llamaperf

NVIDIA CMP 170HX 40GB (unlocked) vs NVIDIA RTX Pro 5000 Blackwell 72GB for local LLMs

3 reports on the NVIDIA CMP 170HX 40GB (unlocked) and 0 on the NVIDIA RTX Pro 5000 Blackwell 72GB

Which is faster for local LLMs?

Nobody has reported the same model at the same quant on both cards yet, so there's no measured head-to-head. Memory bandwidth, which caps how fast a card can write, is 11% higher on the NVIDIA CMP 170HX 40GB (unlocked) (1,493 GB/s on the NVIDIA CMP 170HX 40GB (unlocked), 1,344 on the NVIDIA RTX Pro 5000 Blackwell 72GB). The estimates below compare them on the models people run most.

NVIDIA CMP 170HX 40GB (unlocked)
Memory
40GB
Memory bandwidth
1,493 GB/s
FP16 compute
not on record
Reports
3
NVIDIA RTX Pro 5000 Blackwell 72GB
Memory
72GB
Memory bandwidth
1,344 GB/s
FP16 compute
258 TFLOPS
Reports
0

Estimated by the calculator

The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape.

ModelNVIDIA CMP 170HX 40GB (unlocked)NVIDIA RTX Pro 5000 Blackwell 72GB
Qwen3.8 27B30.4 t/sQ8_013.7 t/sFP16
Qwen3.8 125B · 6B active16.8 t/sQ3_K_S, experts in RAM95.0 t/sQ3_K_M
DeepSeek V4 Flash 284B · 13B activedoesn't fit at these settings9.2 t/sQ2_K, experts in RAM
Qwen3.6 35B · 3B active120.3 t/sQ6_K98.6 t/sQ8_0
Qwen3.6 27B30.4 t/sQ8_013.7 t/sFP16
Gemma 4 26B · 4B active78.3 t/sQ8_032.7 t/sFP16

Change the context, RAM or model in the calculator: NVIDIA CMP 170HX 40GB (unlocked) · NVIDIA RTX Pro 5000 Blackwell 72GB

Every report on each card, with its source: NVIDIA CMP 170HX 40GB (unlocked) · NVIDIA RTX Pro 5000 Blackwell 72GB

More comparisons