Overclocking 3770k on Msi z77 Mpower

[ccaution]

4.3? Ouch!!!

I've seen some really crappy ones, but they could allways do 4.6. Of course it's rather useless if it does 4.6GHz @ let's say 1.46v. I mean, there is no way to keep it cool up there, you need a chiller or something along those lines.

On the other hand there are some real diamonds that can do 32M @ 5GHz @ 1.25v (rock stable @ < 1.33v) and have the really good IMC's (2900MHz+). It's silicon lottery, really...

 

[andrevautour]

To check whether you chip has what it takes for those overclocks, disable all the c1e,eist,etc., enable the internal PLL overvoltage, give the PCH a nice 1.1v and choose a conservative LLC option, in a way for the vcore not to "jump" on load.

Then work your way up, with the multys and the vcore. I have never seen a 3770k that couldn't do 4.6... it might just be one of those c1e "semi-idle" issues, that lots of asus motherboards have, when you're pushing too far with energy saving features enabled

not sure if c1e is called something different one msi boards but i have tried all the stuff listed in master & puppets guide in the fixed overclock section, disable speedstep, turbo, enhanced turbo, pll overvolt is on by default, the v droop compensation i had at 100%

i will trying setting the pch to 1.1v and im not sure about the c1e though


also, have you ever heard of 2133mhz ram not running stably on z77? i have patriot viper xtreme which is xmp 2133mhz and it wont wake from sleep at 2133mhz and gives the same cinebench and prime errors as when the cpu is set above 4.5ghz.

when i set the xmp profile to 1866mhz it works fine

i was very surprised when i noticed the above issue because i though 2133mhz was never a problem with z77

 

[ccaution]

If you experience mem issues only @ 2133, it's a tight "2nd" or "3rd" timming, or low VCCIO voltage, most of the times. Loose the command rate to 2T too

Posting a ss of the cpuz@spd and the memtweakit utility, might also help to pinpoint any "too tight" settings, the bios gives when left at auto.

 

[andrevautour]

If you experience mem issues only @ 2133, it's a tight "2nd" or "3rd" timming, or low VCCIO voltage, most of the times. Loose the command rate to 2T too

Posting a ss of the cpuz@spd and the memtweakit utility, might also help to pinpoint any "too tight" settings, the bios gives when left at auto.

thank you so much for the info! i really don't know much about memory settings at all.

here is the spd screenshot from cpuz. i have just been using the xmp profile. xmp 1866 works fine but xmp 2133 won't wake from sleep and gives errors

i cant restart and fiddle with it now because i am uploading something, but i will as soon as i'm finished.

how do these settings in the xmp 2133 look to you? do you see any obviously issues with it?

i will try setting the command rate to 2T and see about the second and third timings also. i've never even heard of VCCIO voltage to be honest but i will have a look at that also.

one thing i did wonder about this stuff is why the "actual" default speed is so low at 1333mhz, but i recently saw tom's review of the 2400mhz dominator platinum and it was the same the JEDEC speed was 1333mhz and there was an XMP of 2400mhz on it so i guess thats normal.

i have also heard someone one here say for z77 don't get any memory above 1.5v? have you ever heard of that or why that would be the case?

 

[ccaution]

The default 1333MHz is a fallback "profile" that ALL jedec compliant kits ought to have. Nothing to worry there.

The 2133 spd profile seems ok, those Patriots should have Hynix BFR/CFR ICs that can do those timings rather easily.

Forget the 2T command rate comment, since they "prefer" 3T. Stick with 3T for the moment. The VCCIO is your IMC* voltage. If the board sets it too low by default (e.g. <1v) you might want to go @ 1.05v or 1.1v to achieve stability. The other obvious voltage that needs to be right is the DRAM Voltage (vDIMM) @ 1.65 as the SPD wants. Once you sort the voltages out, we will see those timings in detail

edit: *integrated memory controller

 

[andrevautour]

The default 1333MHz is a fallback "profile" that ALL jedec compliant kits ought to have. Nothing to worry there.

The 2133 spd profile seems ok, those Patriots should have Hynix BFR/CFR ICs that can do those timings rather easily.

Forget the 2T command rate comment, since the "prefer" 3T. Stick with 3T at the moment. The VCCIO is your IMC* voltage. If the board sets it too low by default (e.g. <1v) you might want to go @ 1.05v or 1.1v to achieve stability. The other obvious voltage that needs to be right is the DRAM Voltage (vDIMM) @ 1.65 as the SPD wants. Once you sort the voltages out, we will see those timings in detail

edit: *integrated memory controller

i figured those timings look fairly standard, i think that are the same as the g skill ripjaws stuff my dad uses.

thanks very much for the heads up about imc voltage and also changing the pch voltage for the CPU as well. i will try both of those things as soon as i am finished my upload.

it's a bit strange that the board can't set the voltage correctly for a speed like 2133 which is very common now.

speaking of the gskill stuff we have a spare 16gb kit of the xmp 2133 stuff here and i tried it in this board and the same issue. the JEDEC for that kit is 1600 and it worked fine there, but as soon as i set it to xmp 2133 it wouldn't wake from sleep. i never tried 1866 on that stuff since there wasn't a built in profile for it but i suspect it would have worked.

 

[ccaution]

The board sets VCCIO to a standard BIOS pre-defined voltage, when left @ auto. But some processors, having weaker IMCs might start failling @ 2133 or 2400 without this voltage bumped.

Why do manufacturers prefer those modest VCCIO values for their defaults, in the first place? Well I guess it's something to do with the energy efficiency rating for their boards or something along these lines...

Have fun troubleshooting, mate

 

[andrevautour]

The board sets VCCIO to a standard BIOS pre-defined voltage, when left @ auto. But some processors, having weaker IMCs might start failling @ 2133 or 2400 without this voltage bumped.

Why do manufacturers prefer those modest VCCIO values for their defaults, in the first place? Well I guess it's something to do with the energy efficiency rating for their boards or something along these lines...

Have fun troubleshooting, mate

Yes, I figured it might have something to do with my processor and how it's not all that great.

I almost got the same patriot xtreme stuff in 2400mhz i wonder if that would have been even harder to get running right or if the same IMC voltage bump would do the trick for either.

Anyways thanks so much for the help!

 

[Deejeta]

Why would you just plug in someone else's oc bios values in the first place and expect it too work?

 

[andrevautour]

The board sets VCCIO to a standard BIOS pre-defined voltage, when left @ auto. But some processors, having weaker IMCs might start failling @ 2133 or 2400 without this voltage bumped.

Why do manufacturers prefer those modest VCCIO values for their defaults, in the first place? Well I guess it's something to do with the energy efficiency rating for their boards or something along these lines...

Have fun troubleshooting, mate

I was not able to locate any VCCIO voltage settings in my bios? Not sure if they are renamed something else for MSI bios. So, no luck with getting the ram running any faster so far.

I also have another try at getting the CPU to run faster, but I honestly think that the chip is hitting it's limit at the 4.5 mark. The reason I say this is because it can do 4.4ghz at 1.25V no problem and the temperatures are good (i bet it can do 4.4 at even lower volts but I haven't tried that yet) but in order to complete a run of cinebench at 4.5ghz it needs no less than 1.34V, even at 1.33V it would fail and the temperatures at very high at that point. So it seems to me like for this chip 4.4ghz or so is sort of that thermal wall Tom was talking about in his review because it's lovely at 4.4ghz and then the temps just shoot up very high at 4.5ghz

So no luck with CPU clocks either unfortunately , but I did take some screenshots of my bios.

Sidenote: Another thing I noticed is that setting the PCH voltage to 1.1V seemed to dramatically decrease the cinebench score at 4.4ghz and 4.5ghz vs when i had the chip at those speeds with PCH voltage set to auto.


I'm pretty much giving up on getting it past 4.4ghz at this point (i dont need to anyway) but i would still like to get the memory running at 2133mhz if that is possible.

 

[ccaution]

VCCIO is CPU I/O Voltage, in your case. Go for 1.1v and see if that helps.

And yes, it's a rather big jump in voltage for those extra 100ΜΗz. It doesn't worth it imho.

 

[andrevautour]

VCCIO is CPU I/O Voltage, in your case. Go for 1.1v and see if that helps.

And yes, it's a rather big jump in voltage for those extra 100ΜΗz. It doesn't worth it imho.

Alright thanks! I will give that a try and report back. I agree it seems like it actually is the chip that is the limitation.

Thanks for your help again!

 

[andrevautour]

Unfortunately no luck with setting the CPU I/O to 1.1V, still won't wake from sleep at 2133mhz.

I completely understand if you don't want to deal with this anymore, I'm pretty close to leaving it myself, but I did take screenshot of the memory settings in the bios while I was there.

This is not what I was using trying to get it to work, I had it on auto and I was using the XMP profile, but this is what it looked like when i took it off auto and put it onto linked.

 

[ccaution]

This ss is from the 1333 profile (see up & right). I want to see ta 2133 settings

 

[andrevautour]

alright will do

 

[ccaution]

Sweet. I'm about to leave the premises so I'll leave you some rather loose 2133 settings for you to try. The setting are ordered in a way to be put one after another, in your BIOS.

3
11
12
11
32
280
12
8
8
8
42
10
8
36

the advanced 1-3-3-3-3 seems ok. With those loose timings and 1.1v @ IMC and 1.675 @ DRAM you MUST be able to be rock stable through prime/memtest/etc. If the problem insists, it's down to two things: either the patriots fails to run up to specs (chech on another mobo -> RMA) or the MSI needs to get a newer BIOS with more timing options, for us to be able to loose that wrongly tighten timing that bothers your rig :/

But we are talking genuine mem trouble here, things that a memtest and a prime would pick up, right? Waking up from sleep might be another (acpi related) issue that a BIOS update will solve.

Try all the above, and I'll chech the thread again, to see if I can be of any help.

Cheers

 

[andrevautour]

Sweet. I'm about to leave the premises so I'll leave you some rather loose 2133 settings for you to try. The setting are ordered in a way to be put one after another, in your BIOS.

3
11
12
11
32
280
12
8
8
8
42
10
8
36

the advanced 1-3-3-3-3 seems ok. With those loose timings and 1.1v @ IMC and 1.675 @ DRAM you MUST be able to be rock stable through prime/memtest/etc. If the problem insists, it's down to two things: either the patriots fails to run up to specs (chech on another mobo -> RMA) or the MSI needs to get a newer BIOS with more timing options, for us to be able to loose that wrongly tighten timing that bothers your rig :/

But we are talking genuine mem trouble here, things that a memtest and a prime would pick up, right? Waking up from sleep might be another (acpi related) issue that a BIOS update will solve.

Try all the above, and I'll chech the thread again, to see if I can be of any help.

Cheers

Thank you very much!

I should add that the ram does "work" when it's running at 2133MHz but it will never wake from sleep, which is a problem for me because I always put it to sleep every night or some nights anyway.

I had a feeling it might be something to do with how the motherboard wakes from sleep and less with the actual memory itself.

i also had a feeling maybe its something to do with the cpu since it doesnt get very good clock speeds itself but i think it should still be able to do 2133 no problem.

i haven't check this memory with another computer but i have checked the gskill stuff on my dad's rampage iv extreme and it works on there fine but has the same issue on this board, so you might be right it might be the board


unfortunately i'm not a very patient person so RMA ing is not something i would want to do really but we will see


with the gskill ripjawsZ i ran prime 95 testing most of the memory for 16+ hours and it was fine but as soon as i put it to sleep it wouldn't wake up

 

[ccaution]

Ohh I see.

If you can pass memtest86+/prime small/32M then it's not actually memory issue. It migh have something to do with those S1/S3 power options, that determine the way the computer "sleeps" and "wakes". I doubt that messing with timings and voltages will solve that kind of problems.

It's BIOS/chipset (acpi) drivers/OS/crashing drivers that causes those troubles, most of the times...

 

[andrevautour]

Ohh I see.

If you can pass memtest86+/prime small/32M then it's not actually memory issue. It migh have something to do with those S1/S3 power options, that determine the way the computer "sleeps" and "wakes". I doubt that messing with timings and voltages will solve that kind of problems.

It's BIOS/chipset (acpi) drivers/OS/crashing drivers that causes those troubles, most of the times...

i know for sure that the gskill ripjaws stuff would run prime no problem at 2133 and i thought that the patriot stuff would to at one point but it seems to be giving errors in prime now when i try to run prime, so not sure what's going one

either way i'm sick of this and im just going to leave it at 1866mhz for now because i can't really tell the difference anyway certainly not in games at least so its not the end of the world.

 

[andrevautour]

I've set the memory to 1866 MHz and left it there and I found that the CPU can do 4.4 GHz (turbo) at 1.21V/1.22V and the temperatures fairly good there, so that is where I have left it.

I'm running prime now and the temperatures are about 75-76'C across the cores so I'm happy with that.

Thanks very much for all you're help I know we didn't really figure out what the issue is but I'm properly sick of this right now and I'm ready to get back to normal use for now.