[Tunagoblin]

I found some weird thing about the CPU PLL when I was testing last night.

My CPU temp goes UP when I make the CPU PLL LOWER than 1.832.
It passes the test (used quick LinX w/AVX all ram) fine with 1.750.
But it gives me 3c higher in temp.
I though I was doing something wrong so I went back to 1.832 and test it and did the 1.750 again.
But the results were the same. (I tested it 3 times back and forth)
1.832 runs cooler than 1.750.
Weird but that is what I got and also I get better score at 1.832 (133+GFlops) opposed to 1.750 (131~2GFlops).
Someone should try this and see if it does to you, too...

 

[BradleyW]

Isen't 1.8 a danger to the system? I've got it at 1.7v and i am stable at 4.8GHz. 1.38Vcore.

 

[BradleyW]

Quote:

Originally Posted by t00sl0w yeah, 1v = 1000mv for reference. so, are you hitting the vcore you need at load?

Quote:

Originally Posted by donkrx Just go into Windows and see what the voltage is on load, remember that, then add or subtract to it to get it to where you want it. Using BIOS readings is very confusing because of LLC. I like LLC set to the middle option (for me I have 5 different levels and mines on LLC 3), maybe the 2nd highest even. You don't have to use LLC but in theory it could help with random bsod'ing. Also if you switch to Offset voltage be sure you change C3/C6 Report to Disabled and set P-state Limit to Auto (I personally use C2, that can also work just fine). This is the biggest factor in eliminating the random bsods and it works quite well for offset voltage users. If you use Fixed voltage its different.

Well in windows my Vcore is 1.384Vcore full load in at 4.8GHz and 65c Temps. It did hit 71c for a moment.

If i change the C3/6, will i still have reduced Vcore and speed on system idle?

 

[$ilent]

Quote:

Originally Posted by munaim1 could be a background process or driver issue. Did it pass 12hours the first time?

No first time I went with 4.8ghz and it failed after hour or so. I dont think I had any background processes running, except cpu-z and realtemp...i.e nothing special?

What drivers you reckon could be interferring then?

 

[munaim1]

Quote:

Originally Posted by BradleyW Well in windows my Vcore is 1.384Vcore full load in at 4.8GHz and 65c Temps. It did hit 71c for a moment. If i change the C3/6, will i still have reduced Vcore and speed on system idle?

The C states afaik only help with the idle / random bsod problem. Those should be on Auto if you're using manual voltage and should be disabled if you're running Offset voltage.

Quote:

Originally Posted by $ilent No first time I went with 4.8ghz and it failed after hour or so. I dont think I had any background processes running, except cpu-z and realtemp...i.e nothing special? What drivers you reckon could be interferring then?

No I meant any of those could be interferring with prime95, gpu driver is one.

I have 86 processes running in the background. Open task manager to view them. That is why I run prime and bench on a stripped OS.

From the OP

Quote:

The error codes are not 100% and are not ALWAYS correct, with that said, stress testing in your main OS is not a good idea. If possible get yourself a spare HDD and load up windows and run all your stress testing on that. The idea of having another HDD is so that when your running your stress testing, background processes are at a minimum and should help indicate the main source of bsods, disabling the internet connection is also a good idea, same with any type of antivirus. Just remember too many bsods in a OS can cause the OS to become unstable ie corrupted file systems etc. With that said, if you pass 12hrs once you should be able to pass again, however, this does not mean go OCD stress testing.

Seems that both of you guys are making good progress, Im a going to see a prime blend anytime soon?? lol

Just concentrate on getting it stable for 12hours, ie standard blend or custom blend with 90% of your available RAM.

Keep up the good work. I'll be posting my little guide in the OP.

Here's my quick little sandy guide:

Quote:

The only things will that will require multiple changes are the vccio (VTT), PLL voltage and vcore, refer to this: Set the whole thing to stock and start again. This time only change the RAM to XMP (STOCK) and run prime blend for a few mintues to see that your CPU is functioning properly. Then comes the task of determining the voltage for the multiplier, but that comes after you find the correct LLC setting for your mobo. What you want to do is set the LLC to the one that is closet to what you set it to when the cpu is under load, so for example if you set 1.35v and under load it's 1.31v and that's level 3 then you may have to increase to around level 5. The objective is to keep the voltage under load as controllable as possible without it letting it spike. These LLC settings will be different amongst mobo's. For Asus mobo's the Ultra high (75%) LLC seems to work best. Then it comes to that task of finding the actual voltage for the overclock, however before we get to that, I would advise you to reduce PLL voltage to 1.7v (Scroll down or go to sandy stable club about PLL info). Then set the vcore manually to 1.25v, Leave C1E and Speedsteep enabled and run C3 and C6 on Auto if you can, if not leave them enabled. Also leave Spread spectrum enabled, if you find that it disrupts the bclk in cpu-z then just disable it. Additional settings that you need to change from the get go, but won't need to be changed afterwards: Can be found under advanced settings/cpu configuration: Quote: For Asus Mobo's CPU Current Capability - 140% Phase and Duty Control - Extreme EPU Power saving - Disabled VRM Frequency - Manual - 350 Quote: For Asrock Mobo's Turbo Boost Power - Manual Short Duration Power Limit - 250 Long Duration Power Limit - 250 Core current Limit - 250 Quote: For Zotac Mobo's Turbo Boost Power Max - 200 Turbo Boost Short Power Max - 200 Quote: For Gigabyte Mobo's Turbo Power Limit - 200 Quote: For MSI Mobo's Short Duration Power Limit- 250 Long Duration Power Limit - 250 Overvoltage is only needed when a particular multi (usually the high ones) doesn't boot into windows. With that function, you sacrifice sleep mode. You can't have overvoltage and Sleep working together, don't know why, could be BIOS related. This should be a stepping stone to get your rig stable. With those settings you will eventually get to the point where you're stable. Set the multi to 45 and the vcore to 1.25v and increase the vcore each time after you stress test, run a quick custom prime with these FFTs (1344 & 1792) like THIS and go back and change the vcore accordingly, bump it by one not big jumps and that goes for PLL and VCCIO (VTT) and VCORE!!! Work your way up from there, increase multi, test with prime, if it fails up vcore, if not up the multi. Until you are satisfied with the temps and it is stable then continue upping the vcore to stabalise. Just a note: The custom FFT's are not that consistant, making them not all that reliable, however if it works for you, then that's great. What I mean by inconsistant, is that it may pass once with the same settings but may fail the exact same run second time round. In that instance I will recommend you to run a standard blend test to find your overclock, using intervals of 15/30mins. This duration will increase when you're nearing stability. This is a lenthy process, one that takes time and patience, make sure your up to the task Head over to the Sandy Stable Club for more info and tips Here are the additionl info regarding PLL voltage, VCCIO and VCCSA: READ THIS & THIS

 

[cssorkinman]

http://valid.canardpc.com/show_oc.php?id=1969045

I had the voltage set to 1.435 in bios but I see that it automatically added some. It's gonna be a lot of work to get it stable at that speed and I'm not altogether sure my H-60 can manage the heat load .
I have a question for munaim1, I saw your power limit settings for an MSI board in another post- 250 - if i had that set too low what behavior would it result in ? i.e. BSOD code?
Thanks for any replies

 

[donkrx]

Quote:

Originally Posted by $ilent What so your pc just wouldn't put itself to sleep even if it should after x amount of time being idle?

It will sleep, but it will certainly not come out of it properly.

Quote:

Originally Posted by Tunagoblin I found some weird thing about the CPU PLL when I was testing last night. My CPU temp goes UP when I make the CPU PLL LOWER than 1.832. It passes the test (used quick LinX w/AVX all ram) fine with 1.750. But it gives me 3c higher in temp. I though I was doing something wrong so I went back to 1.832 and test it and did the 1.750 again. But the results were the same. (I tested it 3 times back and forth) 1.832 runs cooler than 1.750. Weird but that is what I got and also I get better score at 1.832 (133+GFlops) opposed to 1.750 (131~2GFlops). Someone should try this and see if it does to you, too...

Pretty interesting stuff... I personally did not notice anything significantly different when testing 1.832 or the 1.68ish one, however I do not stress my CPU any more with AVX so its not really relevant information for you.

Are you still sure that 1.823 is more stable? Is that at 1.385v 4.8ghz too?

Quote:

Originally Posted by BradleyW Well in windows my Vcore is 1.384Vcore full load in at 4.8GHz and 65c Temps. It did hit 71c for a moment. If i change the C3/6, will i still have reduced Vcore and speed on system idle?

Disabling C3/C6 will seem to very slightly raise your vcore, maybe something on the order of 1.050 raised to 1.1v. You won't be above 1.2 for sure.

Nice temps

==============

PS the guys are working on my AC right now, hopefully they fix it and I can FINALLY get a blend test for you munaim, lol. I technically did already do it but messed up with the RealTemp window.

 

[munaim1]

Quote:

Originally Posted by cssorkinman http://valid.canardpc.com/show_oc.php?id=1969045 I had the voltage set to 1.435 in bios but I see that it automatically added some. It's gonna be a lot of work to get it stable at that speed and I'm not altogether sure my H-60 can manage the heat load . I have a question for munaim1, I saw your power limit settings for an MSI board in another post- 250 - if i had that set too low what behavior would it result in ? i.e. BSOD code? Thanks for any replies

It's the turbo power limit to stop the mobo from restricting the overclock, ie throttling.

 

[cssorkinman]

Ah, i see - thanks for the enlightenment munaim1

 

[munaim1]

Quote:

Originally Posted by cssorkinman Ah, i see - thanks for the enlightenment munaim1

No worries happy to help.

Take a look at some of the BIOS templates and you should get some sort of idea on some of the things that are really making the differences in overclocking.

 

[$ilent]

Yeah im thinking one more vcore increment should do it for me and I should be ready for another blend sesh tonight. Just thinking about my gpu driver, again its 280.26. Is there anything inherently wrong with this driver?

I remember when 270 was the one to have or 268, and anything higher meant big problems.

 

[Tunagoblin]

Quote:

Originally Posted by donkrx Pretty interesting stuff... I personally did not notice anything significantly different when testing 1.832 or the 1.68ish one, however I do not stress my CPU any more with AVX so its not really relevant information for you. Are you still sure that 1.823 is more stable? Is that at 1.385v 4.8ghz too?

LinX IBT is great for knowing the thermal threshold and also quick way to see your settings give you BSOD or not.
I do trust stability on both LinX and Prime. Not just either one. It's always the best to combine multiple stress tests.
I can make my 5.0GHz prime stable but I can't run LinX so it isn't my 24/7.
I personally want my system to be able to run both and passes fine or otherwise I don't consider my system stable.

Anyway, CPU PLL 1.832 is the default value that all AsRock (maybe for Asus, too) sets to.
Anything lower will give me less score on LinX or unstability.
Above 1.832 won't make much difference but lower will give me problems.
I've tested so many different combination with this setting with VTT/vcore.
1.832 is a must for my system.

 

[munaim1]

Quote:

Originally Posted by $ilent Yeah im thinking one more vcore increment should do it for me and I should be ready for another blend sesh tonight. Just thinking about my gpu driver, again its 280.26. Is there anything inherently wrong with this driver? I remember when 270 was the one to have or 268, and anything higher meant big problems.

[REVIEW] - Comparison of a few of the latest Nvidia drivers inc* Xtreme-G

Check that thread out, lots of reading lol

Short Summary
- Most stable nvidia drivers
- Best benching drivers are 280.19 and 275.33

depends on wether sli or single cards

For gaming anything above the 270.X were quite unreliable, many have reported problems in WOW, BFBC2 etc. The 270.X were problemativ for Folding, not sure about the updated ones and whether or not it's fixed for folders.

 

[donkrx]

Tuna - not sure if you read what I said differently but let me clarify anyway, I mean I don't run AVX so I can't tell you if I also experienced a temp increase.

Again that is interesting because I've never failed IBT or crashed during it, but I can crash very fast in Prime. Agreed that more is better than 1, to each his own anyway

. In the end if the computer runs and never crashes for your own use, that's stability.

 

[BradleyW]

I think you will find all the paperwork is in order.

I know the temp said 77c. It was a strange temp spike because i started effecting the ambients and airflow. The average temp was around 68-70c.

Edit:

munaim1, here ya go. My average temps are also 65-67-68-65.

 

[munaim1]

Sorry bud refer to rule 2:

Quote:

MUST have a screenie WHILE UNDER LOAD

Only reason is that the vcore is different under load. Sorry bud

 

[$ilent]

so your gonna make him do another 12 hour p95 blend?...Sorry I must laugh abit at that

edit: heck donkrx if we only had to go by 20 passes on IBT I would be laughing my tits off at 4.8ghz and well under 1.35v!

 

[munaim1]

Quote:

Originally Posted by $ilent so your gonna make him do another 12 hour p95 blend?...Sorry I must laugh abit at that

It's up to him, im not making him do anything.


Quote:

Just a note to everyone, I post the rules now and again to remind everyone the practices and rules of this thread, if you cannot adhere to these simple rules to enter this club and the spreadsheet then it will be your own fault.

Like I said in the disclaimer:

Quote:

PLEASE READ I take NO responsiblity for the headache / stress / anxiety / emotional discomfort etc lol that this may cause when trying to overclock and stabalize your cpu, this is a lenthy process which requires a lot of patience. If your not upto the task, then this thread may not be for you. This is not an EASY club to join!!! On that note, any damage, degradetion that you may encounter through stress testing or SUICIDE RUNS is your responsiblity and your responsiblity alone. So you have been WARNED, other than that enjoy your 100% stable system and also have fun getting it there!!!

 

[iBlendYourFace]

Hi again, just got back to my computer and I decided to go for a 4.5 GHz overclock. I got it seemingly stable with the two 'hard FFTs' at 1.352 vcore which seems a little high. It's blending 12 hrs now, hopefully it will pass!

I decided not to update the BIOS as I've heard that it can brick your mobo, and since I don't have a spare 200$ lying around I don't really want to do that. Also, I noticed that if I disabled spread spectrum in the BIOS my BCLK would rise from 99.8 to 100, which is good.

 

[BradleyW]

What if i put the CPU under load and then paste the CPU-Z onto here to update my picture?