OC issues on i5-2500K + Gigabyte Z68A-D3H-B3

[bsas]

Hi All,

Here is my build:
CPU: Intel Core i5-2500K Processor
COOLER: Cooler Master Hyper 212 Plus 120mm Sleeve CPU Cooler, RR-B10-212P-G1
MB: Gigabyte Intel Z68 ATX DDR3 2133 LGA 1155 Motherboard GA-Z68A-D3H-B3 (rev1.3 | BIOS: F11)
MEM: Corsair Vengeance Blue 16 GB DDR3 SDRAM Dual Channel Memory Kit CMZ16GX3M4A1600C9B
GPU1: MSI nVidia GeForce GTX550 Ti Cyclone OC 1GB DDR5 2DVI/Mini HDMI PCI-Express Video Card N550GTX-TI CYCLONE OC
GPU2: MSI nVidia GeForce GTX550 Ti Cyclone OC 1GB DDR5 2DVI/Mini HDMI PCI-Express Video Card N550GTX-TI CYCLONE OC
CHASSI: Cooler Master SGC-2000-KKN1-GP Storm Scout ATX/MATX/ITX Mid Tower Case (Black)
POWER: Cooler Master GX Series 750W ATX Power Supply RS750-ACAAE3-US
HD: Momentus XT 500GB ST95005620AS (FW: SD28)

So, I've put both GTX cards on SLI.

Basically I already did all stress tests on the "safe" BIOS config to assure that all parts are stable.
I've did the Memtext86+, FurMark burn test on both GPUs and Prime95 on the CPU.
Everything looks fine.

But now I am trying to OC a little bit and I am confused.

P.S.: Running Prime95 on "Blend" makes my mouse pointer lag a lot. It is common?

A lot of options I select just make blue-screen really fast (sometimes on the boot) on the Win7 Ultimate x64. I've basically gave up of the "Intel ET6 tool" 😀

My current setup is doing 4.3GHz (x43 multiplier) using XMP for the memory on DDR3-1600 and really quick (like with 5 minutes of Prime95) I am already on 60C-65C on ALL cores. At least I am not seeing the temps going 65+ after a while...

So, what I am doing wrong? I would like to be able to do at least 4.5GHz...

The configs I did on the BIOS was:

- AHCI
- Disable HPET
- Disable the iGPU (HD3000)
- Disable C1E, C3/C6, CPU Thermal, Turbo Boost, EIST, Real-Time in OS
- CPU Ratio: 43x
- XMP: Profile1
- DRAM SPD: Quick

The rest I've leaved on "AUTO". Didn't changed any voltage.

P.S.: Running Prime95 on "Blend" makes my mouse lag A LOT. It is common?

 

[koogco]

(Sorry this post got a bit rambling, hope it is usefull anyway)
60-65C is fine. When you stress test your CPU, it is quite normal for the temps to go up very fast, but they should mostly even out shortly after (assuming you have at least decent airflow trough your case) I do not get the mouse lagging you speak of unless i run Furmark. You might want to beware of Furmark by the way, it is highly optimized for making your GPU's use power, and can use much more than a game running at 100% GPU use. The GPU is generally built for a gaming load but not for such an optimized load, and some have been known to die running Furmark.

For getting to 4.5ghz you may or may not have to controll the voltage manually. Sandy bridge chips vary alot, some does 4.8ghz with an air cooler, some will never do more than 4.3ghz stable.
The way I went about things was to put my voltage in fixed mode, set it to 1.35V 4.5ghz test that it was stable, and then start to go down in voltage. I also testet 4.8ghz at 3.1 volt, but that was not stable (and higher voltage at such speed would be too hot)
Now, when you have some idea how high ghz you can go to with "plenty" voltage, it is time to start lowering the voltage in steps. I ended up at 1.24V being short term stable at 4.5ghz. Then I started longer tests and it failed. So I upped the voltage a bit (1.25V) and tried again. This time it was longer before I had an error, but it was time to up it a bit again. Currently running at 1.26V and have yet to do a long enough test to see if it is enough.

I do testing by running prime95 (some say that intelburntest will find errors faster, but I never seem to have any errors with that, yet Prime95 fairly quickly gives me a rounding error if things are not stable)
When going down from 1.35V I would only test for 15minutes of prime95 and about 8 minutes of intelburntest each time. Then, having found the limit for that and moving upwards in voltages I run prime for as long as it will run without error. Some people are really picky at test for 72 hours before they call it stable. But seeing how regular use will never put it under that kind of pressure I will be content with 6-8 hours I think.

At any rate, If you want to make an overclock for 24/7 use, it should definitively be below 70C even during long tests (but don't worry if it gets to 75 during some of your early test)
General opinion seems to be that above 1.350V will hurt the longevity of the CPU (although noone knows for sure since it is not even 1½ years old yet) I personally aim for a voltage of 1.300V at the most as my final result, and might go down to 4.4ghz if nescesary.

By the way, keep a watch on voltage during idle and load (using CPU-Z for instance), if it drops when the load starts, you might need to adjust the load line calibration in the bios (in my case, I needed the strongest version or voltages would drop horribly, now they instead go up a slight bit on load)

Again, I am sorry for the rambling on.

EDIT:
Topic discussing how long to test and other things: http://www.tomshardware.co.uk/forum/page-273073_11_0.html
The spreadsheet for my own tests, if you want to see how I did things step by step (I played around with it a bit before starting to document though): https://docs.google.com/spreadsheet/ccc?key=0Av4YHL2djwkLdGxwNjctbWN2WVB6eV9ZdTlwRTRJc1E

 

[bsas]

Man! Awesome answer! Thanks.

I will do the tests.

One thing that I am really confused about the Z68A-D3H-B3 bios is that I cannot see clearly the "CPU Vcore" option. I don't know if they change the name of it on the F11 update or if it even HAVE this option... That's basically why I didn't changed the voltage.

Anyway, I did 1h of Prime95 and I am below 65C stable on 4.3GHz. I looks fine and I am rethinking if it pays of to change voltages and risk a more unstable system for just 200MHz more... 😀

 

[bsas]

Best answer selected by bsas.