Prime95 30.8 (big P-1 changes, see post #551) - Page 63

kruoli · 2022-09-04, 16:41

My system has hwloc 2.4.1 installed. It seems like this version and also the version you have linked with both report 16 "normal" cores (see my last full hwloc post). I am glad to try another version, though.

LaurV · 2022-09-05, 07:41

What is the best way to play P95 on a i7-1280P, with 16G RAM, and Win 10? This toy is from the 12th generation, it has 6 "performance" cores, which are "normal", I mean, HT, high speed, etc., and 8 "economy" cores (they call them "efficient", but this is stupid and it looks like the others are not efficient

), which are quite slow, low power, non-HT. Hopefully, not ARM or something, haha. Now it is in fashion to put some low power ARM cores everywhere....

Anyhow, Win10 reports 20 cores (which I guess, it is 6*2+8=20) but how to pair them, now idea, so for now I decided to run a single worker, in 4 cores. This seems to always find good cores and not smoke the laptop out (i.e. keep the temperature under control, in normal ranges, no throttling).

Advice?

Xyzzy · 2022-09-05, 12:10

Quote:

Originally Posted by LaurV

What is the best way to play P95 on a i7-1280P, with 16G RAM, and Win 10?

Do you need to lock Prime95 to the performance cores?

James Heinrich · 2022-09-05, 12:50

Quote:

Originally Posted by Xyzzy

Do you need to lock Prime95 to the performance cores?

One of the changelog items in v30.7 is

Quote:

Support for asymmetric processor architectures such as Intel's Alder Lake.

so I would assume that Prime95 is supposed to handle such things automatically(?)

Prime95 · 2022-09-05, 21:37

Quote:

Originally Posted by Prime95

I may try linking with hwloc 2.8.0 or 2.7.1 for you to test.

Since version 2.8 of hwloc did not help let's read the HWLOC documentation on environment variables:

Code:

HWLOC_CPUKINDS_RANKING=default

    change the ranking policy for CPU kinds. hwloc tries to rank CPU kinds that are energy efficiency first, and then CPUs that are rather high-performance and power hungry.
    By default, if available, the OS-provided efficiency is used for ranking. Otherwise, the frequency and/or core types are used when available.
    This environment variable may be set to coretype+frequency, coretype+frequency_strict, coretype, frequency, frequency_base, frequency_max, forced_efficiency, no_forced_efficiency, default, or none.

HWLOC_CPUKINDS_MAXFREQ=adjust=10

    change the use of the max frequency in the Linux backend. hwloc tries to read the base and max frequencies of each core on Linux. Some hardware features such as Intel Turbo Boost Max 3.0 make some cores report slightly higher max frequencies than others in the same CPU package. Despite having slightly different frequencies, these cores are considered identical instead of exposing an hybrid CPU. Hence, by default, hwloc uniformizes the max frequencies of cores that have the same base frequency (higher values are downgraded by up to 10%).

    If this environment variable is set to adjust=X, the 10% threshold is replaced with X. If set to 1, max frequencies are not adjusted anymore, some homogeneous processors may appear hybrid because of this. If set to 0, max frequencies are entirely ignored.

Try HWLOC_CPUKINDS_RANKING=coretype
or HWLOC_CPUKINDS_MAXFREQ=adjust=0

kruoli · 2022-09-05, 21:54

Code:

$ HWLOC_CPUKINDS_RANKING=coretype lstopo --cpukinds
CPU kind #0 efficiency -1 cpuset 0x20002000
  FrequencyMaxMHz = 5083
CPU kind #1 efficiency -1 cpuset 0x10001000
  FrequencyMaxMHz = 5236
CPU kind #2 efficiency -1 cpuset 0x02000200
  FrequencyMaxMHz = 5388
CPU kind #3 efficiency -1 cpuset 0x04000400
  FrequencyMaxMHz = 5541
CPU kind #4 efficiency -1 cpuset 0x40004000
  FrequencyMaxMHz = 5694
CPU kind #5 efficiency -1 cpuset 0x80008000
  FrequencyMaxMHz = 5850
CPU kind #6 efficiency -1 cpuset 0x01000100
  FrequencyMaxMHz = 6003
CPU kind #7 efficiency -1 cpuset 0x08000800
  FrequencyMaxMHz = 6155
CPU kind #8 efficiency -1 cpuset 0x00080008
  FrequencyMaxMHz = 6308
CPU kind #9 efficiency -1 cpuset 0x00020002
  FrequencyMaxMHz = 6461
CPU kind #10 efficiency -1 cpuset 0x00800080
  FrequencyMaxMHz = 6614
CPU kind #11 efficiency -1 cpuset 0x00200020
  FrequencyMaxMHz = 6766
CPU kind #12 efficiency -1 cpuset 0x00100010
  FrequencyMaxMHz = 6919
CPU kind #13 efficiency -1 cpuset 0x00040004
  FrequencyMaxMHz = 7075
CPU kind #14 efficiency -1 cpuset 0x00410041
  FrequencyMaxMHz = 7228

...and...

Code:

$ HWLOC_CPUKINDS_MAXFREQ=adjust=0 lstopo --cpukinds
CPU kind #0 efficiency -1 cpuset 0x20002000
  FrequencyMaxMHz = 5083
CPU kind #1 efficiency -1 cpuset 0x10001000
  FrequencyMaxMHz = 5236
CPU kind #2 efficiency -1 cpuset 0x02000200
  FrequencyMaxMHz = 5388
CPU kind #3 efficiency -1 cpuset 0x04000400
  FrequencyMaxMHz = 5541
CPU kind #4 efficiency -1 cpuset 0x40004000
  FrequencyMaxMHz = 5694
CPU kind #5 efficiency -1 cpuset 0x80008000
  FrequencyMaxMHz = 5850
CPU kind #6 efficiency -1 cpuset 0x01000100
  FrequencyMaxMHz = 6003
CPU kind #7 efficiency -1 cpuset 0x08000800
  FrequencyMaxMHz = 6155
CPU kind #8 efficiency -1 cpuset 0x00080008
  FrequencyMaxMHz = 6308
CPU kind #9 efficiency -1 cpuset 0x00020002
  FrequencyMaxMHz = 6461
CPU kind #10 efficiency -1 cpuset 0x00800080
  FrequencyMaxMHz = 6614
CPU kind #11 efficiency -1 cpuset 0x00200020
  FrequencyMaxMHz = 6766
CPU kind #12 efficiency -1 cpuset 0x00100010
  FrequencyMaxMHz = 6919
CPU kind #13 efficiency -1 cpuset 0x00040004
  FrequencyMaxMHz = 7075
CPU kind #14 efficiency -1 cpuset 0x00410041
  FrequencyMaxMHz = 7228

No luck at all.

chalsall · 2022-09-05, 23:18

Quote:

Originally Posted by kruoli

Code:

$ HWLOC_CPUKINDS_RANKING=coretype lstopo --cpukinds
[SNIP...]

No luck at all.

Please forgive me if I'm telling you how to chew gum.

I have found that sometimes one needs to export a variable before it can be seen by following execution paths.

It is interesting that the report is identical. There might be a bug in there somewhere.

Might be a rabbit hole. Just throwing that out there.

kruoli · 2022-09-06, 08:24

It should work fine this way. Example:

Code:

$ FIRMA="keine" bash # spawn a BASH subprocess
$ env # this gets executed in the subprocess as a subsubprocess
SHELL=/bin/bash
TMUX=/tmp/tmux-1000/default,21541,0
PWD=/home/oliver
LOGNAME=oliver
XDG_SESSION_TYPE=tty
FIRMA=keine
MOTD_SHOWN=pam
HOME=/home/oliver
LANG=de_DE.UTF-8
<snip>
PATH=/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games
DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/1000/bus
SSH_TTY=/dev/pts/0
_=/usr/bin/env

But to be extra sure I tried

Code:

export HWLOC_CPUKINDS_RANKING=coretype
lstopo --cpukinds
unset HWLOC_CPUKINDS_RANKING
export HWLOC_CPUKINDS_MAXFREQ=adjust=0
lstopo --cpukinds

with the same outcome as in my last post.

Prime95 · 2022-09-06, 16:00

Quote:

Originally Posted by kruoli

No luck at all.

Was that lstopo from hwloc 2.8.0?

kruoli · 2022-09-06, 16:24

No, that is not supplied by my package manager; I might have to try it by hand with your mprime version. I will give it a try.

kruoli · 2022-09-06, 16:31

Good news! Using the custom mprime build with hwloc 2.8.0, HWLOC_CPUKINDS_RANKING=coretype gave the correct architecture. HWLOC_CPUKINDS_MAXFREQ=adjust=0 still showed wrong results.

2022-09-05, 07:41	#684
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 10,273 Posts	... got a new toy ... What is the best way to play P95 on a i7-1280P, with 16G RAM, and Win 10? This toy is from the 12th generation, it has 6 "performance" cores, which are "normal", I mean, HT, high speed, etc., and 8 "economy" cores (they call them "efficient", but this is stupid and it looks like the others are not efficient ), which are quite slow, low power, non-HT. Hopefully, not ARM or something, haha. Now it is in fashion to put some low power ARM cores everywhere.... Anyhow, Win10 reports 20 cores (which I guess, it is 62+8=20) but how to pair them, now idea, so for now I decided to run a single worker, in 4 cores. This seems to always find good cores and not smoke the laptop out (i.e. keep the temperature under control, in normal ranges, no throttling). Advice? Last fiddled with by LaurV on 2022-09-05 at 07:46*

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Do not post your results here!	kar_bon	Prime Wiki	40	2022-04-03 19:05
what should I post ?	science_man_88	science_man_88	24	2018-10-19 23:00
Where to post job ad?	xilman	Linux	2	2010-12-15 16:39
Moderated Post	kar_bon	Forum Feedback	3	2010-09-28 08:01
Something that I just had to post/buy	dave_0273	Lounge	1	2005-02-27 18:36

2022-09-04, 16:41	#683
kruoli "Oliver" Sep 2017 Porta Westfalica, DE 10100100111₂ Posts	My system has hwloc 2.4.1 installed. It seems like this version and also the version you have linked with both report 16 "normal" cores (see my last full hwloc post). I am glad to try another version, though.

2022-09-06, 16:24	#692
kruoli "Oliver" Sep 2017 Porta Westfalica, DE 1,319 Posts	No, that is not supplied by my package manager; I might have to try it by hand with your mprime version. I will give it a try.

2022-09-06, 16:31	#693
kruoli "Oliver" Sep 2017 Porta Westfalica, DE 1,319 Posts	Good news! Using the custom mprime build with hwloc 2.8.0, HWLOC_CPUKINDS_RANKING=coretype gave the correct architecture. HWLOC_CPUKINDS_MAXFREQ=adjust=0 still showed wrong results.