$2,700? Hope you all have fun with that.

they maybe also bringing out a dual titan
AMD has fury X and fury X2(dual card)
 
what i have been seeing is hints of a dual Nvidia card for VR use, hi end enthusiast and desktop research configs
VR almost requires dual cards or dual chip card
 
thats if you want to minimize headaches and eyestrain
 
from Nvidia "Pascal will also be available in multi-GPU packaging"
 
 
-------------------------------------------
Taken from a source - take with salt
 
price 899-1000$ usa april-may 2016
GTX 1080 Engine Specs:
2-2.5 x at 4k 980gtx
4096 cuda cores
1950 base clock mhz
2000 boost clock mhz
450 Texture Fill Rate (GigaTexels/sec)
8gb 
185 watt tdp
3DMark Fire Strike Ultra 7000-8000
50- 80 % faster than 980 ti
this goes out first beating amd to the market
 
Gtx titanx 2 dual variant
4-6x gtx980 4k
8192 cuda cores
1750 base clock mhz
1900 boost clock
900 Texture Fill Rate (GigaTexels/sec)
16 gb
350watt tdp
3DMark Fire Strike Ultra 18,000-20000
this takes the crown from fury x2
 
water cooled april-may 2017
1799$+-
Gtx titanx 2x ti dual variant full die
6-8x gtx 980 4k
12,288 cuda cores
1850 base clock mhz
1900 boost clock
1.35 Texture Fill Rate (TeraTexels/sec)
32gb 4096bit 16 x 2
450watt tdp
3DMark Fire Strike Ultra 24,000 +

NVIDIA could sell GeForce for max 1000$, everything more and number of customers will drop more and more.
Difference between 1000 and 1200$ could reject more customers than you think...
1000$ price people could connect easy with 900-950$ and to search for sellers who will offer such price.
But for 1200$ many people will instantly wait AMD on first place to drop price of NVIDIA cards.
You remember what happen with TITAN Z. They didn't sold cards enough to cover marketing and production...
On biggest IT Forums where very rich enthusiasts spend a lot of time I didn't saw 5 people with TITAN Z... 
I can't pay more than 1000$ as well... I'm that's 1100$ I thought on 1000-1050 euro example.
What EVGA plan to offer as variants of reference model GP100/full potential... SC? Hybrid? HydroCopper? FTW?
That would be interesting.

What if AMD raise prices too?
rumor is that Silicon is more expensive with the extra ram and the HBM let alone the die shrink on the GPU (new node is usually more expensive)
could be as much as 30% increase
 
do you think you are gonna get cheaper prices when they are trying to push the technology to keep up with 4k, 8k,12k and 16k TV or monitors coming down the road let alone Holographic TV and VR?
you do realize that going from 2k to 4k quadriples the number of pixels -so look at the progression to 16k
and VR has motion sickness issues - some due to latency lag
 
I had read a paper from Nvidia PDF where they want GPUs to get 100's of  times more powerful in a handful of years just to keep up with lcd density and VR
They are also looking to partner with software and hardware companies because they need efficiencies in software designs - games , OS's, even other tech hardware companies to handle the new hardware
 
because we are gonna hit walls in cable bandwidth,noise,latency,cpu and gpu power,Heat,Motherboard bandwidth(hence NVlink)
 

A $1700 AMD graphics board has just been spotted, potentially the next Radeon 400 series Polaris flagship.

Read more: http://wccftech.com/amd-flagship-polaris-spotted/#ixzz44azaA717 
 
The C98 graphics board made its first appearance on the Zauba shipping database in December. This particular part holds a per unit value that exceeds AMD’s Hawaii chips that were listed in the same database by 33%


 
so like yeah silicon isn't cheap
or they are overinflatting value for insurance puposes

Xavier Zepherious
A $1700 AMD graphics board has just been spotted, potentially the next Radeon 400 series Polaris flagship.

Read more: http://wccftech.com/amd-flagship-polaris-spotted/#ixzz44azaA717 
 
The C98 graphics board made its first appearance on the Zauba shipping database in December. This particular part holds a per unit value that exceeds AMD’s Hawaii chips that were listed in the same database by 33%


 
so like yeah silicon isn't cheap
or they are overinflatting value for insurance puposes

"Just been spotted" over a month and a half ago?

That was February 9th.

Please, everyone, continue with the overpriced speculations just like everyone did when the 970/980 and Titan X was released. We need everyone saying they won't buy due to the price again.. It made such a huge impact last go around and didn't change release prices at all.

New video cards get more and more transistors and more and more RAM every generation though.  We doubled the RAM on the 780 Ti from 3GB to 6GB on the 980 Ti, and the 980 Ti actually cost LESS.
 

It most likely will be in the same range I think, maybe slightly more, that's not out of the question.  I just don't think it will be THAT much more you know?  Especially since these are just going to be the 70/80 models and not the big boy Tis or Titans.  No telling how much the new Titan with HBM2 will be, that might be more.  Hopefully they try and at least stick with just a $1000 price tag (*just* lol).
 
I'm going to try and skip this generation, or at the very least hold out for the Ti version this time.

I'm just curious if the GTX 1060 will go back to a 256bit bus and be a good upgrade from my GTX 960 2gb

ManBearPig
Remember when we had 512bit bus?  Is there some reason why Nvidia dropped 512bit?

The only thing I remember was the excuse that it was overkill and there was no way a video board could saturate that much bandwidth. What do I know? I'd prefer to see a wider bus myself even if it's just a placebo effect.

ManBearPig
 
That is really funny if that was their actual excuse.  No need to go to HBM then right? lol

I remember when ASIC wasn't important either so I stopped paying attention to it or measuring it. What do you know? Apparently it is suddenly important enough to bin the chips and charge more for a higher ASIC score. Harumph!

and a little bit of April 1st fun
Nvidia GTX 1090 Ti & AMD R9 495X Benchmarked & Reviewed – Pascal & Polaris Put To The Test




Read more: http://wccftech.com/nvidia-gtx-1090-ti-amd-r9-495x-benchmarks-review/#ixzz44bnp8Red
 

power and heat consumption - cards eat power - ram eat more
faster and more voltage to drive them causes more leakage and heat
and more failures due to pushing them hard
 
And in HPC with server racks full of them less power is MORE PREFERABLE
 
 
Again i refer to why NVidia published a PDF on this issue and why the switch to HBM this year and VOLTA which will have internal ram in next process node
 
 
 

ManBearPig

candle_86

ManBearPig
Remember when we had 512bit bus?  Is there some reason why Nvidia dropped 512bit?

Because GDDR5
 
GTX 980Ti 384 Bit GDDR5 @ 7000 = 336.5 MB/s
So why would we need a 512bit bus right now?
 
It saves money, a 512bit bus is alot of wires and alot of extra pins. I think 256/384bit is fine until HMB2 enters, then sure why not 4096bit bus since its all one package but until HBM gets on every card its cheaper to stick to lower bus width. 
So why would we need a 512bit bus right now? Though as shown below a GTX 980Ti with a 512bit bus would nearly catch the R9 Fury X in terms of bandwith

So if GDDR5 at 384 bit is more than enough, why pay more for HBM2 which I can only assume would cost even more than a 512 bit.  And sure a 512 bit cost more but they were always reserved for the flagship GPUs where people didn't care about price anyways.

It was enough for everything pre 4k, also HBM is a simpler board design by its very nature. Also consider HBM is clocked at 1000mhz, HBM2 im not so sure so it uses less power = more efficent 

sputnik7913
I figured it would be a nice place to dump all the rumors and stuff...

Rumors are not news.