I just finished my 1st GW from this new batch and the run time is about the sane as before. Now I get to see if this run is anymore reliable in terms of validating.
I decided to make some changes with my two machines. Since (in my case) the AMD/Linux machine seems to have a higher number of invalids on the GW GPU tasks, I have limited that box to crunching FGRP GPU tasks. I have also disabled CPU tasks on that machine, since it only has 2 cores. I would also like to see if disabling CPU tasks allow the GPU side to crunch more efficiently.
My NVIDIA/Win 7 box is limited to GW GPU work. With 12 physical cores, I have plenty of room for CPU work, while running the GPU at x2. This box reports 24 virtual cores and I'm currently running at 65% CPU utilization.
Since the GW and FGRP tasks don't play nicely together, this method solves the incompatibility problem.
Possibly they fixed the problem with Nvidea cards on this run, I have a couple of pending GWs, no invalids and a bunch of valids. I'm sure posting this will jinx it.
My remaining system running v1.07 GW GPU work has usually been reliable, producing days of valid results in a row at times, though it had a brief period of high invalid rates a week or so ago.
It has had a new "bad period".
And now it has had yet another "bad period". Six tasks reported between 3:26 and 9:23 UTC September 7 are currently inconclusive (including four consecutive), so I expect to become invalid.
I've set the project preferences to change from running tasks on this machine at 2X to running at 1X.
<snip>
So this is intended to be an experiment to see whether the "bad periods" recur even when running 1X, not a way to raise real productivity.
That took less than a day. At 1X the machine ran 15 consecutive successful tasks, followed by four successive bad ones (inconclusive at the moment, I confidently expect invalid when the quorum is fulfilled).
So the pattern on my two Windows 10 RX570 hosts running GW GPU v1.07 tasks remains:
At 3X and 4X, 100% invalid (on the one of the two machines tried)
at 2X it runs hours to days at 100% pass rate, then suddenly switches to 100% fail rate (without an obvious trigger), and on one of the two switches back after a few hours, again without obvious trigger.
A 1-day trial at 1X reproduced one instance of the toggle from good to bad and back to good.
So I'm attempting to move that machine back to 2X (work availability has been intermittent, so this may take a while).
I'm also attempting to obtain a few GW GPU tasks for my Radeon VII system, just to add to the v1.07 experience base of systems. That system is expensive, and insanely productive on Einstein GRP, so I don't expect to switch it over to GW for routine work unless a far more effective application emerges.
The upload seems to have fixed itself. I assume the problem is with their servers. At the same time the problem with WCG uploads returned. However there is a thread at WCG warning about the problem unlike here.
I cannot find a way to organize the threads so the newest one is "on top". I looked in community profiles and poked around but didn't see a way to set the newest first. Sometimes, when something is missing, I have to enable the virus transfer protocol active-x to have missing buttons show up.
I cannot find a way to organize the threads so the newest one is "on top". I looked in community profiles and poked around but didn't see a way to set the newest first. Sometimes, when something is missing, I have to enable the virus transfer protocol active-x to have missing buttons show up.
On my system it's: Account>Preferences>Community> scroll down to Forum Settings > "Sort comments in discussions:" and select "Newest post first" in the pull-down menu box. Your optional Signature box is also there under Forum Settings.
Ideas are not fixed, nor should they be; we live in model-dependent reality.
I just finished my 1st GW
)
I just finished my 1st GW from this new batch and the run time is about the sane as before. Now I get to see if this run is anymore reliable in terms of validating.
I decided to make some
)
I decided to make some changes with my two machines. Since (in my case) the AMD/Linux machine seems to have a higher number of invalids on the GW GPU tasks, I have limited that box to crunching FGRP GPU tasks. I have also disabled CPU tasks on that machine, since it only has 2 cores. I would also like to see if disabling CPU tasks allow the GPU side to crunch more efficiently.
My NVIDIA/Win 7 box is limited to GW GPU work. With 12 physical cores, I have plenty of room for CPU work, while running the GPU at x2. This box reports 24 virtual cores and I'm currently running at 65% CPU utilization.
Since the GW and FGRP tasks don't play nicely together, this method solves the incompatibility problem.
Anyone else having issues
)
Anyone else having issues uploading this morning?
Keith Myers wrote:Anyone else
)
Yes, almost rebooted all my system thinking it was me
Possibly they fixed the
)
Possibly they fixed the problem with Nvidea cards on this run, I have a couple of pending GWs, no invalids and a bunch of valids. I'm sure posting this will jinx it.
archae86 wrote:archae86
)
That took less than a day. At 1X the machine ran 15 consecutive successful tasks, followed by four successive bad ones (inconclusive at the moment, I confidently expect invalid when the quorum is fulfilled).
So the pattern on my two Windows 10 RX570 hosts running GW GPU v1.07 tasks remains:
At 3X and 4X, 100% invalid (on the one of the two machines tried)
at 2X it runs hours to days at 100% pass rate, then suddenly switches to 100% fail rate (without an obvious trigger), and on one of the two switches back after a few hours, again without obvious trigger.
A 1-day trial at 1X reproduced one instance of the toggle from good to bad and back to good.
So I'm attempting to move that machine back to 2X (work availability has been intermittent, so this may take a while).
I'm also attempting to obtain a few GW GPU tasks for my Radeon VII system, just to add to the v1.07 experience base of systems. That system is expensive, and insanely productive on Einstein GRP, so I don't expect to switch it over to GW for routine work unless a far more effective application emerges.
The upload seems to have
)
The upload seems to have fixed itself. I assume the problem is with their servers. At the same time the problem with WCG uploads returned. However there is a thread at WCG warning about the problem unlike here.
I cannot find a way to organize the threads so the newest one is "on top". I looked in community profiles and poked around but didn't see a way to set the newest first. Sometimes, when something is missing, I have to enable the virus transfer protocol active-x to have missing buttons show up.
I was. It seems to be fine
)
I was. It seems to be fine now.
Yep my last post worked I got
)
Yep my last post worked I got my first invalid on this run.
JStateson wrote:I cannot find
)
On my system it's: Account>Preferences>Community> scroll down to Forum Settings > "Sort comments in discussions:" and select "Newest post first" in the pull-down menu box. Your optional Signature box is also there under Forum Settings.
Ideas are not fixed, nor should they be; we live in model-dependent reality.