[BroadPwns]

Hey guys, I've got lent a 13900kf with some A-Dies and a garb-o Z790-P Pro from MSI, while I absolutely do not expect any results from this board, I'd like to make sure I've tried all options available. I've had some decent success with DDR4 B-Dies, but with DDR5 I'm completely lost due to the vast amount of possible adjustments. The 7000 C36 XMP seem stable, so I'd already call it a success, given the board is pretty much a renamed Z690 board.

Is there a condensed and usable guide, which focus on how to tune voltages, RTTs and similar settings? This thread is just way too big.

 

[BroadPwns]

SA 1.18V, VDDQ TX 1.25, VDD2 1.18V, VDIMM/VDDQ 1.45V. 7200MHz undoable on Z790-P Pro, although the board boots at 7800MHz without any issues, which is annoying. It's Win10, so TM5 was run with e cores disabled, else it just randomly went bonkers, VST with them reenabled.

Is there any possible trick to force the board to go stable beyond the 7000MHz? I've tried all kind of combinations of voltages and messed a bit with RTTs, but no luck at all. I know the board is pretty much garbage, but I may be missing something.

 

[BroadPwns]

Hello bro! I remember you my last discussion on the Z690-A Pro thread a few good months ago. I got the Z790 Tomahawk and I simply cannot pass anything above 6800 for Y-Cruncher VST. I think that 7000 passes TM5, but will get an error in VST, the furthest I got was playing around with the PLL voltages and extending the error time from 1-2 minute to 20+ minutes. On 6800 I ran for over 3 hours tho. I've spoken with a few people (one that I know personally IRL) and none of them manage over 6800 even on the more expensive Carbon board. It's a shame because MSI Z690 was king for DDR4. At this point I don't know if it's an hardware limitation or a BIOS update that will at least allow us to get 7200 y-cruncher stable.

If you've tried for 7000 on the kit you have (the 5600C46 M-Die), then be happy that you've got 6800MHz with it. With A-Die you should have absolutely no issue breaking 7000MHz and above.

 

[BroadPwns]

The kit I got was highly recommend by Buildzoid in this video here as some of the cheapest A-Die you can get, but it ended up being M-Die for me. A friend of mine has both the MSI Z790 Carbon and Asus Z790 Apex and over 4 kits of A-Die, I tried the A-Die kits on the Z790 Tomahawk and couldn't get over 6800 y-cruncher VST stable, 7000 passes for TM5 though. My friend also couldn't get any of his A-Die kits to pass anything over 6800 y-cruncher VST stable on his Z790 Carbon, while all of them do 7800 y-cruncher VST with ease on the Apex. My kit 5600c46 MDie kit did 7200 VST stable on Apex, but that's a spectacular board so not a realistic comparison. So while as you say 6800 for MDie is lucky be it as it may be or not the main point is a lot of MSI Z790 4DIMM motherboard owners are not getting over 6800 y-cruncher VST stable.

While I don't know any other MSI Z790-P owners, I've spoke 3 other guys on this forum that had the MSI Z790 Carbon and a 7200 XMP kit and couldn't get over 6800. Overall I guess am kinda lucky to have gotten M-Die since at least I can do slightly lower subtimings on some of them.

If you want to feel happier, then y cruncher VST with 7000MHz XMP at A6 bios for the Z790-P resulted in instant error, only newest A92 bios made 7k stable.

 

[BroadPwns]

Oh wow that's awesome! Have you tried messing with the PLL voltages as well? If you've already got 7000 VST stable with stock PLL voltages then I'm rooting for you to get that 7200!
I personally kinda gave up since none of my friend's 7200/7600/7800 XMP kits did even 7000. Anyway there's an H85 beta BIOS version for the Tomahawk, I'll give it a shot later this week. a

I had and they changed nothing to no suprise, since the frequency is so low.

 

[BroadPwns]

Can someone help me find examples of F5-8000J4048F24GX2-TZ5RS on Z690 Apex please. I've been looking but can't find anything. Thanks

It's literally A-Die, copy whoever's attempt and have a blast.

 

[BroadPwns]

Isn't the 48GB kit M-Die?

I hope this help and I hope you get it working at 7600Mhz , this board need some loose timmings but not bad at all , as it is 100% stable for daily 24/7 use. Hopefully this will help the normal guy with a 4 dimm motherboard reach 7600Mhz and gives an idea where to start since 99.99% of the users on this thread are using 2 dimm mobos Apex , Kingpin , Tachyon , Aqua OC etc... and memory on water cooling with huge voltages to reach for the highest Mhz, etc... for the e-penis 😂 , me included 🤣 , have no posted anything memory related in a very long time since i stopped doing extreme overclocking on LN2 😎 .
Go for it ! Mucha Suerte amigo hispano parlante de Argentina 👍.

The e-pen results:

Cyberpunk measured in the city, FFXIV the city fly by benchmark, Witcher 3 Novigrad market, AC Competizione in SPA with 30 cars total. FFXIV max are borked by stopping measurement tad too late, but who cares about max anyway, 0.1% in Witcher 3 with RT enabled probably a random event (I did not run lab tests). The rest are valid, but done with just 2 passes.

Stock clocked 13900kf with just an UV and just 3080 Ti at 1950-2010MHz, so apparently the e-pen isn't strictly e-pen, even if at 1080p - there are gals with 4080, that do more FPS in 1440p, than 3080 Ti at 1080p.
Overclocked timings:

The conclusion is, using up time for just frequency is an absolute waste, you WANT to tune timings. Regardless, finding maximum frequency is a good first step, but timings>>frequency.

 

[BroadPwns]

Anyone here done any testing to see if going above 32GB RAM has any benefit for the newest games? Not talking about max fps here, but loading times and possibly higher minimum fps?

RAM unused is RAM wasted and RAM capacity doesn't affect loading times, unless you lack RAM. Whether or not you need more than 32GB depends on a game, almost none gets close to using up the 32GB.

 

[BroadPwns]

I know, but RAM is good for caching though.

If a game doesn't make your system use more than 30GB, then you're not benefitting from the extra capacity at all, ever.

 

[BroadPwns]

Hello guys, I asking for help. My brand new build has much higher memory latency than expected.

13600KF + MSI Z790 Pro-A and DDR5 G.Skill Trident 6400MHz. Bios newest etc. Clean os install. Disabled core isolation.

On everything stock and just XMP enabled I get astronous 72ns latency.

I attach the examples that this should be much lower. Slower DDR5's geting better latency.

Any ideas what could be wrong before messing with manual timings etc. Is it more failure of DIMM's, or bad internal CPU memory controller?

Second question, i ordered 2 kits of these DIMMS, and one kit was A-die which i use and the second was m-die which I returned. Did I do right keeping A-die if my target is 6800Mhz with low timings? View attachment 2627868
View attachment 2627869

Where's the 72ns? Also, you can't compare scores between different Aida64 versions, as well as you need to close literally every background program for Aida bullshittery.

 

[BroadPwns]

Here is a quick guide/help dedicated to ddr5 overclocking by OCN Moderator and our OCN Team hwbot captain mllrkr88 with the most basic information. hope this helps: https://www.overclockers.com/ddr5-overclocking-guide/

Sorry, but this is useless to anyone, that actually had success with OCing any DDR. Quite frankly, it's completely useless and voltage section is just straight out bad.

Veteran DDR4 bdie OC'er here. Jumped on the DDR5 train which im now questioning if that was a good idea or not 😂

Forgive me team , i understand there is probably thousands of us noobs who find themselves in my exact position here. But at 947 pages, this thread is daunting to hunt through and the search function is....tricky to isolate the info you need. I've gone through a bunch of it and watched countless buildzoid vids that has provided my current understanding.

This is my current setup:

MSI Z790 Edge Wifi (4 dimm board - i suspect very similar to the Tomahawk / A Pro / Gaming etc)
13700KF (Stock)
Kingston Fury Beast 7200Mhz CL38 @ 1.45v

I understand the basic premise and have set the following:

7200Mhz | 40 CAS | 50 TRCD| stock everything else.

1.2v SA | 1.4 TX (CPU VDDQ) 1.4 CPU VDD2 | 1.45 DRAM v (Auto Linked)

I tried every combo I could think of to stabilize the above @ 7200 on Tm5 1usmus V3 and within seconds it was throwing wobblies - a variant of different errors. I increased/decreased each voltage but nothing seemed to be obvious for improving stability at frequency. I've resulted to just leaving the voltages where they are (above) and walking the frequency back. So far my best run has been 6800mhz (same timings as above) that appears to be holding stable for the time being.

Is there something basic I'm missing here?

I am just at the limitation of this motherboard?

I suspect the CPU and ram kit isnt the culprit. I dont care to push the frequency to the moon or anything I just want the damn thing stable at the rated speed of 7200. Please let me know if I'm dreaming or not lol.

Appreciate you taking the time to read and respond.

First of, install newest bios, it's crucial. Then try the XMP again.

 

[BroadPwns]

Just a small update, decided to try the latest Beta (H85) BIOS for the Tomahawk, 6933MT/s passed y-cruncher VST running all night (391+ iterations). Although 7000 still crashes within a few founds.
IMC Voltages for 6933 were on Auto which results in CPU VDD2 1.4v, CPU VDDQ 1.4v, VCCSA 1.3v.
I'm going to try the latest stable BIOS on 6933 again, because to my recollection the stable one didn't pass 6933.

At that point, I'd just buy different RAM sticks instead of wasting my time. 1.3V VCCSA is 8000+ territory, same for VDD2.

 

[BroadPwns]

Oh no I think you may have completely misunderstood me, I simply left all IMC voltages on Auto without tinkering them as an initial first attempt, simply testing if 6933 was feasible on the Beta BIOS for Tomahawk. After seeing that it is indeed feasible, am now testing 6933 on the latest Stable BIOS. Afterwards I'll minimize the voltages.
I already minimized my voltages on 6800c32 setup with fully tuned subtimings to pass VST on 1.1 VCCSA and much lower VDD2, so I highly doubt that 6933 would need much more after minimizing...
This is nothing to do with the kit itself but the actual motherboard as I have already tested a couple of A-Die XMP kits ranging from 7200 to 8000.

When I worked with an Apex, all the A-Die kits I've tried did 7800 VST stable extremely easily on 1.2x VCCSA.

And yet, my garbage P Pro did 7000 easily. Still, it required the most recent, beta bios.

Also, did you mess with RTTs?

 

[BroadPwns]

**** Guess what?! As I'm typing this, turns out the latest Stable BIOS crashes 6933, so I was right to recall 6933 didn't work. So far 6933 only works on the latest H85 Beta version. I even tried my friend's 7200 and 7600 Gskill and TeamGroup kits and same thing.
Well thanks to you at least I gave the Beta BIOS a try, at least now 6933 is doable, still a slight improvement over 6800, lol. Guess I'll fine tune/minimize the voltages for 6933 and call it a day... And maybe replace the mobo later on. It's a shame tho cuz I love the stealthy black aesthetics, fits in very nicely with the rest of the components.

This is why I initially wrote about the bios being crucial. If that won't help, then I'd consider pulling my hair when trying out different ODTs/RTTs, my A-Die kit was happy with 60-48-34-34-40.

 

[BroadPwns]

Been curious if a fan blowing on the back of the motherboard would help temps at all. It appears to make a small difference 0.5-1c cooler RAM/CPU/VRM

What exactly did you expect lol. You're not cooling anything on the back of a motherboard, traces don't benefit from airflow and the 1C drop I expect is not even from the fan.

 

[BroadPwns]

I imagine it’s going to help cool down something. Especially if you use a closed case. Heat always builds up over time. The warmest area on the rear of my Apex Z790 is behind the PCH, so a fan would probably help cool it down.

"Cooling" down the back of a motherboard is pointless, every IC heat source is completely isolated from a motherboard and trying to cool down anything by blowing a fan behind the mobo is ineffective at the very least. The only device, that benefits from having an active backplate is a GPU and even then it's not making that much of a difference, unless you create an oven between GPU's PCB and the backplate.

 

[BroadPwns]

And if you ignore all the maths, get timings low as possible, get full stability and on top of that - progressively and measurably best performance?

 

[BroadPwns]

Well, I’ve seen some motherboards that have active cooling on the back. One was an X299 Dark. It worked pretty well. As for GPU back plates. My 3090 Kingpin would cook without a full coverage Fujipoly thermal pad, backplate, and fans.

I do not cool the back of my motherboard, but I don’t see how someone using a fan is gonna hurt anything.

X299 used terrible mofsets for VRMs, so the entire area was scorching hot, currently VRMs are so overpowered you can use any motherboard for LN2 (putting aside voltage control). Cooling it won't hurt anything, but it's a complete waste of time and money, it does nothing.

 

[BroadPwns]

Where exactly are you getting the "formulas" from?

 

[BroadPwns]

I'm trying to push 7200 and can't get it to stabilize long term, about 3 passes will be ok but the 5th or 6th will always error

Even with AC and the ram temps below 50c, it still errors

I have a fan pointed at the RAM


Are the sticks defective? The Z790 hero says it can do 7600 XMP with these particular sticks

What am I missing here? Please give me some guidance

View attachment 2628121

I'd start with lowering VDD, VDDQ (both CPU and RAM) and VPP.