Maxwell 2 - NVidia GTX 980

[DrPop]

Man, looks like Maxwell is finally hitting its stride. :) Check out the (relatively) low watt burn for this card - 165W. That's sweet considering how many CUDA cores are on this baby for crunching.

http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review

[Maxwell]

At least they named after something awesome...

[John P. Myers]

Check out the (relatively) low watt burn for this card - 165W. That's sweet considering how many CUDA cores are on this baby for crunching.


Yep and they further crippled FP64 down to 144 GFLOPS on the 980. FP32 is 4612 GFLOPS. That is a really nice TDP though.

Edit: The 980 is the most efficient single precision GPU ever made getting 27.95 GFLOPS/W

Edit: As of this launch, Nvidia has discontinued the 780, 780 Ti and 770. Make sure to get one of those models branded as EVGA so when you RMA it they're forced to give you a 900 series replacement :D

[DrPop]

...The 980 is the most efficient single precision GPU ever made getting 27.95 GFLOPS/W...

Do you have any inside scoop on if the the competition (AMD) is going to bring something similar to the table? Or if we need a new GPU in the next few months, is this "da big one"?

[John P. Myers]

Do you have any inside scoop on if the the competition (AMD) is going to bring something similar to the table? Or if we need a new GPU in the next few months, is this "da big one"?

This isn't "da big one". Unfortunately that one will be a bit of a wait. How long is anyone's guess since it's being determined by Nvidia's whims and greed. The chip already exists though.

AMD is working on the R9 300 series, but that'll be awhile too.

[John P. Myers]

"da big one" may possibly be released by the end of the year, maybe. It'll be called the GTX 980 Ti and/or the Titan X or the Titan II, at least as far as anyone can tell right now. Somewhere from 2560 to 2816 cores and a 384 memory bus which is estimated to make it 50% faster than a Titan Black. It's all rumor so we'll see..

[Sarge104]

"da big one" may possibly be released by the end of the year, maybe. It'll be called the GTX 980 Ti and/or the Titan X or the Titan II, at least as far as anyone can tell right now. Somewhere from 2560 to 2816 cores and a 384 memory bus which is estimated to make it 50% faster than a Titan Black. It's all rumor so we'll see..

You mean I'll have to take out a second mortgage by the end of the year? Or will they accept first born? ;) As much as I love my Titan it will take a card with a lot more performance(I'm unrealistically thinking 3x) with lower power requirements for me to be able to justify dropping any more dough.

So probably in the next three years we'll get there, if somebody doesn't buy out all the competition and stagnate the performance increases per year any more.

[Maxwell]

If we want to take up a collection so that I can buy a bunch of these, I'd be OK with that. I've heard rumors* that there's also a 20% performance boost for people who crunch under the name Maxwell...

*Source = that handsome person who lives in the mirror

[John P. Myers]

From what i just saw of the "leaked" images of "da big one", it looks like it's going to be really big. Very, very big. Biggest ever. Based on some image analysis i saw a bit ago, it appears it'll have a die size of about 632mm^2, making it almost 30% larger than the GK110 used in the Titans. This is getting interesting. Remember this is only happening because TSMC failed to deliver 20nm GPUs. Everyone is hitting a wall with the 28nm, but they're still trying to come up something better than the last thing they came up with.

[DrPop]

How can they keep power usage in check when they up the die size so much?:confused:

[MindCrime]

more parallel slower clocks...on top of maxwell's already efficient architecture. Throwing in an uneducated guess based on what they do with dual gpu cards.

[Fire$torm]

From what i just saw of the "leaked" images of "da big one", it looks like it's going to be really big. Very, very big. Biggest ever. Based on some image analysis i saw a bit ago, it appears it'll have a die size of about 632mm^2, making it almost 30% larger than the GK110 used in the Titans. This is getting interesting. Remember this is only happening because TSMC failed to deliver 20nm GPUs. Everyone is hitting a wall with the 28nm, but they're still trying to come up something better than the last thing they came up with.

So..... is there any word from TSMC, on how long it will take them to solve their production issues?

[John P. Myers]

20 nm is hopeless. 16nm next year.

Sent from my SGH-M919 using Tapatalk

[John P. Myers]

Nvidia should officially announce something made with the GM200 at the GPU Conference (March 17-20). Most likely they'll announce the Quadro M6000, but with any luck they'll also announce a 980Ti or maybe the Titan 2 (or whatever they're going to call it) at the same time, or at least soon afterwards.

Not sure if the new Titan will be all that great compared to the original. Nvidia has refused to make Tesla cards with Maxwell because of the lack of FP64 units.

Expect huge FP32 gains though. Comparing the 780Ti to the 980Ti, the 980Ti gets only 1 GFLOP more of FP64 (210 vs. 211), but gets 35% more FP32 (5046 vs. 6758). This assumes an 1100MHz base clock and that all 3072 cores are active, which is reasonable since the 780Ti also had all of its cores active.

[DrPop]

To put it in perspective for me: which projects require the FP64?

[zombie67]

PG Genefer, milkyway. Any GPU project that says only some AMD cards are supported.

[John P. Myers]

To put it in perspective for me: which projects require the FP64?

Milkyway, some Primegrid, GPUGrid uses it just a tad though they don't bother mentioning it due to the insignificance, I remember hearing about Einstein beginning to use FP64 starting maybe a year ago. That's about it really, though there are likely many projects that use it in insignificant amounts like GPUGrid. On Primegrid though, if you're going for your GFN or PSA badge and you're not using a Titan or something from AMD, then i feel bad for you :p

[John P. Myers]

Also the Asteroids@home GPU app requires FP64

[zombie67]

GPUGrid uses it just a tad though they don't bother mentioning it due to the insignificance,

I've wondered about this. What happens if you use a card that doesn't support any DP? Will it just run those calculations in some sort of emulation mode on the CPU or something? Or will the task just flat out fail? I don't know that I have ever had a GPU without at lease some DP.

[John P. Myers]

Not sure about now, but a few years ago it would just fail. I remember the big thing used to be you needed an Nvidia GPU with a compute capability of at least 1.3, which meant it was capable of double precision. Now that doesn't really apply anymore since it seems you can get GPUs with a higher CC rating that don't support FP64 (like the GT 705 through GT 730). I'm thinking it may be emulated, though GPUGrid did drop support for older GPUs that don't have FP64.