At CES NVidia said very officially Pascal for fourth quarter of 2016 for partners and early access partners second quarter which is where the April figure came from. Although to be fair to NVidia end of June is still 2nd quarter. This was in reference to Automotive partners.
-
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
Soldato
- Joined
- 19 Oct 2008
- Posts
- 5,951
Given that stocks of Titan X are running low at OCUK at least and with only a few with an ETA and the OEM(reference) 980 Ti's now uncommon it seems, I reckon new cards are imininininent. Announcement by end of Feb
Last edited:
Yep, nvidia officially announced publicly that automotive partners will get Pascal based Drive PX2in April, with volume production for 2018 model year cars in q4.
That is a plain fact. The only question is if a nvidia will be late, but you would think if they can't make April they wouldn't publicly announce an April release just 3-4 months earlier, they would know by then.
Soldato
By the sounds of it, that would more than likely be a situation with early boards for the automotive partners to test/configure for their car offerings. Considering the volume production for drivePX would be in 2017 for the 2018 car launches.
Yes, most likely a special test board with a lot more I/O ports and supporting infrastructure for diagnosis and testing, but final silicon.
The volume production is forecast in Q4 2016 such that partners have quantity Q1 2017 to put into cars that come out later that year.
The DrivePx units are low priority because the auto industry moves slowly on a year to year release, letting nvidia prioritize their TSMC contract for either compute or gaming products.
All this talk of takeouts and TSMC delays are fairly irrelevant to the actually release date I suspect. The 2 biggest factors are HBM2 availability and market conditions. The latest rumours indicated nvidia would have priority but that might be for a limited quantity at high prices that can be used in the compute parts where margins are high anyway.and as it stands nvidia aren't really facing any competition to get them worried enough to push for an early launch. They won't want to launch any earlier than Polaris.
And I strongly believe Pottsey is right in regards to test chips made on smaller wafers. When I worked at Sun they had a museum and on display it had early test wafers of some SPARC chips alongside the final production wafer which was much larger. Similarly I had a project with Logitech and they brought in a sample wafer in it looked tiny, I asked about it and they said the production wafers much bigger.
300mm wafers are very expensive to produce, so the likes of TSMC themselves don't want to be testing on large wafers, nor do the customers.
But that is just one possible reason why nvidia isn't show any pascal silicon, there are literally dozens of perfectly logical reasons. The reason might be the same as why AMD didn't show final silicon.
300mm wafers are very expensive to produce, so the likes of TSMC themselves don't want to be testing on large wafers, nor do the customers.
But that is just one possible reason why nvidia isn't show any pascal silicon, there are literally dozens of perfectly logical reasons. The reason might be the same as why AMD didn't show final silicon.
Soldato
- Joined
- 11 Mar 2013
- Posts
- 5,451
The whole Nvidia range is much smaller then a month ago. This is always a sign of new range coming in. They did the same thing right before 900 series.
Associate
- Joined
- 1 Aug 2009
- Posts
- 1,625
Waits impatiently with baited breath for concreate news
Soldato
For pattern testing you only need single run 200mm wafer fabrication equipment. But if you worked at sun a few years ago then 100 - 150mm wafers were the norm for mass production in the 80's. 200mm wafers came out in the 90's with 300 in early 00's.
Apparently 450mm is the next big thing in terms of wafers. with fabs gearing up to use them in the next few years. And if you look at pics, 450mm wafers are ****in huge. I would not want to accidentally drop one.
IBM seem to do this often for experimental chips. using 200mm wafers. their graphene transistors are a good example http://www.extremetech.com/extreme/...00-times-faster-using-standard-cmos-processes
Last edited:
Man of Honour
Judging by the price of some of the GTX 980 Ti's I don't think there is anything new from NVidia coming.
Check out the prices of the Asus Matrix and Gold editions for example.
Soldato
Soldato
- Joined
- 19 Oct 2008
- Posts
- 5,951
True, but not sure if that's unusual just before the release of new cards? Pricing doesn't come down?
Keeping fingers crossed. I want to do some gaming again and after owning a Titan X previously I want nothing less (performance wise) but don't want to buy another Titan X new again either so want something a tad faster for approx same price or less.
Keeping fingers crossed. I want to do some gaming again and after owning a Titan X previously I want nothing less (performance wise) but don't want to buy another Titan X new again either so want something a tad faster for approx same price or less.
I think the GP104 will be coming but the 980Ti will stick around a long while yet.
Hell it was probably planned that way all along if you look at 600/700/900.
Soldato
- Joined
- 11 Mar 2013
- Posts
- 5,451
The 700 series didn't move on price right before 900 series and Nvidia still released
Man of Honour
It is not just price, it is the number of Matrix and Gold editions OcUK have in stock.
Soldato
- Joined
- 11 Mar 2013
- Posts
- 5,451
It's all been very quiet on the pascal front.
Are you thinking that full-fat GP104 will slot in above or below the 980ti?
GM104 (980) is about 25% less perf than the 980ti, according to user benchmarks. (e: 980+30%=ti, ti-25%=980)
Depending on how much of an improvement Pascal is over Maxwell, retaining the 980ti could leave you with two products in almost the same perf ballpark.
If GP104 is 50% faster than GM104, that still only amounts to 13% faster than a 980ti.
GP104 would need to be 73% faster than GM104 to be 30% faster than a 980ti.
A 60% improvement over Maxwell would result in being 20% faster than the ti.
But even then you have a cheap to produce, small chip being priced higher than a costly to produce, large chip.
It's the same story with Polaris and Fiji. It doesn't seem to make much sense to retain the largest chips from the current gen.
Last edited:
I changed my mind on replacing my SLI 780Tis, primarily as the latest wave of newer titles have had pretty poor SLI scaling/support. That and the 3GB of VRAM is starting to show it's limitations at 1440P with higher settings.
I've just gone for a 980Ti which I will clock the nuts off under water and will likely skip the first release of pascal, which still seems to be a large unknown.
I've just gone for a 980Ti which I will clock the nuts off under water and will likely skip the first release of pascal, which still seems to be a large unknown.
Soldato
If I was on SLI 780ti's ill be holding off for pascal. I upgraded from a couple of poor clocking 780 non ti to average clocking 980ti sli (1450mhz range) and I say it was a good upgrade as a single OCed 980ti was not far off my old 780 SLI setup.
That and the 980ti landed when I was playing the witcher 3 which I was looking forward to for so long and wanted to play it maxed out at 1440p with a solid 60fps, something I couldn't do with my 780's or a single 980ti without drops.
Its likely I will wait untill there's full fat GP100 card if a GP104 card is only 20% or so faster than a 980ti etc but we shall see