I figured it would be a nice place to dump all the rumors and stuff and in general just to have a mature discussion on the upcoming 1000 series. Please post if you holding out on the current GPU upgrade in favor of Pascal... 
 
As a first rumor, I've heard that new Nvidia  NVLink technology won't be compatible with the current motherboards and we'll have to upgrade... Which is a bit of absurd in my book, many people have upgraded they're motherboards lately, I don't think Nvidia is that stupid to be limiting their Pascal sales...

I still don't 100% believe the hype of pascal, since NVidia bases everything on "deep learning" and just like every single other previous generation, they aren't going to shoot themselves in the foot. I feel like the gaming realm will be much slower than the tesla quadro realm.

I am part of the 2%... Wooo!

I put my vote at 24gb HBM2 on the next titan. Will have more than 2x the performance of titan x.

boylerya
I put my vote at 24gb HBM2 on the next titan. Will have more than 2x the performance of titan x.

Probably almost 2x performance. But for memory it's very hard to say now.
No doubt that would be amazing and expensive card but when you open box worth every dollar or euro.
 
 

Why? If you pay card 1000$ and use 2 years and sell for 400% that mean you use high end card 24 months for 600$.
Anyway nobody beat you to change every series... One card in generation, Kepler, Maxwell, Pascal and that's it, on 2-2.5 years.
Off course I would like to 1000$ cards cost 1000e in Europe, not 1300e... NVIDIA should think about us, poor souls in Europe...
I save 6-7 months money for such graphic card and for me worth to pay... What to tell than for people in USA...
If you put something on marketplace you always find buyers for older stuff, in East Europe is very hard to sell anything and again we manage somehow to get some nice hardware.
For people in USA TITAN X is same as for me 200-300$... Believe me I would change every year with such price high end GPU.... For 200-300e, 100%. 
 

We're not going to see a 24GB consumer card. There is no point. I'm willing to be the highest HBM card we'll see is 12-16GB for the consumer market and 24-32GB for professional series. Even then that is a ton of memory that cannot be saturated outside of use in VDI. Heck I'm currently using GRID K1/K2 cards here at work for My Xenserver with Xendesktop/Xenapp for VDI and I still get tons of density.
 
Getting a 16GB card for gaming is like throwing in 32GB of RAM. The only thing you've achieved is to pad your forum sig to flex that epeen.

have you seen the latest news about AMD line up
 
http://www.techrival.com/amd-greenland-gpu-may-feature-up-to-32-gb-hbm2-memory/ 
 
 AMD is planning to incorporate just as much HBM2 memory as NVIDIA’s professional and high-end solutions with up to 32 GB of VRAM
 
 
PS:GP100 maybe avail Q1 2016 but most likely Q2 - it's ahead of schedule they are ramping up production
from the look of it test silicon was good
 

I still see the NV link being used for research purposes... 8 cards for gaming.. Yeah, Nvidia cares that much... Not likely.

Scarlet-Tech
8 cards for gaming.. Yeah, Nvidia cares that much... about money.

fixed lol. While I agree that 8 cards for gaming is ridiculous and unneeded, I could actually see Nvidia doing it at some point simply because that means there will be people who will buy 8 Titan Nexts (whatever they will be called).

4 cards put out enough heat as is. Scared of the heat 8 would throw out. 

Scarlet-Tech

Sajin
4 cards put out enough heat as is. Scared of the heat 8 would throw out. 

Furnace, maybe?

 

Can't imagine how the scaling of eight cards would be in games. 

Xavier Zepherious
if the card is 1/3 the size you can fit double the cards in the same space - you just need a new interface - and why they are talking nvlink

 
the heat and power consumption going to be a nightmare -bigger PSU's
 
mind you i don't know anyone who would need anymore than a few cards - what you want 8k Gaming???
 

Yes, smaller means they will fit in tighter areas... But that requires more cpu lanes (dual cpu boards) more PCI lanes and a whole new tech to go with it. Tech that universities or research labs could work with, but not gamers.. At least not for now.

Psu's and heat isn't even in the realm of issues.. This would be more super computer directed I would think.. Those things produce some crazy temps from what I have read.. Really neat stuff.

boylerya

Xavier Zepherious
 
The report states that the GP100 ‘big’ pascal chip will hit the professional market first, giving time for green to roll out the GP104 chip to the mainstream consumer segment (followed by GP100 at a later date).  The GM200 taped out in June and hit the shelves in March. That is a total time of approximately 9 months. The GM204 on the other hand took only five months to hit the shelves from its tape out date. At this schedule we can tentatively expect the Pascal architecture GPUs next year in late second quarter. The GP100 chip will also have HBM2 memory, which unlike the HBM1 specification can extend upto 32 GB of total vRAM size.
The report states that the gaming edition of the GP100 will be limited to 16GB of vRAM (which frankly is pretty effing huge considering the current standards


Read more: http://wccftech.com/nvidia-gp100-chip-taped-gp-104-arriving-q2-2016/#ixzz3kbnRGHGB
 
You ask me 32GB of ram will be killer
 
and rumors tout 5000 cuda cores
 
be a good idea to read this
 
http://www.3dcenter.org/news/hbm2-speicher-deutlich-flexibler-bei-den-speichermengen-und-wahrscheinlich-breit-eingesetzt-bei 
 
 
PCIe4 Rev Coming
 
and rumors are you can NV link 8 cards together
 

NVLink running 8 cards for pascal is no rumor, nvidia already announced that as a fact. Hynix isnt giving nvidia their HBM2 until AMD takes theirs first. I get the feeling that GP104 will use HBM1 or GDDR5X so Hynix doesnt delay their release. But GP100 will def have HBM2 considering the longer timeframe to release compared to GP104, however Hynix can always play more games with NVIDIA to help AMD. Also, first release of PCIe 4.0 is on Skylake Xeon, hopefully Skylake-E as well but that is wishful thinking considering I have only seen official details with it being on Xeon.

Nvidia said they are using HBM2 for pascal. Hynix isnt the only one making hbm2, samsung will also make it starting 2016. So AMD will use Hynix and Nvidia can use Samsung.