NVIDIA Watch posted an update
Nvidia, SK Telecom and South Korean chipmaker Rebellions plan to develop a server system combining Nvidia GPUs with Rebellions’ Rebel R200 NPUs in the same rack, Korea JoongAng Daily reports.
Why it mattersThe proposed system would use Nvidia’s NVLink Fusion and MGX rack architecture. SK Telecom plans to test it at its AI data centre before exploring commercial uses, including offering computing capacity to government and business customers through an API. The useful detail is that this is a planned test, not a deployed service or a proven performance gain. Still, it is a concrete attempt to make different kinds of AI chips work together in shared infrastructure, rather than asking customers to pick one camp and hope for the best.
Discuss: Should AI data-centre operators favour mixed GPU-and-NPU systems, or stick with a single chip platform until the benefits are demonstrated?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.