DeskGPU
English

Can NVIDIA GeForce RTX 2080 Ti run Qwen3.5-9B?

This page was written and translated automatically by AI and checked by software against its sources, with no human editing. Specifications come from the sources listed; memory requirements and speeds are estimates calculated with published formulas. How we use AI

Fits By llama.cpp’s default load check, Qwen3.5-9B fits on NVIDIA GeForce RTX 2080 Ti at Q5_K_M with a context of 8,192 tokens. One card needs 7.7 GiB available for “Fits”.

At Q5_K_M, the longest context that fits is 12,032 tokens.

“Fits” means llama.cpp's default --fit load check would place all layers on the GPU: our estimate of llama.cpp's buffers, plus its error margin, plus llama.cpp's 1 GiB per-card reserve, is within the memory counted as available. It does not promise that peak memory while generating fits. How results are decided

Values updated Oct 19, 2026 · Calculator

Run profile: llama.cpp, CUDA back end (NVIDIA discrete GPU), all layers on the GPU, Flash Attention on, micro-batch (-ub) 512, F16 KV cache, context (-c) as shown; “Fits” is judged with llama-server’s default 4 parallel sequences (-np auto, unified KV cache) and “Doesn’t fit” with 1 sequence; idle GPU with no other programs using its memory.

Memory counted as available for “Fits” (C_low): 7.88 GiB per card = 11 GiB nominal − 1,054 MiB (the largest gap between nominal and reported total measured on consumer and workstation GPUs in this profile, used because no public log reports this GPU’s own total) − 2,112 MiB (the largest start-up and runtime use seen in this profile’s public logs), rounded down to a multiple of 64 MiB.

These results assume the card’s default configuration. Settings that reduce usable memory (for example ECC, where the card supports it) are not covered.

Memory used for “Doesn’t fit” (C_high): the full 11 GiB per card. Run profiles and their measurements

Memory needed by quantization and context

Available memory needed for “Fits” (X, GiB per card) and the result on NVIDIA GeForce RTX 2080 Ti.
Quantization4,0968,19232,768131,072
BF16 17.4 GiB Doesn't fit17.5 GiB Doesn't fit18.4 GiB Doesn't fit21.7 GiB Doesn't fit
Q8_0 10.1 GiB Tight10.3 GiB Tight11.1 GiB Tight14.4 GiB Doesn't fit
Q6_K 8.3 GiB Tight8.5 GiB Tight9.3 GiB Tight12.7 GiB Tight
Q5_K_M 7.6 GiB Fits7.7 GiB Fits8.6 GiB Tight11.9 GiB Tight
Q4_K_M 6.8 GiB Fits7.0 GiB Fits7.8 GiB Fits11.2 GiB Tight
Q4_0 6.5 GiB Fits6.7 GiB Fits7.5 GiB Fits10.9 GiB Tight
IQ4_XS 6.3 GiB Fits6.5 GiB Fits7.3 GiB Fits10.7 GiB Tight
Q3_K_M 6.0 GiB Fits6.1 GiB Fits7.0 GiB Fits10.3 GiB Tight

How fast can it generate?

At Q5_K_M, the theoretical ceiling is about 100 tokens per second (616 GB/s ÷ 6.14 GB read per token).

The speed ceiling is memory bandwidth divided by the bytes read per generated token. It is a theoretical upper bound; public llama.cpp benchmarks typically reach about half to four-fifths of it, and small models and mixture-of-experts models often less. Public benchmarks

Alternatives

Other 11 GB GPUs where Qwen3.5-9B fits

Sources