covid moonshot bad wu setup

[Jaqui]

13422 (2191, 3, 1)
My computer has 1.85 days to complete this work unit.

work unit eta: 3.66 days

This wu is only at 4.57% and has been started within the last hour.
This seems to be really badly arranged if it is less than 2 days to complete by design and the radeon r7 gpu will take nearly 4 days.

[JohnChodera]

Eep! Have you noticed other 13422 WUs that take this long to complete? Is it just a few RUNs that seem to have this problem?

These WUs are designed to take ~2 hours to complete, so it's surprising they would consume 3.66 days unless there are a few problematic RUNs that snuck through.

~ John Chodera // MSKCC

[bruce]

work unit eta: 3.66 days

How long have you been processing that WU, and with what GPU?

The eta is truly an ESTIMATE which starts out rather distant from accurate and it's refined as you establish a pattern of usage. Is your power saving setting disabled (processing 24x7 sure helps). Is the POWER slider set to FULL or MEDIUM?

[Jaqui]

Eep! Have you noticed other 13422 WUs that take this long to complete? Is it just a few RUNs that seem to have this problem?

These WUs are designed to take ~2 hours to complete, so it's surprising they would consume 3.66 days unless there are a few problematic RUNs that snuck through.

~ John Chodera // MSKCC

only the very next wu that comp got.
13422 (2211, 22, 2) 3.95 days
It has a dual gpu radeon graphics so runs dual gpu target wu.
Well I currently have 2 13422 wus taking a lot longer than 2 hours.

Bruce,
it started less than 2 hours before I made the original post.
and the time to completion has dropped dramatically, to 2.39 days now.

[foldy]

Maybe your AMD dual GPU is too weak and should be excluded from p13422 assignment?

[Jaqui]

Maybe your AMD dual GPU is too weak and should be excluded from p13422 assignment?

maybe, but the 134xx wus are meant to help refine the wu design and allocation according to another thread about them.
the first one I posted about has dropped the eta so it might finish in time, the second has delayed the eta so it looks like it won't.
either way, the wu and completion time help in setting up the wus in the future.


I posted so the devs can know how the specific work units show on my system and monitor for completion from my comp. This helps them get data to truly build good work units for all.

[muziqaz]

I hope your GPU is not AMD Radeon HD 5970.

[Jaqui]

AMD Radeon R6 dual gpu ( 6 cores )
AMD A10 ( 4 cores )

[muziqaz]

AMD Radeon R6 dual gpu ( 6 cores )
AMD A10 ( 4 cores )

There is no R6 dual GPU you have been caught by AMD marketing. Its very slow integrated GPU. 1 GPU

You you might have one of these A10 7300, A10 7350B, A10 7600P, A10 8700P, A10 8700B, A10 Micro 6700T. All of them are APUs, and their GPUs are super slow

[Yeroon]

If you are folding on the cpu, maybe try pausing that WU to see if the gpu wu picks up ppd. I've noticed some of the 134xx units like to have better cpu access to keep the gpu better fed. (Dual rx470 and r5 3600 cpu)

[markdotgooley]

On my RTX 2060 and RTX 2060 KO (installed on the same motherboard) these COVID Moonshot units usually take just over 3 hours each and I’m usually credited about 200,000 points each. I think that’s a bit too generous but I’m not complaining. I have no clue whether all this is reasonable. If I were seriously hunting points (sort of kind of maybe?) I’d be grateful for these WUs: near (sometimes over) 3 million points a day on cards rated for maybe 1 to 1.1 million each.

But to copy an old advertisement:

“Then why are you crying?”

“Have you seen the size of our power bill?”

[Jaqui]

You you might have one of these A10 7300, A10 7350B, A10 7600P, A10 8700P, A10 8700B, A10 Micro 6700T. All of them are APUs, and their GPUs are super slow

Bingo!
A10 8700P
But there is a dedicated separate GPU. The folding app does find 2 gpus.
1 r7 and 1 r6

[BobWilliams757]

You you might have one of these A10 7300, A10 7350B, A10 7600P, A10 8700P, A10 8700B, A10 Micro 6700T. All of them are APUs, and their GPUs are super slow

Bingo!
A10 8700P
But there is a dedicated separate GPU. The folding app does find 2 gpus.
1 r7 and 1 r6

If I remember correctly, differing versions used the R6 or R7 integrated graphics. But all versions used the R7 drivers, so it appears that both or either may be working. The real differences between the versions were the number of shaders and clock speeds. Really very similar to what they are doing now through the current APU models. As an example, my Ryzen 2400G identifies to FAH as a Vega 8 mobile, even though it uses the Vega 11 graphics. Some of the newer generation chips identify differently, even though many use the Vega 8 at higher clock speeds than the ones of the generation of my APU.

Do you actually have another dedicated graphics card in your setup, or is it somehow folding two work units on just the one APU?







John Chodera,

I've only picked up a couple of work units of this project number, but they perform slowly on my 2400G Ryzen (Vega 11 onboard graphics). This same setup blazed through the 13421 work units in about two hours each. Project 13422 is in the 20 hour range.

PRCG

13422, 2634, 94, 2
13422, 2151, 28, 2
13422, 3164, 42, 1 (currently running)

They are returning fine, but take a bit of time on this APU rig. PPD return is actually just above average, where Project 13421 had freaky high PPD returns, in line with 2-3 times what I see as averages with larger work units. No complaints at all on my end, but your post earlier leads me to think you want these to run quickly, so it might mean excluding some more types of hardware.

[PantherX]

Could you please post the log section of your GPU? Project 13422 is a large Project while 13423 is a smaller one.

[BobWilliams757]

Could you please post the log section of your GPU? Project 13422 is a large Project while 13423 is a smaller one.

Assuming that was directed at me....

Code: Select all

22:57:08:******************************* System ********************************
22:57:08:            CPU: AMD Ryzen 5 2400G with Radeon Vega Graphics
22:57:08:         CPU ID: AuthenticAMD Family 23 Model 17 Stepping 0
22:57:08:           CPUs: 8
22:57:08:         Memory: 14.93GiB
22:57:08:    Free Memory: 12.10GiB
22:57:08:        Threads: WINDOWS_THREADS
22:57:08:     OS Version: 6.2
22:57:08:    Has Battery: false
22:57:08:     On Battery: false
22:57:08:     UTC Offset: -4
22:57:08:            PID: 12460
22:57:08:            CWD: C:\Users\rober\AppData\Roaming\FAHClient
22:57:08:             OS: Windows 10 Home
22:57:08:        OS Arch: AMD64
22:57:08:           GPUs: 1
22:57:08:          GPU 0: Bus:56 Slot:0 Func:0 AMD:5 Raven [Ryzen vega 8 mobile]
22:57:08:           CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': The
22:57:08:                 specified module could not be found.
22:57:08:
22:57:08:OpenCL Device 0: Platform:0 Device:0 Bus:56 Slot:0 Compute:1.2 Driver:2841.19
22:57:08:  Win32 Service: false

My mistake on mixing up project sprints.... I was thinking 13422 and 13421 were together, when in fact it was 20/21 then 22/23. But in my case 13420 and 13422 act the same, taking about 20 hours and giving average or slightly better TPF/PPD returns. Projects 13421 and 13423 act the same as well, but run quickly, give much lower TPF and higher PPD.


A recent 13423, TPF 1 minute 10 seconds.

Code: Select all

23:05:22:WU00:FS01:0x22:Project: 13423 (Run 290, Clone 91, Gen 0)
23:05:22:WU00:FS01:0x22:Unit: 0x0000000112bc7d9a5f388eecb320d989
23:05:22:WU00:FS01:0x22:Reading tar file core.xml
23:05:22:WU00:FS01:0x22:Reading tar file integrator.xml
23:05:22:WU00:FS01:0x22:Reading tar file state.xml.bz2
23:05:22:WU00:FS01:0x22:Reading tar file system.xml.bz2
23:05:22:WU00:FS01:0x22:Digital signatures verified
23:05:22:WU00:FS01:0x22:Folding@home GPU Core22 Folding@home Core
23:05:22:WU00:FS01:0x22:Version 0.0.11
23:05:22:WU00:FS01:0x22:  Checkpoint write interval: 50000 steps (5%) [20 total]
23:05:22:WU00:FS01:0x22:  JSON viewer frame write interval: 10000 steps (1%) [100 total]
23:05:22:WU00:FS01:0x22:  XTC frame write interval: 250000 steps (25%) [4 total]
23:05:22:WU00:FS01:0x22:  Global context and integrator variables write interval: 25000 steps (2.5%) [40 total]
23:05:28:WU00:FS01:0x22:Completed 0 out of 1000000 steps (0%)
23:06:39:WU00:FS01:0x22:Completed 10000 out of 1000000 steps (1%)
23:07:49:WU00:FS01:0x22:Completed 20000 out of 1000000 steps (2%)
23:08:58:WU00:FS01:0x22:Completed 30000 out of 1000000 steps (3%)
23:10:07:WU00:FS01:0x22:Completed 40000 out of 1000000 steps (4%)
23:11:17:WU00:FS01:0x22:Completed 50000 out of 1000000 steps (5%)
23:12:27:WU00:FS01:0x22:Completed 60000 out of 1000000 steps (6%)
23:13:37:WU00:FS01:0x22:Completed 70000 out of 1000000 steps (7%)
23:14:47:WU00:FS01:0x22:Completed 80000 out of 1000000 steps (8%)
23:15:57:WU00:FS01:0x22:Completed 90000 out of 1000000 steps (9%)
23:17:07:WU00:FS01:0x22:Completed 100000 out of 1000000 steps (10%)

A recent 13422, TPF average 11 minutes, 47 seconds.

Code: Select all

20:23:11:WU01:FS01:0x22:Project: 13422 (Run 2151, Clone 28, Gen 2)
20:23:11:WU01:FS01:0x22:Unit: 0x0000000212bc7d9a5f3993146c1bca09
20:23:11:WU01:FS01:0x22:Reading tar file core.xml
20:23:11:WU01:FS01:0x22:Reading tar file integrator.xml
20:23:11:WU01:FS01:0x22:Reading tar file state.xml.bz2
20:23:11:WU01:FS01:0x22:Reading tar file system.xml.bz2
20:23:11:WU01:FS01:0x22:Digital signatures verified
20:23:11:WU01:FS01:0x22:Folding@home GPU Core22 Folding@home Core
20:23:11:WU01:FS01:0x22:Version 0.0.11
20:23:11:WU01:FS01:0x22:  Checkpoint write interval: 50000 steps (5%) [20 total]
20:23:11:WU01:FS01:0x22:  JSON viewer frame write interval: 10000 steps (1%) [100 total]
20:23:11:WU01:FS01:0x22:  XTC frame write interval: 250000 steps (25%) [4 total]
20:23:11:WU01:FS01:0x22:  Global context and integrator variables write interval: 25000 steps (2.5%) [40 total]
20:23:37:WU01:FS01:0x22:Completed 0 out of 1000000 steps (0%)
20:36:30:WU01:FS01:0x22:Completed 10000 out of 1000000 steps (1%)
20:49:19:WU01:FS01:0x22:Completed 20000 out of 1000000 steps (2%)
21:02:33:WU01:FS01:0x22:Completed 30000 out of 1000000 steps (3%)
21:15:19:WU01:FS01:0x22:Completed 40000 out of 1000000 steps (4%)
21:27:59:WU01:FS01:0x22:Completed 50000 out of 1000000 steps (5%)
21:40:39:WU01:FS01:0x22:Completed 60000 out of 1000000 steps (6%)
21:53:17:WU01:FS01:0x22:Completed 70000 out of 1000000 steps (7%)
22:05:56:WU01:FS01:0x22:Completed 80000 out of 1000000 steps (8%)
22:18:35:WU01:FS01:0x22:Completed 90000 out of 1000000 steps (9%)
22:31:13:WU01:FS01:0x22:Completed 100000 out of 1000000 steps (10%)

I run my system with a slight memory overclock, and overclock the GPU a bit on the weekends when I'm home. But running at full stock specs, run time on the 13422 work units would probably be in the 22 hour range. They would make the Timeout deadline, but not by much. I would think that anything that is a Ryzen APU with the Vega 8 or 11 is going to be quite a distance away from a two or three hour turnaround if that is desired for that project.