-
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 **
- Thread starter Boomstick777
- Start date
HBM2 will scale upto 32GB RAM within 2016. What actually comes on Pascal is another matter.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
No, there is no chance of a April release unless it's a paper launch and not available for months.
Far too many times, well, read the semiaccurate article. One guy says something on a forum, one of the usual suspects puts it up as a 'story' and every other questionable tech news site just repeats the same 'story' as if it's fact.
There is nothing at all, no news whatsoever that points to an April release. The only sites pointing to April release are using completely incorrect information.
I can see the new Quadros and Teslas easily using the Max available for HBM2 though.
I kinda doubt we'd see that much on the new Titan graphics card. 16GB possibly, but with HBM2's bandwidth would they even bother with that much again? Could just go 12GB again.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
As far as I know Samsung has only stated they will start to produce 8hi stacks in 2016. I wouldn't be surprised to see 4GB stacks only in 2016, production to start on 8hi stacks maybe Q3/Q4, availability only coming in 2017.
Samsung are supposedly ahead of schedule, if we take their word for it. I can't see 32GB cards in April though - even Samsung's roadmap didn't have 16GB until the end of the year originally and 2-4GB for the first half of the year.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
They'll want a reason to get more money out of their customers, the only differentiator between Titan and x80ti is memory capacity. It's very likely Titan will have more memory than the x80ti and there isn't an easy way to do it other than using double capacity stacks.
Associate
I don't think we will see any top tier cards in 2016 from either side IMHO. What with the delays in manufacture and the probable low yield of the HBM2 chips. This year will be the year of Low (GDDR5X), High-Mid (4 - 8GB HBM/2) cards that will hopefully surpass last generations top tier or more than likely equal them, albeit with lower power consumption levels.
The big guns will come out when HBM2 (8-16GB) are pumping out on a more reugular basis. Probably end of Q4, or Q1 2017.
Just my guess
The big guns will come out when HBM2 (8-16GB) are pumping out on a more reugular basis. Probably end of Q4, or Q1 2017.
Just my guess
Don
Interesting article and probably more factual than any of the other rumours so far. Whilst at time's Charlie appears biased, I don't think anyone can question his journalism - at least his stories are based in some kind of fact, rather than reposting of the same rumours without any investigation.
I was being a bit tongue in cheek. But that doesn't detract from the fact that you are correct
Soldato
I was hoping this would come out early this year, guess i'll have to rethink stuck with my 680gtx at the moment and it's getting a bit long in the tooth. Was going to pick up a 980TI but didn't want to at the end of it's life span. Arghhh decisions!
Last edited:
Can't agree with his "unquestionable" aspect though - the only thing it proves is nVidia ordered some testing hardware with no indication of what they were testing and it certainly isn't the first time they've made such orders recently.
He could be right and its first round of testing on Pascal, it could be replacement parts of stuff that has been used/damaged in testing, it could be for testing a 2nd revision of the product or something completely unrelated.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,342
With the money that nVidia can throw at this particular 'problem', I can't imagine that their will be much in it tbh. Maybe a month after AMD
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
If it's for testing the 2nd revision... it would indicate a second revision was needed, which is really no different to if it was the first revision.
The important part is that ALL of the current "it's coming in April, woo" rumours are based off reading that parts list incorrectly and presuming that is working silicon being shipped back to Nvidia. It is categorically not, more to the point the site that constantly trawls through Zubmaubua whatever it's called looking for shipments didn't spot these shipments before so it's fairly likely to be the first set of testing kit. Because frankly once you see it you simply search on their database and can see if they've made a shipment with similar part numbers/descriptions before.
The more important part of the article is highlighting that absolutely all of the existing articles using that as a basis for their rumours are talking out of their backside.
Nvidia themselves have by way of lying on stage about having Pascal but showing Maxwell, pretty much confirmed that by CES they didn't have any Pascal silicon back from TSMC.
The excuse used that Nvidia wouldn't have spare cores to waste for trade shows is only used by those who have no idea how production works. If Nvidia had working silicon they had most likely hundreds of non working chips back. No one gets 100% yield, even with a single wafer back (which is very impractical and unlikely) you'd have probably a couple of dozen dead cores perfect for show and tell without fear or destroying working cards at trade shows.
The important part is that ALL of the current "it's coming in April, woo" rumours are based off reading that parts list incorrectly and presuming that is working silicon being shipped back to Nvidia. It is categorically not, more to the point the site that constantly trawls through Zubmaubua whatever it's called looking for shipments didn't spot these shipments before so it's fairly likely to be the first set of testing kit. Because frankly once you see it you simply search on their database and can see if they've made a shipment with similar part numbers/descriptions before.
The more important part of the article is highlighting that absolutely all of the existing articles using that as a basis for their rumours are talking out of their backside.
Nvidia themselves have by way of lying on stage about having Pascal but showing Maxwell, pretty much confirmed that by CES they didn't have any Pascal silicon back from TSMC.
The excuse used that Nvidia wouldn't have spare cores to waste for trade shows is only used by those who have no idea how production works. If Nvidia had working silicon they had most likely hundreds of non working chips back. No one gets 100% yield, even with a single wafer back (which is very impractical and unlikely) you'd have probably a couple of dozen dead cores perfect for show and tell without fear or destroying working cards at trade shows.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
It's not a money problem, for 2 years the entire silicon industry has had TSMC 1-2 quarters behind Samsung at 14nm...... the 'money' required to throw at such a problem is in the region of billions, not millions. Nvidia have no ability to accelerate a processes yield/production capabilities.
AMD got silicon back because it was produced on a different process, Nvidia seemingly haven't (or hadn't by lets say Xmas) had silicon back because they are producing at TSMC.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,342
Fairy muff
I was so used to both of them using the same foundry that I'd forgotten about Samsung/GloFlo 14nm, d'oh!!
Total rubbish, his stories is based of the same info as everyone else's, he has just interpreted it differently.
He may be right, but he may just as easily be wrong.
All of the rumours and synopsis from all the various sites and forum dwellers like all of us here, are guessing, we are going by the same piece of shipping info and interpreting it differently.
Myself, I would love to see a Pascal titan shown and released (paper release ?) at GTC, is that possible, well of course it is, is it likely, who knows for sure. Not long to wait and see.
Charlie confirms everything I've been saying (no big Pascal, tapeout rumour was fake, tech sites all copying each other, shipping data = bunk), this is my victory lap.
Everyone will try and shove this under the rug and not give me any credit.
Everyone will try and shove this under the rug and not give me any credit.
Last edited: