Blackwell · Nvidia · Apple · Meta · AMD · Facebook · The Register
Meta hasn't shown how large a cluster its chips can support yet, the slides simply state "multi-thousand accelerator scaling"
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
◌ Single Source
While the MTIA 400 is aimed primarily at LLM training, Meta is already working on an inference-optimized version of the chip.
Key facts
- Meta’s latest accelerator is fed by eight 36 GB HBM3e stacks that deliver 288 GB of memory and about 9.2 TB/s of bandwidth
- But the parts don’t hold up nearly as well when compared to Nvidia and AMD’s latest chips, the MTIA 400 is between 3x and 3.3x slower than Rubin and the Instinct MI455X, respectively
- The MTIA 400 features a pair of the reporter/O chiplets that provide 1.2 TB/s of chip-to-chip bandwidth over RDMA
- Meanwhile, the MTIA 500, which is also slated for 2027 release (probably in H2 if they had to guess), will increase memory bandwidth by another 50 percent, likely using 4 additional HBM4 stacks, and double the number of compute chiplets
Summary
Faster than Blackwell, but still no replacement for AMD or Nvidia. yet. Security Security vets rally around $4 paper password books for sale in Australia. Security AliExpress accused of fingerprinting shoppers with silent audio trick that also muted a dev's headphones. Ai and ML Slop factory bans Russians for using slop factory to create slop. If you’re going to build a custom AI accelerator, inference is a good place to start.