Posts by EG

1) Message boards : Number crunching : Not getting work units (Message 2153)
Posted 6 Dec 2019 by EG
Post:
Never mind.....
2) Message boards : Number crunching : Can the config file handle different models of NVIDIA gpus in the same rig ? (Message 2152)
Posted 6 Dec 2019 by EG
Post:
If the owner of this project see this ... please add the possibility to have a separate config for each GPU :)


Already able to do this....

What you can do is use two separate instances of the client set to use the single GPU in each instance....

One client will read the config for that app from it's own directory and the other will read the config for the second app from it's own directory.

Since it is two complete separate instances of Boinc one will not see (or interfere) with the other....

Then you use an app_config or is it app_info file to point the client to the proper GPU excluding the other GPU...

There is documentation all over the web on how to do this....

It's called "Running Multiple Instances"
3) Message boards : Number crunching : Computations errors (Message 2141)
Posted 2 Dec 2019 by EG
Post:
Every workunit that fails is also sent to another computer up to 6 times before it is marked as bad for the Admin to look at it. Your computersare also hidden so I can't tellthe last digit on the failed files, the closer to _6 it is the better for everyone.


Got one for ya Mikey...

https://boinc.thesonntags.com/collatz/workunit.php?wuid=48309481

Sent out 7 times I was the last one....

Most of the more recent ones that I've looked at are on their 4th or 5th machine......

I'm down to one or two every couple of hours now, it seems like they are making their way through the error/resend system, albeit slowly...

I guess that shares the pain.... If seven different systems call it bad I guess it's bad....
4) Message boards : Number crunching : Computations errors (Message 2138)
Posted 2 Dec 2019 by EG
Post:
Update:

In the last 12 hours, I haven't had a single bad wu.

So I think that those I was getting were bad wu's still in the wu cache of abort resends.... Blackhawk 3 is a dual 1080Ti running at 240 seconds or so, it goes thru a LOT of wu's fairly rapidly so I think it was drawing from the abort resend cache faster than most.... I think a lot of users just aborted their cache's that were full of bad wu's rather than wait for them to clear. Then us guys with a fast machine started getting them in bunches.

All seems good now.... (fingers crossed)
5) Message boards : Number crunching : Computations errors (Message 2137)
Posted 2 Dec 2019 by EG
Post:
Yeah I had a bunch of those two second WU's also, but since everyone else got them I let it fly....... This too shall pass....

But these are taking 15 -25 sec before they error out otherwise they are identical 0 byte files that produce nothing to return....

Same problem, only now they are taking a bit longer..... the rate has dropped since I posted that also. Now down to about 1 an hour.....

This is just my frustrations coming out....

Jon will fix it when he gets around to it, I know this....
6) Message boards : Number crunching : Computations errors (Message 2135)
Posted 2 Dec 2019 by EG
Post:
I don't know what is going on but I am literally getting hundreds of these an hour......

12/2/2019 00:01:11 | collatz | Computation for task collatz_sieve_edf089d8-5f9f-4d8f-89c3-5f7b799cc08e_2 finished
12/2/2019 00:01:11 | collatz | Output file collatz_sieve_edf089d8-5f9f-4d8f-89c3-5f7b799cc08e_2_r151339135_0 for task collatz_sieve_edf089d8-5f9f-4d8f-89c3-5f7b799cc08e_2 absent
12/2/2019 00:01:11 | collatz | Starting task collatz_sieve_f03b5b8d-a09d-4ab5-b1f2-e9832b9ffce5_5
12/2/2019 00:01:12 | collatz | Computation for task collatz_sieve_ade2594c-bf1f-46ae-a41b-2886719a8d0e_4 finished
12/2/2019 00:01:12 | collatz | Output file collatz_sieve_ade2594c-bf1f-46ae-a41b-2886719a8d0e_4_r1942197839_0 for task collatz_sieve_ade2594c-bf1f-46ae-a41b-2886719a8d0e_4 absent
12/2/2019 00:01:12 | collatz | Starting task collatz_sieve_c7b68a80-7837-4e68-a042-bf94ae1ca7ce_3
12/2/2019 00:01:55 | collatz | Computation for task collatz_sieve_f03b5b8d-a09d-4ab5-b1f2-e9832b9ffce5_5 finished
12/2/2019 00:01:55 | collatz | Output file collatz_sieve_f03b5b8d-a09d-4ab5-b1f2-e9832b9ffce5_5_r1981029832_0 for task collatz_sieve_f03b5b8d-a09d-4ab5-b1f2-e9832b9ffce5_5 absent
12/2/2019 00:01:55 | collatz | Computation for task collatz_sieve_c7b68a80-7837-4e68-a042-bf94ae1ca7ce_3 finished
12/2/2019 00:01:55 | collatz | Output file collatz_sieve_c7b68a80-7837-4e68-a042-bf94ae1ca7ce_3_r1930321496_0 for task collatz_sieve_c7b68a80-7837-4e68-a042-bf94ae1ca7ce_3 absent

This is from Blackhawk 3 which has been running for well over a year with no changes.... All of a sudden I'm getting masses of faulty wu's.... It's been happening since the last server issue a few days back and hasn't stopped....

I haven't changed a thing on my end...
7) Message boards : Number crunching : Computations errors (Message 2134)
Posted 30 Nov 2019 by EG
Post:
Thanks mikey,

https://boinc.thesonntags.com/collatz/results.php?hostid=846493
AMD Radeon VII seems to work well
verbose=1
kernels_per_reduction=48
sleep=1
threads=6
lut_size=16
reduce_CPU=0
sieve_size=30
cache_sieve=1

AMD Radeon VII Fails on every task
verbose=1
kernels_per_reduction=48
sleep=1
threads=9
lut_size=18
reduce_CPU=0
sieve_size=30
cache_sieve=1


AMD's do not handle 9 threads, 8 max...... should be erroring all wu's using 9 threads...
8) Questions and Answers : Windows : Windows GPU apps (Message 2120)
Posted 26 Nov 2019 by EG
Post:
Hello,

I have an AMD 2700X Windows 10 x64 system with a Vega 64 GPU running Collatz on the GPU only. I noticed yesterday that there were 2 different apps in my folder...

"collatz_sieve_1.30_windows_x86_64__opencl_ati_gpu"

"collatz_sieve_1.30_windows_intelx86_opencl_ati_gpu.exe"

Now being an AMD CPU, I imagine that the first app ("x86_64") is the proper one to run. But why would the second ("intel_x86") one be used? The Intel app doesn't error and completes fine with the same optimization settings I set for the first app.

Just curious as to why I'm getting both apps and WUs? And more importantly, is there a way to specify only the x86_64 or the intelx86 app to download and run???

Thanks,
blue


Hi Blue, good to see you here...

Simple answer?

Do Nothing.... they are supposed to be there cause there is the potential to use both...

The Reason, collatz has always sent out both 64 bit and 32 bit work units the app marked x86_64 does the 64 bit wu's and the one marked intelx86 does the 32 bit wu's it will issue both types of wu's randomly depending on what work is available.... you will also have two .config files conforming to the app, just place the same optimizations in both files and your good to go...

Why? it keeps collatz backwards compatible with old 32 bit op systems......

BTW: 32 bit is no more faster/slower than the 64 bit wu's, and they both give out the same credits.

Your collatz is as normal.... (just make sure you have your optimizations in both .config files as you will only get the .config file when you download the app needing it)

One more point, you have both apps and .config files cause you already have crunched both, I'll bet without noticing any change in it's crunching performance
9) Message boards : Number crunching : Average work done (Message 1970)
Posted 6 Oct 2019 by EG
Post:
Well, all I can say is thank you for the answer and I mostly wish I hadn't asked the question.


Rac is pretty much a useless measure of performance...., some people are emotionally tied to it though like it's the measure of all great crunchin power...

I have projects I haven't run in over a year that still show some Rac....

Tells me everything I need to know....
10) Message boards : Number crunching : First 100 Billion User! (Message 1948)
Posted 28 Sep 2019 by EG
Post:
Congratulations to Jack Hiker of Sicturadastra......

First to 100 BILLION points!

Way to go Jack!
11) Message boards : Number crunching : Why am I not getting work units? (Message 1939)
Posted 9 Sep 2019 by EG
Post:
Aii I can feel at the moment is a sense of fustration, what do I need an invite nowadays to crunch.

I thought those days were well and truly behind us.

3 x PCs, GPUs ready to crunch and yet after 3 days, not one workunit, yet there are computers in the top 10 who have at least 300 - 400 workunits.

Emailed the admin, but nothing as yet.

This is what I get in my BOINC message log.

8/09/2019 19:45:49 | collatz | update requested by user
8/09/2019 19:45:53 | collatz | Sending scheduler request: Requested by user.
8/09/2019 19:45:53 | collatz | Requesting new tasks for NVIDIA GPU
8/09/2019 19:45:54 | collatz | Scheduler request completed: got 0 new tasks

At my wits end, have crunched here before, and now cannot even help out my team.

Regards


If your running any CPU projects suspend them, If that doesn't work, leave them suspended and restart Boinc.....

I've had this happen to me before.... Collatz will not run or download WU if there is insufficient CPUs to run with the GPUs....

Everytime it has happened, suspending the CPU projects cleared the issue....

When I set up a machine, I reserve two cores for access, one core for each GPU and the rest go to the CPU project which is limited to that number cores by an app_config file limiting the allowed number of simultaneous clients... I let the op system figure out which core gets which work....
12) Message boards : Number crunching : Computation Errors (Message 1936)
Posted 6 Sep 2019 by EG
Post:
I changed my config file to lower the Threads and sieve size and the work units run now.


That's very good!!
Your wu's are doing this:
428.96 0.25 40,290.60

and my 1080Ti is doing this:
257.90 3.00 27,610.74

my settings are:
..............
Your wu's run slower but you get get more credits for each one.


Yours:
27610/257= 107.43 credits per second....

His:
40290/428= 94.14 credits per second....

Your tighter configuration delivers more efficiency resulting in higher point totals even though the points per WU are less....

You and I Mikey have been doing this long enough to understand this implicitly, but it is good to illustrate it clearly from time to time....
13) Message boards : Number crunching : Optimizing the apps (Message 169)
Posted 25 Apr 2018 by EG
Post:
[quote]Thank you for this info. I am successfully running with the default values (no config file) and would like to optimize my settings. What would help greatly is if I knew what the default settings are to use as a starting point, since I have not been able to find any settings for my NVIDIA GT 730 GPU. While I fully realize this is not a powerful GPU, it is nevertheless better than nothing. Any help in this area would be greatly appreciated, since am not a developer or programmer.[quote]

BUMP


Try these:
verbose=1
kernels_per_reduction=48
threads=8
lut_size=15
sleep=0
cache_sieve=1
reduce_cpu=0
sieve_size=30

Adjust from there
14) Message boards : Number crunching : Optimizing the apps (Message 119)
Posted 23 Apr 2018 by EG
Post:
.......
At the moment I am in 3rd place: https://boinc.thesonntags.com/collatz/top_hosts.php


Don't mean to burst your bubble but......

Your basing your ideals of how well your machine is producing based upon one week old Rac?

It's going to take two months of solid steady crunching for Rac to stabilize sufficiently to be any judge of how well a machine is producing.

I've got a machine that is producing 12.8 million a day right now all by it's lonesome, but it's Rac is only 3.2 million.

So your producing at 227 seconds per WU on average. that is roughly 380.5 WU's per day, (86400 sec/day divided by your 227 per WU)

That 380.5 WU per day times the 27200 credit per WU average your getting equals 10.35 million per day.

Pencils out right where a 1080Ti should be,. twice the Rac your now demonstrating.

We have been doing this a while.

That machine will probably be down below 20th position when the dual machine Racs catch up.... Pretty much average for a single 1080Ti....
15) Message boards : Number crunching : Optimizing the apps (Message 105)
Posted 23 Apr 2018 by EG
Post:
I have found that with one WU on an Nvidia 1080 Ti the GPU is only ~65-70% busy.
That's why I let 3 WUs run in parallel, so it is also ensured that the pauses are filled in when terminating/starting the WUs.
one WU needs 5:32, three WUs in parallel need 11:12 each
I control the GPU load/temperature with the MSI Afterburner.


Sounds to me like your not using the configuration file....

Start using it and you will soon drop to one per GPU....

Two (or more) per GPU for me resulted in WU's erroring out at a far greater rate than any benefit gained..
16) Message boards : News : Use at your own risk (Message 79)
Posted 21 Apr 2018 by EG
Post:
Probably a dumb question, but I don't know how to remove and re-attach the project?? The "Remove Project" button in BOINC Manager is greyed out, so I can't use it...Any ideas..Much appreciated!
Mike, Bristol, UK


On the project screen in Boinc Manager highlight the Collatz entry in the project listing section and the remove button should light up.
17) Message boards : News : Use at your own risk (Message 75)
Posted 20 Apr 2018 by EG
Post:
Hi Jon,

As offered I've setup my last dual 7990 machine on this and it's been running for about 6 hours now. I'm using the exact same setup I was using on the old WU's. The WU's seem to be completing in the same amount of time 12.5 minutes on average for both the 7990's and 7970 I've got it on.

The 1080 is doing them in about 5.5 minutes so far without error. Will let you know if I run into anything strange.

One thing to note, the server is giving out a cache of over 1k WU's per machine.(almost 1200 in fact) Previously it was giving out 100 WU's per GPU to ease potential server issues.

I know there are many many details that need to wait cause of more important things. but this is one that you should probably put on the to do list....

And let me know if you want the new badges I designed I still have them if you want to implement them in the future.

Elmer
18) Message boards : Number crunching : Where is the AMD-App ? (Message 16)
Posted 18 Apr 2018 by EG
Post:
There is no AMD OpenCL App ?

On https://boinc.thesonntags.com/collatz/prefs.php?subset=project there is only:
use CPU
use NVIDIA GPU
use Intel GPU
but no AMD GPU !


I'm sure there is one coming.......
Be patient..... :-)
19) Message boards : News : Use at your own risk (Message 15)
Posted 18 Apr 2018 by EG
Post:
What I am noticing as the work gets validated is that the credits seem OK but the accumulated credits are not getting incremented properly. So I see a credit award of 35K for a unit but my total credits get incremented by 350...


Probably a misplaced decimal point...




©2021 Jon Sonntag; All rights reserved