Dell Pro Max GB10 vs. AMD Ryzen AI Max+ Framework Desktop For Llama.cpp, OpenCL & Vulkan Compute

Written by Michael Larabel in Computers on 7 January 2026 at 11:42 AM EST. Page 2 of 5. 90 Comments.

With Llama.cpp the ROCm back-end was used with the AMD Ryzen AI Max+ Framework Desktop and the Dell Pro Max with GB10 using the NVIDIA CUDA back-end.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Text Generation 128. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 1024. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

The Dell Pro Max GB10 performance shined over AMD's Strix Halo with Llama 3.1. Across all prompt processing levels tested, the Dell Pro Max GB10 using the CUDA back-end was significantly faster than the top-end Strix Halo APU with ROCm.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

Not only was the Strix Halo slower but its GPU was consuming more power than the GB10 Blackwell GPU based on the exposed sensors.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

On a total system power consumption basis measured at the AC wall level using the WattsUp Pro, the Dell Pro Max with GB10 ended up consuming 178 Watts on average compared to the Framework Desktop around 130 Watts.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

Even still, the Dell Pro Max with GB10's performance allowed it to boast a superior performance-per-Watt to the Framework Desktop with Strix Halo.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

The strong performance of the NVIDIA GB10 allowed it to also have a slightly better value too even with Dell's $4500 list price currently for this device.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Text Generation 128. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 512. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

👁 Llama.cpp benchmark with settings of Backend: ROCm Or CUDA, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 2048. Dell Pro Max GB10 was the fastest.

Across other models, the Dell Pro Max with GB10 continued offering superior Llama.cpp performance over the AMD Ryzen AI Max+ 395. The performance-per-Watt and performance-per-dollar remained in favor of the NVIDIA GB10 system for this Llama.cpp showdown.

90 Comments - Next Page

Page: 1 2 3 4 5 Next Page

URL: https://www.phoronix.com/review/dell-pro-max-gb10-llama-cpp/2

⇱ Dell Pro Max GB10 vs. AMD Ryzen AI Max+ Framework Desktop For Llama.cpp, OpenCL & Vulkan Compute - Phoronix

Dell Pro Max GB10 vs. AMD Ryzen AI Max+ Framework Desktop For Llama.cpp, OpenCL & Vulkan Compute