LinX stopping me from further overclocking

[Brutality]

I can't get my 2500k to 4.5GHz because the voltages needed to make it stable will put temps above 72*C

 

Is there a way to make the fans go faster on the NH-D14 or do I just need better cooling?

 

BTW, sig is outdated

Edited June 6, 2011 by Brutality

[Nuclear]

There's a couple of tricks you can try that can help lower your vcore or possibly lower your temps. What are your secondary voltages set to? It would be great if you can list out your QPI/VTT, System Agent Voltage, PCH, PLL, and VDIMM.

 

Thanks

[Brutality]

There's a couple of tricks you can try that can help lower your vcore or possibly lower your temps. What are your secondary voltages set to? It would be great if you can list out your QPI/VTT, System Agent Voltage, PCH, PLL, and VDIMM.

 

Thanks

 

I don't know what QPI/VTT is

 

but my "VCCSA" (or System Agent according to http://www.bit-tech.net/hardware/cpus/2011/01/07/how-to-overclock-the-intel-core-i5-2500k/2) is Auto = 0.925

PCH = Auto = 1.062

PLL = Auto = 1.803 (but I've got Internal PLL Overvoltage on Auto)

VDIMM = 1.65 (according to the X.M.P. profile for my 2133MHz HyperX Genesis dimms)

[wevsspot]

I don't know what QPI/VTT is

 

In your BIOS the memory controller voltage is called VCCIO Voltage. If left on Auto you could be running some pretty high voltages through your memory controller with your DRAM speed as high as it is.

 

Open the AI Tweaker menu in your bios and report what the VCCIO Voltage reading is.

 

At DDR3 1333 I believe it defaults to around 1.050v on the Auto setting. But I'll bet your is up around 1.45v or possibly even higher.

 

As an experiment drop your DRAM speed down to the DDR3 1600 or DDR3 1333 range and tighten your memory timings as far as possible while still passing Memtest and LinX. Check VCCIO voltage at those DRAM speeds and see if it is somewhere between 1.05 and 1.20v

 

Now run LinX again and check your temps.

 

Since you note that you're possibly using the XMP profile to run your DIMMs at 2133Mhz (and if that's a true statement) that is one hell of a load on the memory controller. And once you start loading the onboard memory controller it starts kicking out the heat (fast).

 

I'm pretty sure that's where you're at right now.

 

If that proves to be the case, don't worry about running your memory at slower than rated speed. Just tighten the timings as far as you can. I've been through the road wars on DDR3 2000 and above RAM for about a year now, and in day to day activities, I don't notice any difference between running my high speed RAM at DDR3 1600 versus DDR3 2040. Sure the memory bandwidth benchmarks change, but I'm talking real life day to day computer usage.

[Nuclear]

I don't know what QPI/VTT is

 

but my "VCCSA" (or System Agent according to http://www.bit-tech.net/hardware/cpus/2011/01/07/how-to-overclock-the-intel-core-i5-2500k/2) is Auto = 0.925

PCH = Auto = 1.062

PLL = Auto = 1.803 (but I've got Internal PLL Overvoltage on Auto)

VDIMM = 1.65 (according to the X.M.P. profile for my 2133MHz HyperX Genesis dimms)

 

QPI/VTT may be listed as VCIO on some boards, if that helps.

 

Intel overvolts the PLL on the SB chips which actually can have some adverse effects.

 

- Try dropping your PLL from Auto or 1.8v down to 1.7 or 1.72v. You can run even lower if you'd like, I run 1.54v at 5.0GHz on my 2500k. Dropping the PLL will often allow you to become stable using a lower vcore and thus causing lower temps.

 

- PCH voltage is fine and shouldn't be adjusted, just like VCCSA.

 

- VCIO (QPI) shouldn't be any higher than 1.12 or 1.14v. It could be a lot lower but since you are running 2133 RAM it's putting more stress on the IMC and thus probably needs a slight bump.

 

- I would also recommend manually setting every voltage and not leaving them on auto. Even setting them to the default values is a good idea. Auto gives the board the ability to adjust on the fly and it will often overvolt and cause problems when the load gets heavy.

 

What is the vcore currently set at and what type of thermal paste? That cooler should be doing better, but it's not unusual to see 80+C temps on air cooling once these chips started getting clocked.

[Brutality]

QPI/VTT may be listed as VCIO on some boards, if that helps.

 

Intel overvolts the PLL on the SB chips which actually can have some adverse effects.

 

- Try dropping your PLL from Auto or 1.8v down to 1.7 or 1.72v. You can run even lower if you'd like, I run 1.54v at 5.0GHz on my 2500k. Dropping the PLL will often allow you to become stable using a lower vcore and thus causing lower temps.

 

- PCH voltage is fine and shouldn't be adjusted, just like VCCSA.

 

- VCIO (QPI) shouldn't be any higher than 1.12 or 1.14v. It could be a lot lower but since you are running 2133 RAM it's putting more stress on the IMC and thus probably needs a slight bump.

 

- I would also recommend manually setting every voltage and not leaving them on auto. Even setting them to the default values is a good idea. Auto gives the board the ability to adjust on the fly and it will often overvolt and cause problems when the load gets heavy.

 

What is the vcore currently set at and what type of thermal paste? That cooler should be doing better, but it's not unusual to see 80+C temps on air cooling once these chips started getting clocked.

 

The vCore was set at 1.275 when it reached 71*C at 4.5GHz with the above listed values, so I got it stable at 4.4 with a vCore of 1.260

 

The VCCIO was set to Auto = 1.065 (This is the same as VCIO?)

 

I'll try to drop PLL voltage and set default values instead of Auto. Should I disable Internal PLL Overvoltage?

 

EDIT: I used the thermal paste which came with the NH-D14

Edited June 6, 2011 by Brutality

[IVIYTH0S]

Keep in mind also that 99% of other things will never heat your processor up as much a LinX .

[Nuclear]

The vCore was set at 1.275 when it reached 71*C at 4.5GHz with the above listed values, so I got it stable at 4.4 with a vCore of 1.260

 

The VCCIO was set to Auto = 1.065 (This is the same as VCIO?)

 

I'll try to drop PLL voltage and set default values instead of Auto. Should I disable Internal PLL Overvoltage?

 

EDIT: I used the thermal paste which came with the NH-D14

 

You should not be reaching temps that high on that low of a vcore, at least I don't think so. Dropping or increasing my vcore 0.015 usually makes a 1C difference with the max I've seen being around 3C. I think leaving the voltages on auto is allowing them to increase on their own as you go under load and that's what's causing your high temps. At a lower speed like 4.4GHz the system may not play with the voltages as much and should result in lower temps. Open up your motherboard's tuning software and watch all of the voltages under load, especially those set at auto.

 

You shouldn't need Internal PLL Voltage enabled to reach 4.5GHz. The lowest I've seen it needed was 47x multi. I would disable that for 4.5GHz and it should slightly lower your temps as well.

 

Just out of curiousity, what was your max load temp running 4.4GHz under LinX? In the end IVI is right, the highest temp I have seen outside of LinX was almost 20C lower than my max during a 20 run test. I always figured that if my cooling can handle the LinX heat, I'll never have any other problems and can put my cooling system to rest

[Brutality]

You should not be reaching temps that high on that low of a vcore, at least I don't think so. Dropping or increasing my vcore 0.015 usually makes a 1C difference with the max I've seen being around 3C. I think leaving the voltages on auto is allowing them to increase on their own as you go under load and that's what's causing your high temps. At a lower speed like 4.4GHz the system may not play with the voltages as much and should result in lower temps. Open up your motherboard's tuning software and watch all of the voltages under load, especially those set at auto.

 

You shouldn't need Internal PLL Voltage enabled to reach 4.5GHz. The lowest I've seen it needed was 47x multi. I would disable that for 4.5GHz and it should slightly lower your temps as well.

 

Just out of curiousity, what was your max load temp running 4.4GHz under LinX? In the end IVI is right, the highest temp I have seen outside of LinX was almost 20C lower than my max during a 20 run test. I always figured that if my cooling can handle the LinX heat, I'll never have any other problems and can put my cooling system to rest

 

I think it was 68 or 69 at 4.4, but I can't fully remember.

 

Anyways, thx alot so far, I'll try to get it stable at 4.5GHz (at least!) now =)

 

BTW: Room temp is 26*C so that is anoter factor of high temps.. summer is kicking in in Denmark

 

EDIT:

Before LinX I did a 12h p95 blend test followed by a 3h Large FFT test (p95) (according to http://www.clunk.org.uk/forums/overclocking/39184-p67-sandy-bridge-overclocking-guide-beginners.html) finalized by a 2-pass memtest86+.

 

Now I know that the 12h 95 blend test is substituted by LinX, but will it also substitute the 3h Large FFT test? (I think it's meant for testing memory), so that I only need to do LinX and memtest 86+?

 

Thanks =)

 

EDIT2:

 

With the following BIOS settings I got 68*C max temp for 20 pass LinX test;

 

BCLK: 100.0

Turbo Ratio: By Per Core

Ratio Limits: 44

Internal PLL Overvoltage: Disabled

Memory Frequency: DDR3-2133MHz

EPU Power Saving Mode: Disabled

DRAM Timing Control: According to X.M.P. Profile

Load-line Calibration: Ultra High

VRM Frequency: Manual

VRM Fixed Frequency Mode: 350

Phase Control: Extreme

Duty Control: Extreme

CPU Current Capability: 100%

CPU Voltage: Manual

CPU Manual Voltage: 1.260

DRAM Voltage: 1.65000

VCCSA Voltage: 0.92500

VCCIO Voltage: 1.3750

CPU PLL Voltage: 1.72500

PCH Voltage: 1.07000

CPU Spread Spectrum: Disabled

 

Tell me if u think anything needs to be changed. I will now try to go for 4.5GHz by bumping vCore =)

 

UPDATE:

 

Bah reached 71*C again at 4.5GHz with a vCore of 1.285 and above settings.. small improvement though

 

What do you suggest?

 

UPDATE 2:

 

Got it stable at 4.4GHz with a vCore of 1.255, another improvement =)

Edited June 7, 2011 by Brutality

[wevsspot]

BTW: Room temp is 26*C so that is anoter factor of high temps.. summer is kicking in in Denmark

Yep, ambient air temperatures can certainly affect you - even if you're running great air cooling and even water cooling.

[Nuclear]

Just got your PM...

 

Your VCCIO voltages are way too high and you are in danger of degrading the IMC on your chip if it's running at almost 1.4v.

 

If you are unstable or stopping the test due to heat, then it's hard to determine which setting or settings need to be tweaked in order to get everything stable. Vcore, VCCIO, and PLL voltage are the three biggest contributors to heat in these SB chips. Dropping the PLL to below 1.6v and keeping VCCIO as low as possible can change your temps by 2-3C max, but also increase stability in some cases.

 

I'll give you two options / things I would try.

 

1. Buy a box fan or something that can help cool your ambient temps down. I would also perform a case test to ensure your case is getting proper airflow. A case test can be done by simply pulling the side panel off, if your temps go down then you need to increase airflow inside the case and improve your exhaust system. If temps stay the same or get warmer, your case is fine and you can button it back up.

 

2. Tweak these settings and voltages.

 

- Drop your RAM from 2133 to 1600MHz and see if they can run on 1.5 - 1.55v. Decrease your your VCCIO to 1.1v or a max of 1.12v. Your setting of 1.375v is extremely too high and as I stated earlier, could be close to causing to major damage. If you want to keep your RAM at 2133MHz, drop the voltage to a max of 1.14 - 1.16v and see what it does.

 

- Decrease your PLL voltage to under 1.6v.

 

- Decrease PCH back to 1.05v

 

- Increase VCCSA to 0.935v

[Brutality]

Just got your PM...

 

Your VCCIO voltages are way too high and you are in danger of degrading the IMC on your chip if it's running at almost 1.4v.

 

If you are unstable or stopping the test due to heat, then it's hard to determine which setting or settings need to be tweaked in order to get everything stable. Vcore, VCCIO, and PLL voltage are the three biggest contributors to heat in these SB chips. Dropping the PLL to below 1.6v and keeping VCCIO as low as possible can change your temps by 2-3C max, but also increase stability in some cases.

 

I'll give you two options / things I would try.

 

1. Buy a box fan or something that can help cool your ambient temps down. I would also perform a case test to ensure your case is getting proper airflow. A case test can be done by simply pulling the side panel off, if your temps go down then you need to increase airflow inside the case and improve your exhaust system. If temps stay the same or get warmer, your case is fine and you can button it back up.

 

2. Tweak these settings and voltages.

 

- Drop your RAM from 2133 to 1600MHz and see if they can run on 1.5 - 1.55v. Decrease your your VCCIO to 1.1v or a max of 1.12v. Your setting of 1.375v is extremely too high and as I stated earlier, could be close to causing to major damage. If you want to keep your RAM at 2133MHz, drop the voltage to a max of 1.14 - 1.16v and see what it does.

 

- Decrease your PLL voltage to under 1.6v.

 

- Decrease PCH back to 1.05v

 

- Increase VCCSA to 0.935v

 

Whoups that was a typo! VCCIO were at 1.1375

 

Okay I will try what you said, but can you answer the following question meanwhile?

 

Before LinX I did a 12h p95 blend test followed by a 3h Large FFT test (p95) (according to http://www.clunk.org...-beginners.html) finalized by a 2-pass memtest86+.

 

Now I know that the 12h 95 blend test is substituted by LinX, but will it also substitute the 3h Large FFT test? (I think it's meant for testing memory), so that I only need to do LinX and memtest 86+?

 

Update:

 

I'm testing how low PLL and VCCIO can go now, and then I will try 4.5 again.

So since I knew the above setup for 4.4 was stable I changed VCCSA to 0.935 and PCH to 1.05 and PLL to just under 1.6 and VCCIO to just over 1.4 and this was stable

So I tried PLL at 1.54 which wasn't stable.

Now trying to see how low PLL can go.

 

Update 2:

1.55 is as low as my PLL can go without losing stability, and VCCIO will cause instability when less than 1.13750

To sum up, these are my voltages with the above BIOS Settings

vCore: 1.255

DRAM: 1.65

VCCSA: 0.93750

VCCIO: 1.1375

PLL: 1.55000

PCH: 1.05000

 

Will try to get it stable at 4.5GHz now with above voltages (by bumping vCOre)

Update 3: Bah reached 71*C at 1.285 vCore.. improvement in temps though, but dunno what to do now... at least I got a stable setup at 1.255 =)

 

I hope I'm not doing a stupid procedure

Edited October 16, 2011 by Brutality