|
System specs:
MSI P67A-G43 (B3)
i5 2500k LGA1155 3.3GHZ Sandy Bridge 6MB
MSI GeForce N460GTX Hawk Talon
Mushkin Enhanced Silverline Stiletto 8GB 2X4GB PC3-10666 DDR3-1333 9-9-9-24
Seagate Barracuda 1TB 7200
ANTEC|650W EA-650 GREEN RT
Never got around to overclocking this thing.
Just left all the settings stock, bumped the cpu ratio to 40, and the voltage got a little high, 1.396V, where the stock voltage is 1.18V @ 3.3ghz.
EIST and c1e are enabled(I did turn on c1e to try and fix my idle issue described below). Did not see the other C values.
I am using CoreTemp and cpu-z for monitoring, and I noticed that my CPU doesn't idle down at all, it just stays at 3.3 or 4.0, whatever I set it at.
Power settings are set to Balanced and always have been AFAIK.
Also just noticed that neither Furmark or Prime 95 cause my cpu speed to go up. Just stays at 3.3. My cpu used to boost to 3.7ghz.... I am confused.
So these are my questions,
Should turbo boost be on for a 4.2 - 4.5 ghz overclock?
How do I overclock more than 4.0ghz without frying my cpu with too much voltage?
Lastly, how do I get it to turboboost and idle normally?
Thanks a million to this forum and the reg's, helped me build my pc that has been problem free for 2 years.
|
i think you put too much voltage . try lower and run stress test until you have 100% stability for 12 hours on the lowest voltage you can get
edit : mine i5 2500k 4,3ghz - 1,25 v
|
The CPU can both use the turbo multipliers or the normal multiplier to overclock. What you can or have to use depends on your board and its BIOS.
You have to control voltage manually. You can't leave it on Auto. It will do funny stuff by itself! From what I've heard of MSI boards, you will have to use a fixed VCore voltage.
I guess you'd start by first putting it to stock settings, then looking at the VCore voltage you see reported when starting a very stressful program like prime95 (or "IntelBurnTest" or "LinX"). You then go and use that as fixed VCore in the BIOS. It should still run prime95 (etc.) fine after switching over from Auto to fixed VCore.
You also have to look into "LLC" = "Load-Line Calibration". In Intel's specification, the voltage is required to drop if load increases. That's designed to work well with shitty boards. It gives the board more room to maneuver when the CPU's power use changes. In an overclocking board, you can tweak that and remove that drop in voltage. You will want to set LLC to whatever keeps the fixed VCore at a constant voltage when you put load on the CPU by starting prime95.
In practice it looks like this:
You go into the BIOS. You disable all power saving stuff like C1E, C3/C6, ... You punch in a VCore voltage of 1.25V.
You go and boot into Windows. Take a look at CPU-Z (I recommend the program "HWINFO"). Don't touch anything and let the PC idle. It will show something close to those 1.25V as VCore. Now go and start prime95. VCore will drop. It will show 1.22V for example, not 1.25V.
You go back into the BIOS and play around with the LLC option. Then go try again. One of the LLC settings will make it so VCore will stay at 1.25V while running prime95.
After you are done learning how your board and its BIOS works, you go and find the highest multiplier that will run at 1.25V VCore without the PC crashing and prime95 not reporting errors. If temperatures are still looking fine, you go and increase VCore voltage gradually to get higher multipliers to run stable.
At this point you might want to go back into the BIOS and enable the power saving stuff, C1E and C3/C6 and EIST. This will enable Windows to clock down cores and disable them when not needed. With fixed VCore voltage, you can't have voltage drop. I don't know if you need the Windows power profile set to "balanced". It still clocks down the CPU on my PC using a 3570k CPU as "high performance" profile.
Make a note whenever there's a crash or error, what the settings were that lead to that. That documentation will help to see how your particular CPU behaves. You will see a system in how VCore and multiplier relate.
There's guides for Sandy Bridge on the www.overclock.net forums, perhaps also something specific for MSI boards. Ivy Bridge you can overclock solely by tweaking VCore (and LLC). Your Sandy Bridge might be more complicated, so better look at some guide.
EDIT:
Google found this: http://www.overclock.net/t/963798/msi-p67a-gd55-review-and-oc-guide
It seems "LLC" is called "VDroop Control" on your MSI board. You will want low or no vdroop.
|
Yeah 1.4V for 4GHz is a bit silly, that's more for ~5ghz. Higher voltage causes higher temps and degrades your cpu. 1.2-1.3V should get you to about 4.5 GHz with lower temps.
|
Thanks alot for the tips and help guys.
So reset my bios to default settings and it fixed the idle issue, now during idle my cpu does clock down to 1.6. However the turbo boost is not kicking in, even when running prime95 and other programs at once.
Going to update my bios and see if that helps.
|
United Kingdom20300 Posts
Turbo boost set to auto with cpu ratio/multiplier on default should make it boost, but if you're manually setting a multi like 40x @1.1vcore then it'd just sit there
|
Right now there is no increase in the cpu multiplier. Everything is stock settings, 3.3ghz, turbo enabled. I wanted to get everything working before going further with the OC.
It seems like once I mess with the cpu ratio I lose the idle and turbo functions. I used the OC genie once a long time ago, but at some point unknown it stopped working, and I lost the idle and turbo.
Also I did not update my bios because as I understand it that requires a bios flash, and that makes me nervous.
Is there any danger in resetting the bios mutliple times?
|
United Kingdom20300 Posts
No, and if you reset to optimized defaults then you should have turbo working
|
On September 09 2013 09:24 89vision wrote:
Right now there is no increase in the cpu multiplier. Everything is stock settings, 3.3ghz, turbo enabled. I wanted to get everything working before going further with the OC.
It seems like once I mess with the cpu ratio I lose the idle and turbo functions. I used the OC genie once a long time ago, but at some point unknown it stopped working, and I lost the idle and turbo.
Also I did not update my bios because as I understand it that requires a bios flash, and that makes me nervous.
Is there any danger in resetting the bios mutliple times?
Are you getting 3.3GHz with prime 95 on 4 threads? That's potentially normal on a 2500k at stock, default turbo for 4 threads is 3.4GHz, and that is only supposed to be for short periods of time, not extensive stress tests  . Try doing prime95 with one thread perhaps, and see what happens.
Resetting your bios settings is never a problem. Flashing your bios has some chance of failing and bricking your motherboard (extremely unlikely), or if the power cuts out it's probably game over (so don't do it during a storm).
|
Running one thread did cause the ratio to jump up to 3.5ghz.
So it is working now. Next step is to increase the ratio, and I think I am going to need to do it the long way, starting with the volts set low and with low vdroop.
|
Low v droop still allows the voltage to vary. I can not find LLC options. I have speedstep, overspeed protection, c1e, and intel c.state which i should disable also right? Was leaving overspeed protection on.
|
That low vdroop and LLC should be the same setting. It's just a different name on different boards.
What that voltage you see varying is, that depends on what you are talking about exactly. If you use fixed voltage in the BIOS, you should see about what you've set when the PC is idle and you are just staring at an empty desktop. If voltage is then going down somewhat after you start a stress test program, that's vdroop.
If you see very low voltage when idle, but very high voltage when you start a stress test program, that's your voltage setting being on the default setting or you are using offset overclocking. The default is the CPU messaging a "VID" value and the board supplying that. This value changes all the time. Offset is something being added to get this behavior working stable when overclocking.
There's also some sort of automatic voltage setting on boards. This is dangerous as it usually slaps a lot of extra voltage onto the default voltage when overclocking. You should disable any automatic behavior about the VCore voltage if you are experimenting with speeds above the normal turbo speeds of your CPU.
I have no idea what your MSI board can do. For their Z77 boards, they could only do fixed voltage. It might have been the reversed situation with the previous generation which you have, only offset overclocking and no fixed.
I suggest you get yourself the program "HWINFO". It can show a lot of stuff in its sensor window, including that mysterious VID value, instead of only the VCore voltage end result like the usual programs.
I think I've heard about 1.35V or max 1.4V VCore and at all times staying below 85C on core temperatures being suggested as safe guidelines for Sandy Bridge.
|
while i have very little experience with MSI autovoltage efficiency, I do know that sandy bridges can push serious clocks with minimal effort. Generally you'll want to have 1.35V as your "max target" (don't go much higher, lower is of course ideal) as it tends to be the sweet spot. I can usually keep my 2500k at somewhere between 4.7-5.1GHz on air with voltage under 1.4 (of course i live in the south so i'll drop it to like 4.5 in the summertime)
In my opinion there are no consumer scenarios where turboboost should be off. (Leave it on)
If its on you can be pushing high 4.7-4.9s easily
If you want it off, then just set multiplier at like 42 and lock the voltage at something nice like 1.2 and run the usual stability tests
|
Need to clear something up, I was reading VID in coretemp when I was watching my initial OC's voltages. Now using CPU-Z for voltage, CoreTemp for temps(not run in tandem). So leaving everything stock I had more room than I thought before. Got to about 4.2ghz but the voltage ended up at 1.4V after 10minutes under max load with small fft p95 with 94% ram used.
So I set the vCore to 1.25 and just went for 3.9, 10min test p95, then up to 4.2ghz. Everything was fine so I bumped the vCore to 1.275, and got up to 4.6ghz no problem with temps under 70C. Didn't p95 this for long, just about 15min. Kinda got nervous because the OC was so easy and 4.6ghz is getting up there. So I went back to 4.5ghz and did 1344 and 1792 p95 tests for 20min each, no errors. Then blend test with 94% RAM for 8hours 49min no errors. Temps under 70C.
Can I call this a stable clock at this point? I may run another p95 for 12 hours tonight.
Now I should lower the voltage on my 4.5ghz OC and see how low I can get it stable?
Q:-------Still confused on how to set the offset to make the turbo work while OC'd??
STABLE-
4500mhz, 1.275V, MaxTemp: 63-67-68-67, 9hours p95 passed
4500mhz, 1.25V, 10min small fft passed
4600mhz, 1.275V, 15min stable, MaxTemps under 70C
UNSTABLE-
4500mhz, 1.225V, BSOD with p95
I feel based on the temps of the 4.6ghz clock and the fact that my voltage is still pretty low for 4.5-4.6 that I could get around 4.8+ stable under 80C but I don't know if I want to try.
|
|
|
|
|
|
EPT
