Ken Blake wrote:
On Wed, 23 Jan 2019 16:37:31 -0500, Wolf K
wrote:
On 2019-01-23 16:20, wrote:
On Thu, 17 Jan 2019 08:30:32 +1100, Peter Jason wrote:
I am processing movies and the CPU is indicating
100%.
Where can I see the temperature of the CPU?
Peter
I followed the suggestions and discovered that my CPU is running at
99C Is this good? bad? or indifferent?
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
IMO it's bad. Not too bad, in that it will continue to run at that
temperature (obviously), but that temp will degrade it faster, so rather
sooner than later the CPU will fail. FWIW, the CPU on this box is
running at 55-56C.
I have no pretenses on being an expert on this, but 55-56C sounds
pretty hot to me. Mine is around 34C here.
You need to load up the CPU for a decent test.
Having a CPU idling at 55C, means when you use it for
something, the temperature is going to shoot higher, and
it might end up throttling.
CPUs with small silicon dies are harder to cool than larger
silicon dies. The Athlon with the bare die was reasonably
small. And harder to cool when you lowered a heatsink onto it.
Even though at the time, it was a 65W CPU.
A "good temperature for an Athlon" was 65C under load.
They seemed to be happy there. On boards with the
THERMTRIP addon chip (Attansic brand?), the power would get cut at
85-90C perhaps.
*******
There's a claim that "LINPACK" heats up a CPU better
than Prime95, so I tested that out.
There is a tool here that uses a LINPACK executable.
As an example.
https://www.techspot.com/downloads/4...lburntest.html
https://files01.tchspt.com/temp/IntelBurnTest.zip
(
https://www.raymond.cc/blog/test-sys...tem-resources/ )
But what I tried instead, is just getting a LINPACK from
the Intel site.
https://software.intel.com/en-us/art...nchmarks-suite
Windows* package (w_mklb_p_2018.3.011) (.zip)
w_mklb_p_2018.3.011.zip 7,536,943 bytes
Using 7ZIP, look for the "linpack" folder.
w_mklb_p_2018.2.010\benchmarks_2018\windows\mkl\be nchmarks\
linpack
runme_xeon64.bat
In a compiler directory are a couple DLLs and a PDB (symbol table).
w_mklb_p_2018.2.010\benchmarks_2018\windows\redist \intel64_win\compiler\
libiomp5md.dll 1652992 bytes
libiompstubs5md.dll 110848 bytes
Add the DLL kit into the linpack folder, then
Command Prompt (as regular user)
cd /d %userprofile%\Downloads\linpack
runme_xeon64.bat
Then, watch your Speedfan chart for temperature results.
*******
I ran Prime95 as well, just whatever copy I had
handy, and it's sorta in the same ballpark. The
latest is here. At some point, the author of this
tried perhaps, to use AVX512 or similar and some
bug was uncovered. I don't remember the details,
and perhaps the program is back to the old way of
doing it. AVX512 heats up processors enough, to
have its own protection/prevention scheme (the
very definition of useless).
https://www.mersenne.org/download/
Windows: 64-bit 29.4b8 2018-02-09 6.1MB p95v294b8.win64.zip
The advantage of that one, is "double-click to start" on the
EXE, and only answer one prompt ("Join GIMPS" - answer "Only testing").
After that, accepting the defaults starts N threads of test
and makes your CPU heat up. Select Exit from the menu to stop it,
as it'll run in the tray if you iconify it.
These are the charts from my runs. The Intel Coretemps
are used for the charts, and those have relatively large
error bars at low temperature. That's how cores can end up
reading "sub-ambient" temperatures.
The delta on the two tests is about the same.
https://i.postimg.cc/HnxGsc6m/linpack2.gif
https://i.postimg.cc/KcqSpfY8/prime95.gif
One run was drawing 208W, the other 211W, on my
Kill-O-Watt. The power before the test starts
is 95W or so. (Power varies on the machine, according
to whether hard drives or used - in this case, the
SSD boot drive gives the 95W power figure.)
The machine goes higher, with a little help from
the video card.
Paul