7 ms·
Cool tour. I haven't kept up with their developments; what kind of workloads have they been pushing for? Since they don't seem to have any specialized accelera
by girfan 3mo ago
Cool tour.
I haven't kept up with their developments; what kind of workloads have they been pushing for? Since they don't seem to have any specialized accelerators in the Compute Sled, I am assuming they are not targeting AI workloads for now?
- connicpu 3mo agoHard to compete there when most of the margin is going to NVIDIA, pretty much impossible if you don't have the scale of Dell/Supermicro/etc
- adastra22 3mo agoThe Oxide approach, I think, would be to pick an open interconnect standard and work with one or more NVIDIA competitors to make hardware for it. E.g. AMD and a few of the specialized AI hardware companies. Essentially provide an open TPU standard.
- girfan 3mo agoInteresting. What would the incentives be for NVIDIA, for example, to opt into this? They are building their own racks now, like the NVL72 etc.
- adastra22 3mo agoNone, they won’t. But it would out Oxide into the strong negotiating position, and not at the mercy of NVIDIA. NVIDIA has one primary weakness here: GPUs are NOT optimal hardware for training or inference. No competitor has the market reach to challenge them though. An open standard supported by high growth, breakout startup would change that though.
- ironhaven 3mo agoThey already picked out a interconnect standard it is called 100 Gigabit Ethernet. They can interoperate with multiple different vendors using the sfp28 ports on the switch. I bet they are even working on new programmable hardware for a next gen 200 Gigabit Ethernet fabric with oxide controlling the entire stack from hardware to software