Can four Ascend GX10s run large open-source AI at a usable speed?
A four-machine Ascend GX10 or DGX Spark setup has reportedly run GLM5.2 at about 400 to 500 for and about 15 for output, using a 128k context. That is not fast, but it may be usable for personal work or small experiments, especially with .
The main question is whether buying local hardware now makes sense before a stronger arrives later this year or next year. The setup may draw around 1000 watts, but the appeal is being ready to run large models locally instead of depending fully on cloud inference.
Key points
- Four Ascend GX10 or DGX Spark machines were discussed as a local setup for GLM5.2.
- was reported at about 400 to 500 .
- Output speed was about 15 at a 128k context.
- may make large models more practical on this hardware.
- Local hardware can reduce cloud only if usage is high enough to justify the machine and electricity costs.