This wasn't even a fanboy vs thread. He was just asking a general question on 7nm availability. Every GPU thread you go in you post endless AMD fanboy drivel contantly, it's becoming so spam like!
-
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.
Will Nvidia be selling 7nm graphic cards in 2019?
- Thread starter don531
- Start date
This wasn't even a fanboy vs thread. He was just asking a general question on 7nm availability. Every GPU thread you go in you post endless AMD fanboy drivel contantly, it's becoming so spam like!
Intel with its 10nm process tried to skip to second gen 7nm like properties for certain areas of the process without having EUV to help with the goal of getting a 3 year advantage over the competition at current rate looks like they will be lucky just to retain parity.
TSMC by all reports is on a roll with 7nm with nVidia in development for a range of products the only reason we won't see 7nm GeForce cards sooner rather than later is if they hold them back which for financial reasons they might - 7nm is very expensive.
With upfront costs for 7nm estimated between 6 and 10x 28nm for GPU like products I suspect how quickly we see 7nm GPUs from nVidia will depend a lot on the competition especially as 12FF is relatively speaking cheap for them now.
Not just spam like - he endlessly perpetuates anything anti-nVidia even when shown official sources that it is wrong.
Last edited:
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,303
- Location
- Planet Earth
GF has 7NM GPUs out this year for AMD on the commercial side(around 300~350MM2 IIRC),and TSMC is already sampling Epyc for introduction next year. I also think some of the costs bandied around can't be as high people think,otherwise AMD wouldn't be able to ramp to 7NM already,so its not hard to believe than Nvidia won't have a 7NM range out even by the end of next year at the latest(they have more money after all). Nvidia is listing as one of the companies using TSMC 7NM IIRC,which has been in volume production for nearly 6 months:
https://www.anandtech.com/show/12677/tsmc-kicks-off-volume-production-of-7nm-chips
So if we say a year from now,and that means it would be 18 months. The area reduction is purported be 70% so even if all logic won't see that level,you could imagine how much smaller a TU102 shrunk down would be,or even a TU104.
After all the last time Nvidia pushed their large GPU out with smaller ones at launch was during the Fermi era,and those didn't last for two years before replacement,ie,there were two refreshes in under three years. 28NM was an unusually long node due to the GF/TSMC 20NM failing.
Edit!!
Another thing the GF and TSMC 7NM processes are meant to be similar,and TSMC apparently is ahead of GF in volume,so if GF can have a GPU that size already in production,then I would expect TSMC to be able to too,especially since Epyc is probably not a tiny chip either.
https://www.anandtech.com/show/12677/tsmc-kicks-off-volume-production-of-7nm-chips
So if we say a year from now,and that means it would be 18 months. The area reduction is purported be 70% so even if all logic won't see that level,you could imagine how much smaller a TU102 shrunk down would be,or even a TU104.
After all the last time Nvidia pushed their large GPU out with smaller ones at launch was during the Fermi era,and those didn't last for two years before replacement,ie,there were two refreshes in under three years. 28NM was an unusually long node due to the GF/TSMC 20NM failing.
Edit!!
Another thing the GF and TSMC 7NM processes are meant to be similar,and TSMC apparently is ahead of GF in volume,so if GF can have a GPU that size already in production,then I would expect TSMC to be able to too,especially since Epyc is probably not a tiny chip either.
Last edited:
Permabanned
- Joined
- 12 Sep 2013
- Posts
- 9,221
- Location
- Knowhere
They'll both be wanting to hold future shrinks back for as long as possible, Soon they'll have nowhere to shrink too, They both know this and I imagine the long term forecast isn't good because of it, That's why they're bringing things like Ray Tracing to the table, It'll allow them to stretch out how long they have until they simply can't eek out any more performance as we know it.
Haha, really?
Fair bit of mileage with 7nm I'd imagine - easily 3-4 iterations of it as they refine the technology though some might be renamed under other classifications then something around "5nm" - beyond that is going to be a bit of an ouch though as proper 3nm is probably going to cost well over $1bn just for upfront costs on something like a GPU.
Not sure on that - if I have my numbers right it is around $5000 per wafer on 12FF compared to $12-15K on 7nm potentially 16K on some variants. Obviously if you are producing the exact same design shrunk then you get more of them from a wafer but that isn't how GPUs generally work.
Last edited:
The way i see things, is that Nvidia will make the GTX2000 series short lived.
They are likely to respin Turing onto 7nm in a year or so and release a TU200 part as a higher end replacement to TU102 and RTX2080Ti.
This should allow Nvidia to make a 7SM per GPC higher end TU200 part while relegating TU204 and TU202 to the next tiers down.
They are likely to respin Turing onto 7nm in a year or so and release a TU200 part as a higher end replacement to TU102 and RTX2080Ti.
This should allow Nvidia to make a 7SM per GPC higher end TU200 part while relegating TU204 and TU202 to the next tiers down.
Yeah, we're really starting to hit not only the limits of what is physically possible (with silicon), but also what is economically viable.
Various sources have speculated that shrinks beyond "7nm" may never happen...
We're really starting to look towards new materials, new processes, etc, and can no longer rely on "just" making things smaller.
From what I understand the designs are built many years ahead. Not much they can do to change what goes into 7nm now. All that likely will happen is they will reduce price. Suddenly everyone will start praising them again and forget everything
Associate
- Joined
- 27 Apr 2007
- Posts
- 966
An issue there might be that due to low market penetration it made little sense to compare to a product that very few use.
And why would Nvidia even want to mention AMD when they are aiming to take sales off Intel!
Don't even mention the competition when they are small and wanting to grow, let AMD market themselves.
Note: I take anything you say with a tonne of salt so not referring to your performance claims for AMD.
You will see 7nm next year it will be on the equevlent of the GTX 1050 /1060 cards that are availble now. What interests me is will they stick with 75wats and 120 Watts power limit for the new cards or take advantage of the die shrink to cut power usage further.
I guess as a pipe cleaner 7nm on a 1050/60 card next year makes sense it would be a low power chip perfect for laptops etc.
We know AMD have a big gpu already, we just dont know if/when that will translate to a consumer GPU. They are going to be so fat behind the 2080 and 2080ti that I expect them to try it as soon as possible though.
We know AMD have a big gpu already, we just dont know if/when that will translate to a consumer GPU. They are going to be so fat behind the 2080 and 2080ti that I expect them to try it as soon as possible though.
We have no idea what AMD is preparing to launch this year. We only know 7nm Navi (GCN 6.0) is due next year, with Next Gen in 2020.
I do not believe 2080/Ti are "mainstream consumer GPUs". Less than the 5% of the market buys cards above the £700 price mark, let alone £1200.....
Though given Nvidia's greed and pricing, I wouldn't be surprised AMD was getting the 30Tflop Godzilla (Vega 20) into the mainstream market at £1000 just to test the waters.
And I wrote 30Tflop not 21Tflop because we all know that AMD pro cards (like the WX9100) are operating at 2/3 the speed of the consumer cards, at sub 200W power consumption.
However do you, as consumer, believe that seeing overpriced GPUs at £1000 the solution going forward?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
It's pointless, no-one would pay that for an AMD consumer card. Professional where they shine then yes, but I think that price-point for AMD is now dead.
Associate
- Joined
- 15 Oct 2014
- Posts
- 746
- Location
- Somerset England
If you can even call it a Ti at £1200... I’d call it a rebadged Titan
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
Agreed, and for me it seems like a permanent shift (as it separates the Titan nicely as a prosumer card at close to £3k). Considering the price they charge for the new Quadro cards I think the Ti isn't *too* bad! Massive chip though.
I read recently that Microsoft plan to include the hardware within the XBox Live subscription.
Titans do not have part of the chip disabled, although this could be due to yield.