FWIW at stock speeds I think the 680 tech spec of 200W is rather higher than actual, though that bodes well for cooling and power implementation overkill, and OC headroom.

The power ratings are absolute maximum ratings- full GPU, memory, video engine & cooling fan load.
My GTX 750Tis running at mid to high 90% GPU load occasionally hit high 70% of the maximum TDP rating- that's around 42W.
The GTX 680 would probably run closer to 95% GPU load, which would probably be mid 80s for the thermal load- that's around 160W.

---

Grant
Darwin NT

Not sure how crunchability of a 750ti compares to a 580 though...might take a hit there.

Yep.
My GTX 750Tis process less work per hour than the GTX 460 & the GTX 560Ti they replaced. They actually process more longer running WUs per hour, but due to the significant slowdown in processing shorties the WU/hr per card does drop (which is why I'm so interested in the benefits of DX12).
However I can run 3 (maybe even 4) GTX 750Tis and still use less (or only very slightly more for 4) power than a single GTX 460/560Ti.

---

Grant
Darwin NT

Mark

It does take a hit because my numbers are based on a .14/kwh prime day rate.
It gets more complicated because my cost is .14 during business days 10am to 10pm. On holidays, nights, and weekends, my cost is only about .05/kwh.
So the payback gets extended a great deal with that in mind.

That would be
12h x 5d - prime rate
12h x 5d - reduced rate
24h x 2d - reduced rate

0.14(12x5) + 0.05(12x5 + 24x2)
0.14(60h) + 0.05(108h)
8.4 + 5.4 = 13.8/week
divide by 7days and 24h

0.08214 /kwh on a 24/7 running basis

Not counting holidays, so just a little less than that.

---

Mark

It does take a hit because my numbers are based on a .14/kwh prime day rate.
It gets more complicated because my cost is .14 during business days 10am to 10pm. On holidays, nights, and weekends, my cost is only about .05/kwh.
So the payback gets extended a great deal with that in mind.

That would be
12h x 5d - prime rate
12h x 5d - reduced rate
24h x 2d - reduced rate

0.14(12x5) + 0.05(12x5 + 24x2)
0.14(60h) + 0.05(108h)
8.4 + 5.4 = 13.8/week
divide by 7days and 24h

0.08214 /kwh on a 24/7 running basis

Well, regardless of the calculations, I can tell you that my current power bill is $614.00 and change, due in a couple of weeks.
That is what it costs to keep this kit running 24/7 as it currently stands.
About $20.00 per day.
Granted a small portion of that keeps the lights on, the fridge cold, and the waterbed warm for me and the kitties.
But I assure you, that is a minor portion.
I can shut the kit off for a month and see a $150.00 or less bill, I assure you.
Maybe down to a c-note.

Not complaining, really. It is a choice I have made for many years now.
It is simply my devotion to this project. Add it up for almost 15 years, and it is indeed a formidable sum.
It gives me much satisfaction in return, to see the amount of work I have amassed during my tenure here. Granted, we have not found solid results yet, but that is the nature of the task that we have undertaken.
And I am good with that.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

At reduced rates it would take over a year to break even. Regardless of rates, I would choose eight 960s over 14 750Tis due to the costs of the cards.

At reduced rates it would take over a year to break even. Regardless of rates, I would choose eight 960s over 14 750Tis due to the costs of the cards.

Probably a good choice, as the PCIE buss could give better support to fewer cards with less contention issues.

Thanks for the insight.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

At reduced rates it would take over a year to break even. Regardless of rates, I would choose eight 960s over 14 750Tis due to the costs of the cards.

Make it 8*GTX 960 v 16*GTX750Tis.

198*16 = 3,168.
292*8 = 2,336
$832 difference.

120W*8 = 960
60W*16 = 960
Even power consumption.

I don't know what the actual processing abilities for Seti are for the GTX 960, but I suspect due to driver & application limitations it won't be that much better than the GTX 750Ti.
Even with the greater up front cost, you'd get more work per $ spent from the GTX 750Tis, for the same amount of power. And you'd get more work per hour as well.

A couple less GTX750Tis, you'd still get more work per hour, even more work per $ spent, and you'd require less power to do it.

---

Grant
Darwin NT

I only chose eight versus 14 to make the numbers work nicely as the 960 provides 75 percent more flops for only 33 percent more cost but at double the power consumption. Flops don't necessarily equate to performance but I don't have either card in my possession to test so that's what I'm using. Any difficulties in achieving theoretical output would probably apply to both cards. Power consumption is a big deal but if the 960 is faster then you could buy fewer cards which uses less power. So the 14 uses less power but the eight cards are cheaper up front.

You meant slower in MHz when you said:

"Interesting, but according to that article it would be slower than the Titan X which is slower than the current GTX 980s"

GTX 980 > Titan X > GTX 980Ti

That's where I started to get confused, because I was talking about GFLOPS

I wasn't comparing clock speeds, you were.
What I was refering to was the amount of cuda cores in relation with clock speed to calculate sp performance. That is the same for different gpu cores like 980 or titan x.

That order is how I would expect them to be then the 980Ti comes out. It would also fit the pattern from the 700 series. With the x80Ti being faster than the lowest end Titan. Although the DP performance of the 900 series titan is rather... lacking compared to the previous generation. Not that we need DP for SETI@home.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

You meant slower in MHz when you said:

"Interesting, but according to that article it would be slower than the Titan X which is slower than the current GTX 980s"

GTX 980 > Titan X > GTX 980Ti

That's where I started to get confused, because I was talking about GFLOPS

I wasn't comparing clock speeds, you were.
What I was refering to was the amount of cuda cores in relation with clock speed to calculate sp performance. That is the same for different gpu cores like 980 or titan x.

That order is how I would expect them to be then the 980Ti comes out. It would also fit the pattern from the 700 series. With the x80Ti being faster than the lowest end Titan. Although the DP performance of the 900 series titan is rather... lacking compared to the previous generation. Not that we need DP for SETI@home.

Me too. That's why I said that a 980Ti would be better value for money compared to a Titan x.

---

Speaking of value for money, this 780Ti looks to be pretty worthwhile, what do you think?


http://www.overclockers.co.uk/showproduct.php?prodid=GX-208-OK&groupid=701&catid=1914&subcat=1402

---

I have 3 of them, they run really well but are power hungry and generate a lot of heat.

But if you are only getting 1 then it shouldn't be a major issue.

They are rock solid in their performance.

Make sure you have a 6 and 8 pin connectors.

Just made the order. It will make a good replacement for my 670

---