PASCAL SPECSHEET IS OUT
HBM2 CONFIRMED
>>53648252
can't wait to see these likely fake, inflated specs btfo by polaris
>>53648252
>titan
Probably gonna be release 2017
>>53648252
>make fake excel spreadsheet for /g/
>write "nvidia confidential" on the bottom to make it look "real"
lel
>>53648252
>16GB
For what purpose.
>>53648457
Cheap to add, makes amd cucks look retarded after MUH VRAM in the old 970 vs 390 debate
>>53648252
>Titan
>16GB HBM2
That's going to cost $3500 a pop
I don't think nVidia can handle it. Yet.
>>53648510
Price is $1100. Huge capacities are cheap with HBM2
>>53648252
>HBM2 CONFIRMED
On the $2000 Titan. Everything else uses GDDR5.
>>53648528
>implying nVidia isn't going to do it nice and slow
They'll start at 8
16GB of HBM2 is correct
Each HBM2 stack has 4GB
4x4 is 16GB
OP's image is fake though, no GPU will have HBM2 and GDDR5 support
>>53648252
>hbm2 confirmed
so you mean delay confirmed lmao
>>53648252
>1080ti now finally has caught up to the memory bandwidth of the 290x
>>53648528
Stop being a retard. Guaranteed to be >$3000.
>>53648805
U WOT?
290X only had 320GB of bandwidth despite having a 512bit memory bus, retard
Even a 980Ti has 337GB of bandwidth, more than any crappy 290X
>>53648252
>16GB of HBM2 memory
Why? They could have gone with 8 and given the other 8 to the X80Ti.
>>53648899
the only hbm dies being produced currently are 4gb dies from samsung,you need at least 4 4-high stacks to achieve the advertised bandwidth. once hbm2 is matured samsung is going to produce 8-high stacks for 32gb in total, which is going to go into pascal quadro/tesla cards.
if you disable half the chips per stack, or half of the stacks, you get half the bandwidth available to the card. it'd be on par with gddr5 at that point for a much higher cost.
>>53648528
>Huge capacities are cheap with HBM
no it isn't, the more chips you add the more expensive HBM becomes. high capacity would be far cheaper if you're buying gddr5, hbm's advantage is superior performance.
Nvidia : Pascal Is 10x Faster Than Maxwell
Nvidia : Pascal Is 10x Faster Than Maxwell
Nvidia : Pascal Is 10x Faster Than Maxwell
Nvidia : Pascal Is 10x Faster Than Maxwell,
Why did they even call it HBM, "High" is a relative term, once it gets replaced with a newer faster standard it won't be so "High Bandwidth" anymore.
>>53648586
only sane response
>>53649166
it's already happening. HMC will make HBM look like a joke.
>>53649070
So we will get useless 32GB graphics cards before we get useful 8GB graphics cards with the same bandwidth?
>>53649131
(In compute tasks)
>More shader cores on Titan
When has Nvidia ever done this?
When will a titan ever be more than a ti with double the vram, and coming out slightly earlier?
>>53648878
>Bus width and bandwidth are totally the same goy!
>>53649662
Considering it is the same chip count on disabled shaders on the Ti.
>>53649687
Nvidia has never done this before for a titan vs a ti. I'de like to see your sources on this. It's always been just double vram, and sometimes lower clock speed on titans.
What if they add 32gb vram and let load whole game into graphic card?
Will it improve fps?
>>53649765
No 32GB cards except for Tesla & Quadro
Titan Pascal will get 16GB and so will Pascal GTX Ti card, Titan Pascal will have the FP64 switch in the control panel just like the TITAN/TITAN Black for those wanting FP64 performance
>>53648553
>implying
They'll start at 6.5
>>53649844