chonk wrote:
Is that normal? Is it how it works? People are talking about setting it manually. How?
The CH6 tries to apply settings that it finds from training on boot up. These are not always the best for stability or speed. This is a copy and paste from a previous post I did. It gives a half decent method.
syldon wrote:
Try this.
Boot to an OS install HWinfo. Retreive the timings for the following:
Dram Cas# lantency
Dram Ras# to Cas# read delay
Dram Ras# to Cas# write delay
Dram Ras# pre time
Dram Ras# act time
Trc_SM:
TrrdS_SM:
TrrdL_SM:
Tfaw_ SM:
Trfc_SM:
Trfc2_SM:
Trfc4_SM:
reboot back into the Bios
On Main extreme tweaker screen
AI overclocker set to manual
Bclk 100
CPU core leave at 36 for now
Memory frequency set memory to 2800 (since you ran at 3200 for a while this settings should be fine)
Disable CPU boost
Set CPU core voltage to offset mode and +
Set cpu core offset to 0.03125 (I run on a 1800x so this may need to be 0.0375, but not higher, Try to keep your cpu below 1.42v)
I have CPU soc voltage on Offset mode + @ 0.0500. This keeps it below 1.15v for me.
I leave Dram on auto.
Go into Dram timing control and add those setting that HWinfo gave
Dram Cas# lantency
Dram Ras# to Cas# read delay
Dram Ras# to Cas# write delay
Dram Ras# pre time
Dram Ras# act time
Trc_SM:
TrrdS_SM:
TrrdL_SM:
Tfaw_ SM:
Trfc_SM:
Trfc2_SM:
Trfc4_SM:
It should reboot with no issues.
Ok now to fine tune.
In Dram timing control at the bottom add CMD2t to 1T
Reboot. If this works then start to work the PROcODT. In Dram timing control again at the bottom.
Asus recommends 40-96. Work from 96 ohm and work your way to 40. Reboot on each setting. Find the lowest setting and leave it there.
Once you have this increase your CPU in increments till you are happy with the temps. Remember the hotter it is the shorter the life of the CPU. On an 1800x I find 3850 is my sweet spot. It is marginally hotter than 3600 (default). I gain 10 degrees at 3900.
Then increase your memory in increments till you get your max .
I should add that you may need a different ProcODT timing for a change in memory speed. I use 40 for 3200, 43.6 for 3333 and 3466, then 53.3 for 3600. I had multiple choices for 3200 and 3333, but those are the timings that give the best speeds and also lowest CPU temps.
EDIT: changed the voltage offset for CPU from 0.375 to 0.0375 - bad place for a typo.
gupsterg wrote:
Do you have some data regrading ProcODT affecting CPU temps?
I don't no sorry. But if you are setting up then temps is a big indicator things are going well or bad. If memory is being forced then it is going to draw more power to get it to work. This is why you see some set ups with higher voltages to force memory at higher speeds. I could see the difference in my temps when benching speeds.
A stable setup means you can keep those voltage low. I run on 1.35 with memory and around about 1.37 for cpu. I am currently running a cpu offset at 0.03125 and soc offset at 0.0500. Memory is at 3466 and CPU at 38.5 with revision 1403. 9920 will give me 3600 mem speed but it fails stability tests.
gupsterg wrote:
ProcODT for me had no effect on any temperature.
Information from Elmor, Raja@ASUS and The Stilt has not highlighted to this fact either. Their posts of info can be found in OP of this thread, section RAM Info .. > C6H UEFI 0079 onwards ProcODT... > ProcODT/Fail_CNT.
On my last batch of memory I had only 1 option. So on different setting on the ProcODT, it either ran or it didn't.
The reason I mentioned temps was purely down to the fact I had 3 options that would run with a specific procODT at 3200 and 3333. The only thing I have seen is, when I had a few options of ProcODT to choose from, that when benching and stressing the system the fans would kick in a lot slower on certain ProcODT settings. This is indicative that temps are climbing more slowly. There was also a correlation between a good ProcODT adjustment and the way the temp climbed.
I also found that video from AMD. Asus recommend no higher than 96. I haven't seen any memory that would require above 96 yet anyways.
I read through that link, some of it I have seen before. The only reference is that ProcODT could be useful for stability, which is basically what I said.
gupsterg wrote:
The question I posed regarding ProcODT was relating to you highlighting temperature difference on differing settings. You could not provide data, so an assessment can't be made if it is correct methodology or testing. I have in the past use 40ohms to 80ohms for testing if it resolved a rare cold boot issue, I'll endeavour to get the temperature data for CPU on various ProcODT 😉 .
From memory I have seen no change between temps of CPU using various ProcODT. So I do not wish members to have mis-information.
And now you have linked the video which has caused mis-information regarding ProODT. 96ohms is not LN2 territory. Would you think Elmor, Raja@ASUS and The Stilt would be giving incorrect advise?
That video also has some another aspect which is misinformation. I would not waste my time on that video TBH.
Video set at 48 Ohm. Bench starts at 11 seconds and climbs within 2 seconds to 73. Single core output 417.2 Multi core 4383.7.
Video set at 43.6 Ohm. Bench starts at 16 also climbs within 2 seconds to 70.6. Single core output 419.9 Multi core 4410.3.
Video set at 40 Ohm. Bench starts at 13 takes 3 seconds (near 4 but I wont split hairs) to level at 71.8. Single core output 418 Multi core 4456.7.
The temp reading is not exact. My point was the rate of climb is significantly lower when the memory is more stable. These climbs are exaggerated because I have extra processes running to make the video. That being xsplit and Corsair cooler app. With the extras processes taken away and nothing else running, then the difference in climb is much more noticeable.
It is also backed up by aida64 memory benching http://imgur.com/DIKlE5k. Edit In hindsight it would be better to use a numerical meter such as this to back up which would give you the best ProcODT setting.
2nd edit Comment deleted
