Help with CPU overclocking on z77 OC formula

[synack]

OK, so i'm stuck here.
I know Linux is not supported, but I believe it's a BIOS issue.
It seems that the BIOS is reporting that the maximum multiplier is 45x
From the BIOS "main" screen I see it shows i3700k @ 3500MHz, maximum 4500 MHz.
No matter what I seem to select in the OC screen, I ALWAYS get at maximum 3500 (max3600) turbo 4500 Mhz in Linux.

Searching around the net I have seen others with boards from dif manufacturers that report the proper Max multiplier of 63x.

This is what I see from i7z:(linux utility to report cpu frequencies)

Socket [0] - [physical cores=4, logical cores=8, max online cores ever=4]
TURBO ENABLED on 4 Cores, Hyper Threading ON
Max Frequency without considering Turbo 3600.00 MHz (100.00 x [36])
Max TURBO Multiplier (if Enabled) with 1/2/3/4 Cores is 45x/45x/45x/45x
Real Current Frequency 4499.98 MHz [100.00 x 45.00] (Max of below)

Currently with 4500Mhz, I'm only reaching max 60c at full load. I'm pretty sure I can get at least 4700 maybe 4800 out of it safely.
But obviously I can't get it to go any further.

Is the 1.4 BIOS reporting only 45x max? Is there anyway to up that?


Thanks so much for any insight into this issue. I've been bashing my head on this for days, and by all accounts others are able to do this fine within a Linux OS.


Thanks
Syn

[parsec]

So you're saying if you enter a value for the multiplier of 47 or 48, it will change to 45? Do you have one of the predefined OC profiles loaded, or the auto OC feature enabled?

[synack]

So you're saying if you enter a value for the multiplier of 47 or 48, it will change to 45? Do you have one of the predefined OC profiles loaded, or the auto OC feature enabled?

On the "OC" page of the BIOS it says a "target": to whatever I set. The "main" page continues to say "max 4500Mhz".

In Linux it says 45x 100, even though I've set say, 47x in the BIOS.

Linux gets it's info from probing the hardware. Since I know the i3770k CPU reports as 63x as the "max" multiplier (not that i'd ever try that high lol) the only thing left is the BIOS...

So I'm wondering if there is a way to trick either Linux or fix what the BIOS reports? Is there a magic option I missed that gets me past 4500? or is it set to 4500 max as a BIOS failsafe?

Yes, i've used "Nick's" predefined profiles, as well as my own. I've reloaded defaults so many times I couldn't guess. Tried, i'm sure every option in the OC page, as well as the cpu configurations. Nothing I try will stick >4500Mhz.

4.5Ghz was easy lol select Nick's profile for 4500, and set linux to "performance" to be always in "turbo" mode... and tada...
It's getting past 4.5 that's been driving me nuts.

Last thing to try is maybe load up windows.. (yuck) to see if it's a hardware problem. But I would Prefer to have this working in linux.

Maybe someone from ASRock can tell me how they implement the OC in windows, Like what commands are sent to ACPI, msr, the BIOS or whatnot. Maybe I can come up with a patch/fix for Linux. :)

The 4500 itself is not hard coded at least, just the 45x it seems. if i change the bclk to 101Mhz, the "clock speed" will change to 4545Mhz on the main BIOS page as well as Linux, and 4747 as the OC "target" (when I try for 47x that is)

Anyway off to playing with MSRs.. I'm determined to fix this issue. lol

Syn

[synack]

OK, a little bit further... Looks like MSR 0x1ad is set to 0x2d2d2d2d
Which for my processor is 45x 45x 45x 45x for cores 1-4 even though my bios is currently set to 47x

I am able to re-set it to 0x2f2f2f2f2f (47,47,47,47) and set MSR 0x199 to 0x00002f00 (2f=47)

i7z now reports the max turbo multiplier as 47, and the target (msr 0x199) as 47, however the multiplier is still running at 45x (and confirmed with MSR 0x198 = 0x2d00)
I haven't figured out yet how to tell the processor to use the new multiplier.

Anyone here know? And this pretty much confirms to me that the asrock oc formula BIOS is reporting 45x to the OS.
Hopefully some smart ASRock tech can get me over this little hump?

Thanks :D
Syn

[parsec]

Wow, that BIOS OC setup is weird, I suppose they think that's a feature with this board. Every SB or IB CPU board I've used allows direct entry of the multiplier, either by All Cores, or Per Core. If this BIOS is different, who knows, not me...

[synack]

Oh dont get me wrong... You CAN setup the multiplier individually per core or all cores or whatever. you can tweak more than I will over need to, or feel comfortable doing, on this board. frankly I don't even know what half the stuff is for lol... I use the OC profiles for ease of use lol... I select profile 8 which goes and configures the all settings that "Nick Shih" would set it too for 4700MHz (not that HE would personally use those). It makes for a good starting point anyway.

My issue isn't telling the BIOS what I want. My issue is the BIOS NOT telling my CPU/OS what the settings should be. I assume there is some magical windows driver that probes the BIOS and sets everything up while windows is booting. But I can say for sure that the initial boot is not running OC.
It does seem to be setting some values. If I run a 4500Mhz profile the core temps are much cooler than say a 4900Mhz profile at idle. Even though it does not actually run faster. My guess is higher voltage applied to the CPU.

RAM settings take just fine, BCLK works as well. The only thing giving me grief so far is the CPU multiplier. I can't confirm voltages, as I dont know which MSR's they are for Ivy Bridge. And even if I did, it seems to be "interesting" trying to interpret them. (they seem to pack a Floating point number like 1.35 into an integer by applying some math, something like 1/V*2^16)

ASROCK?!? Anything?



So I'm looking for some way to replicate the windows behavior, or have the BIOS set things right in the first place.

[synack]

Thanks to all for the help. I solved my own issue!

Turns out there WAS a problem with the BIOS, however not what I expected.
I discovered just now tweaking my memory settings that they wouldn't stick like they use to... I was thinking.. WTF?!?!?!
the z77 oc formula comes with a VERY HANDY clear cmos on the back of the board. I just tried it out.
Clearing the cmos fixed by memory timing issue AND my multiplier issue!!!

FYI: Apparently Loading default settings is not good enough...

Syn

[profJim]

Clearing the CMOS is S.O.P. when things get too upgemucked.
That's one reason why it's good to have a saved working bios profile (or three), so that you can quickly get back to a stable or test version after the cmos is cleared.

[synack]

LOL yea. usually when things are mucked up bad. But honestly, beside changing the multiplier everything was working just fine. It wasn't until I started changing other things again that I noticed something just wasn't right... ( I try to keep to changing one thing at a time lol)
I did save a profile, however I didn't use it. it took all of 20 seconds for me to manually restore all the settings I had before. I LOVE this mobo lol

Now I'm running my i3770k (3500Mhz) at 4700 Mhz (bclk 100, 47x, 1.35V) and it's super stable.

Did some quick load testing and the CPU reports:
32-39C at Idle
68-73C at full load (thats about 5C warmer than 4500MHz)
My cpu waterblock didn't flinch a single degree (sitting at a comfortable 30c)

I'm a happy man... I wonder if I can get to 4800 stable :D

Syn

[profJim]

You won't know until your try.
BTW, it never ends....hey maybe 4.90GHz is doable.....boy oh boy, I'm real close to 5.0GHz......
ooh, ooh, ooh!!! I wonder how high I can o/c using LN2

OH POOP....did my motherboard just melt???