72
Nvidia Blackwell RTX 5000 GPUs may debut earlier than expected
(www.techspot.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 01 Jan 2024
72 points (92.9% liked)
Games
15822 readers
952 users here now
Video game news oriented community. No NanoUFO is not a bot :)
Posts.
- News oriented content (general reviews, previews or retrospectives allowed).
- Broad discussion posts (preferably not only about a specific game).
- No humor/memes etc..
- No affiliate links
- No advertising.
- No clickbait, editorialized, sensational titles. State the game in question in the title. No all caps.
- No self promotion.
- No duplicate posts, newer post will be deleted unless there is more discussion in one of the posts.
- No politics.
Comments.
- No personal attacks.
- Obey instance rules.
- No low effort comments(one or two words, emoji etc..)
- Please use spoiler tags for spoilers.
My goal is just to have a community where people can go and see what new game news is out for the day and comment on it.
Other communities:
founded 1 year ago
MODERATORS
I dont understand the VRAM cuts.... The RAM fabs have been cutting production because of low prices
I would love more VRAM so that i can have a GPU that can do a bit of gaming and dabble in some AI stuff. 100% agree i'd pay for more VRAM instead of horsepower
More memory means you can do real work with it, and enterprise AI training is a money printer that they'd be scavenging the shit out of with cards that are closer substitutes.
Honestly, the gap between the server parallel compute cards and the home video cards isn't that large. 24GB on video cards, 80GB for a compute card.
That's not even two binary orders of magnitude. That's a narrow window to try to make their money from. Plus, some tasks can be subdivided and run on multiple GPUs, and they can't segment up the market for those.
Like, in general, my bet is that when for most things that fit the above requirements of fitting in that window and having a task that can't be subdivided, there's probably enough room for algorithmic improvements to get two binary orders of magnitude of reduction in memory requirements.