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

[humbug]

I loved the fanboy excuse of Nvidia didn't show real Pascal cores as they would have very few working samples as if that was a genuine excuse.

Break that down a bit, if they have cores back that means they would have made a run of likely min 5 wafers, probably 10 or 20. If you have so few cores working you can't spare one for a big game conference you have a major problem with low yields which would effect when you can launch. Regardless of if you have high or low yields... you will have multiple dead Pascal chips because that is the nature of silicon production. With multiple dead chips they somehow couldn't stick some dead ones on a PX2 pcb to make a realistic final card(though non working) to show up on stage? The only inference to be made from showing non Pascal chips but claiming they were Pascal is that they don't have Pascal silicon back yet.

If they had wafers of Pascal back they have zero excuse not to show a real Pascal on stage. The competition doesn't learn anything from seeing a chip on a board, except if it's real or not. If they had Pascal back from the fab they would definitely have ample chips, working or not, to stick on a card and parade around a trade show.

Sticking a couple maxwell chips on a pascal PCB shows how desperate and far behind they are.

 

[Simon Belmont]

Can we have a link to this two months thing. You are posting dozens of times in this thread saying Pascal not far away, only 2 months to find out. You consistently post these one liners saying info is really soon. Where is your proof Nvidia are going to launch a card or release any information about a specific card at GDC?

Conversely, there are hundreds of news stories about when Samsung started producing 14nm chips, when the improved version when into production, that they have better yields than TSMC, that TSMC have had delays, that their yields are bad, of several massive customers switching from TSMC to Samsung because they are significantly behind.

AMD has had two cores back, tested, got to the point of driver stability they can show demos to journalists, Nvidia hasn't reached the point of showing just a non working but real Pascal and is again resorting to faking a product on stage. You don't fake a product unless you want to mislead people over how far away you are. The last time Nvidia faked products on stage, they were 6 months behind where they claimed to be.

The process industry as a whole agrees TSMC is 3-6 months behind on finfet, the stage Nvidia are at with showing cards to journalists suggests they are a minimum 3 months behind AMD, Nvidia resorting to faking cards(based on previous history) suggests they are 3-6 months behind AMD.

What are the signs Nvidia is close.... despite Nvidia saying nothing of the sort, Nvidia guys are claiming Nvidia will give us loads of information at GDC.

GTC. Then we hope? (I mean hope not fact like everyone seems to be waving around here^) Nvidia release some real pascal news.

 

[Stanners]

actually when it comes to nodes; yes. They've always been behind AMD when it comes to switch to a new node - this is a fact. They have not brought anyone new in to help with node shrink. Again fact. Industry that's watching; has already said Nvidia is at least 1 quarter behind more likely 2 quarters behind AMD on node shrink and that's with small pascal.

Classic Nvidia trick is when they have nothing is to talk about next products coming; now we're starting to hear about Volta; means pascal is no where to be seen.....

You are just going on the past again, we do not know the future. You have thrown a couple of ifs and buts to try and make it fact, you could be right or wrong who knows. Time will tell(in the future )

 

[ICDP]

Until I see actual GPUs available to purchase my estimates are late summer for both Pascal and Polaris. It's quite entertaining to see people from both camps living on nothing but hope and making unsubstantiated "facts" based on this hope. While I am not making claims Nvidia are late I would certainly not be claiming "April" or lol'ing people based on pure blinkered fanboyism.

 

[Stanners]

Until I see actual GPUs available to purchase my estimates are late summer for both Pascal and Polaris. It's quite entertaining to see people from both camps living on nothing but hope and making unsubstantiated "facts" based on this hope. While I am not making claims Nvidia are late I would certainly not be claiming "April" or lol'ing people based on pure blinkered fanboyism.

+1 to this, I would like to see AMD get something out first that starts a shift and gains them more market share as we are stuck with just the two of them. We need a healthy market split to create competition on pricing and performance.

 

[Gregster]

My hopes and dreams are to buy a 16/14nm GPU that performs very quick this year

Edit:

Just to add - People are saying that AMD have a working Polaris chip - Is that the one that was shown behind closed doors to the press but they weren't allowed to see it?

 

[Rroff]

Classic Nvidia trick is when they have nothing is to talk about next products coming; now we're starting to hear about Volta; means pascal is no where to be seen.....

There is a reason we are hearing about Volta... nVidia will be producing it through 2017 - though consumers won't get their hands on it until approx June 2018 according to current roadmaps.

 

[humbug]

My hopes and dreams are to buy a 16/14nm GPU that performs very quick this year

Edit:

Just to add - People are saying that AMD have a working Polaris chip - Is that the one that was shown behind closed doors to the press but they weren't allowed to see it?

It's the one the press were allowed to see but not report on in detail, despite this Anand did make a small writeup on what they saw

 

[hominid]

Just to add - People are saying that AMD have a working Polaris chip - Is that the one that was shown behind closed doors to the press but they weren't allowed to see it?

You know full well what one it is! It's the one that they showed but didn't allow the press to take pictures of it.. It's 100% working 14/16nm silicon in a working board doing testing in a game engine looped on early drivers - impressive for what is a very early board.

 

[Gregster]

It's the one the press were allowed to see but not report on in detail, despite this Anand did make a small writeup on what they saw

So what do we know about Polaris then? I have only read speculation from tech journalists on that but surely for people to be basing this and that on Pascal and getting told off, there must be some hard evidence to back up what Polaris has to offer that was shown to the press behind close doors?

You know full well what one it is! It's the one that they showed but didn't allow the press to take pictures of it.. It's 100% working 14/16nm silicon in a working board doing testing in a game engine looped on early drivers - impressive for what is a very early board.

I have no idea what it is. I have read a couple of articles and some are saying they guess it to be the successor to the 390X but only guesswork. You tell me what it is please as you seem to know.

 

[drunkenmaster]

My hopes and dreams are to buy a 16/14nm GPU that performs very quick this year

Edit:

Just to add - People are saying that AMD have a working Polaris chip - Is that the one that was shown behind closed doors to the press but they weren't allowed to see it?

As I understand it they had a relatively public demo showing the small Polaris vs 950ti system, then in private they showed the medium Polaris, it wasn't clear if that was full card, working demo, etc. Realistically showing it working and given the fact that no possible non 14nm card could run at around the 35W mark and produce that level of performance it couldn't be anything but working 14nm silicon, combined with multiple journos confirming they saw the medium behind closed doors is effectively double source confirmation AMD were demoing on working 14nm silicon.


From experience I would expect the behind closed door showing to be journalists being able to see and hold the final pcb and chip, maybe the cooling, maybe not, and probably a working demo. Ultimately if you have the small card working the driver for the architecture will be pretty much there and if you can get one working the others becomes pretty trivial by comparison.

I would guess the public demo didn't show the card as it had some ghetto mod cooling on it, not final production cooling so they didn't want to show it.

 

[Gregster]

As I understand it they had a relatively public demo showing the small Polaris vs 950ti system, then in private they showed the medium Polaris, it wasn't clear if that was full card, working demo, etc. Realistically showing it working and given the fact that no possible non 14nm card could run at around the 35W mark and produce that level of performance it couldn't be anything but working 14nm silicon, combined with multiple journos confirming they saw the medium behind closed doors is effectively double source confirmation AMD were demoing on working 14nm silicon.


From experience I would expect the behind closed door showing to be journalists being able to see and hold the final pcb and chip, maybe the cooling, maybe not, and probably a working demo. Ultimately if you have the small card working the driver for the architecture will be pretty much there and if you can get one working the others becomes pretty trivial by comparison.

I would guess the public demo didn't show the card as it had some ghetto mod cooling on it, not final production cooling so they didn't want to show it.

Fair but again the press are under NDA (I would assume), so we are left speculating. I can't find anything solid about die size, cores, memory - basically anything.

 

[humbug]

Anand reported on its very small size and low power consumption of the working sample they were shown. You read all this yourself greg.

 

[Simon Belmont]

Fair but again the press are under NDA (I would assume), so we are left speculating. I can't find anything solid about die size, cores, memory - basically anything.

Exactly Greg. The know it alls that type all kinds of silly nonsense are getting boring.

 

[LambChop]

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.

 

[Gregster]

Anand reported on its very small size and low power consumption of the working sample they were shown. You read all this yourself greg.

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?

 

[Orangey]

What are the odds of this:

-Pascal medium (1080, GP104 etc) late 2016
-Pascal abandoned for 10nm Volta
-Volta medium (1180 or perhaps sneakily named 1080ti) 2017
-Volta big (Volta Titan, GV100) 2018

?

Samsung and TSMC are both racing to 10, maybe there's a reason other than bragging rights. Maybe it's a gimme after the switch to finfets with 16 and yield learning from that can be re-applied.

 

[Rroff]

What are the odds of this:

-Pascal medium (1080, GP104 etc) late 2016
-Pascal abandoned for 10nm Volta
-Volta medium (1180) 2017
-Volta big (Volta Titan, GV100) 2018

?

GTX680/GTX980 (regardless of branding/what they are marketed as) type Pascal sooner rather than later - higher end offerings potentially abandoned or late. (Can't see it any other way though nVidia are a bit more optimistic - between TSMC not really optimal for anything above mid power and the timetable for the HBM2 stuff I can't see true high end Pascal before late 2016 possibly into 2017).

Very unlikely to see Volta 2017 (unless nVidia do some dirty renaming of a refined Pascal for consumers), Volta will likely be focused on big parts.

 

[drunkenmaster]

Fair but again the press are under NDA (I would assume), so we are left speculating. I can't find anything solid about die size, cores, memory - basically anything.

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).

 

[Orangey]

Yeah GP204 more likely in 2017.