NVIDIA CMP HX90 mining hashrate
We got our hands on NVIDIA CMP HX90 mining graphic cards. In this article, we are going to benchmark it on the most popular algorithms and we will compare stock versus overclocked settings, you might be surprised.
NVIDIA CMP HX90 is a headless mining GPU. CMP stands for Cryptocurrency Mining Processor. HX90 is very similar to RTX 3080 performance wise, it has the same graphics processor, same bus width, and same amount of VRAM. But it mainly differs from its gaming cousin by its core count, TMUS, and some other specifications not essential for mining.
The current price for the HX90 is around $2000 which makes it quite expensive for a headless GPU. Especially now, when mining software is starting to crack the Hash Rate Limiter on the RTX 3000 series cards.
Below are NVIDIA CMP HX90 specifications compared to RTX 3080
MODEL | GPU PROCESSOR | CORES | MEM. SIZE | MEM. TYPE | BUS WIDTH | BANDWIDTH | BASE CLOCK | TDP |
---|---|---|---|---|---|---|---|---|
RTX 3080 | GA102 | 8704 | 10 GB | GDDR6X | 320 bit | 760.3 GB/s | 1440 MHz | 320 W |
CMP HX90 | GA102 | 6400 | 10 GB | GDDR6X | 320 bit | 760.3 GB/s | 1500 MHz | 200 W |
These are the stock hashrate results of benchmarking HX90 in NiceHash Miner, we have used stock settings for this test. It seems that the GPU is already (at least partly) optimized at 90 MH/s for the DaggerHashimoto algorithm. Power draw was around 250W.NVIDIA CMP HX90 mining hashrate - stock
HX90 stock performance on Autolykos (Ergo) algorithm is 207 MH/s.
NVIDIA CMP HX90 mining hashrate - optimized
After initial tests with NiceHash Miner, we have fired up QuickMiner and OCTune which proves to be the best overclocking utility. We have simply found the highest stable memory clock at +900 MHz. Then we have run the autotune, which found the best core clock limit at 1740 Mhz (absolute value) for performance and 1380 MHz (absolute value) for the best efficiency within 10 minutes of tunning. Interestingly, we were not able to lower the power draw by much. Only down to 220W.
Here are the results after the optimization process:
0 Comments:
Post a Comment