Memory Model F4-3600C16Q-64GTZNC kit compatibility

[LBro]

Hum,
I was hoping for a bit more guidance on this one. You mention this kit as being top notch. I would be pleased with a speed of 4000 or more if possible. Very pricey RAM though, that is for sure.

I have looked at many ASUS Z390 OC guides. Though at the time, my focus was not on pushing RAM, rather CPU. I went back and had a look around from the RAM side of things and ran into a couple of guides that might be of use.

Questions:
If I use XMP to set up the DRAMs, it will over course set up all the voltages and what not. Now if I up the frequency, is XMP going to adjust the corresponding voltages as well for the DRAM kit and selected frequency?

Do you have any rules on how high the CPU VCCSA voltage can go without frying the CPU? The CPU is an Intel Core i9-9900K. I might be able to find this in a guide, but possibly you know?

In "lowering timings to optimize performance" - Are you talking about such things as CAS Latency and what not? I will likely need your assistance with this one... I will let you know.

Thanks for your help!
LB

[GSKILL TECH]

The more you know and understand the better the results can be. These videos are most helpful in understanding the hardware you have.

XMP will only input the DRAM Frequency, timings, and Voltage the memory kit is rated for. With XMP enabled, you can manually adjust all of these values to see what other configurations can work properly.

AUTO Voltages can work fine to some extent, however to truly push the system it will require manual inputs. Find the highest value stable with AUTO settings, note Voltages of this configuration, then you can manually set it higher to see if you can stabilize higher values.

High Voltage is not an issue, unless the CPU cooler can not keep it cool. As long as you scale up step by step then you won't have to worry about frying
the CPU. Once the processor stops responding with stability, you simply stop as any further will likely not improve results.

[LBro]

Hi,
I went through many guides and Google searches. I got the kit up to 3800 speed using a Dram voltage of 1.3875v. One issue I am having is with the Mobo DRAM voltage settings on the Asus z390 Rog Mobo. Manual Dram voltage input sometimes works and most of the time just defaults to about 1.2020v, even though the voltage amount adjustment is still present. Something seems to be influencing this setting. A search via Google shows others have had this issue as well, but no solution was found in those posts.

Any ideas?

UPDATE
Not sure if this setting had anything to do with the DRAM voltage setting but changing it seemed to help. The CPU Chore/Cache Voltage was increased ever so slightly. After this I was able to set the DRAM voltage to 1.43750 and it reports back a voltage of 1.456. The strange thing is that when I try and increase the DRAM frequency from 3800 to 3866 the voltage defaults back to the 1.208v or so setting??? Only by bringing the DRAM frequency back down to 3800, will the DRAM voltage revert back to 1.456v in BIOS.

Do you happen to know what settings affect the DRAM Voltage setting?
At the DRAM speed of 3800 my DRAM Timing is:
CAS Latency - 15
RAS to CAS Delay - 15
RAS to ACT time - 36
Dram Command Rate - Auto

With the DRAM speed at 3800 and the highest voltage per above I tried to lower the above settings to the specs on the RAM and the result is a failed boot.

Thanks,
LB

[emissary42]

From my experience DDR4-4000 with four dual rank modules probably won't be easy or even possible on most Z390 motherboards. Many 4 DIMM boards max out around 3866 with four DR modules. You also might have to test different BIOS versions to even post with 3866/4000 on ASUS models. My old MXIG needed to be downgraded to at least 0905 to be able to run dual rank modules at 4133.

As for Vdimm, some overvoltage is probably manufacturer intended to help with compatibility in edge cases. Most boards don't have Loadline Calibration or Vdroop control for Vdimm, so you have to work around it as good as you can.

If memory settings reset to default frequency, timings and 1.2V, the board probably failed to POST with your manual settings (multiple times). Afterwards your actual settings will still be present in UEFI, but not active unless you save & quit. In case of the board failing to POST you should see the codes on the debug display looping, until the auto recovery kicks in. You can actually disable that feature, but most people prefer not to clear CMOS constantly. So you just have to realize and be aware of what is happening.

[LBro]

@ EM42,
Thanks for chiming in!

Max out around 3866 - Indeed, I am no expert at this. But everything I have tried so far craps out when trying to go up to 3866 as it will not even post. I think at this point I would settle for a stable 3800. 3800 posts well, but failed a MemTest86 with 4 errors at the end of its approximate 8 hour run time. I checked DRAM temps and to me they looked good at 36C.

I am not sure where to go on trying to get the Drams stable at 3800, they seem very close. Here are settings I am using and if you need more let me know. Simply looking for your thoughts on what next to try in stabilizing at 3800.

3800MHz - Dram Frequency
1.385v - Dram Voltage - Note I upped this a bit and had less stability. This made me think the temps might be up, but they seem fine. When going up in voltage, does a CPU voltage need to increase as well?
1.328v - VCCIO
1.424v - SA Voltage
1.064v - PCH Core Voltage
15 - CL
15 - RAS-CAS Delay
36 - RAS# ACT Time
Auto - DRAM Command Rate

Not sure what else you guys need spec wise. just let me know what you may need and I will post it. At this point, would trying a different BIOS be prudent? Or would tweaking some setting values be a better plan?

Thanks so much!
LB

[emissary42]

@ EM42,
3800 posts well, but failed a MemTest86 with 4 errors at the end of its approximate 8 hour run time. I checked DRAM temps and to me they looked good at 36C.

How long did it take until the first error popped up?

Temperature issues usually take a while before they pop up, but will also result in a lot more errors over a long test.
However in general adding a fan for the modules is not a bad idea, if you are overclocking with high Vdimm.

1.385v - Dram Voltage - Note I upped this a bit and had less stability. This made me think the temps might be up.

Seems logical. Luckily temperature issues are the easiest to diagnose, so I would start by ruling them out first.

When going up in voltage, does a CPU voltage need to increase as well?

If your CPU was overclocked with default/lower memory ratio, you might have to readjust your Vcore.

1.328v - VCCIO
1.424v - SA Voltage

Those seem very high. My 9900K needed less than that for DDR4-4400+.

1.064v - PCH Core Voltage

Can be left alone (at default) usually.

Auto - DRAM Command Rate

Set 2T for that, if the board does not default to it. It can sometimes help shorten memory training.

would trying a different BIOS be prudent?

I tested like a dozen versions before staying with 0905 forever. Some of the older ones can definitely be better for memory OC.

[LBro]

Hi,
>How long did it take until the first error popped up?
The test ran for over 8 hours, for 4 passes. The errors showed up in pass 4, at the end, with less than an hour to go.

>Those (1.328v - VCCIO & 1.424v - SA Voltages) eem very high. My 9900K needed less than that for DDR4-4400+.
Yeah they are high, the AI OC & Auto set them high. Any recommendations as to approximately where they should end up? Based upon what you say I will step down closer to your numbers.

Thanks a ton for your other recommendations! Since I had errors at 3800, I thought it wise to drop back and do the Memtest86 again at stock 3600 to make sure that is solid under XPM II. That is running right now. I think the RAM temps at 36C are good? Seems to me the sticking point in all this is the Dram Voltage. If I use XMP II, 3600 speed, the Dram voltage goes to 1.45v, posts fine and is doing good in the mem test. Now if I go to 3800 speed and try to hard set the Dram voltage to 1.45, it won't remain there and reverts back 1.2xx volts or default. Any ideas on that? I know you eluded to many failed posts causing issues. That was the case. How best then to get around that? Will the save and reboot allow the set voltage to take hold?

Thanks again!
LB

[emissary42]

Very late errors might be temperatures, but not necessarily the memory itself. You can also revert your CPU OC for that, which eliminates any CPU related errors as the culprit.

I'd try voltages in the 1.20-1.25V range for SA/IO. If stability gets worse from that, raise them in small(!) steps until stability returns. Keeping them on the low side will also help a bit with CPU temperatures.

If your settings get reverted to defaults, you need to adjust them until your systems reliably manages to successfully POST with them. Also disable MRC Fast Boot while you are still adjusting timings and voltages. It slows down the POST, but makes sure that memory timings get properly retrained upon your changes. You can reenable it, once everything is fully stable.

[LBro]

Ok,
At 3600 speed the mem86 test went fine. I should have done that first, just to establish a good, known baseline...

I had some success at getting back up to 3800. I was able to get the Dram V up to about 1.43xxV, which is higher than I had when it failed the mem test late. I brought the CPU SA/IO down to 1.2v and fired off another memteset86. That promptly failed. I took the CPU SA/IO up to 1.25v and am doing another Memtest86 on that configuration.

Question: What would roughly be faster?
1. 3600 speed at 14-14-14-34 timing
OR
2. 3800 speed at 15-15-15-36 timing
??

You have a lot more experience than I will ever have at this and I am not sure of the speed relationship of the memory timing vs the frequency...

I am starting to really wonder if going from 3600 to 3800 is worth it for the speed gain, then with the higher CAS Latency, there is some tradeoff I would think, but am not sure how much. This is a premium kit, but in what you say about the Z390 MOBO, it seems like it does limit the memory OC options. With a speed of 3800, could it still use the 14-14-14-34 timing?

If this memtest at the higher V on the Dims fails late I will add some cooling. I think I still have some fans for that. But at 36C it still seems to me that is under control. If that fails I will try older BIOS versions.

Thanks again,
LB

[emissary42]

Stability at 3800 with 14-14-14 might require around 1.5V.
The actual IC temperatures will a higher than the DIMM_TS reading, so the 36°C might be a bit misleading.

Usually frequency > timings, to a certain extent. However if you keep having trouble stabilizing 3800 maybe save yourself the headache and just optimize 3733/3600 instead. You can significantly improve both latency and also bandwidth from tweaking timings vs whatever the board trains them at.

Some results from my old Z390 setup for comparison,

1) F4-4000C19D-32GTRS, XMP-Auto


2) F4-4000C19D-32GTRS, 4000 16-16-16 (moderately tweaked sub-timings)


3) F4-4000C19D-32GTRS, 4133 17-17-17 (moderately tweaked sub-timings)


Even though the frequency between (1) and (2) is identical, the timing improvements result in higher bandwidth and lower latency. The jump from (2) to (3), so +1 step in frequency & +1 CL/RCD/tRP, really is negligible in comparison. You trade in some latency for a bit more bandwidth. Even though it was possible, the voltage increase for 4133 at CL16 would have required active cooling for the modules.

[LBro]

I agree on DIMM temps. Since Memtest86 is basically running DOS, I don't have very good temp tools and it's Dram temp reading is NA. I read where DRAM temps in the 40-45C range can start to have error issues. I had thought the temps over 60C were the trouble point. Thus, since the 40C or so range is not that far from 36C, it stands to reason that temps might well be involved.

I found and added the memory fans. I had 2. Both G Skill and sleeve bearings. The blade count and blade size differed. I tried to see if there was a difference in noise or airflow. Hard to say there. Over time I will use both of them as those sleeve bearings may not hold up. Seems a shame to cover up all that fancy looking ram. Though I could do without all the lights, just lower the price and make them like in the past. That would suit me fine.

Gotcha on optimizing the timings @ 3600 or so. That may end up being the way to go. Though I will go up to the 1.5v on the Drams. I was at about 1.41xx and my memtest died. With the new cooling and a Dram V of 1.474v, I will try it again.

You got some nice result on your 390 setup! Thanks for sharing them!

I guess I have the older BIOS(s) that can be tried too. So the jury is still out.

Here is to 9 hours of memory testing, if I get that far.

UPDATE:
Just took a look at the Memtest progress. 9h 10m elapsed and at 91% on final pass 4. This is close to making it at 3800. I will add to this once I know the outcome of the finalized test.

Thanks!
LB

[LBro]

Success! The Memtest @ 3800 made if through the test just fine! Here are some pics of specs and what not:

https://drive.google.com/file/d/1Gb6...ew?usp=sharing

https://drive.google.com/file/d/1V0W...ew?usp=sharing

https://drive.google.com/file/d/1f6D...ew?usp=sharing

I was thinking on this end to go up to the next step @ 3866 and up the Dimm V a bit more to around 1.5v. What you think? Or would it be better to call 3800 good and go after tighter timing? I probably should do more testing in Windows though?

Thanks,
Dave

[emissary42]

I'd recommend at least testing with TestMem5 and the most recent Anta777 config, plus an hour of Aida64 stability test and whatever else is relevant to you. Some demanding games like BF5(MP) can also be useful for stability testing and it is possible to run GSAT inside Windows via the inbuilt Linux subsystem.

In case you want to continue tweaking, definitely save your current settings to a BIOS profile and back it up to USB, just in case you ever have to clear CMOS.

If your system does reliably POST at 3866 it is very likely to stabilize that frequency. With voltages smaller adjustments are recommended, so you don't end up overshooting it, which can start hurting stability (and longevity).

Edit, GSAT: https://www.overclock.net/threads/ru...x-wsl.1747674/

[LBro]

I had already grabbed the BIOS settings, but need to copy to USB.

I will do the test with the tools you mention - Thanks for the tips!

I will hold off on the voltage increases until needed!

I am in the NW corner of the US. Are you in Europe? Just wondering as we both seem to come on at odd hours here and there.

LB

[emissary42]

Yes, I am indeed from Europe. Due to work, my online times tend to be all over the place though