Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016
New Posts  All Forums:Forum Nav:

# [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016 - Page 17

Rumor mills also had GP100 at 450mm², so the math checks out thus far.

(although part of me strongly suspects they crunched the numbers just like you did, then simply threw out a random number within that range so they'd be "in the ballpark" regardless)
Quote:
Originally Posted by Shandl4r

Let's do a pie in the sky Fermi estimation here.

Let's take the Titan X, full GM200.

601mm^2 and 8 billion transistors on the 28nm process.

Now we know the '28nm' and '16nm' are not actually that size, but let's assume the processes are the same relative shrink.

So the full 28nm GPU is capable of 13.31 million transistors per mm^2. Square root of 13.31 = 3648 transistors in a linear mm.

A relative die shrink of 16/28 would increase that to 3648 * 28 / 16 = 6385 transistors per mm

6385^2 = 40.76 million transistors per mm^2 on 16nm ff+.

That estimation matches the rumor mill of 17B transistors for GP100. 17000 / 40.76 = 417 mm^2. A perfectly reasonable size die for a 'big pascal' chip on a new immature process.

So 550mm^2 would get us a whopping 22.42 billion transistors on 16ff+. Now we have to assume per transistor performance losses compared to the Titan X due to FP32 being put back into the architecture. So comparing to the GK110 instead that is 316% of a Titan Black.

So a safe 3x Titan Black performance. Possibly 3.2x. A very comfortable double the performance of a Titan X, and likely more like 215% that of a Titan X.

Exactly like that was what I was looking for

My hopes are very high for Polaris/Pascal, because I think it is reasonable to expect big out of a memory change, architecture change and 14/16nm.
 current rig (4 items)
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja
 current rig (4 items)
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja
Considering that the evolution from maxwell to pascal will be bigger that was from kepler to maxwell, than is almost sure that the sucessor of the gtx 970, the gtx 1070, will have a performance equal or greater than a gtx 980 ti.
Remember: The gtx 970 stock performance is just a bit smaller than a 780 ti, think now about the new pascal gtx 1070...
I'm really considering selling my gtx 970 soon, because when the new tecnology arrive, my card's price will be halved!
Edited by svictorcc - 2/7/16 at 5:48pm
 Akvo Core V1 (11 items)
 Akvo Core V1 (11 items)
Considering 670 (cut down GK104) was anywhere from 10-20% faster than 580 (full GF110), "GTX 1070" needs to be at a bare minimum 10-20% faster than 980 Ti to be considered acceptable.
Quote:
Originally Posted by magnek

Considering 670 (cut down GK104) was anywhere from 10-20% faster than 580 (full GF110), "GTX 1070" needs to be at a bare minimum 10-20% faster than 980 Ti to be considered acceptable.

Absolutely and the GTX 670 was a cut down 294mm² chip. Not all cores enabled. Compared to the 520 or 550mm² (can't remember) GTX 580.

I actually expect the "GTX 1070" to be 25% better than the 980 Ti (stock) atleast. Depending on release time I'd consider anything less a huge failure.
 current rig (4 items)
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja
 current rig (4 items)
CPUGraphicsGraphicsKeyboard
2500k R9 290X Tri-X GTX 680 Filco Majestouch 2 Brown Ninja
Quote:
Originally Posted by svictorcc

Considering that the evolution from maxwell to pascal will be bigger that was from kepler to maxwell, than is almost sure that the sucessor of the gtx 970, the gtx 1070, will have a performance equal or greater than a gtx 980 ti.
Remember: The gtx 970 stock performance is just a bit smaller than a 780 ti, think now about the new pascal gtx 1070...
I'm really considering selling my gtx 970 soon, because when the new tecnology arrive, my card's price will be halved!
have fun using iGPU for 5+months
also, when you buy 1070 hurry and sell it fast because it will also lose value, another 5+months on iGPU.
Pascal is just Maxwell die shrink with HBM + compute improvement, certainly not a bigger change than kepler to maxwell. Heck, Pascal didn't even exist a couple of years back.

The next big change will be Volta.

As to how well the 1070 will perform, let's see what clockspeeds nvidia can wring out of 16nm FF+.
 living on the edge BSODs galore (13 items)
CPUMotherboardGraphicsRAM
e2140@3.2Ghz abit IP35-E HIS IceQ4 4850 4GB 667@800 5-5-5-15
OSPower
win xp 32 bit Corsair 450VX@stock
 living on the edge BSODs galore (13 items)
CPUMotherboardGraphicsRAM
e2140@3.2Ghz abit IP35-E HIS IceQ4 4850 4GB 667@800 5-5-5-15
OSPower
win xp 32 bit Corsair 450VX@stock
Quote:
Originally Posted by Glottis

have fun using iGPU for 5+months
also, when you buy 1070 hurry and sell it fast because it will also lose value, another 5+months on iGPU.

Unfortunately, yes you're right ;(
 Akvo Core V1 (11 items)
 Akvo Core V1 (11 items)
Quote:
Originally Posted by gamervivek

Pascal is just Maxwell die shrink with HBM + compute improvement, certainly not a bigger change than kepler to maxwell.
cause you know what performance improvement Pascal will offer. Please tell us more.

will moving to 16nm allow significantly higher clock speeds?
Edited by Klocek001 - 2/8/16 at 4:17am
if a single card can take my two 980s I will buy two of them
 In Win 904 (18 items) Corsair 250D - mATX (13 items)
 In Win 904 (18 items) Corsair 250D - mATX (13 items)
New Posts  All Forums:Forum Nav:
Return Home
Back to Forum: Rumors and Unconfirmed Articles
• [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016