llamaperf

Qwen3.8 27B

on 2× NVIDIA RTX 5060 Ti 16GB · Strata

Tone: mixedNVIDIA hardware
Nov 6, 2026
Throughput
20.0 t/s gen
Quant
Q4_K_XL (GGUF)
System RAM
80 GB

Use cases

visionlong-context

Summary

User asks whether to upgrade to an RTX 5070 Ti 16GB or a 2x48GB DDR5 kit, running Qwen3.8 27B at Q4_K_XL with vision encoders on an RTX 5060 Ti 16GB plus RTX 4070 12GB with 80GB DDR5. Setup is Strata, with models split across both cards and system RAM; the user says the run fills all RAM and VRAM and reaches about 20 tok/s. The user calls the current setup fine for work and says tinkering is unlikely to raise the speed much; a 3090 is ruled out as used-only with no refunds.

Compare this run

Medians of plain runs: one device, one request, no speculative decoding, the whole model in the device's memory. The report above is one setup, so read it beside them and not as one of them.

Qwen3.8 27B at 4-bit by hardware

Bars are medians on one scale, 0 to 94.0 tokens/s. Each dot is one run and opens its report. Press ▶ to watch text arrive at a row's speed.

Models most run on the NVIDIA RTX 5060 Ti 16GB

Bars are medians on one scale, 0 to 95.2 tokens/s. Each dot is one run and opens its report. Press ▶ to watch text arrive at a row's speed.

Everything reported on the NVIDIA RTX 5060 Ti 16GB →Every Qwen3.8 report →Estimate Qwen3.8 on this card →