• 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

opethdisciple said:
Who's Greta?
Click to expand...

tenor.gif


A Goddess!
 
decto said:
Nvidia is capping RAM as GDDR6X is so expensive and while commodity memory prices in general may be falling, I'd take a bet there isn't a glut of supperfast GDDR6X anywhere to be found.
The Nvidia design forces 8/16, 10/20, these increments would cost $150-$200 to buy from Micron and once you start adding Nvidia margins that will add hundreds to the GPU price.
People would truly lose their minds if the 3080 20GB version cost $1200+, which I expect it would for little measurable performance gain.

I fully expect he 3090 24GB to be $2000 with a cheaper $1400 12GB available later. They want to differentiate the 3090 from the 3080TI which will come out with 11GB.
They have allegedly spent $150 on a cooler to stop the 3090 going nuclear.
I don't just don't see any likelyhood a 3090 with a fully functional die and 2.4x the super expensive memory compated to the 3080 will cost 'just' $600 more.

When GDDR6X gets less expensive, you may see higher memory cards released, or you may even see a 3080 with 20GB of plain old GDDR6.

Until Nvidia sorts out a wider memory bus or a more flexible memory architecture, they just can't make the margins they want even at the latest crazy prices if they add in all that memory.
They'll let the AIB's take the 'gouging' hit.

Can't wait for the announcement, I'll be bringing popcorn.
Click to expand...

Yep. And it's not just the cost of the memory. What Nvidia are doing which is most sensible for them is marketing a primary set of products aimed at what the average person needs in that price bracket, so someone after "high end" like me will drop the cash on a decent high end GPU and most of the people like me are not going to have vRAM problems at 10Gb. So that's a large segment of that performance bracket taken care of by 1 GPU layout/design and manufacturing process. Then there's the people who are basically exceptions, people who are demanding 16Gb or 20Gb for whatever reason, either they genuinely think they need it or will need it or whatever. The thing is it's not just added cost of the memory which is going to be expensive, it will be new board layouts, probably new cooling solution, new set of manufacturing processes etc, that's going to add a premium to those products on top of just the added cost of the memory. If you need to add a new line of manufacturing to pump out a new product every sale of that product covers some of the cost of that overhead and for niche products with very few buyers (the so called halo products) the prices rise exponentially because a tiny few people are paying for the overhead of new production.

And in this more elite space where people are interested in high end, the most significant complaint so far has been, there's no way I'm dropping £1400 on a GPU which honestly is fair enough, I've been buying GPUs since the Voodoo days, the demand for 20Gb variants will be very small and the price is going to be very high. Nvidia is a business, a publicly traded one with a responsibility to make a profit, they're not going to manufacture cards that have more vRAM than the average person needs at higher costs because they'll make less sales.
 
queamin said:
Yes I had a 320mb and it was gimped and soon got over shadowed and drivers took ages to get fixed 6 months orso.
Click to expand...
I'm not sure what potato gfx card I was using to play crysis on but upgrading to a crazy deal on a 8800gts 320mb was a revaluation! 8800gt 512 was good tho, 256 shortlived :(
 
FloatingPoint said:
Interesting all talk is about ram rather than bandwidth. I'm surprised no one wanted to talk about that.
Click to expand...

I will!

My take on that is basically the same as the take on vRAM size. You need enough to service the GPU in it's ability to process data and that's kinda it. You need enough to make sure you're not a bottleneck for the GPU and no more. The way I see this is if you look at the history of video cards and their internal components the struggle has always been the GPU, it's the thing doing the calculations. Amount of vRAM, bus width or total memory bandwidth, these are just things that you either have enough of to service the GPU so there's no bottleneck, or you don't and you bottleneck the GPU which is a waste.

I don't recall any generation where memory size or bandwidth was a limiting factor for overall speed of the video card, it has always been the GPU itself.
 
PrincessFrosty said:
I will!

My take on that is basically the same as the take on vRAM size. You need enough to service the GPU in it's ability to process data and that's kinda it. You need enough to make sure you're not a bottleneck for the GPU and no more. The way I see this is if you look at the history of video cards and their internal components the struggle has always been the GPU, it's the thing doing the calculations. Amount of vRAM, bus width or total memory bandwidth, these are just things that you either have enough of to service the GPU so there's no bottleneck, or you don't and you bottleneck the GPU which is a waste.

I don't recall any generation where memory size or bandwidth was a limiting factor for overall speed of the video card, it has always been the GPU itself.
Click to expand...

The 2080ti's ray tracing performance is bandwidth bottlenecked
 
Grim5 said:
The 2080ti's ray tracing performance is bandwidth bottlenecked
Click to expand...
The only limitation to RT performance on a 2080 Ti isn't just a bandwidth bottleneck. It also doesn't have so many RT cores which is also why Ampere will improve things so much (in addition to bandwidth) when it adds more improved RT cores.
 
Tommy_Here said:
Remember when everyone said the 1080 Ti was going to be a £850 GPU and a cut down titan because there was no competition then it launched at a cheaper price £719 and offered more performance than the Titan because Nvidia were on the fence with Vegas performance? Well yea here we are again big navi is round the corner and I'd be very surprised if Nvidia release the 3080 at £800
Click to expand...

People have short memories :p

Truth is most perceive Nvidia as ultra premium and know they like their profits.


opethdisciple said:
Who's Greta?
Click to expand...

You don’t watch the news much do you? :p

tenor.gif



CAT-THE-FIFTH said:
https://twitter.com/NVIDIAGeForce/status/1300207238586400768?ref_src=twsrc^tfw|twcamp^tweetembed|twterm^1300207238586400768|twgr^&ref_url=https://www.redditmedia.com/mediaembed/ijmn9s?responsive=trueis_nightmode=false


So $1999,then?? :p

Yes!
Click to expand...
Haha. Will be too funny if true. Some people are gonna be squirming if that ends up true :D
 
If the activity in the MM is anything to go by, this must be the least anticipated announcement since Tom Daly came out the closet he thought he was in.

Most people are holding.
 
You must log in or register to reply here.
Back
Top Bottom