Near silent LLM Monster... NVIDIA, take notes

Credibility score: 74/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.

Claims analyzed

Framework Desktop has AMD Ryzen Max+ 395 chip with 128GB RAM — Solid (85/100)

👌✅

Framework might have best cooling; tested Asus Z13 and GMK Evo X2 — Opinion (50/100)

Best cooling? Bold claim for untested hype 🚩😏

Framework board whisper quiet like Mac Studio — Opinion (50/100)

Subjective noise comparison — fair game 👌

Framework announced desktop this year — Solid (85/100)

✅

Qwen 3 Coder 30B Q8 model ~30GB size — Solid (80/100)

👌

AMD chip has dynamic/fixed GPU RAM modes unlike Apple unified — Solid (90/100)

✅

Dynamic BIOS RAM setting faster than fixed for Qwen 3 Coder 30B Q8 on iGPU — Personal Story (70/100)

Dude scripted his own benchmark gauntlet 👌

Ryzen AI Max Plus 395 iGPU has 256 GB/s bandwidth — Solid (85/100)

👌

M4 Pro does 55 t/s on short prompt vs Framework's 47 t/s — Personal Story (70/100)

Personal benchmark — can't verify but plausible ⚖️

M4 Pro Mac Mini maxes at 64GB RAM; 128GB needs M4 Max — Solid (90/100)

✅

Framework AMD 395: Ars review 17k multi-core, DHH got 25k — Solid (80/100)

👌

Got 23k+ on Framework/Evo X2 with Fedora — Personal Story (70/100)

Plausible personal benchmark — OS tweaks matter ⚠️

LM Studio supports ROCm and Vulkan for LLMs — Solid (75/100)

👌

LM Studio auto-installs ROCm/Vulkan for AMD, uses Llama.cpp — Solid (85/100)

👌✅

Framework desktop with 128GB RAM + 1TB drive costs $3699 — Solid (75/100)

👌

Ollama doesn't use Vulkan/ROCm, falls back to slow CPU — Personal Story (65/100)

Personal setup fail — fair, Ollama's AMD support is spotty ⚠️

235B Qwen3 103GB model loads on AMD via Vulkan at 9 t/s — Personal Story (75/100)

Dude's demoing a 103GB beast on AMD live — wild if real 📹🔥

New 20B OpenAI OSS GPT model: 40GB, 63 t/s, crashes on long context — Personal Story (60/100)

OpenAI OSS GPT? Crashes on context bump — sketchy demo vibes 🤔🚩

MoE model uses less space and runs faster on inference — Solid (85/100)

✅👌

GMK Evo X2 beats Framework desktop in every LLM test — Personal Story (65/100)

Personal benchmarks — take with grain of salt ⚠️

Evo X2 beats Framework on 120B param model too — Personal Story (65/100)

More personal tests on new 120B model 👀

M4 Max CPU: 16 cores, 32 threads, near-silent even x4 in rack — Verified (90/100)

✅

M4 Max beats M3 Ultra; Apple uses optimized MLX models — Solid (80/100)

✅ MLX optimization caveat is spot on

Qwen 3 Coder: 65 t/s short, 52 t/s long prompts — OK (60/100)

Specific numbers + GitHub link = verifiable 👌

Framework desktop beats GMK Tech Evo X2 on long LLM prompts — OK (65/100)

Benchmark claim — needs the actual numbers to believe it fully ⚠️

Framework cooling whisper quiet like Mac Studio — Solid (75/100)

👌✅

Easiest and quietest LLM PC he's ever built — Personal Story (50/100)

Personal build experience — take his word as a reviewer 😏

Vulkan best for these chips right now — Solid (80/100)

👌

GPT-4o mini 20B: ROCm 566 PP/62 TG, Vulkan 575/70 & 1163/71 — Personal Story (70/100)

Dude's benching his own hardware — numbers from his screen 👌

Llama 3.3 70B runs at 4.94 t/s in LM Studio, 72/5 t/s Vulcan, 46/5 other — Solid (80/100)

👌 Live demo numbers, can't fake that

See the full analysis with sources and timestamps →