Question Zen 4 builders thread

Page 50 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

In2Photos

Golden Member
Mar 21, 2007
1,629
1,651
136
That's amazing!

You make it sound so easy :D

Going AMD is rewarding when one takes the time to tinker. I guess that's why AMD users are so in love with their CPUs :)

Would you mind if I request you for one run of Geekbench 5, one run of Rapydmark High setting and MemMaxx2 for dessert? :p
In all honesty it is quite easy on Ryzen to undervolt and get great results. You need to know where to look for the settings, but there are several tutorials on YouTube. Optimum Tech has several videos and does a great job explaining what he's doing and showing the results. Setting up PBO+CO is maybe a dozen mouse clicks and entering a value, then testing for stability. I had to do more tinkering with my son's 5600X as it wasn't stable at -30, so it took a little bit longer. My daughter's 7600X worked right away at -30. She played all day yesterday and not a single crash so I think we got a little lucky with the silicon. At -30 I see no need to even try and get better per core results and honestly not sure if you can get any lower. You can see it in action here:

I'm not as familiar with the other benchmarks so hopefully I did them correctly.

Geekbench 5 - https://browser.geekbench.com/v5/cpu/19558578
Geekbench 5.png
Single core 2120, Multicore 11470
This looks to be in line with other scores, although I believe my multicore score looks pretty good in comparison, especially since we are using an air cooler. There are some decent overclocks on there with 5.5+GHz.

RapydMark
RapydMark High.png
285.760s - I have no idea if this is any good or not and it also looks like it tested the GPU in there.

MaxxMem2
MaxxMem.png

Again I have no idea if this is good or not.
 
Jul 27, 2020
16,308
10,333
106
RapydMark
View attachment 73541
285.760s - I have no idea if this is any good or not and it also looks like it tested the GPU in there.
Best score is Det0x's 109.367 for 32 threads so your score with 2.66 times less threads is 2.61 times slower which seems to be great!

13600K GB5 score: https://browser.geekbench.com/processors/intel-core-i5-13600k

Your CPU is beating the 13600K's average ST score easily and per thread MT score comes out to be 955.83 whereas 13600K manages only 758.8 per thread with 20 threads. You got stronger threads! :)
 

In2Photos

Golden Member
Mar 21, 2007
1,629
1,651
136
Best score is Det0x's 109.367 for 32 threads so your score with 2.66 times less threads is 2.61 times slower which seems to be great!

13600K GB5 score: https://browser.geekbench.com/processors/intel-core-i5-13600k

Your CPU is beating the 13600K's average ST score easily and per thread MT score comes out to be 955.83 whereas 13600K manages only 758.8 per thread with 20 threads. You got stronger threads! :)
Looks like I'm in line then. I think I could squeeze more performance out of it if I spent some time playing around. If it was my system I might do that, but I think this is good enough for her. Time to let her enjoy it. She has only played games and watched some Youtube so far. She streams every now and then but maybe with a stronger PC she might do it more often.
 
  • Like
Reactions: igor_kavinski

lopri

Elite Member
Jul 27, 2002
13,209
594
126
Geekbench does vary per run. I also learned that windows power plan plays a role, too, though Geekbench lets you know which power plan was used.

Using Eco mode (65W), I am getting the same ST score and 95% of MT score of default runs. A similar result with Cinebench 30 min test. It is hard to believe AMD pushed this much voltage for that last 5%. (At least for my 7700X)

Edit: Oh and you definitely can squeeze more out of these while limiting the power consumption.
 
  • Like
Reactions: igor_kavinski

Timmah!

Golden Member
Jul 24, 2010
1,419
631
136
Find curve optimizer. choose negative offsetup use something between -5 and -25 (mine are all -25) Then find max temp, and set it to 75 (or your choice)
Set memory to EXPO .

Done.

Thanks, but devil is in detail here :)

Where do i "find" curve optimizer? BIOS or some of those AMD tuning apps - ryzen master i think? Do i set that value separately for each core or for all of them across the board? How do i test for stability, usual way like Prime95 or something else? If it fails, how do i know which core is the culprit? What do you exactly mean by "finding" max temp? I saw some tutorial on youtube by that Spatter? fella on this particular topic and there were some other steps to do you dont mention.

I cant set the RAM to EXPO, as i sadly bought non-expo kit.

It may seem easy to someone with your experience in these matters, but i need to know more specifics.
 

In2Photos

Golden Member
Mar 21, 2007
1,629
1,651
136
Thanks, but devil is in detail here :)

Where do i "find" curve optimizer? BIOS or some of those AMD tuning apps - ryzen master i think? Do i set that value separately for each core or for all of them across the board? How do i test for stability, usual way like Prime95 or something else? If it fails, how do i know which core is the culprit? What do you exactly mean by "finding" max temp? I saw some tutorial on youtube by that Spatter? fella on this particular topic and there were some other steps to do you dont mention.

I cant set the RAM to EXPO, as i sadly bought non-expo kit.

It may seem easy to someone with your experience in these matters, but i need to know more specifics.
Watch the video I linked in my prior post. Curve Optimizer is in the BIOS. The location varies by vendor, but it will be in some version of an overclocking or tweaker settings. There will be an advanced CPU settings option where you will enable PBO. At that point you can apply the settings to "all core" or "per core". Make sure you are applying a "negative" setting and not "positive". It seems that the 7900x and 7950x are slightly harder to do "all core" due to the high number of cores, but Mark was able to set his to -25 so it is possible. Then save the settings, boot into Windows and run some benchmarking software like Cinebench or Prime 95. If it passes at the settings you can try going even lower. If it doesn't then raise it by 5 and try again. Max Temp is a setting that allows you to set the max temperature the CPU can hit. This is also in the video I posted above.

Trust me when I say it's not that hard. I have done 2 systems now and had never done anything like it previously.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
Thanks, but devil is in detail here :)

Where do i "find" curve optimizer? BIOS or some of those AMD tuning apps - ryzen master i think? Do i set that value separately for each core or for all of them across the board? How do i test for stability, usual way like Prime95 or something else? If it fails, how do i know which core is the culprit? What do you exactly mean by "finding" max temp? I saw some tutorial on youtube by that Spatter? fella on this particular topic and there were some other steps to do you dont mention.

I cant set the RAM to EXPO, as i sadly bought non-expo kit.

It may seem easy to someone with your experience in these matters, but i need to know more specifics.
Bios. Since I don't know your motherboard, just look in menus about cpu condif. Look for "Curve Optimizer". Let me look and see of I can find it in the manual for mine.

As far a ram, good luck there. I had a exp kit 6600 cl32, and could not get it to work faster than 5600. But my EXPO kit worked perfects at 6000 cl30.

In this video at about one minute there is a picture


1672257470594.png
 
Last edited:
  • Like
Reactions: Timmah! and Elfear

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
In all honesty it is quite easy on Ryzen to undervolt and get great results. You need to know where to look for the settings, but there are several tutorials on YouTube. Optimum Tech has several videos and does a great job explaining what he's doing and showing the results. Setting up PBO+CO is maybe a dozen mouse clicks and entering a value, then testing for stability. I had to do more tinkering with my son's 5600X as it wasn't stable at -30, so it took a little bit longer. My daughter's 7600X worked right away at -30. She played all day yesterday and not a single crash so I think we got a little lucky with the silicon. At -30 I see no need to even try and get better per core results and honestly not sure if you can get any lower. You can see it in action here:

I'm not as familiar with the other benchmarks so hopefully I did them correctly.

Geekbench 5 - https://browser.geekbench.com/v5/cpu/19558578
View attachment 73540
Single core 2120, Multicore 11470
This looks to be in line with other scores, although I believe my multicore score looks pretty good in comparison, especially since we are using an air cooler. There are some decent overclocks on there with 5.5+GHz.

RapydMark
View attachment 73541
285.760s - I have no idea if this is any good or not and it also looks like it tested the GPU in there.

MaxxMem2
View attachment 73542

Again I have no idea if this is good or not.

Running CoreCycler at -30 will reduce some of that enthusiasm.
 

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
Watch the video I linked in my prior post. Curve Optimizer is in the BIOS. The location varies by vendor, but it will be in some version of an overclocking or tweaker settings. There will be an advanced CPU settings option where you will enable PBO. At that point you can apply the settings to "all core" or "per core". Make sure you are applying a "negative" setting and not "positive". It seems that the 7900x and 7950x are slightly harder to do "all core" due to the high number of cores, but Mark was able to set his to -25 so it is possible. Then save the settings, boot into Windows and run some benchmarking software like Cinebench or Prime 95. If it passes at the settings you can try going even lower. If it doesn't then raise it by 5 and try again. Max Temp is a setting that allows you to set the max temperature the CPU can hit. This is also in the video I posted above.

Trust me when I say it's not that hard. I have done 2 systems now and had never done anything like it previously.
Mark uses -25 because he has one type of load that uses all cores. No deviation from that load. Extreme edge case. If he too tested with CoreCycler he would end up with problems.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
Mark uses -25 because he has one type of load that uses all cores. No deviation from that load. Extreme edge case. If he too tested with CoreCycler he would end up with problems.
I have 2 right now that are only running a video card, using 1 to 2 cores. On 24/7, no problems.
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
How long? Never heard of CoreCycler before but it looks interesting. I will definitely check it out. Right now I am using numbers between -20 and -10.
 

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
How long? Never heard of CoreCycler before but it looks interesting. I will definitely check it out. Right now I am using numbers between -20 and -10.
Took me about a week to tune all 16 cores on my 7950X. Ran it between 18 to 24 hours for each of those days. Y-Cruncher AVX512 two threads per core - 9 minute test per core. Also ran Prime95 large FFT two threads per core - 6 minute test per core.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
CoreCycler does take a long time to run in order to properly find instability per core. I can understand the desire to set and forget in Curve Optimizer.
We few weeks back all 3 ran @100% load for 2 weeks for so 24/7 for the PG race. Not one single problem out of 3 7950x@ CO= -25 and less than 95c temp. Only one has a big 420 AIO, the other 2 are fairly high end air. (one dark rock pro 4 ($90?), and one a deep cool ak620, $50 cooler)
 

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
We few weeks back all 3 ran @100% load for 2 weeks for so 24/7 for the PG race. Not one single problem out of 3 7950x@ CO= -25 and less than 95c temp. Only one has a big 420 AIO, the other 2 are fairly high end air.
It's all about what you are comfortable with accepting.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
My experience with CO is that it isn't the 100% load that's the issue but bursty 1-4 core loads.
Well, then I will probably never have an issue. F@H does not use more than 5% of the resources and is most likely not bursty. The rest of the time its 100% load
 

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
Well, then I will probably never have an issue. F@H does not use more than 5% of the resources and is most likely not bursty. The rest of the time its 100% load
That was the point I was getting at. -25 CO across all cores works for you.

That same CPU at -25 CO all cores wouldn't work for me and the standards I set through testing. Unless, by some miracle, you do truly have a 7950X that passes all tests with a -25 all core CO.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
That was the point I was getting at. -25 CO across all cores works for you.

That same CPU at -25 CO all cores wouldn't work for me and the standards I set through testing. Unless, by some miracle, you do truly have a 7950X that passes all tests with a -25 all core CO.
One of the 2 that are not busy is windows, the other is linux. What would make you believe that my CPU was good at -25 ? give be a test to run, and parameters. Has to be free.
 

AdamK47

Lifer
Oct 9, 1999
15,223
2,843
126
One of the 2 that are not busy is windows, the other is linux. What would make you believe that my CPU was good at -25 ? give be a test to run, and parameters. Has to be free.

Having a hard time deciphering.

The tool I use and mentioned about a dozen times is CoreCycler. Test parameters mentioned a few posts up. There cannot be ANY other load on the CPU when running it. No matter how small.

Running it would be of your own benefit. Nothing else. It's not meant for scoring internet forum points as I really don't give a damn about that nonsense.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,560
14,513
136
Having a hard time deciphering.

The tool I use and mentioned about a dozen times is CoreCycler. Test parameters mentioned a few posts up. There cannot be ANY other load on the CPU when running it. No matter how small.

Running it would be of your own benefit. Nothing else. It's not meant for scoring internet forum points as I really don't give a damn about that nonsense.
Well, the "There cannot be ANY other load on the CPU when running it." kills this test. I have work to do.