i5 2500K OC heat issues and questions

[Charlie98]

I've pretty much got my new build (in sig below) where I want it, hardware-wise; everything is installed and it's been doing fine for the past 6 weeks.

I've attempted to OC it, first using Gigabyte's SmartBoost utility... seemed easy enough. Hit the button for 4.10GHz, restart, and voila'! I get a faster machine, right? Well, yes and no... it get's pretty darned hot at 4.1GHz.

I have recently pulled the mobo out, pulled the 212+ off, cleaned it and the CPU off, reapplied the CM thermal paste (working it into the grooves in the 212's heat sink, applying a small bit and scraping it with a credit card and doing the same on the CPU,) doubled-up on the fans (it is now a push-pull fan setup,) and dropped the whole mess back into the can... a CM HAF922 with 4 fans and plenty of airflow in an ambient 75F room.

Still the same... :'(

I did download and try Gigabytes other utility, EasyTune 6, but I didn't quite understand all the parameters it was allowing me to change, and haven't fooled with it except to use it as a monitor.

I use LinX on max memory as my stress test, I'm not really going to bother with anything else until I can get the heat under control. I will say it is dead stable with LinX testing... I've never gotten one error testing in LinX and have run it for 50 minutes at a time.

I've seen others with basically the same setup in the mid 4GHz's and temps far lower than mine (with air cooling.) I'm just about ready to admit I have one of Those Chips and I'll never be able to OC it very far... but maybe I'm missing something: Voltage management.

2 screenshots...

First is a 41x OC under load, second is after testing @ same OC... what is up with all the different voltage readings?

CPUz: voltage reading hardly changes under load or idle;
ET6: voltage from 1.02v idle to 1.332v under load;
RT3.7: reading VID... 1.4061v, which I understand isn't such a good thing.

This last OC I went into the BIOS and changed the multiplier, not using either of the GB utilities. I have not messed with memory voltage and probably won't. Most of the other settings are on auto...

Questions:

Is the overheating coming from the VID voltage?

CPUz says it's reading core voltage... is it really? Then, what voltage is ET6 reading to have such swings loaded or idle? I assumed they were both reading core voltage...

Much of what I've tried to read online applies to older iCore chips, I don't know how much of it applies to the 2500K.

If someone out there is familiar with the GB mobo's BIOS and settings, I'd love some suggestions, or voltage suggestions in general, or if you see any glaring faults. I don't want a rocketship, but I'd like to run it at 4.1-4.2GHz reliably without fear of cooking my nice, new CPU.

 

[janas19]

SmartBoost is bad. You no want SmartBoost.

 

[dma0991]

I'd consider the CM Hyper212+ as a bare minimum when it comes to overclocking. It is not a beast but it is cheap and gets the job done pretty well. Also to make a direct comparison with others without considering the ambient temps where you live, you're making a bad assumption. They could be living in Antarctica for all we know and they're getting -10C.

As you increase the voltage you could probably expect the heat to rise as well and in the first picture, EasyTune 6 is giving the more accurate voltage reading. In case you're worried about frying your CPU, I had a friend who ran a CPU at 100C for months before I actually have the necessary tools and time to fix the overheating problem. If possible, do a manual overclocking instead of relying on OC utilities.

 

[Charlie98]

SmartBoost is bad. You no want SmartBoost.

...and why would that be? Not arguing, just curious...

As you increase the voltage you could probably expect the heat to rise as well and in the first picture, EasyTune 6 is giving the more accurate voltage reading. In case you're worried about frying your CPU, I had a friend who ran a CPU at 100C for months before I actually have the necessary tools and time to fix the overheating problem. If possible, do a manual overclocking instead of relying on OC utilities

I went back into the BIOS this morning and, lo and behold, at the bottom of the frequency settings page is core volts. It was 1.33v when I was in the BIOS, so that makes sense that EasyTune is giving me better information than CPUz re: core voltage.

VTT is set at 1.05v/auto, I wonder if that's what CPUz is reading...?

 

[janas19]

...and why would that be? Not arguing, just curious...



I went back into the BIOS this morning and, lo and behold, at the bottom of the frequency settings page is core volts. It was 1.33v when I was in the BIOS, so that makes sense that EasyTune is giving me better information than CPUz re: core voltage.

VTT is set at 1.05v/auto, I wonder if that's what CPUz is reading...?

You don't want it to set the Vcore level automatically. Set it yourself.

 

[Charlie98]

You don't want it to set the Vcore level automatically. Set it yourself.

Is 1.33 vCore a good starting point, or should I reduce to...?

I figure I'm going to have to go in and manually set most of the voltage parameters, I just don't know what to use as a starting value.

 

[janas19]

Is 1.33 vCore a good starting point, or should I reduce to...?

I figure I'm going to have to go in and manually set most of the voltage parameters, I just don't know what to use as a starting value.

Actually, the traditional method is to leave your Vcore at stock, then incrementally increase the Bclock multiplier levels until you crash/BSOD. That will tell you the limit for that Vcore. If you don't need additional speed, then you're good.

 

[repoman0]

Is 1.33 vCore a good starting point, or should I reduce to...?

I figure I'm going to have to go in and manually set most of the voltage parameters, I just don't know what to use as a starting value.

If you're happy with 4.1GHz, just reduce the voltage by 0.01 or 0.015 and run Linx for 10 minutes, repeat. When Linx crashes within that 10 minutes you can just add maybe 0.02V and it will probably be stable (obviously test to make sure). You can leave everything but Vcore alone.

You can do it more carefully too but this is what I do to keep it quick

 

[ther00kie16]

Yea, stock voltage OC is the way to go. Power (heat) is proportional to V^3.

 

[Charlie98]

Quick update....

I went in and reset everything to stock and worked my way back into the voltage settings... right now I'm at 4.18GHz at .08 under stock voltage. The best news? My CPU temps dropped 10C in LinX! EasyTune shows voltage around 1.272 under LinX load.

It did BSOD at -.10v, so I bumped it back up .02v. It also did not like anything higher, OC-wise.

Silly question... what's the difference between raising the multiplier vs raising the BCLK? Isn't it just a different means to the same end, or am I missing something... again.

 

[repoman0]

With Sandy Bridge, the clock generator was moved to the chipset I believe and a lot of different system clocks are tied to the 100MHz base clock (PCI-E for one, which has never really liked going much past 100MHz). So you are really risking stability moving much past 100MHz on the base clock, and most likely you won't be able to get past 105-110MHz. So most people leave it at 100MHz and clock with just the multiplier, except for the guys who really care about the extra few MHz they can squeeze out by turning it up to 103-104..

Looks like you're right about where I am - 4.2GHz, about 1.26V at load and ~70C max load temp in Linx (around 50-55C doing less stressful things at full load) with the 212+. Seems to be a pretty common place to be with this cooler. We could easily move up to 4.4-4.5GHz under 1.35V but that would probably mean going over 75-80C in Linx.

 

[janas19]

Quick update....

I went in and reset everything to stock and worked my way back into the voltage settings... right now I'm at 4.18GHz at .08 under stock voltage. The best news? My CPU temps dropped 10C in LinX! EasyTune shows voltage around 1.272 under LinX load.

It did BSOD at -.10v, so I bumped it back up .02v. It also did not like anything higher, OC-wise.

Silly question... what's the difference between raising the multiplier vs raising the BCLK? Isn't it just a different means to the same end, or am I missing something... again.

Nice! Bclck raises the speed of the CPU and system bus, so everything connected on the bus is sped up in tandem. It puts more strain on your system. The multi just affects the CPU. It's better to tweak the multis first, and leave the Bclck alone if it's ok

 

[Charlie98]

The Gigabyte utilities leaned on the BCLK pretty heavy, that's why I bumped mine up to 103.0... but I will probably put it back down to 100 where it belongs... thanks for clearing that up for me. That may have been part of the instability problem at .-10v vCore; I'll have to go back in and try it once more to see.