2 GTX 460s reporting different ETAs for similar work

[Stephenish]

On my 2 GTX 460 rig, running v7.1.52, the ETAs for the two different cards are consistently differing about similar projects.

One card reports what seems to be an accurate ETA of about 9 hours, while the second card reports an ETA of over two days.

I thought I had gotten a problematic WU on one of my cards and not the other. However, both WUs finish in an approximately equivalent time.

Seems as if the ETA reporting is slightly amiss.

[jimerickson]

viewtopic.php?f=67&t=21511&p=215121&hilit=v7+eta#p215121

[Jesse_V]

Similar projects? How so? Please be more specific. Projects differ from one to the other. If the ETA is that different on the same project, then there's a problem somewhere for sure.

[Stephenish]

Both cards are currently running project 8020. Earlier one card reported 2.09 days and the other about 8 hours 46 minutes. Now, one card reports 8 hrs 20 mins and card 2 reports 8 hrs 31 mins.

The cards usually get similar project ETA times, but I often see the incorrect ETA reported by one of the gpus.

I actually thought there was an issue with one of my cards. So I downloaded a fan manager and an overclock manager to test the cards. I found no issues with either card. Excepting that one card is running slightly hotter (but within normal range).

[k1wi]

How are you measuring PPD?

From the client (V7) or from a third party app?

It's possible that the card measuring the longer time was including the period downloading the WU?

The other possibility is that there is something using your first GPU's resources, such as a accelerated web browser.

[Jesse_V]

https://fah-web.stanford.edu/projects/F ... ticket/888

[Zagen30]

IF you see what looks like a wildly unusual ETA, go to the log and check the actual frame times to see if it's actually running slow or if the ETA calculation is just on the fritz again.

Alternatively, I believe most third-party monitoring clients are a lot less likely to mess up their ETA calculations quite that bad; see the 3rd-party subforum to find links to them.

[Stephenish]

How are you measuring PPD?

From the client (V7) or from a third party app?

It's possible that the card measuring the longer time was including the period downloading the WU?

The other possibility is that there is something using your first GPU's resources, such as a accelerated web browser.

The machine is dedicated to v7 folding. ETA is coming straight from the client GUI. The incorrect ETA is an ongoing issue recurrent over several weeks.

[Stephenish]

https://fah-web.stanford.edu/projects/FAHClient/ticket/888

http://foldingforum.org/viewtopic.php?f=67&t=21511

Good link Jesse, guess I should've checked for TPF instead of ETA.

[jcoffland]

@Stephenish I don't yet see enough evidence to prove that this was not a normal difference in run-time between two WUs. If the project numbers are different then all bets are off but it's also possible to have different run-times from one WU to the next even with in the same project. Can you tell us which projects these two WUs were from?