LogViewer for HWINFO is available !

may or may not be relevant but towards the end of you r raw log file yo uhave an extra header row

1626635219439.png



If I delete that row there's still a gap between 6 mi9nutes into the run and approx 38 minutes


1626635358233.png
 
looking at the raw data that Gap starts at 2021-07-18 22:19:33 where the temp jumps from mid 50's to plus 70 then kicks back in at 2021-07-18 22:50:31 when the 79.3c is record

hmm deffo one for @TomWoB
 

chevron

New Member
If I delete that row there's still a gap between 6 mi9nutes into the run and approx 38 minutes
Yes, I believe that was added when I launched GPU-Z a second time and it appended to the log file again with the new header. I think that part can be safely deleted. I can verify that I also see the same gap you see after I delete that header.
 

TomWoB

Well-Known Member
Hi all,

first of all: I think this is not the right place to discuss GPU-Z problems, better will be there.
But OK, let's finalize the discussion here.

Problem is clear:
up to an "End-Time" of "0:16:46" diagram looks normal, starting with End-Time "0:16:47" it doesn't. That's the border where I switch from "showing every log entry" (<=1000 entries) to "showing every second log entry" (1001 .. 2000 entries), because of performance. But that's not the problem, the problem is I have built in a check "how realistic a value can be". In case of GPU-Z it means, the value should not more change than 20°C from one value to another. But so far I don't double this value in case of "showing every second value" is used (should be 40°C).

Why do I do such a value check: because I saw strange spike values, e.g. single log values below absolute zero. If such a case happens, I ignore the spike and take the value from the previous value. I think, this is useful, especially for getiing "useful average values". A spike can ruin the average value. If I switch off the value check, the diagram looks as expected (incl min, ave, max-value):

GPU-Z Logs OK.png
So, what I have to do is at least to double the check value in case of every second value is shown (and so on, every third, etc). Or I switch off the check, actually I'm not sure what is the best solution ...


Some more information:
If you have diagonal lines like this (time range: "some seconds/minutes"):

Diagonal.png
This typically occurs in case of:
  • CPU is running at 100%, proper logging is not longer possible
  • a time jump is within the log-data (logging app was stopped and started again with same log-file)
  • a time jump happens on the PC itself (daylight saving time changeover or time of PC was set manually)

Best regards
Tom
 
Last edited:

chevron

New Member
Thanks TomWoB for the investigation and finding the cause of the problem. Would it be possible to implement a button to turn on/off the check?
 

TomWoB

Well-Known Member
Hi chevron,

yes, could be a solution, I think best will be a CheckBox in the "Edit"-menu. But I had a look at the code and there are less value checks existing as I thought:
  • HWINFO
    only log-categories with unit [FPS] are checked. If value is > 500 FPS, previous value is taken. I do this, because there are bad designed games existing, which have no FPS-limiter in the game menus -> the FPS value will reach 6'000 FPS and more. Such diagrams are unusable, I descibed this problem already in Post #37 (limit was meanwhile increased from 300 to 500).
    BTW: no limit in game menus is still an issue, probably actual in "Amazons New World". There are a some reports of killed RTX 3090 of EVGA existing, it looks like that the EGVAs are overheating in the game menus -> the EVGAs are bad designed too (check Twitch-Streamer Gladd on Twitter or forums on Reddit)

  • Afterburner & AIDA64
    no value checks implemented

  • GPU-Z
    here are the most value checks implemented (perhaps tool has the worst values?). I check temperature values (+/- 20°C change per value), range of percent-values and range of voltage-values.
So ... currently I see no need for action. Best regards
Tom
 
Last edited:
Top