Nvidia Pascal Speculation Thread

Status
Not open for further replies.
... any post-earthquake statement...?

I'm not worried about those at all, since according to TSMC's own statements the "boutique" fabs (where they manufacture the expensive stuff) have completely recovered. I'm rather worried if possible demand from IHVs could be higher than the actual capacities.

***edit:
https://translate.google.de/transla...a-geforce-pascal-till-computex-2016&sandbox=1

smells like a possible paperlaunch, my gut feeling doesn't indicate though if we'll have any woodscrew moments or not....
 
Last edited:
smells like a possible paperlaunch, my gut feeling doesn't indicate though if we'll have any woodscrew moments or not....
We already had a woodscrew-moment with Drive PX2 equipped with Maxwells instead of claimed Pascals
 
We already had a woodscrew-moment with Drive PX2 equipped with Maxwells instead of claimed Pascals

How sure are you really that you don't want more? C'mon admit it.... :yep2::yes:

j/k obviously
 
Last edited:
Question is: were those old chips actually 20nm Paxwells ?

Sorry for the weird idea, but given the length of time Drive PX prototypes must have been evolving in the lab, and the fact that JHH must have learned that no one responds well to wood screws, why screw around? Then there’s the fact that Drive PX comes from the same rabbit hole as Tegra, and that Tegra X1 was billed as the "20nm Super Chip". Makes sense that in-house eleves would want a few wafers of the original next gen arch, given the need to keep ahead in the trillion dollar auto-auto market.

Recall 2014, how excitement turned to frustration as Nvidia kept delaying their next big thing: a cool and sexy new arch, plus 20nm! Then the rumours that something worthwhile on 20nm might not show up until early 2015. And then rumours that some parts of the (mythical?) 20nm project might get retro-fitted into a 28nm gamers card; then along comes 28nm "mini-Maxwell" to help with revenue stream.


Doesn't this fit the 20nm timeline?

How likely do you think it is that NVIDIA cooked up 20nm "paxwell" in week 3 of 2015 with the exact same pcb and die size as GM204?

Good point; but if a form factor is optimized for memory modules of the day, what's to change?
 
From Sweclockers: "Nvidia Geforce "Pascal" for laptops released in mid-June" (original).

Sweclockers has previously been able to uncover that Nvidia is planning to announce Pascal for laptops at Computex 2016 , which takes place May 31 The sharp release, however, is thought to later and allegedly to the editorial, Nvidia announced June 16 to their partner manufacturers.
Since Fermi, we've seen chips from xx4 through xx8 and xx9 (plus one highly cut-down GF100) in the laptop space, so a priori the "Pascal for laptops" could be almost any chip(s). But my guess is that we'll see a low-end Pascal or two to replace the majority of mobile GM107 and GM108 parts.
 
Last edited by a moderator:
BenchLife claims that a Pascal "GTX 1080" will launch in May.

The new generation of NVIDIA GPU codenamed Pascal degree will be center stage in the first half of 2016, which GP104, which is GeForce GTX 1080 will be determined debut in May.
[…]
Although the news has been mentioned, Pascal will import HBM2 memory, but the latest data show, GP104 is the GeForce GTX 1080 This card will remain GDDR5, or a faster GDDR5X memory, and memory capacity is 8GB in size.
 
In a way, that would be a little surprising: didn't some official AMD hack claim that they were months ahead of Nvidia with next-gen process silicon?
On the other hand, it would be totally in character for them to Osborne their own product line (for no good reason) only to see the competition pass them when it matters.

Maybe the mid-year Polaris pronouncements were smoke and mirrors, and it will be introduced soon...

Either way, it's about time silly season starts. Things have been too quiet lately.
 
In a way, that would be a little surprising: didn't some official AMD hack claim that they were months ahead of Nvidia with next-gen process silicon?

http://arstechnica.com/gaming/2016/01/amd-confirms-high-end-polaris-gpu-in-development-for-2016/

"We believe we're several months ahead of this transition, especially for the notebook and the mainstream market," said Koduri. "The competition is talking about chips for cars and stuff, but not the mainstream market."

This was shaded to focus on the product bands that Polaris 10 and 11 go into, which a "GTX 1080" slots higher in the stack and lower in quantity.
It may also depend on what is in quotes about the claimed Pascal schedule. Is it "'GTX 1080' will launch in May", or "'GTX 1080' will 'launch' in May" (or "'GTX 1080 will 'launch' in 'May'", etc.)

AMD seems keen on getting things ready and in volume for back to school device purchases.

On the other hand, it would be totally in character for them to Osborne their own product line (for no good reason) only to see the competition pass them when it matters.
If the focus is in mobile and mainstream, it's an uphill climb already in terms of market and mindshare. So it might be a proportionally smaller Osborne-ing.
 
So much for 'hack'. :) It sounded like somebody with foot in mouth disease like Roy Taylor would say!

This was shaded to focus on the product bands that Polaris 10 and 11 go into, which a "GTX 1080" slots higher in the stack and lower in quantity.
If that was the focus, he definitely went out of his way to not make that subtle distinction.

AMD seems keen on getting things ready and in volume for back to school device purchases.
The thing that surprises me more is not that Nvidia would have something available by mid year, but that AMD was so conservative with their Polaris availability, showing samples to the press in December and all that.

If the focus is in mobile and mainstream, it's an uphill climb already in terms of market and mindshare. So it might be a proportionally smaller Osborne-ing.
A Polaris 11 should be able to challenge most of their current product line, all the way up to a 390X, in terms of performance, right?
 
Either way, it's about time silly season starts. Things have been too quiet lately.
We haven't had a good silly season in years now, and still both AMD and NV are closed up tight like clams without giving us anything substantial. One can only hope there will be some kind of infodump come GDC...
 
BenchLife claims that a Pascal "GTX 1080" will launch in May.

I kinda wanna save myself for the HBM2 enabled big boy. But if this (or more accurately it's GTX 1070 sibling) can outperform a 980Ti and is priced reasonably, I'll be hugely tempted... Whatever I do, I need a nice shiny new monster PC before my OR headset arrives in July!
 
A Polaris 11 should be able to challenge most of their current product line, all the way up to a 390X, in terms of performance, right?
It could threaten that portion of the stack, and possibly higher. AMD could opt to hamstring it in some way so that it chokes at the "4K" marketing line.
The affordable VR bullet point puts Polaris (presumably 11) in the range of the 290X. Officially, that product has been eclipsed by the 390 and Fury lines, which take up 3-4 marketing diagram boxes above the 290X, even if it might be tens of percent difference in many situations.
 
They also, once again, mix GDDR5X into the rumour, even though only GDDR5X manufacturer has said that mass production will start "during summer" (and once they specified August)
I don't think that's a major problem: introduce a 1070 with GDDR5 first, add a 1080 with GDDR5X a few months later.
 
I don't know if the company has commented on this or not, but I don't think nVidia is going to call Pascal the 1000 series because they never progressed through a change in decade when launching a new family.
nVidia did Geforce 5000 - 9000 and then went to Geforce 100-900. Pascal is probably getting a new suffix and do the hundreds again or perhaps tens. Something like Geforce X180 for GP104 and Geforce X160 for GP106.
 
Status
Not open for further replies.
Back
Top