• 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.

The next gen xxnm GPU thread.

Kaapstad

Kaapstad

Kaapstad

Man of Honour
Joined
21 May 2012
Posts
31,922
Location
Dalek flagship
The next gen xxnm GPU thread.



H7eCcS5.jpg

The purpose of this thread is to keep track of all the stories relating to the next gen xxnm GPUs.

If people post the links I will keep the OP updated with the latest stories.:)​
 
Last edited:
Articles.

Date 29/10/2015, Rumoured to be using GDDR5x on new 14nm GPUs. Story Link Post Link

Date 24/09/2015, ‘Pascal’ graphics cards will get 16GB of HBM2 memory. Story Link Post Link

Date 22/09/2015, Pascal to use second-gen HBM memory from two suppliers. Story Link Post Link

Date 22/09/2015, ‘Volta’ is now due in 2018. Story Link Post Link

Date 17/09/2015, "Pascal" GPUs to be Built on 16 nm TSMC FinFET Node. Story Link Post Link

Date 09/09/2015, Synopsys shows working PCIe4 silicon. Story Link Post Link

Date 08/09/2015, Micron to unveil third-gen hybrid memory cube in 2016. Story Link Post Link

Date 04/09/2015, We are actively promoting HBM and do not collect royalties. Story Link Post Link

Date 01/09/2015, Chinese investors eye Global Foundries. Story Link Post Link

Date 29/08/2015, Work on HBM technology started nearly a decade ago. Story Link Post Link

Date 26/08/2015, TSMC growing its 16nm client base. Story Link Post Link

Date 25/08/2015, TSMC obtains orders for 16nm FinFET chips. Story Link Post Link

Date 20/08/2015, Three new GPUs: Greenland, Baffin and Ellesmere. Story Link Post Link

Date 20/08/2015, SS Enters The HBM Market In 1H 2016. Story Link Post Link

Date 14/08/2015, TSMC prepares for 16nm production. Story Link Post Link

Date 05/08/2015, Quantum may never make it to the market. Story Link Post Link

Date 27/07/2015, Inside 1st High Bandwidth Memory. Story Link Post Link

Date 24/07/2015, SS adds 10nm FinFET tech to roadmap, mass production in 2016. Story Link Post Link

Date 24/07/2015, Pascal GPU has 17 billion transistors. Story Link Post Link

Date 21/07/2015, TSMC plans to use EUV lithography for 5nm. Story Link Post Link

Date 18/07/2015, 10nm is on-track for volume production start in Q4 2016. Story Link Post Link

Date 18/07/2015, 7nm tech will use 10nm elements. Story Link Post Link

Date 17/07/2015, We have taped out our first FinFET products. Story Link Post Link

Date 14/07/2015, Priority supply agreement regarding HBM, HBM2. Story Link Post Link

Date 05/06/2015, ‘Big Pascal’ GPU reportedly taped-out, – rumour. Story Link Post Link

​
 
Last edited:
As it seems now the 14/16nm cards wont arrive until Q3 2016, mainly because of HBM2 and interposer manufacturing.
also i'm curious what is the truth in the "AMD has priority on HBM2" news.
 
Funny how GP100 big Pascal will be coming out just as 10nm goes into volume.

Although it sounds like another 20nm, a node that can't be used for GPUs. Let's hope EUV and new materials don't suffer any more delays. If they move to compound III-V transistors instead of silicon there's still hope performance uplifts can be provided via clock bumps.
 
I hope a die shrink isn't to far away I keep staying with this rather slow and loud R9 290 which won't overclock past 1060 core +50mv and I want something much faster but not willing to pay £500 for a GTX 980ti (which is the only meaningful upgrade). Hoping to find a used 290 for £100 but that will never happen.
 
I expect we will be on 16nm for a good few years, probably 2 maybe even 3 generations of GPU before we move to an even smaller node.
 
Orangey said:
Funny how GP100 big Pascal will be coming out just as 10nm goes into volume.

Although it sounds like another 20nm, a node that can't be used for GPUs. Let's hope EUV and new materials don't suffer any more delays. If they move to compound III-V transistors instead of silicon there's still hope performance uplifts can be provided via clock bumps.
Click to expand...

Finfet is supposed to be in volume now, but nvidia and AMD only just taped out, so yeah it will be at least a year later, even Intel have bumped 10nm to 2017 and I find it hard to believe tsmc are spending more on R&D than Intel to overcome the current issues.
 
Orangey said:
Funny how GP100 big Pascal will be coming out just as 10nm goes into volume.

Although it sounds like another 20nm, a node that can't be used for GPUs. Let's hope EUV and new materials don't suffer any more delays. If they move to compound III-V transistors instead of silicon there's still hope performance uplifts can be provided via clock bumps.
Click to expand...


I'm not sure where you get that form. Big Pascal taped out last month so should appear Q2 2016 but could appear as early as Q1 (although those chips likely destined for compute purposes). Volume shipping of 10nm from TMSC isn't until Q1 2017, and could easily slip.

Basically there is a year, maybe more, between Pascal and 10nm shipping. Nvidia can then do a fairly straightforward die shrink to 10nm if the process is suitable and worthwhile, but form TSMC's descriptions it doesn't sound like a big jump.
I doubt Nvidia or AMD would shift their GPUS to 10nm so quickly anyway. Given the costs they will wait. AMD will want to shift APUs to 10nm ASAP though.
 
Mtom said:
As it seems now the 14/16nm cards wont arrive until Q3 2016, mainly because of HBM2 and interposer manufacturing.
also i'm curious what is the truth in the "AMD has priority on HBM2" news.
Click to expand...

Basically how i take that is that AMD get priority supply of HBM 2. So say if other companies order or want x amount of HBM2, however hynix on have x supply of HBM2 then AMD get the priority of the supply. 100 chips and AMD want all of them, then i guess they will get two thirds where the 1 third will be distributed to other companies(theoretical numbers). Or possibly AMD will get all or most of HBM 2 at first then rest distributed to others.
HOWEVER... :) If AMD do not have anything ready or need HBM 2 before other companies such as nVidia then nVidia will get their supply no prob. Just AMD have priority so if Nvidia and AMD want supply at the same time then AMD get the priority.

Thats how i see it and i could have worded that much better but im tired lol.
 
r7slayer said:
Basically how i take that is that AMD get priority supply of HBM 2. So say if other companies order or want x amount of HBM2, however hynix on have x supply of HBM2 then AMD get the priority of the supply. 100 chips and AMD want all of them, then i guess they will get two thirds where the 1 third will be distributed to other companies(theoretical numbers). Or possibly AMD will get all or most of HBM 2 at first then rest distributed to others.
HOWEVER... :) If AMD do not have anything ready or need HBM 2 before other companies such as nVidia then nVidia will get their supply no prob. Just AMD have priority so if Nvidia and AMD want supply at the same time then AMD get the priority.

Thats how i see it and i could have worded that much better but im tired lol.
Click to expand...

It's like Priority seating on buses for the disabled - the ones you never see occupied by a disabled person. AMD is the disabled person, and is too depressed/skint to go out and ride a bus.

Oh no he dit'nt!
 
I bought my 980 Ti expecting it to last 2 years and that looks like a good bet to me. Imo i don't think we will see any top end gaming Gpu's for about 18 months (due to Hbm shortages, which are bad enough now with only AMD using it currently) and by the time they are available in any sort of large numbers i'll be very close to my 2 year mark.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom