210k ppd on a geforce 970

[beer]

Hi
I looks like I finally got a non-15 core for my newly geforce 970.
It is project 9201 with a tpf on 2 min 14 sec and a esitmated ppd on 210.000. Does those numbers look correct?

//Jonas

[Kjetil]

Maybe, i do not have 970 yet, but my 980 have about 330K on 9201. tpf is 1min 38 sec. Running at 1260 base.

[mdk777]

in range.
OC 970 at 1437
2min TPF
242552 estimated PPD

[kscott]

Overclock that 970 and easily get under 2min
I'm rocking 1min53sec on my 970. 1503 core speed. Averaging around 269k PPD

[runpaint]

Is everyone using the latest drivers? I don't want to leave my 750ti unused when I install my 970, so I'll be running them both in the same rig. But the latest driver update will slow my 750ti by 20%. I'm pretty sure it will be better to update the driver for the 970 and take the loss with the overall gain (which I have also done with a 750ti through a PCIEx1 slot @ 40k ppd - not great for that card, but better than 0). But, one of my rigs is running 3 750ti cards which total 192k ppd and only 157k with the latest drivers. So to run that with 1 970 and 2 750s would be slowing down 2 of my cards, but I have to consider that option since it's one of my only computers with a power supply to fit the 970 (which is why I have all these 750s).

[EXT64]

I am using the latest drivers on my GTX 980/GTX 750Ti system. Fortunately I forgot what the 750Ti got with the older drivers, so the performance loss hasn't bothered me

[runpaint]

lol, the 750 is what got me into gpu folding. Any computer with a pcie slot - even a x1 slot - even with a 250w psu - even a Dell refurb that my dad bought for $50 - almost ANY computer can get 40,000 - 60,000ppd just by adding a card for $110 - $150. There has never been anything quite like it, but hopefully the technology will continue to get cheaper with lower power requirements. 

[Nathan_P]

What the power draw on these 980/970's? might be worth investing in a few when BA finishes if PPD/W is good enough

[Kjetil]

At p 9201 and 1 970 Oc to 1242, and 1 980 oc to 1356 on msi z77a-gd55, cpu i7 3770K my computer using 335W.

[billford]

I can't help with the wattage, but my GTX980, o/c'd to 1415MHz with Linux 343.22 drivers, returns ~390K PPD (TPF 1m 28s) on a P9201.

[mdk777]

I would advise reading the thread on core 18 development before making any more investments.
With the completion of the single available core 17 that is running on Maxwell, production (as measured by ppd) will fall to 1/4 the current level.
Hopefully, the issues with drivers, open MM versions, and core 18 on Maxwell will be resolved.
However, be prepared to not see the production discussed here in the near future.

[billford]

Yes, I'm aware of all that.

Core_18 something of an unknown quantity at the moment, the Linux version more so, but hopefully P9201 should be around for a while yet- it only went to Full in July.

[runpaint]

I would advise reading the thread on core 18 development before making any more investments.
With the completion of the single available core 17 that is running on Maxwell, production (as measured by ppd) will fall to 1/4 the current level.
Hopefully, the issues with drivers, open MM versions, and core 18 on Maxwell will be resolved.
However, be prepared to not see the production discussed here in the near future.

I'm not saying I would've followed your advice if I had seen it last week before I ordered my new cards, but unfortunately you might be right.

My 970 and 980 arrived this week, and I immediately started them folding. They each got a core17 to start with, at about 250k and 350k ppd respectively. I already had 13 cards running, but these 2 new cards almost doubled my PPD. So I started planning on when I could afford my next 980.

But my excitement and plans only lasted about 3 hours. 3 hours. ONE core17 each. That's all they got. They each returned their first work unit, and then it was all core15 for the next 2 days. My PPD actually dropped. What a great way to spend $900. I got one more core17 on the 980 yesterday, but since then it's still been all 15s. Oh well, the same thing happened 2 or 3 weeks ago, and then it was back to core17 24/7 for a while so maybe I'll get that million ppd soon.

I actually considered buying 780TIs for my next cards instead of 980s, because 780 is Kepler, but it just doesn't seem like the right move unless they get a lot cheaper. All my other cards are 750 and 750ti, which are also Maxwell like the 900s, so my only hope is to stick with Maxwell and take what I can get. I have to believe that they're going to take advantage of the newer cards, even if there's a little waiting involved.

[bruce]

Yes, {Core_18 + Maxwell + Drivers} is an unknown quantity, but I'm not as pessimistic as mdk777. Neither one of us have any solid facts on which we can base our predictions of the future so either could turn out to be right.

I do know two things:
1) Not unexpectedly, NVidia has a history of providing better support for their proprietary CUDA than for OpenCL (Open standard).
2) The Developer who created Core_18 seems very busy working on SOMETHING but when asked what, has not elaborated. [This is consistent with Stanford's policy of not announcing anything before it's ready and certainly not uncommon for FAH's development team.]

I'm sorry that I can't clear up any of the mystery.

[billford]

I'm sorry that I can't clear up any of the mystery.

No problem- I know I'm a comparative newbie but I've been around long enough to know how the system works

Like runpaint I was thinking of getting another 980 (and eBaying the 650 Ti it would replace), that could give me a million PPD just on three cards… but like most companies NVidia work on an exchange rate of one dollar equals one pound sterling, so they're not cheap over here

Plus, because I use Linux I'd be totally dependant on the supply of P9201's for the Maxwells, if that dries up for any reason the PPD is zero until Linux Core_18 is let loose on an unsuspecting world, and then anything could happen.

So, for a while, I'll delay my combined birthday and Christmas present to myself