Hey guys, if anyone has asus GX10 can you provide any comments on its weak points vs other nodes from other vendors like lenovo nvda, dell, because video reviews don't show any difference at all in performance between all of these. I'm aware of 2242 limitations, yes. Is there anything else anyone of you bumped into?
Asus GX10 (NV GB10) node feedback?
- Thread starter tinfoil3d
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I actually think that they are very good. The reason we have 3 of them is that they were less expensive at the time, faster to arrive, and worked well.
Another item to remember is that when running AI models, the actual workload usually looks like high GPU utilization, high memory capacity utilization, and like 25% CPU utilization. That matters because it is different than what you see when you stress the entire machine using a synthetic tool.
Another item to remember is that when running AI models, the actual workload usually looks like high GPU utilization, high memory capacity utilization, and like 25% CPU utilization. That matters because it is different than what you see when you stress the entire machine using a synthetic tool.
I just set one up in my network a little over a week ago. I don't have any other GB10 machines to test against, but I can't say I've noticed any show-stopping negatives so far, considering I moved from an old test box with an AMD mobile 6600M with 8GB VRAM. Limitations in software definitely exist and if you stick to Ollama to serve models you're likely going to be disappointed. I just moved to llama.cpp with CUDA 13 in Docker with tweaks to improve concurrency and it's cut a lot of my agents' tool calling time by a factor of 3 or 4. Because of this I was able to move from a Qwen 3.6 35B, A3B Q4 quant to a Q5 with vision mods. The tokens per second throughput is slightly lower, but it's more than made up for it with the efficiency in using tools!
Realistically, the community is still pulling more and more performance out of this thing almost weekly, despite its spec-sheet weaknesses. I'm considering picking up a Gigabyte as a second unit, just to keep playing in the sandbox.
Realistically, the community is still pulling more and more performance out of this thing almost weekly, despite its spec-sheet weaknesses. I'm considering picking up a Gigabyte as a second unit, just to keep playing in the sandbox.