• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA ‘Ampere’ 8nm Graphics Cards

Quartz said:
Apart from the power connection, I don't think the PCB is going to be a wacky shape; rather I think it's the cooler that is unusual.
Click to expand...

Nope, the PCB is not rectangular and has a triangular cut-out at the end of it.

Alphacool have confirmed they're making blocks for this odd shaped card.
 
Yaayuh! said:
It’s really hard to see how much VRAM games require considering there’s a load of caching going on, usage does not equal need.
It’s real obvious when you do hit the limit though, performance drops through the floor.

Have we reached that point yet with 8-11GB at 3840 × 2160?
Click to expand...

XeNoNF50 said:
My 11gb on the 1080ti fills up full in a number of games. Needs to be 16gb standard. Don't know if there's a performance hit when it hits that point but there's no room to spare.
Click to expand...

Read Yaayuh's post again. Just because your 11 Gb fills up doesn't mean it hampers performance. Some games might just be using what's there.

Has any actual testing been done to determine what games are VRAM limited?

Surely as long as the transfer between system RAM and VRAM is fast enough, and between the SSD and RAM, then there would be no performance impact?
 
99% likely that all the AIB cards will be a traditional design. It'll just be the founders edition that are ludicrously expensive with the over engineered thermal solution, just so launch day reviews aren't full of "NVIDIA takes note from AMD, RTX3000 series HOT and LOUD!" :P
 
danlightbulb said:
Has any actual testing been done to determine what games are VRAM limited??
Click to expand...

IIRC, there isn't really a definitive measure of the VRAM usage. As mentioned before, some games might just fill and fill as needed, leaving previously used data in the VRAM until that space is needed for something else. A report on usage would maybe include that, but not be a fair representation of whether its VRAM limited.
 
I recently changed my case to a Lian-Li O11 Dynamic XL, I've got another 14cm length spare over my 1080 Ti. Do your worst 3090!
 
tyler_jrb said:
I'm still a little skeptical. The main ampere GA100 is on TSMC 7nm so be very suprising to see the gaming range on 8nm. Hopefully all this power usage talk is for the 8000+ Cudas :p. Time will indeed tell i guess :).
Click to expand...
Well going from 12NM to 7NM and using 50% more power doesn't sound right to me.

My guess is nvidia went with 8NM because it was cheap, they could increase their margins and they were planning a simple Turing like increase across the stack but their spy's over at AMD brought some worrying news and this has prompted them to push the power limits sky high as this is the only way they could boost performance to counter what they now see as a genuine threat.
 
Joxeon said:
Well going from 12NM to 7NM and using 50% more power doesn't sound right to me.

My guess is nvidia went with 8NM because it was cheap, they could increase their margins and they were planning a simple Turing like increase across the stack but their spy's over at AMD brought some worrying news and this has prompted them to push the power limits sky high as this is the only way they could boost performance to counter what they now see as a genuine threat.
Click to expand...
Ahh, so maybe the inside joke could be in the nomenclature?
 
Woodsta888 said:
I remember 7970Ghz edition and it was about £350 and I bought when they came down to £300. Cant remember the cost of the 7990 but I remember they sent me a higher Hz version than the one I'd paid for, by accident.
Click to expand...

7990’s were a grand when they first came out, they couldn’t sell them at the time because of the micro stutter issue AMD had with Crossfire. If I remember rightly they knocked them down to below half price only to fix the micro stutter a few months later.. I’ll go through my old order history to see what I paid for my ones
 
Last edited:
Joxeon said:
Well going from 12NM to 7NM and using 50% more power doesn't sound right to me.

My guess is nvidia went with 8NM because it was cheap, they could increase their margins and they were planning a simple Turing like increase across the stack but their spy's over at AMD brought some worrying news and this has prompted them to push the power limits sky high as this is the only way they could boost performance to counter what they now see as a genuine threat.
Click to expand...

I think that 7nm is volume constrained at TDCM and NVidia are renowned for being a pain to work with. TDCM probably told them to jump.
NVidia absolutely had to move to a smaller node as their RTX 20XX dies are huge, costing a lot to fab and have low yields. Need more RT and tensor cores to push their closed ecosystem onto everyone else...
 
You must log in or register to reply here.
Back
Top Bottom