Cooler Master H500M Build Log

[TechFreak1]

Update:

Needed to do some more testing and had to reinstall Windows again for testing.

This is the highest score I've gotten before the latest undervolting revisions with PBO enabled.

12th August.


But, I started getting random shut downs even with PBO on default setting which is now auto - the default setting for PBO is AUTO as opposed to disabled.

However, ever since disabling PBO I have experienced zero random shut downs. That's with a clean install with latest AMD drivers for GPU and Chipset. Additionally, did a clean install with no drivers, no av, did a clean install with latest drivers + AV/Firewall installed, clean install with latest drivers with no av/firewall installed. No shut downs with PBO disabled, shut downs with PBO enabled in certain work loads. Verifiable crashes when running R20.

So I decided to go very aggressive on the undervolting as the difference is about 120 points with PBO enabled.

Revision Test with no PBO 13th August:

1) 2) 3) 4)

There is one caveat that is really not mentioned by folks who do benchmarking videos - you are going to lose points on any daily driver PC as installed Apps like steam, firewalls, antiviruses also take up CPU cycles.

16th August with 13th Aug Undervolt Settings - installed Applications - Steam, MSI afterburner, RivaTuna, Mytube Beta, AV+Firewall, Deezer, Latest Chipset drivers, No GPU Drivers.


Disclaimer: If you do copy the settings below - you do so at your own risk - I nor this forum will bear in responsbility for any issues or damage caused to your components. These settings are for a reference guide for those who know how to recover a none booting PC after an aggressive undervolt via a CMOS reset or worse case scenario - a bios flashback.

Took me sometime to figure out that the first partition on my usb stick had to Fat32 for the bios to save screenshots... As I never had issues saving screenshots before. I'm guessing Rufus did something wonky to the USB stick as the old versions didn't use to create a seperate UEFI partition on the stick and you don't have disable secure boot to use a bootable W10 install using rufus - all you have to do is delete this small UEFI partition using diskpart or any other partition manager (it does not show up in disk manager however).







Idle with above settings.



Continued in next post.

 

[TechFreak1]

Continued:

These are the lowest values that can be set in the Aorus Elite X570.



However it's not all sunshine and rainbows.

Hwinfo has stopped working for some strange reason and I'm at a loss how to fix it. It's not a big deal just an annoyance as it's broken the Rainmeter skin I was working on T_T.

The readings below are inaccurate.



Sensors are working fine:



The other issue I'm still trying to resolve is games crashing. Superposition at 8K (to completely soak up all the VRAM on the GPU), 1080p extreme or medium, Heaven on lowest setting or highest settings, MSI Kombustor etc all run fine on several hour loops. I even ran Memtest 86+ no errors, Ran Prime95 to stress CPU / Memory - completely stable. Also tried a combination of tests with fans at 500 RPM, 600, 650, 800 RPM for both intake, 780 for exhaust - Nada.

GTA 5 crashes always at this point, the other game I've tried is Shadow of Mordor and same thing crash after a few minutes.


I've tried stock bios settings, GPU drivers from last year April / December / both latest drivers released this month + combination of old and chipset drivers, , stock ram settings, GTA 5 on all default settings, 800x600, page file on SSD, Page file on NVME. But GTA 5 always crashes at this point in the benchmark run. Prior to using ingame benchmark to test I was using in gameplay - online solo and story mode - game would crash after a few minutes.

The only other things I can think of is going back down to 16 Gigs of RAM and switching out the PSU.

 

[TechFreak1]

Update: I feel I've hit the point where I can't squeeze any more out of the CPU without the risk of stability at the level of Undervolting. I was able to run a single pass of R20 at 4100 Mhz all core, multiple at 4075 Mhz all core before crashing. Now stable at 4050 Mhz.
I have Corsair link and VMware installed. PBO is still disabled.




Above score tested today, without Hwinfo. I've used the PC for about two days, leaving it on idle for several hours to see if it crashes. Thankfully, I haven't had a single crash yet be it running full loaded CPU or idle.

---------------------------------------

These tests were run 2 days ago.

These scores are with the same settings as today, front intake fans at 550 RPM, rear at 530 RPM. As you can see there is clear fluctuations due to ambient temperature fluctuations.

With Hwinfo running.


Without Hwinfo.




These are with RAM at 2733, Fans as above.

Without Hwinfo


With Hwinfo


Scores with CPU at 4075 Mhz All Core.



Fans as above


All fans at 550 RPM.


Scores at 3.9 Ghz All core, Fans at 500 RPM
Without Hwinfo


With Hwinfo


Scores at 4.0 Ghz All core, Fans at 500 RPM
without hwinfo


With hwinfo


Next to tweak is ram timings.

Additionally, there is one thing I need to mention - after the PC crashed running R20 second run at 4.1 Ghz - (Did not finish as crashed just soon after it finished at a score 4086 - so not counted) Hwinfo started automagically working again, after this crash I reduced it 4075.





After seeing it was extremely stable at 4.0 Ghz. I knew 4.2 ghz would be unstable from previous testing last year, so I went straight to 4.1 Ghz to see how much performance I can squeeze out the CPU. If it was somewhat stable at 4.1, that meant I could hit stability at between 4.0 and 4.1 Ghz. I was hoping it was closer to 4.1 but I can't complain as at 4075 the CPU would hit 81.x degrees when running R20 with fans at 500 RPM. around 80.x degrees at 550 RPM.

Now seeing this amazing performance from this CPU, I'm leaning more to water cooling with this CPU and upgrading a lot later down the line. That is unless I need PCI-E 4.0 for GPU performance.

Speaking of the GPU I haven't fixed the crashes whilst running games, that's next on the list along with RAM timing tuning.

 

[TechFreak1]

I've reached the point that I don't think I can squeeze out any more performance from my current set of components.



This is with the CPU at current undervolt settings, RAM at 3000 Mhz with loose timings with a completely clean install.

Settings:



This is the second best R20 result I've gotten to date.

This is with the F21 Bios, CPU with undervolt set of 1.316 and -108 mv offset, oced to 4050 all core with XMP (3200 Mhz) and again a clean install.



I did notice some really odd behaviour once I sort of let the CPU loose.





I saw this once but I didn't have a usb stick at hand to screenshot it ergo why I was contemplating watercooling this 2700X. By the time I upgrade the CPU I've got a custom loop at hand but life had other ideas.

Due to the current pandemic I no longer have my job just a little more spare time. So I can't really justify a large expense like that when I don't really need to water cool the CPU. As I plan to upgrade the CPU a few years down the line.

This is the CPU completely undervolted as previous post screenshots, RAM set at 2733 Mhz, Ram set 16-15-15-36 at 1.290 Volts.
17.09.20


CPU completely undervolted as previous post screenshots, RAM set at 2733 Mhz, Ram set 16-15-15-36 at 1.236 Volts (Auto - stock volts for 2133).
19.09.20


Strangely the motherboard sets Cas to 16 as opposed to 15.



This started occurring after I completely messed up the memory controller, as no matter what I set the memory controller to it wouldn't overclock or under clock. Even with two sticks in dual channel.

I was able to resolve this by resolve this by running all four sticks in single channel and then enabling clear DRAM timings and setting DRAM scrub time to 1 hr. I honestly did panic for abit and I thought I completely broke the memory controller on the CPU :grincry::grincry::grincry:.

On the 17th, the RAM clock was stuck at 1345.6 and wouldn't gear up.

Now it's all fine, sometimes you just have let the system rest and it will work out the bugs itself.

You may wondering why I am running the RAM at 2733 Mhz as opposed to using XMP or higher clocks.

There are three primary reasons:

1) Any clock higher than 2733 results in memory errors in Memtest 86+
2) Given the infinity fabric and memory controller is on the CPU as well as on Zen + they are both tied together. Thus overclocking the RAM over clocks the CPU. As a result the temps of the CPU increase as well as resulting in stability issues.
3) Higher clocks on certain cpus don't net you much performance gain and introduce issues into the system.

Lastly I managed to get hwinfo working again... not sure how but it's all working fine.



The current temps with a ambient room temp of 24.6 celcius, with all fans at 500 rpm.

Finally, GTA 5 never got round to fixing the crashes however when I do I'll post an update here.

I'm probably going to wait for the 6000 series cards to drop before buying another GPU to test, I could buy a 980 Ti for a about 80 quid but that's just wasting money I could spend on a better a GPU.

There is one small project I am working on - making a make shift bracket for the Arctic 34 cpu cooler as the current fan mounting system is extremely frustrating to install back on the cooler after cleaning out the dust from the CPU heatsink. Once I've figured it out, I'll post an update.

 

[TechFreak1]

Update: the latest GTA 5 update has fixed the crashing issues which is a relief as that means I can continue using the RX 480 for longer. Additionally, the latest radeon drivers have increases FPS in other games as well by a few percent.

 

[TechFreak1]

Update:

The CPU heatsink bracket is on the back burner for now.

Tried Noctua's fan brackets but they were too big, so got some cheap fan clips but also marginally too big. The cheapest option is to fashion brackets from a wire hanger but every time it's twisted there is small residue metal flakes so not taking that risk lol.

Installed the F32 Bios today and did some basic tweaks.

All tests done with ambient temp of 22.3 degrees celcius, hwinfo not running, all system tray items closed other than AMD Radeon software and ant-virus along with the usual windows 10 install (not a clean install - fully updated) and other background services from apps / programmes installed.






Yes, you've read that right... 1.476 volts on idle!?

On stock CPU will start to thermal throttle on sustained work loads due to the insane amount of voltage pumped through, CPu was hitting around 85 degrees plus just running R20....





RAM timings manually set at 2733 Mhz

Hard limiting the Vcore to 1.316 volts would cause the system to crash when trying to run R20. Therefore the only way to get PBO to run is to use negative offsets.

Enabling PBO, the CPU would fluctuate between 40.99 to 4.1 Ghz all core and at around 1.440 volts on the vcore even with the negative offsets which is way too high. I'm not sure what the peak temp was since Hwinfo was not running and I didn't want to run the CPU at these voltages any longer than necessary.

These are my final settings now. RAM Timings as above. PBO manually disabled. I haven't drastically tried to undervolt like last time... sometimes it's just best to keep it simple lol.



As you can see I got better results with the all core over clock at 4075 Mhz and at a lower voltage to boot.




CPU temp test with final settings, multiple programmes running, including browser with multiple tabs, hwinfo with CPU polling at 50 ms (lowest setting):



Peak Temp:


As you can see there is a massive difference in terms of performance as well as thermals when undervolting.

 

[TechFreak1]

Damn, it's been a long while - haven't been able to login for awhile and then life just happened and work has been super exhausting. Now moving to another post at work.

The Build has come a long way since my last post, I've since upgraded to AM5. Most of my components except the CPU has been given to a family member. As that build has been upgraded to a 5600.