5950X boost clocks

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Sesse
Posts: 274
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 5950X boost clocks

Post by Sesse » Wed Dec 16, 2020 10:54 pm

I have no idea what the Vcore was—I didn't modify it. The temperature was around 60 degrees for the single-thread tests.

corres
Posts: 3657
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: 5950X boost clocks

Post by corres » Thu Dec 17, 2020 8:37 am

Sesse wrote:
Wed Dec 16, 2020 10:54 pm
I have no idea what the Vcore was—I didn't modify it. The temperature was around 60 degrees for the single-thread tests.
If you use Ryzen Master it displays the actual Vcore for every each core.
Vcore is a very important parameter because it determines the lifetime/degradation of the given CPU.
For Zen 2 and 3 type CPUs (7 nm technology) the "official" max.safe Vcore for full load (every cores are on the set
clock frequency) is only 1.325 Volts. If only one core is boosted the Vcore at most 1.470 Volts.
The working temperature of the the processor also influences the lifetime, for Zen3 this is < 90 grade Celsius.
So the cooling is also important and in general the overclock of Zen2-3 type CPUs are rather limited and a dangerous act.

Sesse
Posts: 274
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 5950X boost clocks

Post by Sesse » Thu Dec 17, 2020 10:15 am

corres wrote:
Thu Dec 17, 2020 8:37 am
If you use Ryzen Master it displays the actual Vcore for every each core.
For the Nth time—I don't use Windows.

User avatar
Nordlandia
Posts: 2743
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: 5950X boost clocks

Post by Nordlandia » Thu Dec 17, 2020 10:28 am

Sesse wrote:
Thu Dec 17, 2020 10:15 am
corres wrote:
Thu Dec 17, 2020 8:37 am
If you use Ryzen Master it displays the actual Vcore for every each core.
For the Nth time—I don't use Windows.
1. Can i ask why not.
2. Also the time is ripe to upgrade the sesse hardware in addition to invest in SSD that can store whole 7-piece WDL. 7-piece DTZ can be stored on HDD or external drive.

These two combinations together will be very great boost in analytical performance.

Sesse
Posts: 274
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 5950X boost clocks

Post by Sesse » Thu Dec 17, 2020 11:33 am

Nordlandia wrote:
Thu Dec 17, 2020 10:28 am
1. Can i ask why not.
No, you cannot. :-)
Nordlandia wrote:
Thu Dec 17, 2020 10:28 am
2. Also the time is ripe to upgrade the sesse hardware in addition to invest in SSD that can store whole 7-piece WDL. 7-piece DTZ can be stored on HDD or external drive.
I don't have anything to announce at this point.

Sesse
Posts: 274
Joined: Mon Apr 30, 2018 9:51 pm
Contact:

Re: 5950X boost clocks

Post by Sesse » Fri Dec 18, 2020 12:37 am

I see now that I've been benchmarking the wrong thing for the Ipman test (I didn't see the text at the top of the page); I just took the default start position, so if anybody has wanted to compare these numbers against anything, here are my values:

Total time (ms) : 1669
Nodes searched : 130256121
Nodes/second : 78044410

78 Mnps puts it nicely near the top of the 5950X group, with no overclocking (unless you count XMP as overclocking). And yes, I enabled huge pages manually just to be 100% sure (with such a short test, it's not given that THP can kick in).

User avatar
MikeB
Posts: 4614
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: 5950X boost clocks

Post by MikeB » Fri Dec 18, 2020 2:00 am

corres wrote:
Wed Dec 16, 2020 3:02 pm
MikeB wrote:
Tue Dec 15, 2020 9:57 pm
Also , keep in mind that you may get a huge jump with Large Pages set to true

Large Pages Off:
# asmFishW_2017-05-22_popcnt.exe bench 1024 64 26
asmFishW_2017-05-22_popcnt
*** bench hash 1024 threads 64 depth 26 realtime 0 ***
info string hash set to 1024 MB no large pages

Large Pages On:
# asmFishW_2017-05-22_popcnt.exe
asmFishW_2017-05-22_popcnt
setoption name LargePages value true
bench 1024 64 26
*** bench hash 1024 threads 64 depth 26 realtime 0 ***
info string hash set to 1024 MB page size 2048 KB
Based on your memory there is anybody who marked his result as it gotten with large pages??
This is one of cause of the disturbed results!
Even Ippmann did not deny the use of large pages but this would be more correct, because the effect of using large pages the result will be rather uncertain.
I believe Ipman requests results without large pages.
Image

corres
Posts: 3657
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: 5950X boost clocks

Post by corres » Fri Dec 18, 2020 8:09 am

MikeB wrote:
Fri Dec 18, 2020 2:00 am
I believe Ipman requests results without large pages.
I am afraid you are not the only one who know that trick...
And the human vanity is an important factor.

Post Reply