Gtx 980 f@h!!!

[PCTC2]

Did you guys see the GTX 980 performance for Folding@Home? I think I might just be in the Holiday Race this year!

http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/20

Seriously. Wish I was still at my old job so I could toss 4 to 8 of these in a system.

 

[biodoc]

Impressive!

 

[Ken g6]

Yeah, I think there's a very significant chance I'll have a 970 in the race this year.

 

[Ken g6]

Tom's Hardware has done some power tests showing high power consumption and/or throttling for compute loads. How long is the F@H bench? It might not be long enough to get a long-term speed.

 

[biodoc]

Check out this thread for some early GTX980 numbers for project 9201 (core 17).

One user is reporting ~330,000 ppd and the other ~269,000 ppd.

Looks encouraging but time will tell.

 

[matszpk]

Indeed, performance of the single precision is amazing. However this review shows a bad news: an embarasing performance on the double precision. New geforces has 1/32 performance in DP of performance in SP .
There are not good news for other volunteers whose are crunching for example Primegrid genefer.

 

[Assimilator1]

Or Milkyway@Home!

 

[biodoc]

Darn, couldn't resist buying a GTX980

Seems to do well on FAH. Got the fan set at 60% and temp is 59C at 98% gpu utilization. This is linux by the way.

 

[GLeeM]

@biodoc
Nice! To which rig did you add it? Or did you replace the 460?

Have you tried it on GPUGrid or anything else yet?

Anyone know how the GTX970 does in comparison?

 

[Assimilator1]

333,570 ppd!

 

[biodoc]

@GLeeM

I'm just running 2 rigs right now. The GTX980 is in the 3930K and the 780Ti is in the 2600K.

We are waiting for a cuda 6.5 app at GPUGrid. The standard apps don't work on the "BigMaxwells" See thread for more details.

There are some links to Einstein work in the above thread for both the 970 and 980.

 

[biodoc]

I have data comparing my 780Ti with the 980 at Folding@home.

hardware:
780Ti (only gpu in system) in 2600K, PCIE2 slot, 64-bit linux mint 17 LTS, nvidia driver 343.22
980 (only gpu in system) in 3930K, PCIE2 slot, 64-bit linux mint 17 LTS, nvidia driver 343.22

Project 9201 (core 17)
780Ti@1106MHz (+100MHz OC): TPF=117 seconds, PPD=255,750
980@1352MHz (+100MHz OC): TPF=93 seconds, PPD=360,510

Looks like a 20.5% reduction in TPF. Seems to correlate with difference in clock speed (22%)?

It's not a perfect comparison but it's the best I can do.

EDIT: I have SMP Wus running on both systems (SMP7 on the 2600K (1 free core) and SMP10 on the 3930K (2 free cores)

Total PPD for both systems: 666914

 

[Ken g6]

The AnandTech 970 review is up!

That last one probably applies to double-precision projects like GPUGrid and PrimeGrid GFN.

 

[Assimilator1]

F@H is single precision only, right?

Milkyway@Home is also Double precision.

 

[biodoc]

Yep, F@H is single precision.

At $180 cheaper, it looks like the GTX970 is the way to go.

I can't believe how cool these cards run and low power consumption is another big bonus.

For DP floating point calculations, AMD is the way to go.

I wonder when we will see 980Ti and maxwell titans?

 

[biodoc]

https://foldingforum.org/viewtopic.php?f=38&t=26757&start=15

PPD for a 970 is 309K.

 

[GLeeM]

At $180 cheaper

I thought the 970 was $329 and the 980 was $549 for a difference of $220?

 

[biodoc]

I thought the 970 was $329 and the 980 was $549 for a difference of $220?

I guess I was looking at a superclocked version. Newegg is selling them for $329-369.