RTX 5070 Ti vs RTX 4080 SUPER
RTX 5070 Ti · 16 GB VRAM and RTX 4080 SUPER · 16 GB VRAM side by side for local AI.
Facts from both hardware pages
| Specification | RTX 5070 Ti | RTX 4080 SUPER |
|---|---|---|
| Memory | 16 GB VRAMDedicated · system RAM does not add · source | 16 GB VRAMDedicated · system RAM does not add · source |
| Memory bandwidth | 896 GB/sManufacturer spec · not model speed · source | 736 GB/sManufacturer spec · not model speed · source |
| Text · run well | 12536 run slowly · within 16 GB | 12536 run slowly · within 16 GB |
| Vision · run well | 125within 16 GB | 125within 16 GB |
| Image · within memory | 30within 16 GB · documented paths and minimums | 30within 16 GB · documented paths and minimums |
| Video · within memory | 27within 16 GB · documented paths and minimums | 27within 16 GB · documented paths and minimums |
| Audio · within memory | 116within 16 GB · documented paths and minimums | 116within 16 GB · documented paths and minimums |
| Embeddings · within memory | 89within 16 GB · 8-bit estimate | 89within 16 GB · 8-bit estimate |
| Published speed runs | 99 on this machine | 32 on this machine · 1 naming only the chip |
| Launch price (MSRP) | USD 749US · dated 2025-02-20 · not a tracked offer · source | USD 999US · dated 2024-01-31 · not a tracked offer · source |
Counts use the same memory estimate as each hardware page: text and vision run well within the fast memory (tight fits counted apart); image, video and audio use documented paths and publisher minimums. Formulas
Top text models each runs well
The first 5 rows of each hardware page's Text list, in its order: LiveBench score, then reviewed models, then Hub downloads. Memory is the planning estimate on that machine.
| Rank | RTX 5070 Ti | RTX 4080 SUPER |
|---|---|---|
| #1 | Qwen3 14BAlibaba · ~14 GB | Qwen3 14BAlibaba · ~14 GB |
| #2 | Qwen3 8BAlibaba · ~9 GB | Qwen3 8BAlibaba · ~9 GB |
| #3 | Qwen3.5 9BAlibaba · ~9 GB | Qwen3.5 9BAlibaba · ~9 GB |
| #4 | Llama 3.1 8B InstructMeta · ~8 GB | Llama 3.1 8B InstructMeta · ~8 GB |
| #5 | Qwen3 1.7BAlibaba · ~4 GB | Qwen3 1.7BAlibaba · ~4 GB |
Measured speed
Published output tokens per second, as measured. A speed is compared only when the same checkpoint, quantization, runtime and lengths ran on both machines themselves; same-chip references are not compared.
| Same test | RTX 5070 Ti | RTX 4080 SUPER |
|---|---|---|
| gpt-oss-20b MXFP4 · llama.cpp CUDA | 189.45 ± 0.09Not stated by source · llama.cpp gpt-oss guide · maintainer result | 186.51 ± 0.33Not stated by source · llama.cpp gpt-oss guide · maintainer result |
| Llama 2 7B Q4_0 · CUDA | 182.43 ± 0.09Not stated by source · llama.cpp CUDA scoreboard · community result | 147.48 ± 1.41Not stated by source · llama.cpp CUDA scoreboard · community result |
All published measurements · RTX 5070 Ti
| gpt-oss 20B Q4_K_M | 236.89 ± 2.18 | Q4_K_M | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
|---|---|---|---|---|
| gpt-oss 20B Q8_0 | 223.82 ± 2.47 | Q8_0 | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
| gpt-oss-20b MXFP4 | 189.45 ± 0.09 | MXFP4 (native gpt-oss weights) | llama.cpp CUDA | llama.cpp gpt-oss guide · maintainer result |
| Llama 2 7B Q4_0 | 182.43 ± 0.09 | Q4_0 | CUDA | llama.cpp CUDA scoreboard · community result |
| Qwen3.5 9B Q4_K_M | 126.03 ± 0.80 | Q4_K_M | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
| Gemma 4 12B Q4_K_M | 89.15 ± 0.31 | Q4_K_M | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
| Qwen3.5 9B Q8_0 | 86.39 ± 0.24 | Q8_0 | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
| Gemma 4 12B Q8_0 | 58.41 ± 0.18 | Q8_0 | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
| Mistral Small 3.2 24B Q4_K_M | 52.85 ± 0.07 | Q4_K_M | llama.cpp CUDA | Data: Local LLM Advisor (localllmadvisor.com) · CC BY 4.0 |
All published measurements · RTX 4080 SUPER
| gpt-oss-20b MXFP4 | 186.51 ± 0.33 | MXFP4 (native gpt-oss weights) | llama.cpp CUDA | llama.cpp gpt-oss guide · maintainer result |
|---|---|---|---|---|
| Llama 2 7B Q4_0 | 150.20 ± 0.26 | Q4_0 | Vulkan | llama.cpp Vulkan scoreboard · community result |
| Llama 2 7B Q4_0 | 147.48 ± 1.41 | Q4_0 | CUDA | llama.cpp CUDA scoreboard · community result |
Verdict
Both have 16 GB VRAM. By the memory estimate, 125 text models run well on each. RTX 5070 Ti lists 896 GB/s of memory bandwidth, 1.2× the 736 GB/s of RTX 4080 SUPER (a manufacturer spec, not a measured speed). In the same published test (gpt-oss-20b MXFP4, llama.cpp CUDA), RTX 5070 Ti measured 189.5 and RTX 4080 SUPER 186.5 output tokens per second. Launch prices on file: RTX 5070 Ti USD 749 (2025-02-20) and RTX 4080 SUPER USD 999 (2024-01-31). The same number of text models run well on both; RTX 5070 Ti measured faster in the shared test.
Limits and dates
Model counts are memory estimates, not speed measurements or guaranteed fits; verify runtime and checkpoint support. Bandwidth is the manufacturer's figure. Prices are launch or list prices with their dates, not current offers.
Page updated · Model catalog checked · LiveBench 2026-06-25