AMD Radeon PRO W7800
The AMD Radeon PRO W7800 has 32 GB VRAM and 576 GB/s memory bandwidth. It can run 53 of our 94 tracked models natively in VRAM at 8k context.
With 32 GB GDDR6, the AMD Radeon PRO W7800 is a workstation-tier GPU that can run 53 models natively. It comfortably runs 7B–32B models at Q4; 70B-class models typically need CPU offload.
The AMD Radeon PRO W7800 is an RDNA 3 professional workstation GPU with 32GB ECC-capable GDDR6 on a 256-bit bus at 576 GB/s, backed by 64MB of Infinity Cache and 70 compute units. Its 32GB VRAM enables 34B models at Q4_K_M and 27B models at Q8_0 fully in-memory, on par with the NVIDIA RTX 5000 Ada on capacity. ROCm support applies on Linux; Windows users should use the Vulkan backend via llama.cpp.
AMD Radeon PRO W7800: 2023 RDNA 3 professional workstation GPU with 32GB ECC-capable GDDR6 on a 256-bit bus at 576 GB/s, 64MB Infinity Cache, 70 compute units.
34B models fit at Q4_K_M; 27B models fit at Q8_0 entirely in VRAM, capacity on par with the NVIDIA RTX 5000 Ada. ~18-25 t/s for 7B Q4.
ROCm on Linux for full acceleration; Windows users should use the Vulkan backend via llama.cpp. AMD's mid-tier answer to NVIDIA's Ada workstation lineup.
| Vendor | AMD |
| Architecture | RDNA 3 |
| VRAM | 32 GB |
| Memory type | GDDR6 |
| Memory bandwidth | 576 GB/s |
| Compute backend | ROCM |
| Tier | Workstation |
| Released | 2023 |
| Models (native) | 53 / 94 |
| Models (offload) | 9 / 94 |
Software: ROCm is Linux-only; on Windows use the Vulkan backend instead. Requires llama.cpp compiled with ROCm support.
Popular models for this GPU
Models this GPU runs natively in VRAM (53)
- Mixtral 8x7B Instruct v0.146.7B · MMLU-Pro 29.7Q3_K_M · ~17.2 t/s
- Command-R 35B35B · MMLU-Pro 33.0Q2_K · ~15.6 t/s
- Qwen 3.5 35B-A3B (MoE)35B · MMLU-Pro 85.3Q5_K_M · ~51.4 t/s
- Qwen 3.6 35B35B · MMLU-Pro 85.2Q5_K_M · ~13.8 t/s
- Ornith 1.5 35B-A3B (MoE)35B · MMLU-Pro N/AQ5_K_M · ~51.4 t/s
Show 48 more
- Yi 1.5 34B Chat34.4B · MMLU-Pro 37.0Q5_K_M · ~14.1 t/s
- Qwen3 32B32.8B · MMLU-Pro 65.5Q5_K_M · ~15.2 t/s
- Qwen 2.5 32B Instruct32.5B · MMLU-Pro 69.0Q5_K_M · ~14.8 t/s
- Qwen 2.5 Coder 32B Instruct32.5B · MMLU-Pro 62.3Q5_K_M · ~14.8 t/s
- DeepSeek R1 Distill Qwen 32B32.5B · MMLU-Pro 65.0Q5_K_M · ~14.8 t/s
- Nemotron 3 Nano 30B32B · MMLU-Pro 78.3Q6_K · ~43.3 t/s
- Gemma 4 31B30.7B · MMLU-Pro 85.2Q6_K · ~14 t/s
- Qwen3 30B-A3B (MoE)30B · MMLU-Pro 61.5Q6_K · ~41.5 t/s
- Nemotron 3.5 Lightning 30B-A3B30B · MMLU-Pro 81.6Q6_K · ~45.3 t/s
- Muse Glimmer 30B27.8B · MMLU-Pro N/AQ6_K · ~16.3 t/s
- Gemma 2 27B Instruct27.2B · MMLU-Pro 38.0Q6_K · ~14.7 t/s
- Gemma 3 27B Instruct27B · MMLU-Pro 67.5Q6_K · ~15.8 t/s
- Qwen 3.6 27B27B · MMLU-Pro 86.2Q6_K · ~16.5 t/s
- UI-Mate 27B27B · MMLU-Pro ~86.2Q6_K · ~16.5 t/s
- Bonsai 27B27B · MMLU-Pro 81.5Ternary (Q2_0) · ~42.5 t/s
- Qwen 3.8 27B27B · MMLU-Pro N/AQ6_K · ~16.5 t/s
- Gemma 4 26B (MoE)25.2B · MMLU-Pro 82.6Q6_K · ~34.7 t/s
- Mistral Small 3.1 24B Instruct24B · MMLU-Pro 66.8Q8_0 · ~13.9 t/s
- Mistral Small 22B22.2B · MMLU-Pro 49.2Q8_0 · ~14.7 t/s
- GPT-OSS 20B21B · MMLU-Pro 67.9Q8_0 · ~28.9 t/s
- Qwen3 14B14.8B · MMLU-Pro 61.0Q8_0 · ~21.9 t/s
- Qwen 2.5 14B Instruct14.7B · MMLU-Pro 63.7Q8_0 · ~21.7 t/s
- Phi-4 14B Instruct14B · MMLU-Pro 70.4Q8_0 · ~23.1 t/s
- Mistral Nemo 12B Instruct12.2B · MMLU-Pro 35.6BF16 · ~14.5 t/s
- Gemma 3 12B Instruct12.2B · MMLU-Pro 60.6BF16 · ~14.7 t/s
- Gemma 4 12B (Unified)12B · MMLU-Pro 77.2Q8_0 · ~23.4 t/s
- Gemma 2 9B Instruct9.2B · MMLU-Pro 32.0BF16 · ~17.6 t/s
- Qwen 3.5 9B9B · MMLU-Pro 82.5BF16 · ~20.5 t/s
- Ornith 1.5 9B9B · MMLU-Pro N/ABF16 · ~20.5 t/s
- Llama 3.1 8B Instruct8B · MMLU-Pro 48.3BF16 · ~21.9 t/s
- DeepSeek R1 Distill Llama 8B8B · MMLU-Pro 41.0BF16 · ~21.9 t/s
- Qwen3 8B8B · MMLU-Pro 56.7BF16 · ~21.8 t/s
- Qwen 2.5 7B Instruct7.6B · MMLU-Pro 56.3BF16 · ~23.9 t/s
- Mistral 7B Instruct v0.37.25B · MMLU-Pro 30.0BF16 · ~24 t/s
- Gemma 3 4B Instruct4B · MMLU-Pro 43.6FP32 · ~22.7 t/s
- Gemma 4 E4B4B · MMLU-Pro 69.4FP32 · ~22 t/s
- Phi-3.5 Mini Instruct3.8B · MMLU-Pro 47.4FP32 · ~20.3 t/s
- Phi-4-mini Instruct3.8B · MMLU-Pro 67.3FP32 · ~23 t/s
- Llama 3.2 3B Instruct3.2B · MMLU-Pro 24.0FP32 · ~27.2 t/s
- Qwen 2.5 3B Instruct3.1B · MMLU-Pro 32.4FP32 · ~29.5 t/s
- Gemma 2 2B Instruct2.6B · MMLU-Pro 17.8FP32 · ~33.2 t/s
- Gemma 4 E2B2B · MMLU-Pro 60.0FP32 · ~44.6 t/s
- SmolLM2 1.7B Instruct1.7B · MMLU-Pro 19.0FP32 · ~44.5 t/s
- Qwen 2.5 1.5B Instruct1.5B · MMLU-Pro 16.8FP32 · ~60 t/s
- Llama 3.2 1B Instruct1.24B · MMLU-Pro 12.5FP32 · ~71.6 t/s
- Gemma 3 1B Instruct1B · MMLU-Pro 14.7FP32 · ~86.5 t/s
- Qwen 2.5 0.5B Instruct0.5B · MMLU-Pro 10.0FP32 · ~178.2 t/s
- SmolLM2 360M Instruct0.36B · MMLU-Pro 8.0FP32 · ~210.9 t/s
Models that fit with CPU offload (9)
These use system RAM for layers that don't fit in VRAM, so expect much slower inference.
- Qwen 3.5 122B-A10B (MoE)122B · MMLU-Pro 86.7Q2_K · ~4 t/s
- Nemotron 3 Super 120B120B · MMLU-Pro 83.7Q2_K · ~4 t/s
- Llama 4 Scout 109B109B · MMLU-Pro 74.3Q2_K · ~2.9 t/s
- GLM-4.5 Air 106B106B · MMLU-Pro 81.4Q2_K · ~4.6 t/s
- GLM-4.6V 106B106B · MMLU-Pro 79.9Q2_K · ~4.6 t/s
- Qwen 2.5 72B Instruct72B · MMLU-Pro 71.1Q4_K_M · ~1.4 t/s
- Llama 3.3 70B Instruct70B · MMLU-Pro 68.9Q4_K_M · ~1.5 t/s
- DeepSeek R1 Distill Llama 70B70B · MMLU-Pro 70.0Q4_K_M · ~1.5 t/s
- Llama 3.1 70B Instruct70B · MMLU-Pro 66.4Q4_K_M · ~1.5 t/s
Too large for this GPU (32)
- Mixtral 8x22B Instruct v0.1
- Llama 3.1 405B Instruct
- DeepSeek V3 671B
- DeepSeek R1 671B
- Llama 4 Maverick 400B
- Qwen3 235B-A22B (MoE)
- MiniMax M1 456B
- GPT-OSS 120B
- GLM-4.5 355B
- GLM-4.6 355B
- GLM-4.7 358B
- MiniMax M2.5 229B
- GLM-5 744B
- MiniMax M2.7 229B
- Kimi K2.6
- GLM-5.1 754B
- DeepSeek V4 Pro 1.6T
- DeepSeek V4 Flash 284B
- Mistral Medium 3.5 128B
- GLM-5.2 753B
- Nemotron 3 Ultra 550B-A55B
- Step 3.5 Flash
- Step 3.7 Flash
- MiMo V2.5 Pro
- Kimi K2.5
- MiniMax M3
- Inkling
- Kimi K3
- DeepSeek V4 Flash 0731 284B
- Qwen3.8 2.4T-A95B
- DeepSeek V4 Pro 0813 1.6T
- Ornith 1.5 397B (MoE)
Frequently asked questions
- How much VRAM does the AMD Radeon PRO W7800 have?
- The AMD Radeon PRO W7800 has 32 GB of GDDR6 with 576 GB/s memory bandwidth.
- What is the AMD Radeon PRO W7800 best for?
- With 32 GB of VRAM, the AMD Radeon PRO W7800 is well-suited for running 7B–32B models at Q4 with room for context, making it a great all-rounder for local LLM inference.
- What LLMs can the AMD Radeon PRO W7800 run locally?
- The AMD Radeon PRO W7800 can run 53 of the 94 open-weight models tracked by CanItRun natively in VRAM at 8k context. Top options include: Qwen 3.8 27B at Q6_K, Muse Glimmer 30B at Q6_K, Ornith 1.5 9B at BF16.
- Can the AMD Radeon PRO W7800 run Gemma 4 31B?
- Yes. The AMD Radeon PRO W7800 runs Gemma 4 31B natively in VRAM at Q6_K quantization, achieving approximately 14 tokens per second.
- Can the AMD Radeon PRO W7800 run Qwen 3.6 27B?
- Yes. The AMD Radeon PRO W7800 runs Qwen 3.6 27B natively in VRAM at Q6_K quantization, achieving approximately 16.5 tokens per second.
- Can the AMD Radeon PRO W7800 run Qwen3 8B?
- Yes. The AMD Radeon PRO W7800 runs Qwen3 8B natively in VRAM at BF16 quantization, achieving approximately 21.8 tokens per second.