NVIDIA GeForce RTX 3050 (6 GB) Specs Leak: Reduced Cores, Memory Capacity, Bandwidth, and TDP

[Tsing]

The specifications for NVIDIA's revision of the GeForce RTX 3050 have made their way online, revealing a new version of the Ampere GPU that features not only fewer cores, but also less memory and lower memory bandwidth. That said, the TDP is also lower—70 vs. 130 watts.

See full article...

 

[Zarathustra]

What would one even use this for, when many new titles in 2023 (and presumably 2024) are swapping to system RAM on 8GB GPU's at 1080p low settings?

Old games and esports?

Why would one buy a new GPU for that?

I mean, in 2013 I bought my Titan and it had 6GB back then. (granted I never got anywhere near using it all before taking it out of service, but...)

 

[Grimlakin]

Corporate workstations that need SOME 3d acceleration to be useable in modern OS's with modern Apps but not enough to game on.

 

[Zarathustra]

Corporate workstations that need SOME 3d acceleration to be useable in modern OS's with modern Apps but not enough to game on.

Isn't that what low end Quadro's are for? (Sorry "RTX x000", "RTX Ax000" and "RTX x000 Ada Generation"*)

(*Nvidias marketing team is failing big time. They had a recognizable workstation brand name, and they just went and ****ed it all up.)

Most of the explicit enterprise models are a little bit on the pricy side, but an A2000 would probably do the trick if those were your needs.

But businesses are used to paying more for professional versions because... reasons? (ECC? Better validation? Driver support for professional applications?)

I think the likes of Dell use (used?) the A2000's in their workstations, but don't quote me on that. I vaguely remember seeing it, but I am too lazy to google it right now.

Edit: I lied. I went to the Dell Precision Workstation page. It's a mix of A2000's and Radeon Pro W6300's.

Most Enterprise systems are just going to use IGP's. The ones that aren't are almost certainly going to be using Radeon Pro or Nvidias new confusingly named Quadro replacements.

Doing AI research or something like that? You probably have a VM in a rack somewhere remotely, and it likely has some sort of dedicated Tesla compute card in it.

 

[Grimlakin]

Doing AI research or something like that? You probably have a VM in a rack somewhere remotely, and it likely has some sort of dedicated Tesla compute card in it.

Man those cards are just SO **** expensive. It's cheaper for me to build a computer with a high end video card in it and 256 gig of ram and a bunch of compute to work on first and do my POC then move to a real distributed virtual computing system.

 

[Zarathustra]

Man those cards are just SO **** expensive. It's cheaper for me to build a computer with a high end video card in it and 256 gig of ram and a bunch of compute to work on first and do my POC then move to a real distributed virtual computing system.

I mean, I could do it for a lot less money, but I'm willing to do things Enterprise IT buyers aren't

I hardly paid full price for my server