Important news on BRP7 and FGRPB1 work on E@H

TRAPPIST-713
TRAPPIST-713
Joined: 13 May 20
Posts: 12
Credit: 2486692942
RAC: 1457481

Bernd Machenschalk wrote: It

Bernd Machenschalk wrote:

It looks to me that the validation rates got a lot better with time, without us changing anything on the project side...

In part this could be because users have better adapted to BRP7. For example, my RX 580 (Computer 12832734) generated 100% invalid results until I uninstalled AMD Adrenalin software and let windows 10 reinstall drivers. Now “validate error” dropped to normal rate.

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3953
Credit: 46799002642
RAC: 64206202

TRAPPIST-713 wrote: In part

TRAPPIST-713 wrote:

In part this could be because users have better adapted to BRP7. For example, my RX 580 (Computer 12832734) generated 100% invalid results until I uninstalled AMD Adrenalin software and let windows 10 reinstall drivers. Now “validate error” dropped to normal rate.

It's more than that. a lot of users who had no issues with their systems, and made no changes to their systems (primarily Linux users) saw a large decrease in the invalids of their own systems after the new apps were released.

_________________________________________________________________________

Ereignishorizont
Ereignishorizont
Joined: 17 May 21
Posts: 19
Credit: 3013592861
RAC: 1067045

I can confirm that. The last

I can confirm that. The last days, I had max. 1 invalid per day or 0,3% without any changes on my computers. Before the invalid rate was about ~10% or even more.

lunkerlander
lunkerlander
Joined: 25 Jul 18
Posts: 46
Credit: 31464094
RAC: 0

Quick question: Is this work

Quick question: Is this work unit unusual because it has 1 validation error, 1 invalid result, and finally a 2 valid results, or does this happen from time to time? https://einsteinathome.org/workunit/763986594

The reason why I'm asking is that I just reinstalled BOINC and got 2 validation error tasks (BRP7) on my first 2 work units and was wondering if it was something with my specific PC. 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4964
Credit: 18720019099
RAC: 6403337

No, not unusual. Just the

No, not unusual. Just the same old problem of two hosts running the same app and OS ganging up on the other wingmen.

 

DF1DX
DF1DX
Joined: 14 Aug 10
Posts: 105
Credit: 3855226854
RAC: 4936685

For about two or three days

For about two or three days now, the rate of invalid / inconclusive BRP7 WUs on my host has been rising again at an unusually high rate.

In the last few weeks I regularly had about 17-20 invalid and 25-30 inconclusive WUs with the optimized app, which works very well here on the NV 4090.

Today, however, there are already 48 invalid and 126 inconclusive WUs.

Has anything changed in the validator or the data?

(As always, I lose with Linux against BRP7 0.12/Windows.)

Boca Raton Community HS
Boca Raton Comm...
Joined: 4 Nov 15
Posts: 240
Credit: 10562085586
RAC: 24523851

DF1DX wrote: For about two

DF1DX wrote:

For about two or three days now, the rate of invalid / inconclusive BRP7 WUs on my host has been rising again at an unusually high rate.

In the last few weeks I regularly had about 17-20 invalid and 25-30 inconclusive WUs with the optimized app, which works very well here on the NV 4090.

Today, however, there are already 48 invalid and 126 inconclusive WUs.

Has anything changed in the validator or the data?

(As always, I lose with Linux against BRP7 0.12/Windows.)

 

It's not just you. My invalid rate went up a good amount over the last few days. It is also not just your 4090, our 4090's and other GPUs also have increased (on Windows 11 machines, Linux machines with the optimized app). Maybe the data being analyzed has changed leading to a slightly higher invalid rate? I could be completely wrong about that though.

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3061
Credit: 4966177686
RAC: 1414371

Boca Raton Community HS

Boca Raton Community HS wrote:

It's not just you. My invalid rate went up a good amount over the last few days. It is also not just your 4090, our 4090's and other GPUs also have increased (on Windows 11 machines, Linux machines with the optimized app). Maybe the data being analyzed has changed leading to a slightly higher invalid rate? I could be completely wrong about that though.

I'll chime in too, it is not just the 4090's.  My 3080 & 2070Super also have received 44 invalids since Nov 14.  The last invalid previously was on Oct 28.  And I am running Linux Ubuntu 22.04 LTS with the optimized app also.

George

Proud member of the Old Farts Association

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3953
Credit: 46799002642
RAC: 64206202

Boca Raton Community HS

Boca Raton Community HS wrote:

DF1DX wrote:

For about two or three days now, the rate of invalid / inconclusive BRP7 WUs on my host has been rising again at an unusually high rate.

In the last few weeks I regularly had about 17-20 invalid and 25-30 inconclusive WUs with the optimized app, which works very well here on the NV 4090.

Today, however, there are already 48 invalid and 126 inconclusive WUs.

Has anything changed in the validator or the data?

(As always, I lose with Linux against BRP7 0.12/Windows.)

 

It's not just you. My invalid rate went up a good amount over the last few days. It is also not just your 4090, our 4090's and other GPUs also have increased (on Windows 11 machines, Linux machines with the optimized app). Maybe the data being analyzed has changed leading to a slightly higher invalid rate? I could be completely wrong about that though.

Bernd did say before that some of the invalids are data driven. IE, it's some difference that derives from the data itself and propagates to a different final result, and not necessarily a problem with the app. 

i think we will occasionally see this happen with invalids waxing and waning.

_________________________________________________________________________

HWpecker
HWpecker
Joined: 27 Jan 22
Posts: 25
Credit: 77748827
RAC: 0

I have a good chunk of

I have a good chunk of invalids gained on nov 14 too, BRP7 (cuda102/test)

Also felt I had a lower credits gained per day on nov 11/12.

Could it be that some validated units are rerun (it's a testapp)?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.