Thread

M5 Ultra Mac Studio Review

269 points · 262 comments · piotrgrabowski

simonw · 1790001122 · focus · HN ↗
The numbers I was most interested in are tucked away in a chart towards the bottom - the speed comparison of the Mac Studios v.s. a RTX 5090:
```
  Qwen3.8 27B tokens&#x2F;sec generation speed

  Prompt size    8K    64K   128K   256K
  RTX 5090 PC    59    51    44     n&#x2F;a
  M5 Ultra       48    39    32     24
  M3 Ultra       31    23.5  20     15
```
A whole bunch more comparison numbers in this section: <a href="https://www.macstories.net/stories/m5-ultra-mac-studio-review-the-dream-mac-for-local-ai-agents/#mx-pc" rel="nofollow">https://www.macstories.net/stories/m5-ultra-mac-studio-revie...
1. karmakaze · 1790014798 · focus · HN ↗
  I really appreciate seeing these dense model numbers. For a large unified memory system though I expect that MoE numbers are what people are more interested in.
  These numbers could and should get much better. As an example I can run Qwen3.8-27B-MXFP4 (W4A8) on 2x AMD R9700 that gets 260+ tokens/sec to start and slows down to ~110 tokens/sec over 128k context and can do the max 256k. These are for batch size 1 and throughput goes higher with batching. This is due to speculative decoding, efficient all-reduce inter-gpu compression, and custom GEMM kernels for the specific hardware. Note each R9700 only has 644 GB/s memory bandwidth.

Unofficial Hacker News client; not affiliated with Y Combinator.