NEED HELP WITH MEMORY OVERCLOCKING VOLTAGE ON Z690 HERO

[tomer20vt]

hello everyone i have asus z690 hero motherbord and i9 13900k cpu

MY MEMORY KITS IS G SKIL 2X16 DDR5 6400 CL32-39-39-102 1.40V

I OVERCLOCK THE MEMORY TO 6800MHZ WITH THE SAME CL

I ADJUST THE VOLTAGE

IMC VOLTAGE 1.2 DOES IT SAFE FOR 24\7?

VDD IS 1.4 VOLTAGE DOES IT SAFE FOR 24\7

VDDQ SET TO 1.35 VOLTAGE IT SAFE FOR 24\7?

I FOUND IN HWMONITOR VDDQ TX VOLTAGE 1.4 WHAT IS USE FOR I CANT FIND IN MY BIOS TO ADJUST THIS VOLTAGE DOES 1.4 IS SAFE FOR 24\7?

THANKS FOR THE HELP

 

[P4-630]

Perhaps @ir_cow can you tell more about it.

 

[plastiscɧ]

In my opinion it is safe.

I chase 1.6 volts through the RAM it has not yet had a negative braking effect.
However, it MUST BE COOLED

XMP I ___realtime clock toggeling GSKILL TridentZ RGB 4000@ASUS Z590-E wifi ____further/deeper adjustments

aight guys, greetings again! this time i have a quite (maybe) crispy question! here are the stats so far trying to describe the present state as good and precisely as i can: When i run XMP II profile i adjust the ram to 19-19-39 2T 700 as recomended by the manufacurer! fair enough. I have an...

www.techpowerup.com

maybe this chart helps u a bit to get more confident in mind

 

[Deleted member 185158]

VDDQ TX is Vdimm if vdimm isn't available with your version of Hwmonitor.

IMC voltage 1.2v is probably too high.

System Agent also can run lower than 1.2v.

Anything more than 1.45vdimm is considered high, will unlock high vdimm capable in bios for more.

And probably don't use DDR4 specs as reference either. Lower voltages are better when applicable.

Searching for performance, lower memory divider and use bclk instead of increasing the divider. 1:19 bclk:memclk is pretty decent. My system favors 1:14 divider actually.

 

[ir_cow]

Memory voltage is will depend on the ICs used. Its not universal. That G.SKILL kit is the same as I have (should be SK HYNIX). The datasheet says 1.5V for that, but will need active cooling on the memory (aka fan). No need to worry about the XMP 1.4 though.

As for supporting voltages, you have System Agent (SA), VDD2 (IMC) and VDDQ_TX. Each vendor likes to rename these.. I'll have to check the ASUS Hero I have because instead of VDD2 its called Memory Controller If I remember correctly. Which is correct, but even VDD2 isn't followed always. I've seen labels flipped before as well. The point I'm trying to make here is don't trust the naming conventions in the BIOS.

In any case, I don't know the 24/7 safe numbers. SA seems to be fine at 1.35V, VDD2 would be 1.35-1.4V and TX 1.4-1.5V. I have noticed if the SA is too high, memory becomes unstable. VDD2 at 1.45V will eventually lead to the IMC overheating and cause the system to not boot for a while.

If you check what auto is, its usually 1.35 SA, TX 1.35-1.4 and VDD2 1.3-1.35. It will change per MB as well.

I FOUND IN HWMONITOR VDDQ TX VOLTAGE 1.4 WHAT IS USE FOR I CANT FIND IN MY BIOS TO ADJUST THIS VOLTAGE DOES 1.4 IS SAFE FOR 24\7?

This is IVR Transmitter VDDQ Voltage...unless ASUS renamed it again.

 

[Deleted member 185158]

Memory voltage is will depend on the ICs used. Its not universal. That G.SKILL kit is the same as I have (should be SK HYNIX). The datasheet says 1.5V for that, but will need active cooling on the memory (aka fan). No need to worry about the XMP 1.4 though.

As for supporting voltages, you have System Agent (SA), VDD2 (IMC) and VDDQ_TX. Each vendor likes to rename these.. I'll have to check the ASUS Hero I have because instead of VDD2 its called Memory Controller If I remember correctly. Which is correct, but even VDD2 isn't followed always. I've seen labels flipped before as well. The point I'm trying to make here is don't trust the naming conventions in the BIOS.

In any case, I don't know the 24/7 safe numbers. SA seems to be fine at 1.35V, VDD2 would be 1.35-1.4V and TX 1.4-1.5V. I have noticed if the SA is too high, memory becomes unstable. VDD2 at 1.45V will eventually lead to the IMC overheating and cause the system to not boot for a while.

If you check what auto is, its usually 1.35 SA, TX 1.35-1.4 and VDD2 1.3-1.35. It will change per MB as well.


This is IVR Transmitter VDDQ Voltage...unless ASUS renamed it again.

System Agent 1.35v?

Whatever board you have, I'm glad I'm not using it. Should be no more than 1.1v on auto....

 

[ir_cow]

System Agent 1.35v?

Whatever board you have, I'm glad I'm not using it. Should be no more than 1.1v on auto....

I guess you shouldn't use any Z690/Z790 MB ever lol. Default is generally 1.3-1.35V. If you go to low like 1.1, it simply won't boot depending on your memory freq and configuration.

I'm at SA 1.3, TX 1.5 and VDD2 1.45V for DDR5-8000 right now. Go much lower and it just doesn't boot.

 

[Deleted member 185158]

I guess you shouldn't use any Z690/Z790 MB ever lol. Default is generally 1.3-1.35V. If you go to low like 1.1, it simply won't boot depending on your memory freq and configuration.

Yeah, that's just too high for any XMP profile. Sounds like a Asus board. Which my Maximus likes to do the same shit.

But no, 1.35v is way too high either way. (In my humble opinion)
Running 1.1v SA on mine. Seems good, I'm not getting gains though.

Current board is a B660-G actually. I'd have to check on auto if it's pushing that much, but by default, I set these manually from the start. Just something I've gotten used too because of older Intel platforms and Asus boards I've used.

I'll digress here. I know nothing of memory tweaking, ignore the Y-Cruncher I posted in the benching threads!!

 

[ir_cow]

1.35V SA is my cutoff. Above this, the CPU just rolls over and starts to give me problems.

 

[Johnthedon1]

1.35sa is fine for everyday use, most board default to this as it is what’s needed for high xmp kits and as sa is not controlled through xmp it uses this as a base to make sure all kits work

 

[ir_cow]

1.35sa is fine for everyday use, most board default to this as it is what’s needed for high xmp kits and as sa is not controlled through xmp it uses this as a base to make sure all kits work

Yes 1.3 - 1.35 is default. But It can have a negative impact if it's too high. I certainly wouldn't suggest going above this value.

I'm using SA 1.25v for DDR-8000 forget reference. Newest BIOS lowered it.

 

[Deleted member 185158]

Yes 1.3 - 1.35 is default. But It can have a negative impact if it's too high. I certainly wouldn't suggest going above this value.

I'm using SA 1.25v for DDR-8000 forget reference. Newest BIOS lowered it.

Because nobody with a kit 6000mhz or less should need more than just 1.1v.

If you feel the memory degraded after some short time and serious benching, it's more likely the memory controller degraded and not the ram.

Everyone under- volts cpu v-core, but insists SA and Vccio needs to be 1.35v cause the board autos it.

Will someone practice SA and Vccio under-volt at TPU? The world may never know....

 

[ThrashZone]

Hi,
Auto will do wild things especially ROG boards
If temperatures are okay auto can be to to a point
It's best to calm/ lower some voltages
VID can be pretty high seeing bad chips will request more than really needed but that's where "silicon lottery" comes from
It's best to go by core voltage and lower it depending on cooling and stability

Asus boards do come with the ability to rate your chip with a Prediction and SP# higher number for example 100+- need a lot less core voltage/ VID than worse chips rated 60+- plus V/F offsets to play with

 

[ir_cow]

Because nobody with a kit 6000mhz or less should need more than just 1.1v.

If you feel the memory degraded after some short time and serious benching, it's more likely the memory controller degraded and not the ram.

Everyone under- volts cpu v-core, but insists SA and Vccio needs to be 1.35v cause the board autos it.

Will someone practice SA and Vccio under-volt at TPU? The world may never know....

On the ASUS Hero for example, I straight up cannot run 6000+ without 1.2v . I've heard this time and time again that 1.1 is all you need, but I have yet to find a MB that works with 1.1 with anything above 5600.

 

[Deleted member 185158]

On the ASUS Hero for example, I straight up cannot run 6000+ without 1.2v . I've heard this time and time again that 1.1 is all you need, but I have yet to find a MB that works with 1.1 with anything above 5600.

Interesting.

Mines locked at .90v being a non K sku and seem to have no issues running over 5600mhz.

Same thing for that G6900. You've even seen the screen shot for Y-Cruncher. That was done at .90v.

Every system is different indeed.
My 8700K, flash back example was fine 4000mhz 1.18v SA.

 

[ir_cow]

Every system is different indeed.
My 8700K, flash back example was fine 4000mhz 1.18v SA.

I'm strictly talking about DDR5 and 12th / 13th gen. I agree they are all different. At one point with a early ASUS Z690 BIOS, it wouldn't boot if you manually set it to 1.1 even with DDR5-4800.

 

[Deleted member 185158]

I'm strictly talking about DDR5 and 12th / 13th gen. I agree they are all different. At one point with a early ASUS Z690 BIOS, it wouldn't boot if you manually set it to 1.1 even with DDR5-4800.

Yes, 12400F and G6900 are both 12th gen socket 1700 and both are non adjustable at .90v

So instead of repeating myself....

Google system agent voltage non K sku 12th gen and please read your findings.

Then maybe we'll be on the same page.

Maybe.

 

[ir_cow]

Yes, 12400F and G6900 are both 12th gen socket 1700 and both are non adjustable at .90v

What does HWInfo say? I get its non-adjustable. Maybe it has a lower cache ratio too? .90v is unfathomable. But those are locked CPUs, so maybe Intel is doing some magic...or straight up ignoring the limit and the BIOS is actually applying 1.35v.

 

[Deleted member 185158]

What does HWInfo say? I get its non-adjustable. Maybe it has a lower cache ratio too? .90v is unfathomable. But those are locked CPUs, so maybe Intel is doing some magic...or straight up ignoring the limit and the BIOS is actually applying 1.35v.

I'm at work for many more hours. So this will have to wait.

But there is not a setting for "system agent".

I do however have adjustable "IMC voltage" but I don't normally set that higher than 1.2v. In fact I think it's at 1.1v actually.

V-Dimm 1.280v Samsung dies 5200mhz xmp 36-36-36.

But if system agent was available, I'd be sure to pass 6000mhz which this system will not do with ascribed hardware unfortunately.

But my point is, 1.35v system agent is rather high. Maybe try the IMC voltage instead??

 

[ir_cow]

If the SA is too high, why is 1.3-1.35 default for all the vendors? If you don't enable XMP, it usually is 1.1 and that is 4800.

 

[Deleted member 185158]

If the SA is too high, why is 1.3-1.35 default for all the vendors? If you don't enable XMP, it usually is 1.1 and that is 4800.

Cause it's not a low end board I have to guess.

My Maximus Hero X would push 1.45v with XMP enabled. Which is crazy high on ddr4 (example for what the board does on auto).

Any how, use of less when possible. 1.20v for xmp at 5600mhz is fine.

But why use 1.35v???
Cause the board defaults to that???
Wasn't a good excuse 10 years ago, still isn't one today.

But I'm sure there's some Overclocking guide that can give us some pointers on the matter....

Here, give this thread a read.

CPU VDDQ / CPU VDDQ- TX Voltage @ 1.490v or higher...

I looking for users here that are having to use CPU-VDDQ /CPU DDQ-TX (sets the same on MSI boards) 1.49v or higher to achieve ram stability. Im having very conflicting replies with wether it is safe to use this much voltage for this setting or if its okay and not worry about it... im very...

www.overclock.net

What does HWInfo say? I get its non-adjustable. Maybe it has a lower cache ratio too? .90v is unfathomable. But those are locked CPUs, so maybe Intel is doing some magic...or straight up ignoring the limit and the BIOS is actually applying 1.35v.

OK homeslice.

Here's that requested screen shot.

 

[ir_cow]

VID is what the CPU is requesting, not the actual voltage (which can be found further down). Anyways, I'll accept what you have. Must learn the ways of your System Agent mastery.

If you read through that thread, everyone has a different answer. Only thing I found so far is 1.4+ SA will degrade the IMC, even though VDD2 is the actual IMC voltage (per Intel data sheet).

 

[Deleted member 185158]

VID is what the CPU is requesting, not the actual voltage (which can be found further down). Anyways, I'll accept what you have. Must learn the ways of your System Agent mastery.

If you read through that thread, everyone has a different answer. Only thing I found so far is 1.4+ SA will degrade the IMC, even though VDD2 is the actual IMC voltage (per Intel data sheet).

I have that set manually (SA *VID) manually set in bios to 1.1050v. This is locked on the cpu side, not the board side.

So it's requested and stuck at .92v and the memory frequency is 300mhz higher than the claims 1.20v is needed for 5600mhz and higher.

It would be seemingly an issue not of system agent voltage, nor the request because I cannot force it higher.

Not sure how HWinfo64 reads " memory controller, *IMC" voltage which is exactly how it's described in my bios.

Which at some point I can take a picture of the bios, these settings.

But I'm not arguing which voltage is described as which, but more so to clarify the need to run 1.35v System Agent "because the bios defaults it at XMP" which really isn't accurate statements.

Most people's stability is probably just low cpu v-core, not actually their memory settings.

But no. SA doesn't need to be 1.35v or 1.25v to run these mid ranged frequencies. At least so far, that's the experience I have.

 

[ir_cow]

Lots of variables that are worth looking into to.

• Do primary timings affect the SA voltage
• Do secondary timings affect SA voltage
• Does Ring Ratio (Cache) affect SA voltage
• 2 DIMMS vs 4 DIMMS
• Command Rate 1T vs 2T
• DDR4 Gear 1 vs 2

I can't get DDR5 1T to work with 13th Gen, but 12th needed a good voltage boost for sure. I'll need to check my notes.

Anyways, I agree 1.3-.135V is unnecessary for a majority of memory setups. I just don't agree that 1.1V works above 5600 (if that). I see your screenshots. I see it works for you. Instead of asking you to do mem stability tests and y-cruncher. I'll do those myself and find out because I want to know for certain. Just going to be a few weeks. Much to busy with current workload.

 

[Deleted member 185158]

Lots of variables that are worth looking into to.

• Do primary timings affect the SA voltage
• Do secondary timings affect SA voltage
• Does Ring Ratio (Cache) affect SA voltage
• 2 DIMMS vs 4 DIMMS
• Command Rate 1T vs 2T
• DDR4 Gear 1 vs 2

I can't get DDR5 1T to work with 13th Gen, but 12th needed a good voltage boost for sure. I'll need to check my notes.

Anyways, I agree 1.3-.135V is unnecessary for a majority of memory setups. I just don't agree that 1.1V works above 5600 (if that). I see your screenshots. I see it works for you. Instead of asking you to do mem stability tests and y-cruncher. I'll do those myself and find out because I want to know for certain. Just going to be a few weeks. Much to busy with current workload.

I don't know about DDR4 because I don't have a 1700 LGA ddr4 board to test this theory. This could be a difference for sure.

Also find it amazing that I've been able to run such fine memory clocks with such a low System Agent voltage myself.

My down fall is having a 4 slot board. So I'm restricted at 6000mhz as a result. Have a B660-I however that will run the memory 6400mhz CL40's but has only 2 slots... I don't use the board because it lacks an external clock generator to accomplish an overclock which would require additional v-core.

Also, along with this testing is running on a 600 chipset (vs) or a 700 chipset, will this make a difference (ddr5). I don't think this would make anny difference with my processor because the SA voltage is locked at the CPU no by the board. My B660-G does have the option to increase as I've mentioned, but seems to do nothing at all, it's always pinned .92v.

The last screen shot was the timing set I was trying for efficiency running Y-Cruncher. My time so far accomplished with a 12400F is 111s and some change. That is a 2.5b run, not 1b. I do not have more than 16gb of memory to test with at this time, but I can pick up some more memory at some point as well.

The kit I do have is actually Samsung die. I don't know if this makes any difference, but it may very well.

I'd be happy to collaborate with you on these testing efforts also. Was hoping to nab a 12100F for some benching in the very near future, but I could also nab a 12600K (13600K) as well. Just let me know where I can help, and I'll do my best!