Nvidia reveals key GP100 Pascal details at GTC 2016 (610mm2, 15.3B transistors, 300W TDP, 3584 cores, 1480MHz boost clock)

[Samfisher]

3 minutes ago, GidonsClaw said:

And like I replied... Ok... let's for arguments sake agree this is a binned GP100... possibly the next Ti?

I would think yeah, the P100 is looking like the Ti for Pascal.  However, we've had 2 proper Titans so far (ignoring the dual-GPU ones) and both Titans had different core configs.  The Kepler Titan was cut down, but the Titan Black was the full Kepler same as the 780Ti.  The 980Ti however reversed it, with the Ti being cut down and the Titan being the full Maxwell.

[Xorbot]

On 4/6/2016 at 11:27 PM, shermantanker said:

Well I'm ready to upgrade my 680s. They have been struggling on my UW screen.

Same boat, bro.

Center screen: ROG Swift PG279Q (capped at 120Hz due to Kepler)

Side screens: PB278Q, game mainly only on center screen anyway.

 

Would love to push the 2560x1400 to a constant 165 Hz instead of this CRAPPY 120 Hz peasant limit.

 

[patrickjp93]

1 hour ago, Notional said:

Indeed. For NVidia to use cut down chips for their high end business market should tell you something about their yields at the moment. A high end consumer model is far away for sure.

It tells you nothing. The Tesla K80 uses 2 GK 210s which are cut down as well, and that's on the 28nm process with very high yields. If anything, this says Nvidia is confident AMD won't be able to compete.

[zMeul]

45 minutes ago, patrickjp93 said:

If anything, this says Nvidia is confident AMD won't be able to compete.

absolutely!

 

AMD hasn't shown anything notable, yet

their Polaris GPUs are maybe mid-tier and quite possibly only TDP optimized GPUs for use in laptops

[Samfisher]

51 minutes ago, patrickjp93 said:

It tells you nothing. The Tesla K80 uses 2 GK 210s which are cut down as well, and that's on the 28nm process with very high yields. If anything, this says Nvidia is confident AMD won't be able to compete.

Not in the way you think, but Nvidia has said themselves that Pascal is 60 SMs, and the P100 only has 56 which means the consumer version will eventually have 60, Big Pascal.

[don_svetlio]

300W TDP? THERMI 2.0!

[Senzelian]

9 minutes ago, don_svetlio said:

300W TDP? THERMI 2.0!

My 980s already run near tj max

I definitely need a custom loop... 

[don_svetlio]

Just now, Senzelian said:

My 980s already run near tj max

I definitely need a custom loop... 

Imagine what 300W with a reference Nvidia cooler will be like. Can't wait. My frying pan is ready!

[zMeul]

3 minutes ago, Senzelian said:

 

considering how humongous the GP100 is, it's actually impressive

the Titan X is rated for 250W power draw and it thermal throttles - I guess nVidia has to come up with a new cooling design for their top desktop Pascal

[don_svetlio]

1 minute ago, zMeul said:

considering how humongous the GP100 is, it's actually impressive

the Titan X is rated for 250W power draw and it thermal throttles - I guess nVidia has to come up with a new cooling design for their top desktop Pascal

Most of the size comes from HBM2 tho - it was the same with the Fury X

[patrickjp93]

25 minutes ago, don_svetlio said:

300W TDP? THERMI 2.0!

1480MHz clockspeed. Clearly not Thermi 2.0

 

10 minutes ago, don_svetlio said:

Most of the size comes from HBM2 tho - it was the same with the Fury X

HBM and interposer do not factor into the quoted die sizes for either company.

[don_svetlio]

Just now, patrickjp93 said:

1480MHz clockspeed. Clearly not Thermi 2.0

 

HBM and interposer do not factor into the quoted die sizes for either company.

Highest single GPU thermal output in history - Thermi 2.0

[patrickjp93]

2 minutes ago, don_svetlio said:

Highest single GPU thermal output in history - Thermi 2.0

By 10W over the reference 290X. Big fat whoop!

[don_svetlio]

Just now, patrickjp93 said:

By 10W over the reference 290X. Big fat whoop!

15W* - it was 285W

[don_svetlio]

1 minute ago, patrickjp93 said:

By 10W over the reference 290X. Big fat whoop!

I guess the 290X could be Thermi 2.0 making Pascal Thermi 3.0. Happy now?

[D-Dos Dan]

1 hour ago, don_svetlio said:

300W TDP? THERMI 2.0!

 

 

[RexinOridle]

Imagine proper windows and gaming support on NVIDIA® DGX-1™ Deep Learning Supercomputer.

Not saying it's real but just imagine. Now wipe that cum off your chair.

[D-Dos Dan]

10 minutes ago, RexinOridle said:

Imagine proper windows and gaming support on NVIDIA® DGX-1™ Deep Learning Supercomputer.

Not saying it's real but just imagine. Now wipe that cum off your chair.

Minecraft at 256 *chunk* render distance + shaders 

[zMeul]

2 hours ago, RexinOridle said:

Imagine proper windows and gaming support on NVIDIA® DGX-1™ Deep Learning Supercomputer.

too bad those Teslas don't have outputs 

they're called co-processors and not video cards for a good reasons

[Senzelian]

3 hours ago, don_svetlio said:

Imagine what 300W with a reference Nvidia cooler will be like. Can't wait. My frying pan is ready!

Maybe thats why we get a new cooler design? o:

[ACatWithThumbs]

5 hours ago, don_svetlio said:

Imagine what 300W with a reference Nvidia cooler will be like. Can't wait. My frying pan is ready!

It will need a new design anyways regardless of heat output because HBM makes the card far smaller than GDRRR5 cards.


 

[don_svetlio]

1 minute ago, JasonRoGo said:

It will need a new design anyways regardless of heat output because HBM makes the card far smaller like with the Fury.


 

Not necessairly - remember the penis extension of the 670?

[ACatWithThumbs]

1 minute ago, don_svetlio said:

Not necessairly - remember the penis extension of the 670?

True totally forgot about cooler extension on the reference card.
 My guess would be that they'll go with something similar to the GTX690 but in small.

[don_svetlio]

1 minute ago, JasonRoGo said:

True totally forgot about cooler extension on the reference card.
 My guess would be that they'll go with something similar to the GTX690 but in small.
 

That particular cooler was practically the worst - it had the cores thermal throttle, was loud and bulky. Not to mention NVidia always make blower cards so I'm expecting either a long-ass card or a thermally-limited reference design

[i_build_nanosuits]

2 hours ago, zMeul said:

too bad those Teslas don't have outputs 

they're called co-processors and not video cards for a good reasons

02:27PM EDT - Chip on Wafer on Substrate, the largest such chip ever made blah blah blah deep learning

02:28PM EDT - And that's a wrap

 

02:31PM EDT -  ''Bolt me a freaking display port on the back of that thing i need to play arkham knight on that shit'' --- Jen-Hsun Huang

02:33PM EDT -  ''i still can't run that fucking smoke at 60FPS man!! you fucked up! '' --- Jen-Hsun Huang

02:35PM EDT -  ''also... 300W? really?! '' --- Jen-Hsun Huang