Announcement

Collapse
No announcement yet.

Device Memory TCP Nears The Finish Line For More Efficient Networked Accelerators

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Device Memory TCP Nears The Finish Line For More Efficient Networked Accelerators

    Phoronix: Device Memory TCP Nears The Finish Line For More Efficient Networked Accelerators

    A year ago Google engineers posted experimental Linux code for Device Memory TCP for more efficient transferring of data from GPUs/accelerators to network devices without having to go through a host CPU memory buffer. After going through many rounds of review, Device Memory TCP appears to be nearing the finish line...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    How is this diffetrent than RDMA?

    Comment


    • #3
      Originally posted by S.Pam View Post
      How is this diffetrent than RDMA?
      Rdma requires nics that support it and switches in DCB mode while this is basically any nic on any network

      Comment


      • #4
        so will this help all the programmers be put out of jobs when their responsibilities are taken over by hyper-efficient AI? i dont really see the benefit otherwise

        a big part of crypto is making hashing and signing really computationally slow in order to prevent misuse. AI must have something similar.

        Comment

        Working...
        X