Sign in with
Sign up | Sign in
Your question

Concerned about voltages (core i5/GA-P55-UD3R)

Last response: in Overclocking
Share
December 2, 2009 7:57:33 AM

This is my first time overclocking and I just want to post about my results and find out if the voltages I've used are appropriate. I may be interested in pushing my new i5 further in the future, but right now my goal is simply to establish a BIOS profile that maximizes the performance of my system without sacrificing stability or lifespan in any serious way.

Relevant components:
CPU: core i5-750
Memory: 4GB of F3-15000CL9D-4GBRH (rated for 1866 at 9-9-9-24)
Mobo: GA-P55-UD3R
Heatsink: Cooler Master 212 plus

I first tried getting it to 3.2GHz, no problems. Then I tried cranking it up to 3.8 GHz - problems. So I raised the core voltage from 1.19v to around 1.25v (that's what i had read was the normal voltage anyways - what gives?). Now windows ran fine but prime95 quickly had errors. So then I tried tweaking the voltage some more and found that it ran stable for a few minutes at 1.30v. When I tried backing it down, it quickly had problems. So I tried running it at 1.30v for a long while: after around 45 minutes, prime95 reported a problem. (it has always said something to the effect that rounding produced an unexpected value, and it has always been on the third core, #2 of 0-3)

My reading has suggested that the voltage should only be increased a little past 1.25 (maybe to 1.27) for 3.8GHz, so what am I doing wrong? I don't want to keep raising and raising the voltages without advice and damage my CPU or socket (btw, I do have a foxconn socket, so I may want to be extra careful).

Some other relevant details:
- CPU multiplier at 20x (turbo turned off)
- base-clock at 190MHz
- memory turned down to 8x (so I don't confuse memory problems with cpu problems, i'd like to turn this up to 10x and see if I can get my memory just a hair over spec - thoughts?)
- memory timings at defaults (9-9-9-24)
- core voltage currently at 1.30v (prime95 error after 45 min)
- QPI voltage raised from 1.1v to 1.17v for the hell of it (what should I be doing with this?)
- memory voltage raised from 1.5v to 1.66v (my memory was running fine at 1.5 despite claiming to need 1.6, but i figured the higher base-clock may require higher voltage.)
- I'm getting a temperature range between 30C and 59C (idle, vs max-load) at 3.8, is this safe? (was seeing 25C - 49C at stock speed)

Please advise! My knowledge is a rough assembly of many vague bits collected from all over, I await your better judgment ;) . I really don't want anything unstable or hazardous for my brand-new components! Just looking to learn about overclocking and max out my performance safely...

Thanks in advance.
December 2, 2009 8:15:37 AM

Also, CPU-Z seems to be reporting my core voltage as 1.264! But my motherboard has it set to 1.300!
December 2, 2009 1:22:17 PM

You should not raise vCore over 10% of stock voltage for longevity of the processor.
Related resources
December 2, 2009 5:23:28 PM

1.375v is the max no-risk voltage, so crank up them voltage! Maximum max temp at load should be around 75c
December 2, 2009 6:50:26 PM

mizzl said:
1.375v is the max no-risk voltage, so crank up them voltage! Maximum max temp at load should be around 75c


Ok then, I'll crank it up more and report back - when I get home. I'm still curious though: why would CPU-Z be reporting a different vcore than my bios? which do I trust?

Also, should I just increase the vcore? Or should I change the QPI voltage too?

Thanks!
December 2, 2009 6:52:32 PM

I think pumping vCore is good enough.

CPU-z is displaying another voltage because it IS at another voltage. This has to do with a little thing called VDroop. Google it ;) 
December 2, 2009 8:05:00 PM

mizzl said:
I think pumping vCore is good enough.

CPU-z is displaying another voltage because it IS at another voltage. This has to do with a little thing called VDroop. Google it ;) 


Thanks for the answer. I've done my due diligence I think and i'm still a bit confused. vdroop is where the motherboard supplies less actual voltage to the cpu than it says it will... for safety reasons. Correct?

Does this mean I should follow the vcore given by CPU-Z for overclocking? Or should I keep both in a safe range?

Thanks again.
December 3, 2009 1:15:37 AM

Okay. I've raised the vcore to 1.325 and ran all 3 prime95 tests for 30 min each - seems fine sofar. I raised my memory multiplier to 10x (10x190) and ran memtest86+ for a few passes. Now i'm running prime95 again and all seems well. I do plan to run these tests for extended periods of time once I've settled on final settings though.

I noticed there is a setting for load line calibration in my bios. It says I can disable it to use intel's specs or enable it to reduce vdroop: should I enable it? If so, can I then reduce my vcore voltage in the bios?

Also, I would really appreciate a brief *practical* explanation of vdroop - my CPU-Z is now at 1.232 (even through i raised it to 1.325 in the bios) and i'm pretty confused by what I've read.
December 3, 2009 11:47:43 AM

Don't reduce vdroop (load line calibration). It can affect your processor's life.

December 4, 2009 2:16:52 PM

evongugg said:
Don't reduce vdroop (load line calibration). It can affect your processor's life.


Okay - I suppose I will leave that alone and not worry about the vdroop. I have it running at 3.8GHz (190x20, and the memory at 1900MHz - above spec) now and it seems fine (with vcore at 1.325v). It's prime95 stable for >12 hours straight with a max temp of 57-61C. I'll do more prime95 and memtest86+ when I have the time.

Thanks for the advice everyone.
a b K Overclocking
December 17, 2009 5:54:17 AM

quintusfabius said:
It's prime95 stable for >12 hours straight with a max temp of 57-61C. I'll do more prime95 and memtest86+ when I have the time.


You really should! I just failed prime95 Large FFTs after 23hours-11mins.
December 17, 2009 6:06:38 AM

andy5174 said:
You really should! I just failed prime95 Large FFTs after 23hours-11mins.


Thanks for the advice. I'm surprised to hear that it failed only after 23 hours!

Two questions then:
- How long should prime95 be run?
- and, does this need to be in one continuous span of time? I ran mine for around 12 hours, and then for around 10 hours. But these were two separate instances. Is running it for 12 hours, 4 times the same as running it for 48 hours and not finding any problems?
a b K Overclocking
December 17, 2009 6:20:00 AM

quintusfabius said:
Thanks for the advice. I'm surprised to hear that it failed only after 23 hours!

Two questions then:
- How long should prime95 be run?
- and, does this need to be in one continuous span of time? I ran mine for around 12 hours, and then for around 10 hours. But these were two separate instances. Is running it for 12 hours, 4 times the same as running it for 48 hours and not finding any problems?


Personally, I will only consider a system as stable after running 24hours CONTINUOUS LARGE FFTs test!
a b K Overclocking
December 17, 2009 6:27:14 AM

BTW, core voltage will be very stable with LLC turned OFF. Just download OCCT and run OCCT test for ten minutes with LLC on and off and then observe their Vcore variation graphs if you don't believe it!
December 17, 2009 6:50:21 AM

andy5174 said:
Personally, I will only consider a system as stable after running 24hours CONTINUOUS LARGE FFTs test!


What is your reasoning behind 24 continuous hours being more stressful than two sets of 12 hours?

Also, why 24 hours? If yours failed after 23:11 hours, couldn't it just as easily have failed after 24:11 hours? And if you stopped at 24 hours you'd never know? Doesn't your experience seem to indicate that 24hours might not be enough?
a b K Overclocking
December 17, 2009 7:03:29 AM

quintusfabius said:
What is your reasoning behind 24 continuous hours being more stressful than two sets of 12 hours?

Also, why 24 hours? If yours failed after 23:11 hours, couldn't it just as easily have failed after 24:11 hours? And if you stopped at 24 hours you'd never know? Doesn't your experience seem to indicate that 24hours might not be enough?


1. It's just like a car or any other electronic devices. A 30-year old car that are fine for 1-hour daily drive to work for a month can easily stalled during a 24-hour travel!

2. Why 24 hours? Since I sometimes use my computer for more than 18 hours, I want to be sure that it will be very stable within 18 hours and so a 24-hour task given to it.
One month test would be required for professional users who might use their PC for some calculations that can take more than 3 weeks.
December 17, 2009 7:06:49 AM

andy5174 said:
1. It's just like a car or any other electronic devices. A 30-year old car that are fine for 1-hour daily drive to work for a month can easily stalled during a 24-hour travel!

2. Why 24 hours? Since I sometimes use my computer for more than 18 hours, I want to be sure that it will be very stable within 18 hours and so a 24-hour task given to it.
One month test would be required for professional users who might use their PC for some calculations that can take more than 3 weeks.


Hehehe. Well, I rarely need to run my computer at capacity for any length of time, but I do usually leave my computer on for days at a time between reboots and rarely have it off. Will 24 hours be good for my usage?
a b K Overclocking
December 17, 2009 7:10:29 AM

24-hour Large FFTs should be enough unless your computer is mostly at full load and ON for more than one day!

By the way, it means that my PC is fairly stable if it fails only after 23 hours. Hence, I will be very confident of its stability if it can stabilize for more than 24 hours after upping the voltage!
a b K Overclocking
December 17, 2009 7:20:07 AM

Although small FFTs test is well known for testing CPU's stability, I personally don't trust it at all.

I can easily get 24-hour small FFTs stable and fail Large FFTs only after 4 hours SEVERAL times.

It is definitely due to the CPU but not other parts as I can get it stable for 23hours-11minutes Large FFTs after upping my Vcore a notch!

Now another 24 hours Large FFTs with PCH(responsible for main I/O interfaces) increased from 1.05 to 1.08V. Hope that it won't fail after 23hrs-59mins-59secs.....
a b K Overclocking
December 20, 2009 9:49:12 PM

I have that same board and CPU. I ran for awhile at 3.9GHz with no problems but prime95 did fail after running at 4.0Ghz for a little while. It was not mentioned above but is there any benefit to reducing the QPI multiplier?

As for how long to run Prime95, I guess that is up to the owner of the machine and what they intend to use it for. We run CAD and need all the power we can get, but even at that we only need the speed for a little bit of time here and there, nothing for long continuous lengths of time. Who's to say that prime95 might not occasionally crash after 24 hours running at stock speeds? I did quick and dirty testing of both of my overclocked machines when I built them and then I put them into use. Now they have run for almost 12 months straight with no problems in actual use.
a b K Overclocking
December 21, 2009 1:41:09 AM

cadder said:
1. I have that same board and CPU. I ran for awhile at 3.9GHz with no problems but prime95 did fail after running at 4.0Ghz for a little while. It was not mentioned above but is there any benefit to reducing the QPI multiplier?

2. As for how long to run Prime95, I guess that is up to the owner of the machine and what they intend to use it for. We run CAD and need all the power we can get, but even at that we only need the speed for a little bit of time here and there, nothing for long continuous lengths of time. Who's to say that prime95 might not occasionally crash after 24 hours running at stock speeds? I did quick and dirty testing of both of my overclocked machines when I built them and then I put them into use. Now they have run for almost 12 months straight with no problems in actual use.


1. Yes, set the QPI multiplier to the lowest. Even the lowest QPI is more than sufficient, so most people do this. BTW, also true for 920.

2. Passing a 12-hour test ONCE can easily fail within 5 hours next time in the SAME test! For example, my system is stable for more than 15hours with 1.20625V CPU-Z 100% load Vcore first time, but then it failed in ALL the 4 successive tests within 5 hours. Now I get Large FFTs stable for more than 23 hours in 2 SUCCESSIVE test, so I am immensely confident that it is 99.9999999999% stable which is good enough to me.

My Large FFTs test with 1.216V CPU-Z 100% load Vcore: (p.s. Even 1.201V should be enough, but the 1.216V is the closest to 1.20V with Level 2 LLC on my board)
Quote:
1st: Failed after 23hrs-11mins
2nd(up the PCH to 1.08V): Stable for 25hrs-1mins and stopped here


a b K Overclocking
December 21, 2009 1:55:06 AM

Hope that the upcoming BIOS version if any will make it down to at least 1.208V.
December 22, 2009 6:17:12 AM

didn't know about a new bios version.
!