3 ms·
This doesn't seem right at all, given that DeepSeek reported massive performance increase because the Huawei team helped them port their LLM to HW infra. I was
by rjzzleep 1y ago
This doesn't seem right at all, given that DeepSeek reported massive performance increase because the Huawei team helped them port their LLM to HW infra. I was willing to put it in the "I don't know, maybe, let's see" category, but that comment specifically makes it read like a propaganda piece.
- maxglute 1y agoAscend 910A was Q1 2022. Qwen 2.5 / Deepseek v3 was Q1 2025. Implied timeline seems to be it took ~2.5 years developement to figure out how to use Ascend somewhat competitively and HW may rival Nvidia with proper support. Where proper support in authors opinion is building a team and treating team well. My guess is Huawei can keep burning talent at problem. The Huawei+Deepseek > Nvidia for inference claims is based on HW CloudMatrix 384 supernode using all of the optical interconnects and power to allegedly out perform Nvidia cluster. Bottleneck of Nvida cluster is switching. Bottleneck of PRC cluster is chips on old node size (slower or more power hungry). CM384 workaround is 5x more chips 910c chips and 4x more power consumption and connecting cluster with full optical to compete vs GB200 on pure cluster performance. Hard to say what actual economics of that solution is.