Tracked topic
…I also wanted to dig into the new “Virgo” scale out, datacenter-scale Ethernet fabric that Google has created for linking together racks of machinery, including but not limited to TPU pods…
…By our math, it looks like a coherent cluster domain for the Maia 100 was 576 nodes with a total of 2,304 compute engines – not too shabby for a glorified Ethernet…
…Each server node has 400 GB/sec of Ethernet bandwidth (eight ports at 400 Gb/sec), and the Rebel100 cards are hooked to each other and to the host CPUs with a…
…Earlier this month, I mused about how Broadcom might be the only sizable counterbalance to the hegemony of Nvidia right now , considering its dominant position in Ethernet switch ASICs and its fast…
…As far as we know, this Ethernet spine not use the Spectrum-6 ASICs with co-packaged optics, but it could us optics in the spine and copper in the chip-to…
…The Helios rack has UALink tunneling over Ethernet and the latency of the GPU fabric is a lot higher – and in part because the copper cables in the Helios racks are a…
…includes optical DSPs, active electrical cable DSPs, PCI-Express retimers, datacenter interconnect optical modules, and Ethernet switching ASICs. Conversely, Marvell is going to be using EDA simulation software on the AWS cloud…
…and likely won't, because I think the world is moving towards NVM-Express or Ethernet.” And, he added, it wanted to focus on where customers are using its platform. “A lot…
…This I/O may link out over DPUs to Ethernet switches, with both running the ESUN memory coherency protocol that Meta Platforms is pushing. (Or rather, pulling.) Those PCI-Express 6.0…
…connected via a low-latency, near-time interconnect , like RDMA over Converged Internet (ROCE), Ultra Ethernet, and NVQLink, among others. The last are partner scale-out systems, either in the cloud or…