Skip to content
Run LLM Locally

Gemma 3 27B IT on the Arc B580

Gemma 3 27B IT at Q4_K_M needs about 18.4 GB, more than the 12 GB of the Arc B580: part of the model runs from system RAM, which slows generation. At Q3_K_M it does not fit in VRAM.

ShareFacebookXWhatsAppLinkedInEmail

Gemma 3 27B IT

Q4_K_M · 8K context

CPU/GPU offload

The model needs about 18.4 GB but only 11.6 GB of VRAM is usable: 6.8 GB must be loaded into system RAM. It runs, 63% on the GPU, but more slowly.

Your hardware
Arc B580 12 GB · 64 GB RAM
Estimated memory
~18.4 GB estimate
VRAM (usable memory)
11.6 GB
GPU only
No
CPU/GPU offload
Required (6.8 GB)
Generation speed
2.5–5 tok/s estimate
Largest context in fast memory
None
Confidence: low

Based on: memory bandwidth (456 GB/s) and model size · system RAM bandwidth (~51 GB/s)

Memory breakdown

~18.4 GB (17.6 GB–19.2 GB) needed · estimate

6.8 GB must be offloaded to system RAM

  • Model weights15.4 GB
  • KV cache1 GB
  • Runtime overhead1.1 GB
  • Safety margin0.9 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (SYCL) · flash attention on · weights = real file size · sliding-window layers keep 1,024 tokens

EstimateData updated 24 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card

What makes this estimate less reliable

  • no benchmark for this hardware and model yet
  • split CPU/GPU runs vary a lot with the CPU and PCIe link
  • system RAM speed unknown (dual-channel DDR4-3200 assumed)

Other quantizations on the Arc B580

Gemma 3 27B IT on Arc B580 + 64 GB RAM · 8K context
Gemma 3 27B IT Q8_08K~30.3 GB · 1–2 tok/s estimatePossible but slow
Gemma 3 27B IT Q6_K8K~23.9 GB · 1.5–3 tok/s estimatePossible but slow
Gemma 3 27B IT Q5_K_M8K~21.1 GB · 1.9–3.5 tok/s estimatePossible but slow
Gemma 3 27B IT Q4_K_M8K~18.4 GB · 2.5–5 tok/s estimateCPU/GPU offload
Gemma 3 27B IT IQ4_XS8K~16.7 GB · 3–6.5 tok/s estimateCPU/GPU offload
Gemma 3 27B IT Q3_K_M8K~15.4 GB · 4–8 tok/s estimateCPU/GPU offload

Alternatives

GPUs that run Gemma 3 27B IT Q4_K_M entirely in VRAM

Other models that fit the Arc B580

Hardware that runs Gemma 3 27B IT Q4_K_M entirely in memory

Prices and links follow the market you choose.

Radeon RX 7900 XT

AMD

CompatibleRuns Gemma 3 27B IT Q4_K_M entirely in memory

Price not available

Availability unknown

View on amazon.com

Affiliate link

AMD Ryzen AI Max+ 395

Unified memory PC

CompatibleRuns Gemma 3 27B IT Q4_K_M entirely in memory

Price not available

Availability unknown

View on amazon.com

Affiliate link

No current offer imported for this market: the link opens a marketplace search.

Frequently asked questions

Can the Arc B580 run Gemma 3 27B IT?
The model needs about 18.4 GB but only 11.6 GB of VRAM is usable: 6.8 GB must be loaded into system RAM. It runs, 63% on the GPU, but more slowly.
Which quantization should I use on the Arc B580?
None fits entirely in VRAM at 8K context; the smallest, Q3_K_M, still needs offload.
What is the largest context on the Arc B580?
At Q4_K_M, the model does not fit entirely in VRAM even at a short context.
How fast is Gemma 3 27B IT (Arc B580)?
We estimate 2.5–5 tok/s (estimate, confidence low). Based on: memory bandwidth (456 GB/s) and model size, system RAM bandwidth (~51 GB/s).