[Carillo]

Previously 4266C17D-32GTRS can reach 4700 16-17-28-280 with the VDimm around 1.65V. At 4600 16-17-280 the best VDimm was around 1.55V with tightened timings and MT7.0 passed.

Just might come up with some training issues (err 55) and need some tweaks.

Interesting. My Patriot Vipers is water cooled, and can do up to 1.67V HCI. Over that is benchmarking stability. Hope this kit can do 1.65 24/7 on water BTW, if anyone wants a very well binned Viper kit, ( the kit that can do 5000 cl17.17.17.37) DM me. I can send it anywhere

 

[Ichirou]

4600 16-17 1.6V should be easy with that kit

Now I feel tempted to give it a shot, if only there was a way for me to sell my old RAM sticks, lol. No point having multiple kits.

Also, my ASUS Prime Z390-A has issues POSTing over 4,200 MHz. Not sure if it's just my kit, or a BIOS issue. Doesn't seem to be a VCCSA issue. It's probably because it's a mid-tier motherboard instead of a Rampage. I think somebody else here mentioned a similar situation.

 

[OLDFATSHEEP]

Now I feel tempted to give it a shot, if only there was a way for me to sell my old RAM sticks, lol. No point having multiple kits.

Also, my ASUS Prime Z390-A has issues POSTing over 4,200 MHz. Not sure if it's just my kit, or a BIOS issue. Doesn't seem to be a VCCSA issue. It's probably because it's a mid-tier motherboard instead of a Rampage. I think somebody else here mentioned a similar situation.

Asus Z390s are T-Topology boards. You have to install 4 dimms to get higher than 4200. I only managed to get 4500 with 4 bdie sticks on M11H. If with 2 dimms, the max I've got on M11H was 4200.

MSI Z390i should be a good choice if you would look at other z390 boards.

 

[Ichirou]

Asus Z390s are T-Topology boards. You have to install 4 dimms to get higher than 4200. I only managed to get 4500 with 4 bdie sticks on M11H. If with 2 dimms, the max I've got on M11H was 4200.

MSI Z390i should be a good choice if you would look at other z390 boards.

Well that's ironic; I'm already using four DIMMs on it right now, all from the same batch (4x16 GB). Micron B-die.
It simply refuses to POST above 4,190 MHz or so. Not really sure why. I have no choice but to just keep tightening tCL instead.
Any idea for what to try?

On a side note, I found this off of Google; it's an old thread that details something similar (although the thread lead nowhere):

Memory controller sucks on new i9 9900K!

Hello, I just upgraded from a i7 8700K to the i9 9900K and I'm very much regretting it. My overclock was stable with the i7. Now my system won't make it past POST with the newly inserted i9. The boot process always halts at debug code 61 which is "NVRAM Initialization" according to the manual...

www.overclock.net

It seems that certain CPUs could just fail to allow the RAM to go higher.

 

[OLDFATSHEEP]

Well that's ironic; I'm using four DIMMs on it right now, all from the same batch (4x16 GB). Micron B-die.
It simply refuses to POST above 4,190 MHz or so. Not really sure why. I have no choice but to just keep tightening tCL instead.

View attachment 2467202

Ummm, maybe try to play with RTLs. That's the only thing I can come up with. Also, increase VccIO should help a lot. For a reference, 4 Samsung bdies running at 4300 17-18-36 needs around 1.4V IO and SA.

For SR micron 16GB with 4 DIMM config you can try ODT 80-40-48 (WR-PARK-NOM). DLLBwEn=1 might also help.

 

[Ichirou]

Ummm, maybe try to play with RTLs. That's the only thing I can come up with. Also, increase VccIO should help a lot. For a reference, 4 Samsung bdies running at 4300 17-18-36 needs around 1.4V IO and SA.

For SR micron 16GB with 4 DIMM config you can try ODT 80-40-48 (WR-PARK-NOM). DLLBwEn=1 might also help.

Memory controller sucks on new i9 9900K!

You should never have those on auto when overclocking RAM. Auto typically runs them way too hot, often 1.3v or more for 3200MHz RAM when 1.15v or so is usually stable. At your frequency VCCIO and VCCSA could be much higher. You should set them manually to the lowest stable voltage as one of...

www.overclock.net

^ If you take a look at that specific post and a few posts down (and the next page), it seems that apparently, in order to break the barrier, I have to set an atrociously high IMC voltage.
So as you say, perhaps it's really just that my IO/SA is too low (1.23V/1.25V respectively)

What is the point of high frequency RAM then, if they need to run at high IO/SA voltages that can potentially kill the CPU? That truly baffles me.

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

Anyway, what are those settings you told me to adjust? I've never seen them before.
I can't touch my RTLs or IOLs pretty much at all, or I risk potentially failed POSTs. My motherboard already tightens them as low as they can go, and the minimums fluctuate every so often. Thus, they're just on Auto right now.

Perhaps I'll leave IO/SA on Auto just to see how much it requires for the system to POST.

 

[OLDFATSHEEP]

Memory controller sucks on new i9 9900K!

You should never have those on auto when overclocking RAM. Auto typically runs them way too hot, often 1.3v or more for 3200MHz RAM when 1.15v or so is usually stable. At your frequency VCCIO and VCCSA could be much higher. You should set them manually to the lowest stable voltage as one of...

www.overclock.net

^ If you take a look at that specific post and a few posts down, it seems that apparently, in order to break the barrier, I have to set an atrociously high IMC voltage.
So as you say, perhaps it's really just that my IO/SA is too low (1.23V/1.25V respectively)

What is the point of high frequency RAM then, if they need to run at high IO/SA voltages that can potentially kill the CPU? That truly baffles me.

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

Anyway, what are those settings you told me to adjust? I've never seen them before.
I can't touch my RTLs or IOLs pretty much at all, or I risk potentially failed POSTs. My motherboard already tightens them as low as they can go, and the minimums fluctuate every so often.

Perhaps I'll leave IO/SA on Auto just to see how much it requires for the system to POST.

They need to be set according to specific rigs, can just play with them.

RTLs need to be exact to post. Too high IO/SA sometimes might not help.

Also you can arrange your sticks according to their qualities: best-worse-better-worst (A1-A2-B1-B2)

 

[Ichirou]

They need to be set according to specific rigs, can just play with them.

RTLs need to be exact to post. Too high IO/SA sometimes might not help.

Also you can arrange your sticks according to their qualities: best-worse-better-worst (A1-A2-B1-B2)

No I mean, what are those settings you are asking me to change, exactly? I've never seen them before. Are they part of the DRAM Timings section (I'm on an ASUS)?

Wouldn't the RTLs set at Auto already fulfill that requirement? Or can there be potentially bad training?
My IO/SA are relatively low right now, so if you have a suggestion for what I should try running them at, that would be nice.

Not sure how I can figure out which stick is better or worse. Should I just randomly plug them in and test?

Memory controller sucks on new i9 9900K!

What's even funnier is this. The OP posted a thread over on reddit about the same thing. https://www.reddit.com/r/overclocking/comments/b18oaw/absolutely_awful_memory_controller_on_new_i9_9900k/ If you search his comment history though, you'll see that he has a "Bachelor's Degree in Computer...

www.overclock.net

^ This post was also insightful. Apparently it's really just a lottery, though sometimes a BIOS change can help.

Memory controller sucks on new i9 9900K!

what bios are you on.. did you use XMP? ? 1105 BIOS. I did not use XMP, just set all the timmings and volts manually.

www.overclock.net

^ And this one implies a VCCIO sweet spot issue, as you mentioned. I guess I'll have to set it to Auto just to see if it'll POST.

 

[LunaP]

Oh right question on the uh Etreme1 by @anta777 , is there a length it should run for or does it eventually end? going on almost 2 hours now and still on cycle 1, is it like HCI and just keeps going till u tell it to stop?

 

[Ichirou]

Oh right question on the uh Etreme1 by @anta777 , is there a length it should run for or does it eventually end? going on almost 2 hours now and still on cycle 1, is it like HCI and just keeps going till u tell it to stop?

anta777's config is kind of long indeed; it's part of the reason why I just stick with 1usmus' config and do normal usage tests instead once it completes.

 

[OLDFATSHEEP]

Not sure how I can figure out which stick is better or worse. Should I just randomly plug them in and test?

They are all under timing config.

Can use my suggestions as initial values, and tune up and down.

 

[Betroz]

Asus Z390s are T-Topology boards. You have to install 4 dimms to get higher than 4200. I only managed to get 4500 with 4 bdie sticks on M11H. If with 2 dimms, the max I've got on M11H was 4200.

The Z390 Apex should be able to do better though?

 

[TheBoom]

Oh, so you ramped up VCCIO to 1.38V just to get tRCD+tRP+tRAS working, and haven't bothered to tighten it yet. Do keep me posted then, since I'd like to find an optimal VCCIO to try as well.

Have you experienced any sweet spot issues with VCCIO? Or has it worked for the most part so far just setting VCCIO that high?

Yep there is definitely a sweet spot, but it only shows when doing imc or combined stability tests.

I’ve found linpack the easiest as the closer you get to the sweet spot the more residuals match without getting WHEA errors as well.

If you see the official page for the Prime A they do mention up to 4200mhz memory. Could be a bios limitation.

As for RTL/IOLs someone just posted a nice guide last week here on how to properly train them. Seems to work for me.

Set RTL init to CLx2 + 37 and IOL inits to 4. IOL offset to 21.Your IOLs should settle around 7-8. If they don’t, set them manually and subtract the same difference from the corresponding RTL.

 

[LunaP]

Ok this is weird, if I use BLCK OC and set mine over BLCK 105 I can easily push 3800 freq to over 4000 w/ under 1.46v and timings 16-17-17-36, yet if I set it manually to 4000 base w/ 0 blck it won't boot even if I go up to 1.6v and VCCIO/VCSAA ( up to 1.3 each ), also note 3800 only requires VCCIO @ 1v and VCCSA at .800 Only had to bump VCCIO to 1.1 for 31 cache which I took back down to test with in the event it was voltage related. Tossing voltage at the CPU also doesn't help, so thinking its def a bios issue.

 

[munternet]

I think this is about as far as I can push these sticks F4-4266C17D-32GVKB
I've tried everything I can think of but I just get BSOD or no POST if I push anything further
Ran some GSAT and TM5 Extreme and Played some BFV
Going to run it for a daily for a while

4600-18-19-19-36-2
vdimm 1.45
vccio 1.42
vccsa 1.47

 

[TheBoom]

I think this is about as far as I can push these sticks F4-4266C17D-32GVKB
I've tried everything I can think of but I just get BSOD or no POST if I push anything further
Ran some GSAT and TM5 Extreme and Played some BFV
Going to run it for a daily for a while

4600-18-19-19-36-2
vdimm 1.45
vccio 1.42
vccsa 1.47
View attachment 2467233

Dayum if I ran my IO and SA that high my chip would probably fry itself getting past windows.

 

[Imprezzion]

Dayum if I ran my IO and SA that high my chip would probably fry itself getting past windows.

Nah, i've run way higher then that for benches / short tests. 1.55v SA and 1.50v IO survived just fine for several hours. Not that I'd recommend doing it but it won't cause immediate damage.

 

[munternet]

Dayum if I ran my IO and SA that high my chip would probably fry itself getting past windows.

I'm pretty confident those voltages are fine for z490. Well below what a few other people run 24/7
Luumi took it to the limit and killed a CPU @ 1.65 sa or io according to this

It's worth watching right through if you haven't seen it

 

[TheBoom]

I'm pretty confident those voltages are fine for z490. Well below what a few other people run 24/7
Luumi took it to the limit and killed a CPU @ 1.65 sa or io according to this

It's worth watching right through if you haven't seen it

I meant literally, temps wise. I have a dud so I have to run it at 1.35vcore load to be stable at 5.0ghz. With IO and SA at 1.4v the chip peaks over 90c in normal gaming use with some avx instructions.

Also we have no idea what kind of long term degradation these voltages might cause to z490 yet, so that’s another factor.

I remember reading somewhere that earlier skylake architectures were prone to degradation and motherboard traces getting damaged by imc voltages in excess of 1.35v.

 

[Ichirou]

Ummm, maybe try to play with RTLs. That's the only thing I can come up with. Also, increase VccIO should help a lot. For a reference, 4 Samsung bdies running at 4300 17-18-36 needs around 1.4V IO and SA.

For SR micron 16GB with 4 DIMM config you can try ODT 80-40-48 (WR-PARK-NOM). DLLBwEn=1 might also help.

Well, I tried leaving VCCIO on Auto, and it didn't help. I couldn't find the "ODT" setting, but I did test out DLLBEWN a little, fiddling with values up to 4, and I can't say that I had great results.
The RAM started feeding the PC bad checksum data... Very scary really; looked like data was getting corrupted on boot. Sure, it allowed certain configs to boot, but I can't say it was worth it.
That is definitely a "tweak at your own risk" setting that I wouldn't recommend anyone changing, lol.