You don't have to read any of this thread as it's nothing to do with "validate errors" and the answer to the above question is in the message I linked in my previous post.
To explain in more detail, your task can't validate because the task it was supposed to validate against has been deleted on the server. As Christian says, this had to be done because of errors in the results making them worthless from a scientific viewpoint. The problem has now been corrected. As he also says, they intend to do a manual adjustment so that all returned tasks will be credited even though there is now no other task to validate against. In an imperfect world, this is about as good as possible to give people the recognition for the work they have done.
There hasn't been much discussion here recently, but I have seen a 27% "Validate error" rate on my RX 570 on Win7 64-bit with the 18.9.3 drivers. That is too much for me, so I am dropping out for now.
Would other drivers work better?
Would a GTX 980 (or 1070) work better?
How about Linux? I can do Ubuntu.
Something is not consistent with something else, but I don't know if it is drivers, GPU type, or OS.
Something is not consistent with something else, but I don't know if it is drivers, GPU type, or OS.
In your case I'd try lowering both the core clock and the memory clock rate on the GPU card by 10%. Then run for a while to see if it makes a difference. You'd need to give it some time, as the comparison of your result with another returned result often must await the other result.
27% is crazy high. I think typical runs between 0.2% and 2%, depending.
I, personally, see levels far below 27% on all of my cards, all of which currently run under Windows 10, and as to type are all Nvidia, including 1050, 1060, 1070, and (on low-pay WUs) the 2080.
It could be the GPU core voltage. I lowered it to save power ...
The fact that you only get 27% validate errors tends to suggest there's no real hardware problem but that the operating conditions (eg core voltage) are just too marginal for stable running. The easiest way to prove that is just put it back on stock voltage for a while and check that the validate errors go away.
If so, just be somewhat less aggressive with the voltage reduction next time :-).
I need the lower voltage to prevent overheating. I am validating fine on MilkyWay, even running two at a time. When the new AMD Navis come out, I will try Einstein again. I need something I can run on Linux anyway, but you can't really undervolt the present AMD cards on Linux. I am hoping the new ones won't need it. Thanks for the input.
I'm not sure what to look for here. GPU has the mildest of overclocks, so I reverted to stock settings in hopes that would help. I can't find any error in any of the logs but again, not sure exactly what to look for / what would actually be causing the errors in the first place.
gemini8 wrote:I get an 'Error
)
You don't have to read any of this thread as it's nothing to do with "validate errors" and the answer to the above question is in the message I linked in my previous post.
To explain in more detail, your task can't validate because the task it was supposed to validate against has been deleted on the server. As Christian says, this had to be done because of errors in the results making them worthless from a scientific viewpoint. The problem has now been corrected. As he also says, they intend to do a manual adjustment so that all returned tasks will be credited even though there is now no other task to validate against. In an imperfect world, this is about as good as possible to give people the recognition for the work they have done.
Cheers,
Gary.
Thanks a lot for explaining.
)
Thanks a lot for explaining.
- - - - - - - - - -
Greetings, Jens
There hasn't been much
)
There hasn't been much discussion here recently, but I have seen a 27% "Validate error" rate on my RX 570 on Win7 64-bit with the 18.9.3 drivers. That is too much for me, so I am dropping out for now.
Something is not consistent with something else, but I don't know if it is drivers, GPU type, or OS.
Jim1348 wrote:Something is
)
In your case I'd try lowering both the core clock and the memory clock rate on the GPU card by 10%. Then run for a while to see if it makes a difference. You'd need to give it some time, as the comparison of your result with another returned result often must await the other result.
27% is crazy high. I think typical runs between 0.2% and 2%, depending.
I, personally, see levels far below 27% on all of my cards, all of which currently run under Windows 10, and as to type are all Nvidia, including 1050, 1060, 1070, and (on low-pay WUs) the 2080.
Thanks. It could be the GPU
)
Thanks. It could be the GPU core voltage. I lowered it to save power, but maybe too much.
(It runs without error on Folding though, but that is a different matter.)
Jim1348 wrote:27% "Validate
)
Has your card been running them one at a time or 2x, 3x ... ? It would be good to try 1x for comparison if that wasn't the choice so far.
Richie wrote:Has your card
)
Always 1x. I think it loaded the card to well over 80%, maybe 90%.
I think I will try it on Milky Way, but I don't recall if they even do a validity check or just an error check. It might work.
Jim1348 wrote:It could be the
)
The fact that you only get 27% validate errors tends to suggest there's no real hardware problem but that the operating conditions (eg core voltage) are just too marginal for stable running. The easiest way to prove that is just put it back on stock voltage for a while and check that the validate errors go away.
If so, just be somewhat less aggressive with the voltage reduction next time :-).
Cheers,
Gary.
I need the lower voltage to
)
I need the lower voltage to prevent overheating. I am validating fine on MilkyWay, even running two at a time. When the new AMD Navis come out, I will try Einstein again. I need something I can run on Linux anyway, but you can't really undervolt the present AMD cards on Linux. I am hoping the new ones won't need it. Thanks for the input.
I was just going through my
)
I was just going through my tasks. Started crunching last night and I already have 10 invalid.
https://einsteinathome.org/host/12775393/tasks/5/0
I'm not sure what to look for here. GPU has the mildest of overclocks, so I reverted to stock settings in hopes that would help. I can't find any error in any of the logs but again, not sure exactly what to look for / what would actually be causing the errors in the first place.