Wrong CPU Speed detected w/25.8

[TheOutcaste]

Found a couple of threads mentioning the Completion times being off, but none that mention the CPU Speed, which will obviously affect the completion time calculation.

I just upgraded 3 systems from version 24.14 to 25.8 build 5. Version 25.8 is incorrectly reporting the CPU speed, which is affecting Estimated Completion times. (by 7 years on the PIII system)

Processor          Version 24.14  Version 25.8
  Type               Reports        Reports
Pentium D 3.00 GHz  2992.49 GHz     200 MHz
Celeron 1400 MHz     935.47 MHz     449 MHz
(underclocked to 933 MHz)
PIII 700             784.96 MHz     100 MHz
(OC to 785 MHz)

The assignment for the PIII should take 15 months, but this version is reporting 87 months. My concern is that this exponent will eventually be re-assigned because the server thinks it will take too long to complete, resulting in lost time and duplicated work.

Jerry

[Prime95]

What is weird is that the cpu speed detection code did not change between version 25 and version 24.

I'll look at it, but in the meantime look at undoc.txt. There should be a CPUSpeed entry you can use to override the detected speed.

[TheOutcaste]

Checked the old local.ini file for each of those PCs, they all had the CPUSpeed=??? entry matching the speed V 25.8 is reporting. The local.ini files were copied over from old 486/PII systems to these newer ones that replaced them.
Deleted that entry* and they are all reporting the correct speed now.

Seems V 24.14 is ignoring the setting in the local.ini file, whereas V 25.8 is using it.
Perhaps V 24.14 only uses that setting if it can't determine a speed, and V 25.8 uses it regardless?

Thanks,

Jerry
*EDIT: Deleted in local.txt of course

[ET_]

Originally Posted by TheOutcaste

Found a couple of threads mentioning the Completion times being off, but none that mention the CPU Speed, which will obviously affect the completion time calculation.

I just upgraded 3 systems from version 24.14 to 25.8 build 5. Version 25.8 is incorrectly reporting the CPU speed, which is affecting Estimated Completion times. (by 7 years on the PIII system)

Processor          Version 24.14  Version 25.8
  Type               Reports        Reports
Pentium D 3.00 GHz  2992.49 GHz     200 MHz
Celeron 1400 MHz     935.47 MHz     449 MHz
(underclocked to 933 MHz)
PIII 700             784.96 MHz     100 MHz
(OC to 785 MHz)

The assignment for the PIII should take 15 months, but this version is reporting 87 months. My concern is that this exponent will eventually be re-assigned because the server thinks it will take too long to complete, resulting in lost time and duplicated work.

Jerry

Did you set your machine to run 24/7?

Luigi

[TheOutcaste]

Originally Posted by ET_

Did you set your machine to run 24/7?

Luigi

Yes.

The difference is that Version 24 doesn't seem to use the CPUSpeed entry in local.ini, whereas Version 25 does use that entry in local.txt.

Jerry

[joblack]

Originally Posted by TheOutcaste

Yes.

The difference is that Version 24 doesn't seem to use the CPUSpeed entry in local.ini, whereas Version 25 does use that entry in local.txt.

Jerry

The estimated time of my Intel Atom has also increased by the factor 3 ... I´m trying 29 and will see if it will changes ...

[joblack]

Originally Posted by joblack

The estimated time of my Intel Atom has also increased by the factor 3 ... I´m trying 29 and will see if it will changes ...

Ok - I switched again to the 27 (linux 64 bit). Instead an estimated year of completion it shows 2.5 months.

Thou I think between v27 and v28 something changed in the cpu detection and that leads to wrong time estimation

[Prime95]

Originally Posted by joblack

Ok - I switched again to the 27 (linux 64 bit). Instead an estimated year of completion it shows 2.5 months.

Thou I think between v27 and v28 something changed in the cpu detection and that leads to wrong time estimation

Bad news for you. The CPU detection code did change and the estimated completion date in 25.8 is much more accurate. Atom = not-good-for-LL.