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

[Orangey]

Youtube is not a valid source end of.

 

[drunkenmaster]

I liked where he had zero solid information and also kept mentioning AMD in a oh, and then AMD type of fashion.

Nvidia are moving to 16nm and yeah, I guess AMD are going to smaller process too. Then he said everything points to Nvidia coming first to market. We've got some samples going out for GP100 in the next several months but will only ship in volume in Q1/17, those are the actual dates Nvidia has talked about. AMD has said flat out Polaris will be available this summer. More to the point Polaris 11 and 10 have been shown in working demos as far back as December, GP104 we've seen qualification samples with manufacturing dates from all the way back in.... mid April.

Notice that the qualification samples didn't have the mounting ring around the edge yet, nor a cooler, no sign it was running anywhere and pre-production memory on it, all of which do nothing to point to imminent release. AMD showed working cards fully 4 months earlier than Nvidia... but every sign points to Nvidia releasing first. It comes across as a Nvidia PR puff piece.

The only indication Nvidia will 'launch' first is because what Nvidia is calling a Tesla P100 launch is what AMD calls normal sampling of products before release to customers. Nvidia paper launched and is pretending the normal 3-9 month sampling period is a real launch.

 

[DarrenM343]

Didn't NVidia stop producing cards recently? Some of the more popular cards too, ie mid range? Has to point to new cards incoming pretty soon, not Q1 2017 or they're going to lose out to AMD in that area once their stocks run out and they won't let that happpen.

 

[Seanspeed]

I liked where he had zero solid information and also kept mentioning AMD in a oh, and then AMD type of fashion.

Nvidia are moving to 16nm and yeah, I guess AMD are going to smaller process too. Then he said everything points to Nvidia coming first to market. We've got some samples going out for GP100 in the next several months but will only ship in volume in Q1/17, those are the actual dates Nvidia has talked about. AMD has said flat out Polaris will be available this summer. More to the point Polaris 11 and 10 have been shown in working demos as far back as December, GP104 we've seen qualification samples with manufacturing dates from all the way back in.... mid April.

Notice that the qualification samples didn't have the mounting ring around the edge yet, nor a cooler, no sign it was running anywhere and pre-production memory on it, all of which do nothing to point to imminent release. AMD showed working cards fully 4 months earlier than Nvidia... but every sign points to Nvidia releasing first. It comes across as a Nvidia PR puff piece.

The only indication Nvidia will 'launch' first is because what Nvidia is calling a Tesla P100 launch is what AMD calls normal sampling of products before release to customers. Nvidia paper launched and is pretending the normal 3-9 month sampling period is a real launch.

Weren't you telling me some months back how AMD were like 6 months ahead of Nvidia and they wouldn't be able to put anything out til like the end the year while AMD would have **** out by early summer?

No wonder you're pushing back against this. All that posturing would look mighty foolish at this point if that weren't all true.

Anyways, this is no Nvidia 'puff piece'. Digital Foundry have no bone in this fight whatsoever. Your paranoia in thinking that's what is happening speaks more than anything. At worst, Richard here just turns out to be wrong and is basing his information on faulty sources. It happens to all of us. But your attitude towards all this tells me you've got a dog in this fight, which makes it rich when you accuse somebody else of it simply because they are saying things you dont like.

 

[LoadsaMoney]

NVIDIA Pascal Editors’ Day To Take Place Next Week

http://videocardz.com/59459/nvidia-pascal-editors-day-next-week

 

[Seanspeed]

http://videocardz.com/59459/nvidia-pascal-editors-day-next-week

You're gonna give drunkenmaster a heart attack man, come on now. Give him time to relax before adding this to the fire.

 

[EL_JOCK]

Didn't NVidia stop producing cards recently? Some of the more popular cards too, ie mid range? Has to point to new cards incoming pretty soon, not Q1 2017 or they're going to lose out to AMD in that area once their stocks run out and they won't let that happpen.

They may have stopped but, if you have a say 1 million units worldwide knocking about, they may not want more knocking about if the sums tell you that stock will last until the new cards, trust me they would have done the sums best they can...

now im not saying it will be Q1 2017, but if the process is not ready to go, its not ready and while they may not want to let it happen they may have no choice, as AMD will not wait

 

[drunkenmaster]

Everyone in the industry knows AMD showed cards in December behind closed doors and slightly more publicly at CES in Jan... Nvidia showed samples made mid April in the past week or so. Anyone claiming to know a lot about the industry while saying everything points to Nvidia coming first is talking out of their backside.

In no way can showing their cards working 4 months before your competition shows bare bones unfinished cards interpretable and looking like launching later, none.

The same sources that are saying GP104 will be available in July were saying don't forget that Titan would be out in April. Everyone on here is suddenly convinced GP104 will come out in July when they were convinced of the entire consumer lineup being shown at GTC, Titan being launched in April and on and on, all proven to be completely false. But the same sources making up new clickbait articles about July launch are suddenly implicitly believable?

 

[muziqaz]

You're gonna give drunkenmaster a heart attack man, come on now. Give him time to relax before adding this to the fire.

I think you guys are already having some of that AMD had an Editorial last December

I am not saying nvidia will or will not release anything, but I'm with DM, AMD showed working stuff, while nvidia was completely silent, and all the events they attended they talked about Tesla.
Maybe nvidia is playing some game, maybe they are not. Unless someone from inside can say what is happening none of you guys have any leg to stand on, unless you believe all the echo chamber so called 'news'.

 

[Seanspeed]

Everyone in the industry knows AMD showed cards in December behind closed doors and slightly more publicly at CES in Jan... Nvidia showed samples made mid April in the past week or so. Anyone claiming to know a lot about the industry while saying everything points to Nvidia coming first is talking out of their backside.

In no way can showing their cards working 4 months before your competition shows bare bones unfinished cards interpretable and looking like launching later, none.

The same sources that are saying GP104 will be available in July were saying don't forget that Titan would be out in April. Everyone on here is suddenly convinced GP104 will come out in July when they were convinced of the entire consumer lineup being shown at GTC, Titan being launched in April and on and on, all proven to be completely false. But the same sources making up new clickbait articles about July launch are suddenly implicitly believable?

I'm not 'convinced' of anything. I take *everything* with a pinch of salt. I have no idea what will or will not happen. I'm not so arrogant to think that my predictions are worth so much more than anybody else's.

Truth is, none of us here, sitting with our thumbs twiddling, have ANY clue what is actually going on. We can make educated guesses based on experience, but in the end, companies change strategies all the time. There's a reason you cant reliably use a past precedent as a predictor and have it come true all the time.

Word to the wise - refrain from acting like you are sure of anything. You may get things right here and there and revel in your brilliance and gained notoriety among meaningless internet denizens, but you're also very likely to end up looking completely foolish at times, ruining the credibility you so obviously crave.

A man who says "I dont know" gains nothing, but they also dont lose anything either.

 

[Rroff]

I think you guys are already having some of that AMD had an Editorial last December

I am not saying nvidia will or will not release anything, but I'm with DM, AMD showed working stuff, while nvidia was completely silent, and all the events they attended they talked about Tesla.
Maybe nvidia is playing some game, maybe they are not. Unless someone from inside can say what is happening none of you guys have any leg to stand on, unless you believe all the echo chamber so called 'news'.

DM may be right but also AMD have typically been much more forward with working hardware and tend to do a lot more work using physical prototypes while these days nVidia seem to a lot more development using emulation, etc. and less use of actual hardware so it may not be that reliable an indicator of which will come to market first. Infact we may see a somewhat staggered release schedule where they are competing less than normal in any given tier.

 

[drunkenmaster]

I'm not 'convinced' of anything. I take *everything* with a pinch of salt. I have no idea what will or will not happen. I'm not so arrogant to think that my predictions are worth so much more than anybody else's.

Truth is, none of us here, sitting with our thumbs twiddling, have ANY clue what is actually going on. We can make educated guesses based on experience, but in the end, companies change strategies all the time. There's a reason you cant reliably use a past precedent as a predictor and have it come true all the time.

Word to the wise - refrain from acting like you are sure of anything. You may get things right here and there and revel in your brilliance and gained notoriety among meaningless internet denizens, but you're also very likely to end up looking completely foolish at times, ruining the credibility you so obviously crave.

A man who says "I dont know" gains nothing, but they also dont lose anything either.

You're entire post is based on assuming I said I was sure of anything, I said nothing of the kind.

I gave my opinion of a video and you jumped all over it with the usual "I'm not biased but here is some angry reaction to someone saying something about the brand I like" tripe and attacking my post.

The video said nothing new and took an extremely ridiculous pro Nvidia stance on which cards would come first.

In your totally unbiased state you said I thought I'd have a heart attack because Nvidia would finally show off the cards behind closed doors.... because you still don't have a clue what you're talking about.

As someone pointed out, AMD had a similar day in December, and hence the time scale that looks more likely from all information is that AMD would be first. Actually all information points to that and saying the exact opposite really does show an extremely biased point of view in that video. Having hardware and showing it 4 months before the other guys is an indication of one thing.

 

[drunkenmaster]

DM may be right but also AMD have typically been much more forward with working hardware and tend to do a lot more work using physical prototypes while these days nVidia seem to a lot more development using emulation, etc. and less use of actual hardware so it may not be that reliable an indicator of which will come to market first. Infact we may see a somewhat staggered release schedule where they are competing less than normal in any given tier.

It now costs more than ever to tape out a chip, and putting together a few pcbs once you have chips back is honestly extremely minor. In designing chips and working on drivers simulated hardware is key to everyone now, back when it was a few million transistors even the low hundreds of millions and they were drastically less complex well even layouts were done by hand. These days it just isn't feasible and simulation is a huge part of the process. But once you get physical hardware back there is a huge amount of work to be done going from first chips back to release. Qualification/certification can't be emulated or simulated, it takes time. Finding any errors, coming up with microcode fixes or finding out you need a respin all has to be done by checking the physical hardware.

There is precisely no way to go from emulated chips to first A1 chips back from fab and launching a few weeks later. Does it have to take as long as AMD have so far, maybe, maybe not. The more complex the chips and the process the longer everything is taking. The smaller chips get and the more transistors per chip the harder testing is even from a basic fault finding and fixing.

It doesn't mean Nvidia can't come first, AMD may find a fault need a respin and get put back 6 months, the fab might find a fault and need to fix something causing a large delay (Tsmc had a big problem with a nitrogen leak which tanked their increased capacity line for a few months, can't recall if that was 40nm or 28nm now). Every little bit of timing can't be predicted and delays can come up from anywhere. But in general if a chip is being shown 4 months earlier then it's likely got a significant lead in time to market.

 

[Seanspeed]

You're entire post is based on assuming I said I was sure of anything, I said nothing of the kind.

I gave my opinion of a video and you jumped all over it with the usual "I'm not biased but here is some angry reaction to someone saying something about the brand I like" tripe and attacking my post.

The video said nothing new and took an extremely ridiculous pro Nvidia stance on which cards would come first.

In your totally unbiased state you said I thought I'd have a heart attack because Nvidia would finally show off the cards behind closed doors.... because you still don't have a clue what you're talking about.

As someone pointed out, AMD had a similar day in December, and hence the time scale that looks more likely from all information is that AMD would be first. Actually all information points to that and saying the exact opposite really does show an extremely biased point of view in that video. Having hardware and showing it 4 months before the other guys is an indication of one thing.

Your desperation and paranoia in accusing everything and everyone else of being biased is really telling man. I've got no bone in this fight. And Digital Foundry dont, either. They've proudly been praising the 390 lately for how it has been beating the 970 in most of the latest big-name games. I also dont care either way. I am excited about the upcoming cards and am hoping to upgrade based on whoever provides the best value in the £250-300 range, whoever that is. I dont care which. I'll gladly sit here and give a speech about how dumb brand loyalty is if you need me to, cuz I believe all that 100%.

I just think people who seem to damn sure of themselves and make predictions based on what they seemingly *want* to believe, end up ruining their credibility more often than they bolster it. Whereas a person who just plays it cautious and doesn't try and read too much into anything probably just walks away integrity intact.

You do what you need to do, though. Maybe you're right man. I dont know. I'm not saying I disagree with you or that you're wrong. If you're right, freakin thumbs up bro. But you're playing a game of odds and I'm sorry if I prefer to just sit back and wait for certain evidence before I start to latch onto any certainty bandwagon.

 

[drunkenmaster]

Except once again, show me the prediction I made in the last few posts? None, that would be correct. So you've again gone off on another post about how I'm sure of this and that because you can't seem to comprehend something.

I'm arguing against that guy saying everything points to Nvidia coming first. This is entirely and completely separate from who will launch first, that is an entirely different subject.

ALl information, all of it, with no exception, points to AMD being first. They might not be, they could come out one year after Nvidia, I'm not claiming or predicting AMD are first. I'm saying the information we have all points to AMD being first so stating everything points to Nvidia being first is quite provably factually wrong.

AMD showed the first non working cores first, they showed the first working cards first, they showed a working private and public demo first, their cards were shown 4 months before Nvidia showed a non working card in a card that didn't look ready for shipping.

They showed a fake Pascal in the same event AMD showed a real Polaris in January. They 'launched' P100 while stating no real availability till 2017 and their one demo product had I think it was 8 Pascal cards in, none that we know were working and of those cards several had HBM1 in. So they were fake or even at that time when 'launching' they didn't even have 8 full Pascal's with HBM2 to put in the single demo(as in for show) system.

The last time Nvidia showed fake cards and 'launched' with no availability a high end card just coming up to a known AMD launch was Fermi. The industry says Samsung and Glofo are ahead on 14nm LPP compared to TSMC's 16nm ff+ process.

Every single thing we've heard since December quite clearly points towards AMD being months ahead of Nvidia, from the cards shown, to dates shown on cores that have been seen, to Nvidia faking cards and to previous history of faking cards of paper launching to appear ahead. Which part of any of this points to Nvidia launching real cards in real volume first?

They still might launch first, once again, that is entirely separate. Of all the information available AMD looks months ahead so if someone does a video and claims that not one or two recent rumours but 'all information' points to Nvidia coming first I will call them on that information. He could have said "I have a source that says Nvidia is coming first", I wouldn't argue with that because I couldn't. But he claimed that everything points to Nvidia being first, it's such complete BS that I can only assume it's a PR puff piece.

As for recommending 390's, seriously people need to learn about the media. Look up the issues on youtube with paid reviews on a 'normal' site and reviewers getting crap for not specifically stating they were paid for said review. A website can do 100 'normal' reviews on their own then MS comes along and says review the Surface Pro for us, we'll pay you $1000 for every 100k hits for this review, then the guy suddenly does a completely absurd pro MS piece on the surface.

Media isn't one thing or another, it's not every single review or piece of information must be biased or completely believable. This is precisely the problem, trusted people (particularly on youtube) who out of the blue do insanely biased reviews. MS pay people to use their trust of their audience to sell products. It's dishonest, misleading and legally they are supposed to put a disclaimer up when they take money for such a review. AS such having one video recommending AMD and another video trashing AMD, or the same for any other company/product, is not only common it's fairly normal these days.

 

[bru]

I reckon that once one side sets an launch date, the other side will pip them at the post and launch a few days earlier.

As for the Fake P100 stuff, some one said NVidia had a DGX up and running at the launch event.

 

[drunkenmaster]

Just for reference for those who talk about both leaks and how Nvidia are always quiet and are seeing die shots and thinking about when that means a card launch. The first GM204 die shots had a production date of 1421, or mid May in 2014, Gm204 then launched 4 months later mid September.

So when dies are first seen and the first dates you see on them to launch tends to take a while. Also since AMD started talking about Polaris in Jan and Nvidia were silent this forum started to suggest a new idea that Nvidia never shows cores or talks early about cores and they just magically appear out of the blue, yet there we have both Gm204 and Gm200 die shots out months before release.


On DGX yeah, I've seen people reference a demo system, I can't find anything definitive about a live demo on any of the big sites. I wonder if it's a case of people not in attendance doing their own story and assuming other sites talking about a demo system meant working demo rather than just a box with cards in. It was a Quanta box I think that was on show at the event and that they didn't make an effort to show one running on stage in the presentation despite it being the only box on sale till Q1/17 hints that there wasn't one but hard to know. But in terms of fake I was more talking about the CES Pascal/drive px2 and they 'showed' a supposed GP100 with HBM2 at CES 2015... which was quite clearly not real though they claimed it was the first HBM/3d memory card being made at the time.

 

[eddyr]

http://techreport.com/review/30048/exploring-nvidia-pascal-architecture

A dive into Nvidia's Pascal (P100) whitepaper and commentary by Rys Sommefeldt.

 

[CAT-THE-FIFTH]

On AT forums,a poster mentioned they saw the following on Chiphell forums:

1. GTX1080 GP104-400-A1 DDR5X 8G
2. GTX1070 GP104-200-A1 DDR5 8G
3. GTX1060 GP104-150-A1 DDR5 6G

Nvidia is using a 310MM2+ part for the GTX1060??

 

[TheRealDeal]

On AT forums,a poster mentioned they saw the following on Chiphell forums:

1. GTX1080 GP104-400-A1 DDR5X 8G
2. GTX1070 GP104-200-A1 DDR5 8G
3. GTX1060 GP104-150-A1 DDR5 6G

Nvidia is using a 310MM2+ part for the GTX1060??

They did this with the gk104 and gtx680/670/660ti. I think gk104 was also a similar sized chip if not slightly smaller. We could be in for a very similar release to Gk104 from nvidia. That's no bad thing it was a decent chip.