• 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 Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

Soldato
Joined
22 Sep 2011
Posts
10,575
Location
Portsmouth (Southsea)
But in that time the games will get harder to run which puts you back to square one ?
Not by that much really I don't believe this time, until the next generation of consoles the core complexity isn't likely to increase as quickly as the demand from 1440p/1080p to 4k, at the moment this is one huge hurdle which requires the next generation more than usual.

Normally I'd agree that 'waiting' is fruitless, but even games a couple of years old are not running flawlessly at 4k with the current generation. This makes perfect sense as none of the current generation were really made with 4k in mind, this of course will change as Nvidia & AMD attempt to be the first to fill this gap in the market.
 
Associate
Joined
27 May 2007
Posts
1,346
Location
Kent
It will be on the motherboard and looks to need significant modification for power delivery and trace length. You can disregard it unless you are intending on building some form of supercomputer and are willing to pay through the nose for it. It wont be marketed at the PC consumer, closest it will get is servers and they will probably have to co-support both pci-e/Nv-link.

An overview here:
http://www.anandtech.com/show/7900/nvidia-updates-gpu-roadmap-unveils-pascal-architecture-for-2016

Thanks for your reply eddyr, so the first couple of releases of Pascal will work at its best through the PCI bus?

Reading up on Pascal it just seemed they will max out the PCI bus oh well all is good then I would will be waiting for Pascal it just feels that that card is going to be mental.
 
Associate
Joined
29 Aug 2004
Posts
2,381
Location
Alpha centauri
Thanks for your reply eddyr, so the first couple of releases of Pascal will work at its best through the PCI bus?

Reading up on Pascal it just seemed they will max out the PCI bus oh well all is good then I would will be waiting for Pascal it just feels that that card is going to be mental.

Yes pascal performance could very well be mental and AMD`s new Fury GPU & HBM will give us an idea of what to expect going forward.
 
Soldato
Joined
22 Aug 2008
Posts
8,338
Xilinx has taped out a 16FF+ chip.

http://www.kitguru.net/components/a...-out-worlds-first-16nm-finfet-system-on-chip/

Doesn't mean much but may lend credence to the rumour of Pascal tapeout last month. We should be expecting various bits and bobs to slowly feed into the channel as NV perfect the various techs used (HBM/new process/new arch). Unfortunately the 950Ti is confirmed to be using Maxwell so perhaps nothing for a while yet.
 
Man of Honour
Joined
13 Oct 2006
Posts
91,533
Timing is about right for the claims that they would be ramped up to volume production late this year... (dot, dot, dot for I'll believe it when I see it when it comes to TSMC).
 
Caporegime
Joined
20 Jan 2005
Posts
45,751
Location
Co Durham
Ive seen it said the gaming card will have 16Gb ram. I dread to think what the Nvidia price will be.

I can easily see these cards been £1000 cards again :(
 
Soldato
OP
Joined
2 Jan 2012
Posts
11,932
Location
UK.
^^ Yup.

Flagship Titan Pascal GP200 could well have 12GB / 16GB of HBM. With the normal Flagship GP204 having 6GB / 8GB of HBM.

They are going to be pricey for sure, but performance is going to be mind-blowing.

We're getting Gen 2.0 HBM along with a massive die shrink either 16nm/14nm down from 28nm. Plus a new architecture. These cards are going to be scary !
 
Soldato
Joined
5 Sep 2011
Posts
12,827
Location
Surrey
^^ Yup.

Flagship Titan Pascal GP200 could well have 12GB / 16GB of HBM. With the normal Flagship GP204 having 6GB / 8GB of HBM.

They are going to be pricey for sure, but performance is going to be mind-blowing.

We're getting Gen 2.0 HBM along with a massive die shrink either 16nm/14nm down from 28nm. Plus a new architecture. These cards are going to be scary !

Didn't take you long to get over the Fury X hype I see. :p
 
Associate
Joined
28 Jan 2010
Posts
1,547
Location
Brighton
Thinking small pascal may be 15% or so faster than Titan X, hoping I'm wrong.

Pascal Titan will be where its at.

I'd have thought it should be a lot better than that.

If you're comparing this to the GTX 780 Ti --> GTX 980, you have to remember that was on the same process node (28nm) and there was no upgrade in memory tech (infact technically it was downgraded since it went to 256-bit from 384-bit).

Titan X --> GP104 will be similar to GTX 580 --> GTX 680 in terms of what's changing.

We are getting HBM2 and 14/16nm.
 
Last edited:
Associate
Joined
25 Jul 2014
Posts
413
Location
Leicester
I think we will see a big leap forward into the next generation and further power efficiency reductions but the die shrink may also come at the price of less overclocking.

I also think the cards coming in 2016 will give a big clue as to what to expect from the next generation of consoles that come out towards the end of the decade (and I expect a focus on 4k gaming there).
 
Soldato
Joined
22 Aug 2008
Posts
8,338
A normal leap is very roughly about 80%, you take away the shrink you get something more like half that, which is what TX uplift vs TB provides (~40%).

GP100 AKA "Titan Y" :D should be almost twice as fast as GM200, leaving aside any problems they may run into.
 
Man of Honour
Joined
13 Oct 2006
Posts
91,533
Pascal should be quite a big leap - don't forget even with the refinements to the 28nm process we've still been stuck on it for quite a long time - just taking the TX and reworking it on 16nm FF+ (AFAIK there is no direct (optical type) shrink path from 28nm to FF+) would be a straight up ~45% performance increase - never mind a new architecture and designing it to get optimal results from the process.

For DX12 titles I'd be unsurprised if the high end (not the inevitable mid-range masquerading as high end offerings) Pascal GPUs are 2.5x the performance of the TX in some cases.
 
Last edited:
Soldato
Joined
9 Nov 2009
Posts
24,898
Location
Planet Earth
AMD and Nvidia are going to hit the same issues of cost which they also hit with 28NM. So I would expect sub 400mm2 dies for the first high end products available for consumers IMHO.

They have realised that they can get away with relatively small increases in performance at 28NM while charging a decent amount,so I see no reason why they won't milk 14NM/16NM to maximise profits especially with the decreasing number of discrete cards being sold.

You only need to look at the jump from 40NM to 28NM to see where things will end up in the first year.
 
Last edited:
Back
Top Bottom