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

[Rroff]

Its always kind of been the way though - nVidia often showing off dummy or hacked together random PCBs rather than actual cards even fairly close to launch while AMD have often had actual boards on show.

 

[D.P.]

Some of the facts are big pascal taped out first back in summer 2015 and there have been no rumors of re-spins or issues. Doesn't mean there weren't any but in previous occasions NVidia or AMD had issues there were several reliable leaks, e.g. 480.

Its also a fact that Nvidia have to supply big pascal chips for compute purpose with some of their contracts. This may mean that there can be no spare cores for gaming, or that there will be a limited release Titan Pascal where the markup is suitably high and volumes low.


Its a fact that various auto partners will be getting Pascal based PX2 units in April, with volume shipping of production in Q4 (aiming for 2018 model years).

GP104 was always scheduled for the summer. GP100 was scheduled Q2.


Anything else is speculation and none of the above proves anything about Pascal being on time or delayed. I'm simply saying there isn't a shred of evidence that pascal is late.

 

[humbug]

I wish I could read it but looked at the anand site and nothing about it. I skimmed read an article on polaris but that was prior to the press seeing it, so instead of telling me, show me the link and I will read it.

Very small size and power - Is that a 950 competitor then?

You Quoted Boom here

Cheers Boom and good to see some solid info (even if it is from wccftech). Hopefully we can get to see some die sizes and specs soon so the mathematicians can do the sums for performance. It would be great to see a release soon. My machine needs an overhaul

[UPDATE 3 : 2016 January 8, 12:53 AM ET]
We’ve reached out to AMD and they have confirmed that the Polaris GPU demoed was in fact built using Globalfoundries’ 14nm FinFET process. Some publications have reported that Polaris will be a mix of both TSMC 16nm and Globalfoundries 14nm GPUs, which is where some of the confusion could potentially have stemmed from. However, according to AMD Polaris “is only 14nm”.

Furthermore, they’ve asked us to refer back to the company’s press release — which you can find quoted in its entirety in the original article below — which clearly states that the Polaris Architecture is 14nm based. Here’s the the relative segment of the press release :


Quote:
AMD’s Polaris architecture-based 14nm FinFET GPUs deliver a remarkable generational jump in power efficiency. Polaris-based GPUs are designed for fluid frame rates in graphics, gaming, VR and multimedia applications running on compelling small form-factor thin and light computer designs.

Read more: [URL="http://wccftech.com/amd-polaris/#ixzz3we4dfHYB"]http://wccftech.com/amd-polaris/#ixzz3we4dfHYB[/URL]

Anand...

In any case, the GPU RTG showed off was a small GPU. And while Raja’s hand is hardly a scientifically accurate basis for size comparisons, if I had to guess I would wager it’s a bit smaller than RTG’s 28nm Cape Verde GPU or NVIDIA’s GK107 GPU, which is to say that it’s likely smaller than 120mm2. This is clearly meant to be RTG’s low-end GPU, and given the evolving state of FinFET yields, I wouldn’t be surprised if this was the very first GPU design they got back from Global Foundries as its size makes it comparable to current high-end FinFET-based SoCs. In that case, it could very well also be that it will be the first GPU we see in mid-2016, though that’s just supposition on my part.
For their brief demonstration, RTG set up a pair of otherwise identical Core i7 systems running Star Wars Battlefront. The first system contained an early engineering sample Polaris card, while the other system had a GeForce GTX 950 installed (specific model unknown). Both systems were running at 1080p Medium settings – about right for a GTX 950 on the X-Wing map RTG used – and generally hitting the 60fps V-sync limit.


The purpose of this demonstration for RTG was threefold: to showcase that a Polaris GPU was up and running, that the small Polaris GPU in question could offer performance comparable to GTX 950, and finally to show off the energy efficiency advantage of the small Polaris GPU over current 28nm GPUs. To that end RTG also plugged each system into a power meter to measure the total system power at the wall. In the live press demonstration we saw the Polaris system average 88.1W while the GTX 950 system averaged 150W. Meanwhile in RTG’s own official lab tests (and used in the slide above) they measured 86W and 140W respectively.
Keeping in mind that this is wall power – PSU efficiency and the power consumption of other components is in play as well – the message RTG is trying to send is clear: that Polaris should be a very power efficient GPU family thanks to the combination of architecture and FinFET manufacturing. That RTG is measuring a 54W difference at the wall is definitely a bit surprising as GTX 950 averages under 100W to begin with, so even after accounting for PSU efficiency this implies that power consumption of the Polaris video card is about half that of the GTX 950. But as this is clearly a carefully arranged demo with a framerate cap and a chip still in early development, I wouldn’t read too much into it at this time.

At that size its smaller than a HD 7770. Which is the smallest GCN GPU currently on the market, which is about a third the performance of a GTX 950.

 

[D.P.]

Its always kind of been the way though - nVidia often showing off dummy or hacked together random PCBs rather than actual cards even fairly close to launch while AMD have often had actual boards on show.

And a lot of times its done on purpose to prevent disclosing information they don't want out in the public.

The Drive PX2 unit at CES clarly was't pascal but select industry partners have seen pascal running.
There couple be numerous reason why a Maxwell based prototype was shown:
1) Nvidia don't have enough working engineering samples
2) Nvidia doesn't want anyone not under NDA to actually see the chip.
3) Mid-szied Pascal doesn't exist.


The AMD fanboys jump on 2, completely ignore possibilities 1 and 2, when possibility 2) is exactly the same status as AMD. AMD haven't shown the public any working chips either. T=AMD have shown journalists under NDA. NVIDIA have shown auto partners under NDA and its possible hey have shown compute partners big pascal, e.g. sent IBM engineering samples. There is no source o the latter, i'm just putting it out there because when partners under NDA then they keep quiet.

AMD's showing was slightly different on nature, they invited tech journalists so wanted to generate some noise, but AMD were also show casing different products for different markets with different intended consequences.

 

[Orangey]

Big Pascal taping out was just a message board rumour.

Furthest thing from fact there is.

And so what if they want to sell new Teslas. If they can't make it they can't make it.

 

[humbug]

And a lot of times its done on purpose to prevent disclosing information they don't want out in the public.

The Drive PX2 unit at CES clarly was't pascal but select industry partners have seen pascal running.
There couple be numerous reason why a Maxwell based prototype was shown:
1) Nvidia don't have enough working engineering samples
2) Nvidia doesn't want anyone not under NDA to actually see the chip.
3) Mid-szied Pascal doesn't exist.


The AMD fanboys jump on 2, completely ignore possibilities 1 and 2, when possibility 2) is exactly the same status as AMD. AMD haven't shown the public any working chips either. T=AMD have shown journalists under NDA. NVIDIA have shown auto partners under NDA and its possible hey have shown compute partners big pascal, e.g. sent IBM engineering samples. There is no source o the latter, i'm just putting it out there because when partners under NDA then they keep quiet.

AMD's showing was slightly different on nature, they invited tech journalists so wanted to generate some noise, but AMD were also show casing different products for different markets with different intended consequences.

is exactly the same status as AMD. AMD haven't shown the public any working chips either.

Only if you ignore AMD's public demo of a working chip.

 

[D.P.]

Only if you ignore AMD's public demo of a working chip.

Where have AMD shown a public demo. AFAIK they have only provided a private demo to people under NDA, the same as nvidia has done. They have made all sorts of marketing spiel as expected, but I haven't seen any public demonstration, benchmarks or anything.

Please post the source.

 

[Rroff]

I love that this board has members with so much factual knowledge and inside information proven to be factually correct. I'm looking forward to referring back to this thread once spec/cards are released.

It is interesting looking back

 

[humbug]

Where have AMD shown a public demo. AFAIK they have only provided a private demo to people under NDA, the same as nvidia has done. They have made all sorts of marketing spiel as expected, but I haven't seen any public demonstration, benchmarks or anything.

Please post the source.

CES.

Its where they compared it to the GTX 950


http://gadgets.ndtv.com/laptops/news/amds-next-generation-polaris-gpu-demoed-at-ces-2016-785742

 

[Gregster]

What exactly is your point? Is it the usual deflection? Has anyone claimed Nvidia hasn't given this information but AMD has, nope, but you're pushing the discussion away from the current different stages AMD and Nvidia are at.

I say Nvidia haven't shown Pascal at all and AMD have which indicates very different stages of development, but you try to make this but AMD and Nvidia haven't shared die sizes.

Architecture talk comes before specific product talk. Before the 480gtx was launched properly the Fermi architecture was 'launched' with a fake 480gtx card shown.

Again, sequence of events leading up to a card being available to buy is the same every generation. Shown off cards and talk about basic architecture stuff, more behind closed doors than public. Specific cards get talked about very very late in the day usually right before release, ie within 2 weeks usually.

AMD are talking about the architecture and showing parts. Nvidia were talking about PX2, not much about Pascal itself and didn't show any parts.

Die size info is almost never talked about till release reviews. Estimates from press briefings such as Anand guesstimating that Polaris 10 is 100-110mm^2. But nothing official from AMD/Nvidia will be stated till review press packets(if then, often the die sizes come from measurements done by the reviewers).

No deflection but you are having a pop at people who are just going on rumors and then going on rumors yourself. Maybe step off that pedestal and let people speculate. It is harmless and enjoyable reading for me. Pascal is coming - Polaris is coming. We know nothing about cores or die sizes on any of them as yet but the click bait sites will do what they have done for years and speculate and we will do the same. You have a real bad habit of having a dig at people and no need at all. If you have something solid, post it. If you want to speculate like the rest of us, speculate but don't get all humpty with people because they are speculating please. Not needed at all.

Humbug, yer I remember that wccftech article but missed the Anand one. Still nothing there (and why would they if they are under NDA) but makes no odds to me, as the first with the big guns will see my cash. If it is Polaris, I am interested and if it is Pascal I am interested. Bring it on I say

 

[bru]

Anand

In any case, the GPU RTG showed off was a small GPU. And while Raja’s hand is hardly a scientifically accurate basis for size comparisons, if I had to guess I would wager it’s a bit smaller than RTG’s 28nm Cape Verde GPU or NVIDIA’s GK107 GPU, which is to say that it’s likely smaller than 120mm2. This is clearly meant to be RTG’s low-end GPU, and given the evolving state of FinFET yields, I wouldn’t be surprised if this was the very first GPU design they got back from Global Foundries as its size makes it comparable to current high-end FinFET-based SoCs. In that case, it could very well also be that it will be the first GPU we see in mid-2016, though that’s just supposition on my part.
For their brief demonstration, RTG set up a pair of otherwise identical Core i7 systems running Star Wars Battlefront. The first system contained an early engineering sample Polaris card, while the other system had a GeForce GTX 950 installed (specific model unknown). Both systems were running at 1080p Medium settings – about right for a GTX 950 on the X-Wing map RTG used – and generally hitting the 60fps V-sync limit.


The purpose of this demonstration for RTG was threefold: to showcase that a Polaris GPU was up and running, that the small Polaris GPU in question could offer performance comparable to GTX 950, and finally to show off the energy efficiency advantage of the small Polaris GPU over current 28nm GPUs. To that end RTG also plugged each system into a power meter to measure the total system power at the wall. In the live press demonstration we saw the Polaris system average 88.1W while the GTX 950 system averaged 150W. Meanwhile in RTG’s own official lab tests (and used in the slide above) they measured 86W and 140W respectively.
Keeping in mind that this is wall power – PSU efficiency and the power consumption of other components is in play as well – the message RTG is trying to send is clear: that Polaris should be a very power efficient GPU family thanks to the combination of architecture and FinFET manufacturing. That RTG is measuring a 54W difference at the wall is definitely a bit surprising as GTX 950 averages under 100W to begin with, so even after accounting for PSU efficiency this implies that power consumption of the Polaris video card is about half that of the GTX 950. But as this is clearly a carefully arranged demo with a framerate cap and a chip still in early development, I wouldn’t read too much into it at this time.

You missed the important bit.

As a quick preface here, while RTG demonstrated a Polaris based card in action we the press were not allowed to see the physical card or take pictures of the demonstration. Similarly, while Raja Koduri held up an unsoldered version of the GPU used in the demonstration, again we were not allowed to take any pictures. So while we can talk about what we saw, at this time it’s all we can do.

So in fact all the press have seen is a closed computer running a demo and an unsoldered card and been told it is the one running the demo. Now I don't for one minute suspect that it wasn't the same card running the demo and I'm sure the power numbers are indeed what AMD say they are, but you have to love how AMD are instantly believed and because NVidia show a stage piece that the internet has claimed is exactly the same as an earlier part, even though we only have a shaky screen grab to go on, they cannot be telling the truth.

Both sides could be lying through their teeth or telling the truth, to be honest who cares, we just want the cards.

Anyway, my tuppence worth on the currant April release rumour.

NVidia show the Titanesque card at GDC saying it will be available in a coupe of weeks for a lot of monies ($1500 maybe) then surprise everyone with a 1070 and 1080 in may that just straddle the 980ti in performance for £300 and £400, therefore repeating the absolutely colossal success they have had with 970 ( even with the whole lying thing). The 1080ti with HBM then follows in August/September, thus avoiding the issue with 980ti coming so close to the after the release of the TitanX.

 

[drunkenmaster]

No, Nvidia did only show Maxwell and they did actively lie and said it was Pascal, you can't just explain around that because you want to.

Likewise AMD showed a demo that could ONLY be a working 14nm GPU, this establishes a basis to not believe but KNOW they have working Polaris GPUs and they've had them long enough to get them working in a demo which implies a relatively decent amount of time spent working on drivers.

So no, AMD can't be lying through their teeth because they have proved to have 14nm gpus running and no Nvidia can't be telling the truth because they specifically lied about having Pascal.

There are no journalists suggesting they've seen Pascal behind the scenes in demo, pictures of the core, nothing. We have one instance where we know Nvidia lied. AMD showed stuff behind the scenes but already established the fact they have 14nm gpus publicly so there is no reason to believe they would lie in a behind the scenes journo session.

Who cares, you do obviously because you're suggesting Nvidia might not be lying and AMD could be lying even though we know this isn't true. As for the AMD are believed and Nvidia aren't, that's simply because that is what this situation is. Also, a shaky screen grab... in which multiple journalists appear to have confirmed from their own pictures or the ones available on the web that all the numbers on the GPU match Maxwell? It doesn't matter how shaky or bad a picture is, just that it has enough information. The modules are clear as day Maxwell, it looks like a Maxwell MXM module, it has the same parts and the silicon itself states it's Maxwell. Again I said even before it was confirmed as Maxwell and before I'd seen a video of the conference, if he was just saying "this is what it WILL look like", it's fine, and normal. I also said if he claims it IS Pascal but isn't, that is very dodgy, after watching the video it's very clear he claimed it was Pascal several times.

Just to add to it, multiple journalists also said they showed medium Polaris in that behind the scenes event, again I've not seen a clear claim on if this was a bare core, card, working demo or what, just that medium Polaris was definitely there as well as small Polaris. Though judging from not showing the small Polaris on a working card it's being used on a non final pcb and/or non final cooling.


My take, GDC nothing much, maybe they'll talk about Pascal, maybe not. Whenever AMD launch a card that is midrange and probably 15-25% faster than Fury at £350-450 for the full and salvaged parts and beats a Titanx/980ti for much cheaper, Nvidia will wait a month max before paper launching their cards but will only have real availability 2-3 months behind AMD. Titan/1080ti might be talked about this year but probably won't see it till Q2 2017.

 

[FoxEye]

My take, GDC nothing much, maybe they'll talk about Pascal, maybe not. Whenever AMD launch a card that is midrange and probably 15-25% faster than Fury at £350-450 for the full and salvaged parts and beats a Titanx/980ti for much cheaper, Nvidia will wait a month max before paper launching their cards but will only have real availability 2-3 months behind AMD. Titan/1080ti might be talked about this year but probably won't see it till Q2 2017.

dm, what do you think AMD will have at the £200 - £250 price point, out of curiosity? Even more cut-down Polaris or something else?

 

[bru]

So sorry DM please forgive me, how dare I post that the vile evil NVidia might have actually shown what they said they did, I forgot that you are all knowing and all seeing and can never be wrong, it was so foolish of me to put across one of many possible explanations, even though I did say that AMD probably did show what they said. Please forgive me, as I never meant to cause offence.


Disclaimer this post is very sarcastic in nature and is not meant to be taken seriously.

 

[D.P.]

CES.

Its where they compared it to the GTX 950


http://gadgets.ndtv.com/laptops/news/amds-next-generation-polaris-gpu-demoed-at-ces-2016-785742

So, exactly as i said just some more marketing material without a giving the chips to the public for testing, just a closed test set up by AMD.

Nvidia did the much same thing with Pascal in PX2 with select auto partners.

 

[D.P.]

Anand




You missed the important bit.




So in fact all the press have seen is a closed computer running a demo and an unsoldered card and been told it is the one running the demo. Now I don't for one minute suspect that it wasn't the same card running the demo and I'm sure the power numbers are indeed what AMD say they are, but you have to love how AMD are instantly believed and because NVidia show a stage piece that the internet has claimed is exactly the same as an earlier part, even though we only have a shaky screen grab to go on, they cannot be telling the truth.

Both sides could be lying through their teeth or telling the truth, to be honest who cares, we just want the cards.

Anyway, my tuppence worth on the currant April release rumour.

NVidia show the Titanesque card at GDC saying it will be available in a coupe of weeks for a lot of monies ($1500 maybe) then surprise everyone with a 1070 and 1080 in may that just straddle the 980ti in performance for £300 and £400, therefore repeating the absolutely colossal success they have had with 970 ( even with the whole lying thing). The 1080ti with HBM then follows in August/September, thus avoiding the issue with 980ti coming so close to the after the release of the TitanX.

Exactly, there is really no difference between what AMD and Nvidia have shown.

 

[Boomstick777]

Pascal looking less interesting now, if the rumors hold any truth that is.

So the rub is that the only card with HBM 2.0 will be the Titan Pascal card, and performance will be huge but so will the price. The GTX 980 replacement will offer 980 Ti performance with GDDR5/X but will likely be priced at GTX 980 launch prices which was ~ £500, basically it's going to be a more power efficient GTX 980 Ti level card for the same price.

No increase in price VS performance at all, only power consumption. To get any real performance boost you will need the Titan flagship, until Nvidia launch a Ti variant which will usurp the Titan at some point..

I hope this isn't the case.. Please deliver AMD, we need competition !

 

[Troezar]

I will never for the life of me work out why people read this stuff as fact. Not least when it's so obvious they are trolling for hits and the rumours randomly get made up right after any real news.

Don't shoot the messenger

 

[drunkenmaster]

Exactly, there is really no difference between what AMD and Nvidia have shown.

The level of delusion is severe. Journalists saying AMD have shown working Polaris on one side, same guys saying Nvidia showed Maxwell and pretended it was Pascal..... but AMD and Nvidia showed the same thing... wow.

 

[humbug]

So, exactly as i said just some more marketing material without a giving the chips to the public for testing, just a closed test set up by AMD.

Nvidia did the much same thing with Pascal in PX2 with select auto partners.

Don't be coy, you have a conspiracy theory, AMDs live public Polaris demo was a lie....... It must be right, it's the only way your argument can be right. :d