can you tell me something about error rates in the various einstein tasks?
22 Nov 2014 10:17:49 UTC
Topic 197817
(moderation:
can you tell me something about error rates in the various einstein tasks? I don't have a clue as to what is acceptable from my machines.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
can you tell me something about error rates in the various einst
)
I have one system with 2x HD 7970 running BRP4G WUs and the error rate was about 15-20%. It was running at 950Mhz core with temps around 60C. Downclocked it to 925 and turned the fans up to keep temps in low 50s. Haven't had an invalid in over 12 hours now. Not sure that was it, just what I'm seeing right now. I have 2 other HD 7970s in other systems that had no errors. So it appears to be that system, not the app/WUs.
No errors at all would be the
)
No errors at all would be the goal but that might not be achievable.
If you go to the server status page and then look at the right hand side and a bit down there are 2 sections with "search progress" data for BRP4 and BRP5, there are also "Details" links for the two searches, those links takes you to pages with statistics for amongst others error rates. You can click on the graphs to open them as .pdf files so you can zoom in.
thanks fellas, both tips were
)
thanks fellas,
both tips were very helpful.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
On one computer I get near a
)
On one computer I get near a 10% error rate. The gpu temp's are mid-sixties.
I have a r9 270x and a r9 280x and a discrete igpu 4600. I can no longer get a correct reading on CCC for purposes of underclocking. MSI afterburner is equally confused as to what is going on and is useless. In other words I can't find a way to underclock my amd gpu's. Any suggestions?
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
Not sure what you mean by
)
Not sure what you mean by can't get a correct reading. Do you mean it's showing you values you don't think are correct, can't see the values or can't change the values?
RE: Not sure what you mean
)
All three.
What happened is sapphire in their quick install guide said disable my igpu before I install the software for the new graphics card which I did but then I was left with my 2 amd gpu's but without my igpu. So then I reinstalled my igpu and now I have all 3 and it runs much smoother but I can only get readings from gpu-z and HWinfo64 which are mostly correct but they don't let me change the values. If I try and reinstall CCC with the igpu installed I don't know if I'll mess up the whole works and be worse off.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
If you have an ipgu and pcie
)
If you have an ipgu and pcie gpu, CCC and afterburner won't show or let you control the pcie gpu unless either a monitor is connected to the gpu or , what I do, is make a dvi dongle and plug that into the pcie gpu. Here's how to make a dvi dongle...https://rumorscity.com/2013/12/06/how-to-create-dummy-plugs-for-your-graphics-cards/
In my experience the 'best'
)
In my experience the 'best' way to install drivers typically is iGPU first, NVidia second, and AMD always last. AMD seems the most finicky, unless you're trying to use the iGPU for crunching and then you really need to pay attention to the version number as they don't all work here at Einstein - but that's a topic in a different thread :-)
I have had to physically plug the monitor into the AMD card, and the AMD card needs to be in the slot closest to the CPU and leave it there in order to get various GPU OC tools to read / work correctly. I have had to switch the monitor back and forth between the different cards to set OC and now I just always do that out of habit. Once all OC is set I typically make a final switch of the monitor to the iGPU.
For others reading who have both AMD and NVidia ... at one point I had to put the AMD card in the first slot to get all the tools to work properly. Nvidia doesn't much care which slot it is in, unless it is the only GPU and then first slot is best (actually MOBO limitation rather than GPU itself). I still use this configuration out of habit - not sure it is really necessary but it does always work for me.
I've noticed that if I crash the AMD driver then if I suspend GPU tasks and then 'Reset' the AMD driver (most OC tools have either reset or apply defaults button) I can start crunching again just fine. On the other hand, if NVidia driver crashes then it requires a system boot to get things back in order.
- Have a crunching day :-)
--------------------------
- Crunch, Crunch, Crunch -
--------------------------
thanks guys, I now have work
)
thanks guys,
I now have work to do on my system.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
I normally don't get errors
)
I normally don't get errors on BRP4G and BRP5 at all.
http://einsteinathome.org/host/11368189/tasks
This is with two HD 7790s on a Haswell Z87 motherboard, each card supported by a CPU core. The temps run in the low-to-mid 70's during the cooler months, up to about 80C in the summer since the heastsinks on these cards (Power Color) are small, but I do have a 120mm side fan. They run at the default GPU clock of 1030 MHz; I never overclock, and avoid the factory-overclocked cards, since the effect is the same.
However, I have seen occasional errors in the past, I think mainly on the BRP4Gs. They were due to my anti-virus, which apparently didn't like something every couple of days. I was using Microsoft Security Essentials at the time, though I have tried others. You might find one that works with the right exclusions, but exclusions don't always help (I would try Panda Free next if I were doing it; it is very light on resources and generally does not cause trouble). At the moment, I am just using Windows Defender, which is anti-spyware only on Win7 64-bit.