TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Bolt Graphics Zeus a New GPU Architecture with Up to 2.25TB of Memory and 800GbE

27 pointsby Teeverabout 2 months ago

4 comments

Jasper_about 2 months ago
Every year they drop a &quot;new GPU!&quot; announcement with absolutely no details. All their numbers are &quot;in simulation only&quot;.<p>Last year it was their &quot;Thunder&quot; architecture being the world&#x27;s fastest GPU, now it&#x27;s Zeus. Neither one actually exists. <a href="https:&#x2F;&#x2F;bolt.graphics&#x2F;bolt-graphics-unveils-thunder-the-worlds-fastest-graphics-processor&#x2F;" rel="nofollow">https:&#x2F;&#x2F;bolt.graphics&#x2F;bolt-graphics-unveils-thunder-the-worl...</a><p>All their blog posts and marketing material are just generic hype about the concept of raytracing. I don&#x27;t think these guys have an actual product.
评论 #43524503 未加载
semessierabout 2 months ago
I don&#x27;t get why more than 2 years after the ChatGPT release moment there is not a plethora of high-mem matrix-matrix matrix-vector hardware for the high-end available. Both high bandwidth and commodity DRAM. Both operations are very well understood for dense and sparse cases. There were FPGAs and ASICs early on but nothing really caught on relative to the GPU behemoth which is tons of silicon on the die that is not needed for base matmult. Hence, it&#x27;s unbelievable how Nvidia continues to charging for memory to be one of the highest value companies in the world.
评论 #43517479 未加载
评论 #43518161 未加载
gymbeauxabout 2 months ago
Certainly neat, especially the SFP port and Ethernet port next to the display ports.<p>- How much will it cost?<p>- How compatible will it be with existing AI assets (eg models, code for training and inference)?<p>- Will there have to be a translation layer between RISC and CISC (eg my CPU)? What’s the performance penalty?<p>- Will I actually be able to get one, or is this for “enterprise” customers only, who must buy a minimum of 100 at a time?
jl6about 2 months ago
Hard to know who this is aimed at until we see the price. I guess they are going for the “slow memory, but lots of it” market that is less sensitive to how fast their very large LLMs run as long as they run at all. Hobbyists will rejoice if they can afford it, but is there a commercial use case that can tolerate low tokens per second?
评论 #43525463 未加载
评论 #43517381 未加载
评论 #43517299 未加载
评论 #43517370 未加载