree shipping | ||
| Open box | $935.00or best offerfree shipping | ||
| Open box | $1,045or best offer+$38.88 shipping | ||
| Used | $1,117or best offer+$38.88 shipping |
Auctions ending soon
No auctions ending soon right now.
Prices include shipping and were read from eBay at the time shown. They can change: check the listing before you buy.
Owner reviews
No owner reviews yet. Own one? Tell others how it runs.
From signed-in owners, checked before they're published. PerGB doesn't pay for or edit reviews.
How it compares
Live on eBay- #4 of 33 PCIe GPUs with 32 GB or more for the money: $986.10 per GeForce RTX 3090 of performance.
- Software: Not on ROCm's support list (older ROCm or Vulkan).
Faster, for the least
- Intel Arc A770 8 GB: $259.99, 103% of the GeForce RTX 3090, 8 GB
- NVIDIA Tesla V100 PCIe 12 GB: $261.00, 123% of the GeForce RTX 3090, 12 GB
- NVIDIA Tesla V100 PCIe 16 GB: $325.00, 123% of the GeForce RTX 3090, 16 GB
From the live listings of every GPU, shipping included; performance against the GeForce RTX 3090 (FP16 or FP32 compute and memory bandwidth). Updated with every build.
Local AI: what it runs
| Model (Q4, 8K context) | Instinct MI50 32 GB needed | Costs now | Speed (est.) |
|---|---|---|---|
| Llama 3.1 8B | 1 card | $633.68 | 74–127 tokens/s |
| Qwen3 14B | 1 card | $633.68 | 40–68 tokens/s |
| gpt-oss 20B | 1 card | $633.68 | 99–169 tokens/s |
| Gemma 3 27B | 1 card | $633.68 | 22–37 tokens/s |
| Qwen3 32B | 1 card | $633.68 | 18–31 tokens/s |
| Llama 3.3 70B | 2 cards | $1,329 | 8.4–14 tokens/s |
| gpt-oss 120B | 3 cards | $2,027 | 70–119 tokens/s |
| Qwen3 235B-A22B | 5 cards | $3,426 | 16–28 tokens/s |
| DeepSeek R1 / V3 671B | more than 8 | — | — |
Cost: that many of the cheapest live listings. Speed: an estimate from the memory bandwidth for one person chatting. Try other models, sizes and contexts.
Decision models
| Model (Q4, 8K context) | Instinct MI50 32 GB needed | Costs now | Per question (est.) |
|---|---|---|---|
| Julia-10.144B | 1 card | $633.68 | 16–34 ms |
| Laya0.421B | 1 card | $633.68 | 26–56 ms |
| Kev-4B4B | 1 card | $633.68 | 63–135 ms |
| lev4B | 1 card | $633.68 | 189–406 ms |
| OpenJev27B | 1 card | $633.68 | 226–485 ms |
| Clef27B | 1 card | $633.68 | 226–485 ms |
| Perplexity Decider v1.227B | 1 card | $633.68 | 226–485 ms |
| mammoth-127B | 1 card | $633.68 | 226–485 ms |
Decision models score the options you give them instead of writing text, so they're timed per question. The time is scaled from each model's reported speed on an RTX PRO 6000 by this card's compute and memory bandwidth, within about half either way. About the models.
Compare elsewhere
Search other shops for AMD Instinct MI50 32 GB. These open the shop's own search; we don't show their prices here.

































