Last update :- February 19th, 2012

Quick index:

Click on any image below to see a larger version.

Core i5 - Overclocking

Warning: if you overclock your processor or graphics card you will invalidate any warranty

Expectations

From my earlier section on overclocking the Athlon XP 2600+ to 2.412GHz you'll recall that all components are not made equal. Unless all your system components are picked especially (if you can find them) the level of overclocking cannot be guaranteed. In particular, the level of overclocking you can achieve depends upon the following components, in no specific order:

For instance, you may have a highly overclockable CPU (processor) but you won't see the best performance if the memory isn't up to it or the motherboard chipset won't overclock very well - if indeed the BIOS allows it. Therefore, when you overclock you cannot expect anything but only hope for some kind of performance gain. Unless you want extreme performance you're advised to set yourself a goal in terms of what you want to achieve and what your limits are. Also do the research and see what others are achieving with similar configurations.

From my past experiences with overclocking, when trying to overclock the Intel Core i5-2500K things seem to have become both simpler and more complicated at the same time:

Testing Utilities

During the overclocking process I used the following utilities to test and monitor the system:

Cooling

Before detailing the overclock I achieved and what I had to do to get there I explain how and why I modified the cooling as supplied with the system I bought. Throughout testing the room temperature was around 24°C.

As supplied

As stated in the section on selecting the components the chosen case was the Cooler Master Silencio 550 and the CPU cooler was the Corsair Hydro Series H60 sealed water cooling system - although I opted for Computer Planet's "Silent Edition" option for a further £10 where they replaced the stock Corsair 1700 RPM, 30 dBA fan with a Sharkoon "Silent Eagle" 1000 RPM, 19dBA fan (note that the specification suggests it to be the 2000 RPM version but testing showed it to be the 1000 RPM version).

I was interested to see how this would perform without overclocking the system as the combination isn't ideal - but I did want a silent (or near silent) system under normal circumstances. The reason for this is that most of the cases people use for overclocking these days have at least 5 fan positions and some more - 2 inlet fans at the front that also cool hard disks, 1 or more inlet fans on the side that also cools the graphics card(s), 1 or more exhaust fan at the rear to get rid of hot air and 1 or more exhaust fans at the top. In this scenario they normally recommend mounting the H60 in one of the exhaust positions at the rear with the fan pulling cold air in from the outside and pushing it through the radiator - with the other exhaust fans expelling the hot air. With the Silencio 550 I didn't have this option so had to mount it with the fan pulling air through the radiator and out of the case - as demonstrated below:

Case Cooling

Case Cooling

I was actually surprised by the results, given that a slower speed fan was used in a less than ideal configuration (note - the background changes because I have it cycling):

Supplied - Idle Supplied - Load
Supplied - Idle Supplied - Load

So, with the Sharkoon "Silent Eagle" 1000 RPM, 19dBA fan fitted that's a core temperature of 44°C when idle and 62°C under load. Given that the Core Temp FAQ recommends that the core temperature is kept around 20°C lower than the junction temperature (Tj.Max) of 98°C under full load then 62°C vs 78°C gives a headroom of 16°C! It would have been interesting to see what the stock 1700 RPM, 30 dBA fan produced. You can see that the pump runs at ~4400 RPM, the CPU fan (i.e., the H60 fan) at ~1030 RPM and the case fans at ~720 RPM - and these are maxed! With these speeds the Silencio 550 lives up to it's name - it is practically silent!

Modified cooling

In the end I regretted buying the "Silent Edition" of the H60 though - having paid £10 for this option - as I doubted whether it would cope with overclocking the processor. I therefore decided to look into buying some more fans for both the case and the H60 and research indicated that one's with a high static pressure would be beneficial. In simple terms (you can find out more for yourself if you want to), the higher the static pressure rating of a fan the better it is at pulling in the surrounding air and pushing it through the thin but dense fins of a radiator or heatsink. The standard one supplied with the H60 for example is rated at 3.20 mm/H2O for that reason.

I found an excellent review of 120mm fans at Xbitlabs giving detailed specifications and results, including what they regarded as a comfortable noise level of 36 dBA - see their review of 140mm fans for the methodology. One of the best performers in their tests and therefore one of their recommended products was the one I opted for - the Cooler Master SickleFlow. Running at 12V, full speed they measured it as producing an airflow of 33.3 CFM (in their tube) at 46.2 dBA whilst it was still producing 20.3 CFM at their comfortable noise level of 36 dBA (3rd in test, running at 1230 RPM, 6V). The specs below tend to disagree with some of these values - but then again they're probably tested differently so I've also included those for two of the other highly rated fans in their list for a comparison:

Fan Speed Noise Air Flow Static Pressure
Corsair 1700 RPM 30.2 dBA 74.4 CFM 3.20 mm/H2O
Cooler Master SickleFlow 2000 RPM 19.0 dBA 69.9 CFM 2.94 mm H2O
Scythe Kama Flow 2 1900 RPM 33.8 dBA 63.2 CFM Unknown
be quiet! SilentWings 1500 RPM 17.0 dBA 50.5 CFM 1.63 mm H2O
Sharkoon "Silent Eagle" 1000 RPM 19.0 dBA 36.8 CFM Unknown

So, with the fans chosen, I had to (a) decide how to mount them and (b) how to control them. The diagram below shows the mooting options available to me, compared to the standard method - with the fan pulling air from inside the case through the radiator, pushing air through the radiator and a push-pull arrangement which most regarded as the best option. You could also mount the fan outside the case but I tested the "pull" and "push-pull" arrangements.

Hydro H60 Options

Corsair Hydro H60 Mounting Options

To control the fans I initially considered using the ASUS Fan Xpert utility that's part of AI Suite supplied with the P8P67. Whilst it appeared to be abled to control the CPU fan it didn't seem to affect the chassis fans. I may have been doing something wrong here but instead I opted for the convenience of a fan controller. In the past I'd used the Akasa Fan Controller Junior 3.5" but it wasn't practical this time around as the control knobs would clash with the foam door of the Silencio. There were options for touch-screen models but these appeared to be overkill and cost £40+ so I opted for the NZXT Sentry Mesh for £26 which has 5 sliders and is reviewed at Hardware Secrets:

NZXT Sentry Mesh

NZXT Sentry Mesh

When using the push-pull arrangement (which I currently use) I used a 3 pin fan splitter cable from Akasa to combine both fans on one slider and you will also need 4 off 6-32 x 1 1/4" screws to mount the second fan. As I was using a controller I had to find some way of monitoring the fan speed. The SickleFlow fans include all 3 wires on their connectors and therefore include the yellow "rotation" wire. I had some spare 3-pin fan headers lying around so I moved the yellow wire from the SickleFlow connectors to the spare ones and connected these to the CPU_FAN, CHA_FAN1 (via an extension cable) and CHA_FAN2 headers on the P8P67. The results are shown below, with a summary at the end:

Pull - Min speed - Idle Pull - Mini speed - Load
Pull - Min. Speed - Idle Pull - Min. Speed - Load
Pull - Same speed - Idle Pull - Same speed - Load
Pull - Same Speed - Idle Pull - Same Speed - Load
Pull - Max speed - Idle Pull - Max speed - Load
Pull - Full Speed - Idle Pull - Full Speed - Load
Push-Pull - Min speed - Idle Push-Pull - Min speed - Load
Push-Pull - Min. Speed - Idle Push-Pull - Min. Speed - Load
Push-Pull - Same speed - Idle Push-Pull - Same speed - Load
Push-Pull - Same Speed - Idle Push-Pull - Same Speed - Load
Push-Pull - Full speed - Idle Push-Pull - Full speed - Load
Push-Pull - Full Speed - Idle Push-Pull - Full Speed - Load

I found the minimum speed of the SickleFlow to be ~880 RPM - which is obviously higher than the 720 RPM fans supplied with the case but the impact on the internal case temperature should be minimal, maybe 1 or 2 degrees. As well as running the CPU fan(s) at the minimum speed I also ran it at roughly the same span as the Sharkoon for a comparison - both tests with case fans at minimum. When running the CPU fan(s) at full speed I also ran the case fans at full speed - as I wouldn't do one without the other and it would give a direct comparison with the overclocked system.

Configuration Core Temp CPU Case
Idle Load
Pull - Minimum 39°C 66°C 910 RPM 859 RPM
Push-Pull - Minimum 35°C 55°C 928 RPM 854 RPM
As supplied 42°C 62°C 1028 RPM 720 RPM
Pull - Same 36°C 64°C 1021 RPM 868 RPM
Push-Pull - Same 33°C 52°C 1021 RPM 863 RPM
Pull - Maximum 34°C 51°C 1869 RPM 1807 RPM
Push-Pull - Maximum 31°C 46°C 1903 RPM 1820 RPM
4.5 GHz OC 30°C 65°C 1911 RPM 1837 RPM

As you can see above, the push-pull arrangement gives a significant reduction in core temperature - especially under load conditions. I only revert to overclocking when playing more demanding games so most of the time I have the sliders on the NZXT set to minimum and the core temperature will therefore fluctuate between 35°C and 55°C and with these settings it's virtually silent. With the fans maxed out it's still acceptable when playing the games that might need it. You can only really notice a difference in the fan noise when it they get to around 1380 RPM.

Processor Overclock

As you may have noticed from the table above I can run my Core i5-2500K at 4.5GHz compared to the stock 3.3GHz - that's a 36% overclock! Research shows that ~50% of these processors can reach 4.4~4.5GHz and I was happy to do so. I could have tried pushing it further but left it at that. Now I'll explain the process I used to achieve this - which is straight forward compared to previous methods.

With the P8P67 and other ASUS motherboards a good starting point that gives you an idea of what can be achieved is their TurboV EVO utility that's part of AI Suite:

TurboV EVO - TurboV TurboV EVO - Auto Tuning
TurboV EVO - TurboV TurboV EVO - Auto Tuning

The TurboV tab (strange as the utility is called TurboV EVO) allows you to change the base clock (BCLK) and CPU/DRAM voltages. If you select More Settings → CPU Ratio → it will make a change to the BIOS and activate this. This may not work though - in my case, the BIOS feature that this is supposed to enable is noyt available. Note that changes are not retained in the BIOS and need to be saved to a profile to load after Windows starts and as with all such utilities all the options you may need to get the best overclock are not available.

The Auto Tuning option is very useful as this will give you a starting point for your overclock. If you don't plant to run your system overclocked all of the time then first reboot, enter the BIOS and save the current settings to an "OC Profile" under "Tools". Then, the best option to choose is "Extreme" and upon doing so the system will change CPU and memory settings, save them to the BIOS, reboot a few times and eventually display the results. You can then reboot, enter the BIOS and save these changes as a different "OC Profile" as your start point. In my case, the following changes were made to the BIOS for a 4.326 GHz overclock:

Property Default Updated
BCLK Frequency 100.3 MHz 103.0 MHz
CPU Voltage 1.165V 1.195V
DDR Voltage 1.5V 1.65V

If you're happy with this simple method and don't want to push it further then you can leave it there. If you want to have a go at manual overclocking, first of all, have a look at the generic i5-2500K overclocking guide at bit-tech (which has a dedicated P8P67 section) and the official P8P67 overlclocking guide at [H]ard|Forum as they're invaluable.

Voltage

Understanding the core voltage (VCC) on the 2nd generation Intel Core range is difficult - as the minimum and maximum values aren't specified anywhere in the datasheet. It does use a new serial interface to replace the parallel on on older processors for VCC Voltage Identification (VID) and this is used to control external voltage regulators and it states that "Individual processor VID values may be set during manufacturing so that two devices at the same core frequency may have different default VID settings". The range of values covers from 0.25V to 1.52V and these represent VCC_MAX - so you won't be able to set VCC higher than this. If you use the "TurboV" tab as shown above it will tell you what Ryour defaults are - in my case VCC is 1.165V with the DRAM at 1.5V.

The [H]ard|Forum guide states that for (unlocked) K series parts "the stock voltage supplied will allow for consistent overclocking generally up to a multiplier of 43x" and that the default voltage range in this case is approximately 1.24~1.26V under load. To be able to use multipliers between 44x and 47x it will generally require 1.3~1.375V and for 48x to 50x it's 1.4~1.5V. The guide states that the best potential overclock for most systems is 1.4~1.425V - but I opted to stick with a maximum of 1.375V and accept a possible 47x result.

Note: If you use Core Temp the FAQ states that "Core Temp's VID readings should be ignored" for the newer Core series.

Temperature

As stated earlier, the Core Temp FAQ recommends that the core temperature is kept around 20°C lower than the junction temperature (Tj.Max) of 98°C under full load.

Clock

As mentioned in the previous section (and noted in the bit-tech article on overclocking) Intel chose to use a single clock generator on the Sandy Bridge processors with all clocks derived from this. For this reason the old method of adjusting processor and memory clocks won't yield much. You can potentially adjust this base clock (BCLK) by up to 10 MHz and most motherboards supporting overclockimg will allow you to adjust this by increments of 0.1 MHz - but unless you want to get the absolute best out of your system you don't generally have to touch it.

Results

Over a few days (using the automatic results as a starting point) I made changes to the BIOS, rebooted and ran the Prime95 "Torture Test" on all 4 cores overnight to ensure stability and the results I got are shown below:

4.5GHz Overclock - Idle 4.5GHz Overclock - Load
4.5GHz Overclock - Idle 4.5GHz Overclock - Load

To achieve this I had to use the following BIOS settings on the "AI Tweaker" page - with all other settings left unchanged:

Property Default Updated
Target CPU Turbo-Mode Speed 3700 MHz 4500 MHz
Target DRAM Speed 1333 MHz 1600 MHz
AI Overclock Tuner Auto Manual
BCLK/PEG Frequency   100.0
Turbo Ratio Auto By All Cores (Can Adjust in OS)
By All Cores (Can Adjust in OS)   45
Memory Frequency Auto DDR3-1600MHz
DRAM CAS# Latency Auto 9
DRAM RAS# to CAS# Delay Auto 9
DRAM RAS# PRE Time Auto 9
DRAM RAS# ACT Time Auto 24
DRAM COMMAND Mode Auto 2
Load-line Calibration Auto Ultra High
VRM Frequency Auto Manual
VRM Fixed Frequency Mode   350
CPU Voltage Offset Mode Manual Mode
CPU Manual Voltage   1.285V
DRAM Voltage Auto 1.65V

I was lucky enough to be able to achieve 4.5GHz with a core voltage of only 1.285V compared to the 1.3~1.375V mentioned in the [H]ard|Forum guide - which helped keep the maximum temperature under load to 65°C, 33°C lower than the junction temperature (Tj.Max) of 98°C. When you consider the less than ideal cooling with the Silencio 550 this isn't bad. As I chose the Corsair 8GB XMS3 PC3-12800 1600MHz memory the DRAM voltage and timings above are in the SPD and therefore within specification.

Graphics Overclock

As I'd chosen the Gigabyte GeForce GTX 550 Ti OC (GV-N550OC-1GI) I already had a graphics card that was factory overclocked and thanks to the excellent EVGA Precision free utility by EVGA Corporation (that can be used with any make of modern graphics card) - it is easy to adjust the card's core, shader and memory clocks and fan speeds. Once complete you can save the settings as a profile and quickly apply them when you want to.

Apart from a graphics benchmark like 3DMark 11 aborting or showing graphics anomolies (also known as artifacts) the best way I know of checking for errors is the old ATITool utility by TechPowerup. This hasn't been updated since December 2006 and doesn't support overclocking of modern cards but does include a handy artifact test. Simply run it, select "Show 3D View" and "Scan for Artifacts". The results in the bottom left hand corner may well show "No errors for xx:xx.xx sec." but the artifacts are fairly obvious - look at the image on the right below:

Overclock - Clean Overclock - Error
Overclock - Clean Overclock - Error

As you can see from the image on the left above I maanged a decent overclock for this board that had already been factory overclocked - and the increased fan speed wasn't even noticed with the case and processor fans on minimum.

Configuration GPU Memory Shader
Stock GeForce GTX 550 Ti 900 MHz 4100 MHz 1800 MHz
Gigabyte GeForce GTX 550 Ti OC (stock) 970 MHz 4200 MHz 1940 MHz
Gigabyte GeForce GTX 550 Ti OC (oclock) 1015 MHz 5040 MHz 2030 MHz

Benchmark Results

Rather than showing the benchmark results throughout I thought it would be easier to show them in one table here - including the older 3DMark Vantage for reference:

Benchmark Stock CPU OC GPU OC CPU/GPU OC
PCMark 7 2945 3479 3183 3561
3DMark 11 P2743 P2795 P2914 P2940
3DMark Vantage P11632 P12232 P12259 P12757

If you want to advertise on this page please E-mail me at paul_at_paulslounge_dot_co_dot_uk

Copyright © Paul Collins, 2001 - 2012
Paul's Lounge
All rights reserved

Valid XHTML 1.0 Transitional

Privacy Site Map Home About Paul Blog News Links
Computers Emulators Software Games Tips & Tricks Fishing Golf