Nvidia's 3000 Series RTX GPU [3050, 3060, 3070, 3080, 3090 now with TIs]

@Frenetic Pony 8gb at what resolution? I don't feel like combing a 16 minute video to figure it out.

 

8GB+ VRAM usage, but that is not a good indicator of the game "requiring" over 8GB. I have seen that game using over 16GB of VRAm on 3090 on particular scenes but 3080 has no problem handling that scene with mere 10GB of VRAM.

 

Using +8GB...or allocating +8GB?

 

uh.. i just looked at RTX 3000 prices, yikes! ridiculously high! Heck, all GPU got price hike. My gtx 1660 super was 3.4mill IDR brand new, now brand new its 4.6mill IDR......

yep, using more than 8GB ram doesn't automatically means requiring over 8GB. Some games automatically hogs VRAM, maybe to improve streaming.

 

Thats certainly good to know, and more can only be better, all other things equal, however as others have noted, that doesn't necessarily mean the game needs that much vram. That's demonstrated here at about 3:25 with the 3070 matching the 6800's performance at 4K without RT. With RT it beats the 6800 with ease but RT is skewing the comparison. That said, looking at the DF video, the game is allocating over 10GB at 4K RT and yet the 8GB GPU is comfortably beating even the 16GB 6800XT in average and minimum framerates.

 

Reminds me a bit of the time when Windows Vista was released and people were complaining it would suck up all RAM in the system it could find. No more free RAM *doh*

 

Yup, unused ram is wasted RAM...we really need a segmentation of the VRAM like with System RAM (used, cached etc.)

 

No, it's not. Call of Duty games are well known to use VRAM to cache streaming assets, that hardly means that they need that much VRAM.
In Cold War there's a setting which allow you to directly control the percentage of VRAM which the game will allocate.
And from what benchmarks there is I see zero indication that it actually require more than 8GB of VRAM, even with RT.

Example one:


https://www.pcgameshardware.de/Call...-Cold-War-Benchmark-Release-Review-1362383/2/

Example two:


https://www.overclock3d.net/reviews...r_pc_performance_review_optimisation_guide/11

It does seem to require more than 6GB in 4K with RT Ultra. But considering that you get <60 fps in such mode even on a 2080Ti and even with DLSS that seems like a non-issue.

 

LOL. Even if memory was issue are we arguing 4k is valid option with those framerates when using ray tracing? 3070 is a 1440p card that can run limited 4k. Especially so if you look into future games 2 years from now requiring more juice.

 

Bring on the 5000 series. Then I can finally play last year's games properly at actual 4K. I bet they have more RAM too.

Maybe call them the Geforce RTFX 5000 so I can reminisce about that fine old series of the past too. Perhaps AMD can launch a new Radeon 9700 by then as well.

 

We will have next gen before we can buy current gen

 

Doubtful. Next gen is more than a year away.

 

I was not 100% serious (But their is no sign of solution for wafer shortages right now...)

 

Looking at the first example at least, the cards' 1st percentile of fps go down to 60% of avg. on the 8 GByte card and only to 85% on the 24 GByte card. Without memory consumung raytracing in your second example, 8 GByte cards seem to fare a bit better.

 

The 3000 series numbers still going up:


The difference between 3090 and 3060 Ti is strangely low...I suspect we need one or two more datapoints, but I guess the explanation could be that NVIDIA is selling every single chip they put to the market (due to the supply constraints).

 

Not really if you consider the difference in their launch dates.

 

Puget Systems Professional application reviews:

DaVinci Resolve Studio - NVIDIA RTX A6000 48GB Performance
Written on January 28, 2021 by Matt Bach

Adobe Premiere Pro - NVIDIA RTX A6000 48GB Performance
Written on January 27, 2021 by Matt Bach

Adobe After Effects - NVIDIA RTX A6000 48GB Performance
Written on January 26, 2021 by Matt Bach

Adobe Photoshop - NVIDIA RTX A6000 48GB Performance
Written on January 25, 2021 by Matt Bach

 

Looks neat

 

A100 vs V100 Deep Learning Benchmarks
January 28, 2021
A100 vs V100 Deep Learning Benchmarks | Lambda (lambdalabs.com)

NVIDIA RTX A6000 Deep Learning Benchmarks
January 4, 2021
RTX A6000 Deep Learning Benchmarks | Lambda (lambdalabs.com)

 

Best GPUs For Workstations: Viewport Performance of CATIA, SolidWorks, Siemens NX, Blender & More – Techgage
February 5, 2021