picksbycard
October 7, 2026No. 158
Arc B580 ยท 12GB GDDR6

Can the Arc B580 run Gemma 3 1B IT?

Fits comfortably at Q4_K_M.

Gemma 3 1B IT is 1B parameters. At Q4_K_M the weights are about 0.6 GiB; with the 1.5 GiB runtime reserve that is 2.1 GiB against the card's 12GB, leaving about 9.9 GiB for context.

All three quants

12GB card
QuantWeightsWith reserveVerdictHeadroom
Q4_K_Mthe everyday quant0.6 GiB2.1 GiBFits comfortably9.9 GiB
Q8_0near-lossless1.0 GiB2.5 GiBFits comfortably9.5 GiB
FP16full weights1.9 GiB3.4 GiBFits comfortably8.6 GiB

Verdict rules: fits comfortably when weights plus reserve sit within 85% of VRAM; tight when they fit with little left for context; CPU offload when they exceed VRAM by up to 2x (runs, slowly, with layers in system RAM); no beyond that. The KV cache per token depends on the model's architecture and is not modelled here.

The numbers for this pairing

At Q4_K_M the weights take about 0.6 GiB; with the 1.5 GiB we set aside for the runtime that is 2.1 of the card's 12 GB, leaving about 9.9 GB for the context cache.

The highest-quality version that fits entirely in memory is FP16 (full quality).

The cheapest card on our list that runs it comfortably at Q4 is the Arc B570 at about $260, against about $330 for this card.

At about $330, this card costs roughly $27 per GB of VRAM.

Running it at its full 190 W board power for 4 hours a day would use about 23 kWh a month, about $4.20 at the US average residential rate of 18.4 cents per kWh (2026-05); real use sits below that ceiling.

Prices are the lowest current listings our tracker saw on Oct 7, 2026, since averages include premium variants; launch prices are marked where we have no listing.

Good match. Get the card.

The Arc B580 clears Gemma 3 1B IT at Q4_K_M and at Q8_0. Live listings below; this page was rebuilt 2026-10-07.

Amazon links may earn us a commission. Prices move daily; the links open live listings.

About the model

Gemma 3 1B IT by Google: 32,768-token native context, text, released 2025 under the Gemma Terms of Use. Model card →

The Silicon Note

One short read on the GPU market, most mornings. No spam, one-click out.