Maximum PC

Hey all. I've ordered an Intel i5-2500k and an Asrock Extreme3 Gen3 board and am looking into how to overclock them appropriately.

I'm new to overclocking so I'm a bit nervous about doing something to fry my CPU. I'd like advice on settings that would help me establish a safe and stable overclock (nothing risky at all please). In doing some reading it seems like reaching 4.5Ghz is quite possible with the 2500k.

I'd be using Asrocks UFEI to change the settings and would like suggestions as to what settings I should change and what programs I should use to test for stability and temperatures.

Thanks

I think the other thing you might need to have would be a decent cooling solution for your CPU and for your case.
Other than that set the multiplier to 45 and your BCLK to 100 and your done. Though you might want to start with a multi of 43 and go up by one increment at a time after doing a stress test on it for at least 24hrs. By allowing the 24hrs to go by you should be able to see how high the temps get and also to see if you get a BSOD. I also would recommend grabbing CPU-Z and HWmonitor so that you can keep a eye on system temps and what clock your CPU is at.

yeah, its pretty stupid simple now. You really don't have to mess with the voltages till you start to go beyond 4.5ghz. See what your max stable multiplier is with your chip. It varies widely between batches.

I bought a cheap but well reviewed and commonly used aftermarket heatsink for the CPU (Coolermaster Hyper 212+). I've installed aftermarket heatsinks before so that shouldn't be a problem (although applying the TIM always makes me a bit nervous).

Thanks for the program suggestions, in looking into it I had downloaded OCCT and Real Temp for stress and monitoring respectively. I will check out CPU-Z and HWmonitor as well.

I agree that I should start low and take it slow, doing stress tests with each step up. I also bought some good quality low CAS latency 1600mhz memory (http://www.newegg.com/Product/Product.a ... -_-text-_-) and an already overclocked Geforce 560X Ti (http://www.newegg.com/Product/Product.a ... -_-text-_-). I was only going to bother OCing the processor though as I figure thats the safest place to get good gains.

As for the CPU so just change the multiplier and leave the voltage the same or on AUTO? Or should I step the voltage up accordingly via offset or manual. That was my main question I suppose is what voltage to use with what multiplier for a 2500k. I know it can vary batch to batch but was looking for a good starting point that will likely be stable and work from there. So like 43x with 1.32 volts or something like that.

I left mine in auto since I wasn't going for a crazy OC but @ 4.5GHz you may need to adjust the voltage a tiny bit. for anything below that you should be fine to leave the voltage in auto.
Another thing to look for is if you do a auto OC the settings typically change the BCLK and the Multi. so just be aware of that. you seem to be wanting to adjust the settings yourself so leave the BCLK @ 100 and just adjust the Multi. For some reason the Sandy bridge chips haven't been happy with BCLK changes at least typically more than say 103 at least for me.

Also I would suggest a second fan for the Hyper 212+, I have one to and it helps! For me it was on average about a 5-7C change over one fan.

Thanks for all the advice guys.

I just got my parts and assembled my computer. Hilariously it turns out my optical drives are so old that they don't have SATA ports and my motherboard doesnt have one of the old ribbon style connectors so I couldn't install Windows 7 on my new SSD and wont be able to until I get a new DVD drive. Instead I booted up an old copy of windows xp on an old hardrive and started overclocking.

First I set my clock and latency timings for my memory then checked the temps on my processor. My initial temps were around 25 degrees idle for 3 of the 4 cores and then 30 for the 4th (I've heard that one core typically runs hotter on sandy bridges)

I started at 4.0GHz with a fixed 1.25 Vcore and have been slowly stepping up the multiplier and doing repeated tests with LinX monitoring with Real Temp and CPU-Z. I'm up to 4.2GHz and its only gotten to about 52 degrees on the hottest core at max loads so lots of room to spare. Should hit 4.5GHz easy possibly more.

I've heard you can change BCLK very slightly (by like 0.5MHz) and get away with it but I don't think I'll bother.

Interesting about the extra fan Wiked I may do that. My case has a back exhaust fan that is right near that side of the heatsink already though so I'm not sure if it would make that much difference for me. Still, I may give it a try.

Yeah I have a fan back there too they are probably 2in apart but it did help. What are you using to stress the CPU for the OC? I got mine up to 4.7GHz but after stressing it for 2 hours it wasn't happy and I had to back it down. I honestly believe you could hit 4.7 if you adjust the Vcore to say 1.35-1.4V. I also wanted to keep my temps down below 70C.

Good to know about the fan, I'll see if Vcore or temp ends up being the limiting factor for the overclock first though.

Still working on it, right now i have it stable at 4.5GHz and 1.375 VCore for 10 loops of LinX running at about 64 degrees on my "hot" core. Monitoring all four cores and one is about 7 degrees hotter than the other three. I bumped to 4.6GHz but it froze almost immediately upon testing.

Going to try 4.6GHz at 1.4V then probably call it quits.

Settings are
Vcore manual set to 1.4V
Multiplier set to 46
Turbo Boost Power Limit set to manual
Short duration Power limit at 250
Long duration Power Limit at 200
Long Duration maintained (sec) at 1
Core Current Limit at 300
Spread Spectrum Disabled

CPU Load-line Calibration set at max (Level 5)

*EDIT*
4.6GHz was stable at Vcore 1.4 for 10 loops of LinX. The "hot" core peaked briefly at 65 but mostly was at 64 again while the others sat around 57. The actual measured Vcore peaked at 1.38 abut mostly sat at around 1.3. Gave 62 GFlops of power. Think I will keep it there. Not sure I want to go above 1.4V on the Vcore as I'd prefer to play it very safe. 64 degrees is still not to hot though so I may not bother with the extra fan. I'll keep an eye on it though.

After more testing I had to bump the Vcore up to 1.41 before 4.6GHz was stable. That is starting to get a bit higher than I'd like as I set my personal limit at 1.4 (perhaps for no good reason I just want to be cautious). Stepping up from 1.375 to 1.41 Vcore just to go from 4.5GHz to 4.6GHz seems perhaps not worth it so I might bump it back down to 4.5GHz and be happy.

This is where your guys expertise would come in handy though. Are there other settings I could change besides Vcore that might give me better stability at 4.6GHz? I'd love to be able to drop the Vcore, change something else and have a stable system. Any suggestions?

Thanks

Got 4.6GHz stable by going to offset mode with +0.05 which lets the Vcore go as high as 1.36 at peak. I also changed my PLL to 1.750V, put Core current limit at 230, short duration power limit at 230 and long duration power limit at 200. I also disabled C3 and C6 states. Load Line is set to level 4. My CPU temp maxes at 70 degrees under heavy load on LinX or IBT.

With the same settings I can get a stable 4.5GHz with an offset of only +0.01 which peaks the voltage at 1.32 which is just above stock. CPU temp peaks at 64-65 then.

I also was able to get to 4.7GHz by using an offset of 0.07 which hits a Vcore of 1.39 at peak. I also bumped the core current limit and short duration power to 250. My temps then peak at about 71 degrees.

I think I might go with the 4.5GHz still with offset mode. I like the fact that the settings are so close to stock and I'm not sure the difference between 4.5GHz and 4.7GHz is enough to warrent changes.

Anyone have an opinion, is the settings I used for 4.7GHz safe?

Wow! Nice job! It's nice to see someone finally show a lot of what they did and get a stable clock bump on the 2500K. I think you would be fine at the 4.7GHz since the core is still below the threshold voltage of 1.4V I don't think I would push it above that. 1.4V would be my personal threshold I personally like to keep my temps around 60C but that is me, I like to keep a buffer between 60C and 70C to allow for dust build-up. Typically once I see one of the cores hit 65C I know its time to clean the system out. It may not be the bast way to do things but it works for me. All-in-all great job on the documentation of the OC. If I get a AIO water set up for my CPU then I will try to boost it up to 4.7GHz. I just like to see a 2500K get to where most people can get the 2600K though I am sure some can get theirs to 5GHz. VID max I think is 1.52V,
though that isn't Vcore max I would say keep the Vcore at/below 1.40V on air and 1.45V on water.

Well I just discovered that although I am stable in LinX and IBT I do not appear to be Prime95 stable as it errors out or freezes in the first hour. I guess I have to ask myself if I care since this is not a workstation its a gaming computer and I doubt that will cause me any issues. If I do care I have wiggle room with the voltages and could probably bump them up to compensate.

My only real concern is the fact that of my 4 cores one of them is a good 10 degrees hotter on idle and usually 5-6 degrees hotter under full load and one that is extremely cool (like when the hot one is 66 the cool one is 56 even under load while the others are 60). I've read some posts to suggest with the quad core sandy bridges that one of the cores is usually dominate and therefore will be hotter but I've also read threads suggesting you should be concerned if it is 10 degrees hotter that you may have an airbubble in your TIM or that the heatsink isnt quite seated properly. Once again I am not sure I care as even under full stress test loads I am only hitting something like 66 on my hottest core and 60 on the others.

I agree with the 1.4V threshold that was it for me as well. Ive read its safe to push to 80 degrees but I feel the same way as you as I want something under 70.

Ah well as it turns out all those settings went out the window when I got my SSD and installed Windows 7 (I had been running Windows XP). Now Linx isn't limited to 2gig of RAM and can use the full 7gig free that I have plus features associated with windows 7. That meant that what was stable on Win XP was not stable on Win 7. So I redid the overclock and found that 4.5GHz is the max I am comfortable with as with a PLL of 1.709, PLL Overvolt Enabled, a Vcore offset of +0.02 and C3 and C6 states disabled I am hitting core temps of 70 degrees under max load.

Then again with windows XP my max overclock at 4.8GHz produced 64 GFlops of processing power in LinX. Now with Win 7 my 4.5GHz overclock produces 109.5 GFlops. The results carry over to more real-world benchmarks such as PCMARK05 where my score went up considerably with Windows 7. So I'm not getting stable at those higher clocks but I seem to be getting more bang for my buck.