Error: Round Off???

edorajh · 2003-11-24, 09:02

I just came to work and find several errors in my results file. It is "Error: Round Off" error.

Prime95 continue to run with message: "Continuing from last save file". How serious is this error? Is there any point to continue to run Prime95 on this machine, or will it just produce bad results?

garo · 2003-11-24, 17:40

If the continuing from save file is followed in a little while with a "Disregard last error" then it is okay. These errors are harmless. Check your results.txt

edorajh · 2003-11-24, 18:13

I'm now at home, and I don't remember if there is a message "Disregard last error". Will check it first thing tomorow morning.

What if there is no such message?

garo · 2003-11-24, 19:02

If there is no such message then either there was a transient error which was taken care of when it continued from the last save file. Or the result is bad. I'd recommend putting the machine on doublechecking for a little while just to make sure that all your results are correct.

edorajh · 2003-11-24, 19:50

Thanks garo.

Can I stop current LL test and pick some double checking assignments? If yes, what files I must copy in order to continue from current LL test (once doublechecking is over?).

Thanks.
Edo

garo · 2003-11-24, 20:27

I'd say let the current test run its course and then pick up doublechecks. Just set the work preference in the menu to doublechecks. If you do want to do the doublechecks first, stop Prime95 and exit. Then open the worktodo.ini and rearrage the exponent lines as you wish. Save and close before starting prime95 again. The savefiles will not be delted and the LL test will pick up from where it left off.

BTW, you can get doublechecks now by increasing teh "days of work" field after changing the worktype.

pakaran · 2003-11-25, 01:36

Quote:

Originally posted by garo
I'd say let the current test run its course and then pick up doublechecks. Just set the work preference in the menu to doublechecks. If you do want to do the doublechecks first, stop Prime95 and exit. Then open the worktodo.ini and rearrage the exponent lines as you wish. Save and close before starting prime95 again. The savefiles will not be delted and the LL test will pick up from where it left off.

BTW, you can get doublechecks now by increasing teh "days of work" field after changing the worktype.

Also, you'll have to wait for George's next DB to see if your results are correct (and if they don't match, the error may be in the first test).

We can help you with analyzing the DB.

Nathan

edorajh · 2003-11-25, 09:50

There is no Disregard last error message

. I find only several these lines:

ERROR: ROUND OFF (0.4985351563) > 0.40
Continuing from last save file.

It doesn't look good to me. I suspect that this happens because machine run during the weekend (first time for so long) and CPU was overheated. Is it possible?

Is there any chance it will be OK result, or should I stop processing this exponent?

lycorn · 2003-11-25, 12:33

Analysis of past data shows that for that type of errors the chance that the result is bad is in the order of 50 to 60%. It is your decision to stop, or not, the test, as the result will be DCed anyway. How far you have gone in this test may also be a factor to consider.
I don´t know the conditions of the room where the suspect machine is, is the temperature likely to rise during the weekend, because the room is closed or whatever? Check the CPU and mobo temperature, and let us know the values. What type of CPU/mobo are you using?
But in any case, my advice is to switch the machine to DC for the next couple of assignments. If errors still persist, and mismatches with the previous LL tests, you could switch it to Trial Factoring, which is less stressful to the ensemble CPU/chipset/memory. And besides, TF is needing to be pushed forward, before first-time LL catches up...

edorajh · 2003-11-25, 14:21

I perform 21% on this exponent.

There is no temperature changes during weekends in my office. Not sure how to check CPU and mobo temperature.

I'll definitelly run some doublecheckings.

So, what others suggest to give up on this exponent or to stick with it? I know it's my decision, but I'm just interested in your opinions.

50-60% that result will be bad... it's sounds bad.

Reboot It · 2003-11-25, 14:51

If it were me in this situation at 21%, I would do almost what garo suggested, but in a slightly different order:
change work type to DoubleChecks,
extend number of days of work to collect 2 or 3 DCs,
manual connect to get them,
stop Prime95 and exit,
edit WTD to put all the DCs at the top and leave the 21% first time LL at the bottom of the list
and restart.

I would wait for a couple of weeks until at least the first 2 DCs had come back as either Confirmed or Triple Check Required. If all were confirmed I would change back to first time LLs and allow the interrupted LL test to complete. If any failed, I would consider dropping the 21% test and sorting out the PC (torture test, cooling, etc).

YMMV, but this is what I would do. HTH.

2003-11-24, 09:02	#1
edorajh Oct 2003 Croatia 2³·3·19 Posts	Error: Round Off??? I just came to work and find several errors in my results file. It is "Error: Round Off" error. Prime95 continue to run with message: "Continuing from last save file". How serious is this error? Is there any point to continue to run Prime95 on this machine, or will it just produce bad results?

2003-11-24, 19:02	#4
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	If there is no such message then either there was a transient error which was taken care of when it continued from the last save file. Or the result is bad. I'd recommend putting the machine on doublechecking for a little while just to make sure that all your results are correct. Last fiddled with by garo on 2003-11-24 at 19:03

2003-11-25, 14:51	#11
Reboot It Aug 2002 London, UK 5C₁₆ Posts	If it were me in this situation at 21%, I would do almost what garo suggested, but in a slightly different order: change work type to DoubleChecks, extend number of days of work to collect 2 or 3 DCs, manual connect to get them, stop Prime95 and exit, edit WTD to put all the DCs at the top and leave the 21% first time LL at the bottom of the list and restart. I would wait for a couple of weeks until at least the first 2 DCs had come back as either Confirmed or Triple Check Required. If all were confirmed I would change back to first time LLs and allow the interrupted LL test to complete. If any failed, I would consider dropping the 21% test and sorting out the PC (torture test, cooling, etc). YMMV, but this is what I would do. HTH. Last fiddled with by Reboot It on 2003-11-25 at 14:52

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Skylake FMA3 round off error	tha	Hardware	17	2016-02-07 04:50
Round off error	Androx72	Software	2	2013-02-28 00:00
mprime ROUND OFF ERROR: Triple-check advised?	Bdot	Software	5	2012-12-22 22:34
HDT55TWFK6DGR voltage and round off error	RickC	Hardware	2	2011-02-19 04:07
Another Round Off Error Issue	PhilF	Software	12	2005-07-02 19:03

2003-11-24, 17:40	#2
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	If the continuing from save file is followed in a little while with a "Disregard last error" then it is okay. These errors are harmless. Check your results.txt

2003-11-24, 18:13	#3
edorajh Oct 2003 Croatia 710₈ Posts	I'm now at home, and I don't remember if there is a message "Disregard last error". Will check it first thing tomorow morning. What if there is no such message?

2003-11-24, 19:50	#5
edorajh Oct 2003 Croatia 2³·3·19 Posts	Thanks garo. Can I stop current LL test and pick some double checking assignments? If yes, what files I must copy in order to continue from current LL test (once doublechecking is over?). Thanks. Edo

2003-11-24, 20:27	#6
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	I'd say let the current test run its course and then pick up doublechecks. Just set the work preference in the menu to doublechecks. If you do want to do the doublechecks first, stop Prime95 and exit. Then open the worktodo.ini and rearrage the exponent lines as you wish. Save and close before starting prime95 again. The savefiles will not be delted and the LL test will pick up from where it left off. BTW, you can get doublechecks now by increasing teh "days of work" field after changing the worktype.

2003-11-25, 09:50	#8
edorajh Oct 2003 Croatia 710₈ Posts	There is no Disregard last error message . I find only several these lines: ERROR: ROUND OFF (0.4985351563) > 0.40 Continuing from last save file. It doesn't look good to me. I suspect that this happens because machine run during the weekend (first time for so long) and CPU was overheated. Is it possible? Is there any chance it will be OK result, or should I stop processing this exponent?

2003-11-25, 12:33	#9
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 7×13×17 Posts	Analysis of past data shows that for that type of errors the chance that the result is bad is in the order of 50 to 60%. It is your decision to stop, or not, the test, as the result will be DCed anyway. How far you have gone in this test may also be a factor to consider. I don´t know the conditions of the room where the suspect machine is, is the temperature likely to rise during the weekend, because the room is closed or whatever? Check the CPU and mobo temperature, and let us know the values. What type of CPU/mobo are you using? But in any case, my advice is to switch the machine to DC for the next couple of assignments. If errors still persist, and mismatches with the previous LL tests, you could switch it to Trial Factoring, which is less stressful to the ensemble CPU/chipset/memory. And besides, TF is needing to be pushed forward, before first-time LL catches up...

2003-11-25, 14:21	#10
edorajh Oct 2003 Croatia 2³×3×19 Posts	I perform 21% on this exponent. There is no temperature changes during weekends in my office. Not sure how to check CPU and mobo temperature. I'll definitelly run some doublecheckings. So, what others suggest to give up on this exponent or to stick with it? I know it's my decision, but I'm just interested in your opinions. 50-60% that result will be bad... it's sounds bad.