Qwen on Intel Battlemage

Can I run Qwen2.5-Coder 32B on an Arc B580?

Not at Q4_K_M — it needs 21.5 GB against 10.5 GB available. You would need 3 of these cards.

Won't fit

21.5 GB of 10.5 GB · 205%
022 GB
Weights 18.6 GB
KV cache 2.0 GB
Runtime overhead 0.8 GB
Over the limit 11.0 GB

Short by 11.0 GB. You can run it with 26 of 64 layers on the Arc B580 and the rest in system RAM, at roughly 2.54 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation2.54tok/s
Prompt processing295tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Qwen2.5-Coder 32B on a Arc B580

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 61.0 GB 63.9 GB Won't fit — 0.59
INT8 / W8A8 8.50 32.1 GB 34.9 GB Won't fit — 1.23
Q8_0 (GGUF) 8.50 32.1 GB 34.9 GB Won't fit — 1.23
FP8 (E4M3) 8.00 30.5 GB 33.3 GB Won't fit — 1.31
Q6_K 6.56 25.0 GB 27.9 GB Won't fit — 1.67
Q5_K_M 5.67 21.6 GB 24.5 GB Won't fit — 2.03
Q5_K_S 5.52 21.1 GB 23.9 GB Won't fit — 2.13
Q4_K_M 4.85 18.6 GB 21.5 GB Won't fit — 2.54
AWQ 4-bit 4.25 18.3 GB 21.2 GB Won't fit — 2.57
GPTQ 4-bit 4.25 18.3 GB 21.2 GB Won't fit — 2.57
MXFP4 4.25 18.3 GB 21.2 GB Won't fit — 2.57
Q4_K_S 4.58 17.6 GB 20.5 GB Won't fit — 2.73
Q4_0 4.55 17.5 GB 20.4 GB Won't fit — 2.74
IQ4_XS 4.25 16.4 GB 19.3 GB Won't fit — 3.04
Q3_K_M 3.91 15.2 GB 18.0 GB Won't fit — 3.51
IQ3_M 3.70 14.4 GB 17.3 GB Won't fit — 3.77
IQ3_XXS 3.06 12.1 GB 15.0 GB Won't fit — 5.33
Q2_K 2.63 10.6 GB 13.4 GB Won't fit — 7.33
IQ2_XXS 2.06 8.5 GB 11.3 GB Won't fit 5K 14.5
IQ1_M 1.75 7.4 GB 10.2 GB Fits, but tight 9K 27.4

Also worth checking

Qwen2.5-Coder 32B on every GPU
Full hardware tableEverything that runs on a Arc B580
Full model tableWhy the KV cache is the part that bites
Guide