eFMer Seti Performance results GTX 580 standard 772GHZ / Driver 306.97

Started by Caprio, November 14, 2012, 05:23:50 AM

Previous topic - Next topic
Starting automatic test: (x41g)
14 November 2012 - 21:03:02 Start, devices: 1, device count: 1 (1.00)
--------------------------------------------------------------------------- Results: Device: 0, device count: 1, average time / count: 135, average time on device: 135 Seconds (2 Minutes, 15 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:05:19 Start, devices: 1, device count: 2 (0.50)
--------------------------------------------------------------------------- Results: Device: 0, device count: 2, average time / count: 192, average time on device: 96 Seconds (1 Minutes, 36 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:08:35 Start, devices: 1, device count: 3 (0.33)
--------------------------------------------------------------------------- Results: Device: 0, device count: 3, average time / count: 265, average time on device: 88 Seconds (1 Minutes, 28 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:13:11 Start, devices: 1, device count: 4 (0.25)
--------------------------------------------------------------------------- Results: Device: 0, device count: 4, average time / count: 347, average time on device: 86 Seconds (1 Minutes, 26 Seconds)
>> The best average time found: 88 Seconds (1 Minutes, 28 Seconds), with count: 0.33 (3)

Now with 2 devices Slot 0 GTX 580 and Slot 1 GTX670.

670 is markably slower than 580 to complete task. And GTX 670 is clocked according to EVGA FTW version 1200MHz. The 580 is stock reference 772MHz.

But boy does GTX 580 run hot at 90 degrees. And fan spinning over 3000 rpm.

Starting automatic test: (x41g)
14 November 2012 - 21:23:40 Start, devices: 2, device count: 1 (1.00)
--------------------------------------------------------------------------- Results: Device: 0, device count: 1, average time / count: 128, average time on device: 128 Seconds (2 Minutes, 8 Seconds)
Device: 1, device count: 1, average time / count: 187, average time on device: 187 Seconds (3 Minutes, 7 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:26:51 Start, devices: 2, device count: 2 (0.50)
--------------------------------------------------------------------------- Results: Device: 0, device count: 2, average time / count: 194, average time on device: 97 Seconds (1 Minutes, 37 Seconds)
Device: 1, device count: 2, average time / count: 249, average time on device: 124 Seconds (2 Minutes, 4 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:31:09 Start, devices: 2, device count: 3 (0.33)
--------------------------------------------------------------------------- Results: Device: 0, device count: 3, average time / count: 267, average time on device: 89 Seconds (1 Minutes, 29 Seconds)
Device: 1, device count: 3, average time / count: 351, average time on device: 117 Seconds (1 Minutes, 57 Seconds)
Next :---------------------------------------------------------------------------
14 November 2012 - 21:37:11 Start, devices: 2, device count: 4 (0.25)
--------------------------------------------------------------------------- Results: Device: 0, device count: 4, average time / count: 347, average time on device: 86 Seconds (1 Minutes, 26 Seconds)
Device: 1, device count: 4, average time / count: 456, average time on device: 114 Seconds (1 Minutes, 54 Seconds)
>> The best average time found: 103 Seconds (1 Minutes, 43 Seconds), with count: 0.33 (3)

x41g was designed & built long before the Kepler GPUs were released.  I would recommend using x41zc on the 670, Cuda 4.2 build as a compromise that'll work reasonably on a 580, while the 670 would prefer Cuda 5.0 it'll still perform under 4.2.

Any particular reason you're comparing an originally more expensive, higher power GPU, to a cheaper more efficient one ?  For absolute comparisons I would suggest total cost of ownership (including initial cost and power cost, really need to be factored in, as well as if the application was designed for the card(s)

JAson
It is not the strongest of the species that survives, nor the most intelligent that survives. It is the one that is the most adaptable to change.
Charles Darwin
---
Chaos: When the present determines the future, but the approximate present does not approximately determine the future.
Edward Lorenz

The reason is to compare Next generation GPU with older generation GPU, by changing only one variable at the time. At the time of test the x41z client was not widely available, so comparison was made witht he x41g.

One variable is the absolute performance per card, second is the investment price, 3rd one are the running costs. Before the x41z client was available I was mildly dissapointed with the performance of the 680 GTX compared to 560Ti TOP. I mean if you shell out nearly 500€ you did not get more than double performance to the 200+€ 560Ti.

If you do not know the performance of each card you cannot make a sound choice of which model to buy, or what is the real crunching value of cards on comparable terms between the 500 and 600 series.

The 4th and at this stage last parameter is that if you are limited on PCIe 16x physical slots, what kind of deck is cost efficient to have installed. Maybe a 480 is something that I would currently not put on my shopping list of approved cards due to temperature and power consumption issues.

The key takeaway of this study is then that a 600 series card will not bring you additional value over a 500 series card, unless you run it with a x41z client. I.e. the price premium of a 600 series card is hard to justify over the 500 series unless you run with with the x41z client. Although I am writing this in February 2013, back in summer 2012 the pricing and availability of cards was more in line with my reasoning. Nowadays, you are not likely to find 500 series cards for sale, except on the second hand market. However, you can still make some decent second hand picks in a 500 series card.



I think all that's fair comment.  In my cases the 560ti and 680 have both been stellar performers for different reasons.  What I'd like to see along those lines going into V7, and hope to setup some tools for, is testing that better mirrors live running, and allows submission and online comparison by various approaches with some sortof database.  At the moment it can be difficult to judge many of the factors due to one single uncommon task being used often for benches. 
It is not the strongest of the species that survives, nor the most intelligent that survives. It is the one that is the most adaptable to change.
Charles Darwin
---
Chaos: When the present determines the future, but the approximate present does not approximately determine the future.
Edward Lorenz

Powered by EzPortal