AMD Instinct MI355X
Hacker ApprovedRanked 3 of 11 datacenter GPUs by recommenders
Strong memory and performance per dollar for large AI inference and training; critics say its software stack lags and it trails Nvidia's B300 on some benchmarks.
Strong memory and performance per dollar for large AI inference and training; critics say its software stack lags and it trails Nvidia's B300 on some benchmarks.
19 people gave an opinion
- recommend it
- 14
- criticize it
- 5
- approval share
- 74%
Recommended in 7 of 13 months
- November 2025: recommended
- December 2025: recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: recommended
- May 2026: recommended
- June 2026: not recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it carries the badge
- 14 people recommend itNeeds 10 or more, met.
- 2.8 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 7 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Capable hardware, competitive raw performance
“AMD compute cards are sold under the Insinct line and are vastly more powerfull than their GPUs.”
Comment from 8 Dec 2025 in “AMD GPU Debugger” “Our hardware is arguably better (spec for spec) apart from critical areas like memory bandwidth”
Comment from 15 Nov 2025 in “HipKittens: Fast and furious AMD kernels” “their Instinct parts are actually, at a hardware level, very very capable pieces of kit”
Comment from 15 Nov 2025 in “HipKittens: Fast and furious AMD kernels” - Comment from 14 Jul 2026 in “S&P downgrades Oracle to BBB – only one notch above junk level”
- Comment from 15 Nov 2025 in “HipKittens: Fast and furious AMD kernels”
- Comment from 22 Apr 2026 in “Our eighth generation TPUs: two chips for the agentic era”
Lots of VRAM for the price
“The metric where AMD usually comes out on top is perf/$. Or with their instinct cards VRAM/$”
Comment from 13 Apr 2026 in “Taking on CUDA with ROCm: 'One Step After Another'” “With 288GB per GPU the MI355X can fit the model within one 8 GPU node.”
Comment from 6 Aug 2026 in “Running Kimi K3 on MI355X at Better Performance per Dollar Than B300” “leads in VRAM and throughput but trails in accuracy (slightly)”
Comment from 28 Nov 2025 in “TPUs vs. GPUs and why Google is positioned to win AI race in the long term” - Comment from 27 Jul 2026 in “Kimi-K3 Technical Report [pdf]”
Better performance per dollar and watt
“The operating cost is so much lower it's paid for itself plus an additional $40k in the first year.”
Comment from 2 Sep 2026 in “My local model setup on an M4 Pro Mac Mini” “The metric where AMD usually comes out on top is perf/$. Or with their instinct cards VRAM/$”
Comment from 13 Apr 2026 in “Taking on CUDA with ROCm: 'One Step After Another'” “they are better per dollar, better per watt, and out competing on raw performance”
Comment from 14 Jul 2026 in “S&P downgrades Oracle to BBB – only one notch above junk level” Good for inference and training
“AMD has been doing inference chips for many years and are the leader for HPC.”
Comment from 22 Apr 2026 in “Our eighth generation TPUs: two chips for the agentic era” “running our training workloads on the beefy Mi355x is awesome”
Comment from 23 Aug 2026 in “Where are people finding GPU capacity?” “MI355 are quite good for inference.”
Comment from 7 May 2026 in “Google to sell TPU chips to select customers”
What people warn about
Each mark is one comment
Trails B300 on benchmarks
“leads in VRAM and throughput but trails in accuracy (slightly)”
Comment from 28 Nov 2025 in “TPUs vs. GPUs and why Google is positioned to win AI race in the long term” “On every row the B300 beat the MI355X”
Comment from 2 Aug 2026 in “Running Kimi K3 on MI355X at Better Performance per Dollar Than B300” Weak developer tools and ROCm software
“Benchmarks are unreproducible, power costs are missing, ROCm was patched”
Comment from 2 Aug 2026 in “Running Kimi K3 on MI355X at Better Performance per Dollar Than B300” “Developer ux on amd is terrible.”
Comment from 14 Jul 2026 in “S&P downgrades Oracle to BBB – only one notch above junk level” Power costs and draw unclear
“Benchmarks are unreproducible, power costs are missing, ROCm was patched”
Comment from 2 Aug 2026 in “Running Kimi K3 on MI355X at Better Performance per Dollar Than B300” “A single MI355 is ~230W, just idle.”
Comment from 25 Dec 2025 in “JEDEC developing reduced pin count HBM4 standard to enable higher capacity”
Other datacenter GPUs people recommend
The four most recommended other datacenter GPUs, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| NVIDIA H100 | 20 | 16 | 56% | 11 | — | — |
| NVIDIA Tesla V100 | 16 | 4 | 80% | 8 | — | Approved |
| AMD Instinct MI355XThis page | 14 | 5 | 74% | 7 | — | Approved |
| Nvidia A100 | 10 | 4 | 71% | 8 | — | Approved |
| Intel D270 GPU AI Accelerator | 10 | 10 | 50% | 1 | — | — |
What people said
23 of the 24 comments that mention it, newest first. Each links to the original comment.
“Wow impressive optimizations on the AMD GPUs”
“this is amazing speedup on AMD”
“The operating cost is so much lower it's paid for itself plus an additional $40k in the first year.”
“you are literally burning money by not having one in your building”
“running our training workloads on the beefy Mi355x is awesome”
“With 288GB per GPU the MI355X can fit the model within one 8 GPU node.”
“We own & operate our AMD GPUs US data centers & serving layer.”
“Benchmarks are unreproducible, power costs are missing, ROCm was patched”
“On every row the B300 beat the MI355X”
“Those cards have a lot of memory for their price”
“Developer ux on amd is terrible.”
“they are better per dollar, better per watt, and out competing on raw performance”
“being extremely successful selling their AI chips to much bigger customers”
“MI355 are quite good for inference.”
“AMD has been doing inference chips for many years and are the leader for HPC.”
“The metric where AMD usually comes out on top is perf/$. Or with their instinct cards VRAM/$”
“AMD Instinct is quite different from what regular folks can easily put in their workstation”
“A single MI355 is ~230W, just idle.”
“AMD compute cards are sold under the Insinct line and are vastly more powerfull than their GPUs.”
“leads in VRAM and throughput but trails in accuracy (slightly)”
“I was already aware the hardware was good from my own experiences”
“Our hardware is arguably better (spec for spec) apart from critical areas like memory bandwidth”
“their Instinct parts are actually, at a hardware level, very very capable pieces of kit”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.