…who knew overclocking took so long…
If you love graphs, then you’ll love today’s post. Following on from last weeks post about the installation of my new Corsair H50 CPU Cooler it was time to explore probably the biggest benefit of better cooling, overclocking. Overclocking is where you adjust the settings so that the component (generally the CPU, but it’s also possible to overclock RAM, graphics cards and some other small bit and pieces) runs above it’s normal operating levels. A bad overclock can cause significant damage to the component or components in question, however a careful overclock only has two possible downsides, increased energy requirements and greater heat generation, which is where the increased cooling comes into play. Components can handle certain levels of heat without too much problem, anything above will cause them long term damage. But supposing your cooling system is good enough to cool the component far below their maximum tolerance level, let’s say the maximum tolerance is 90°c but your cooling brings the temperatures down so that even when it’s under load (when it’s working) it only reaches 50°c. That’s fourty more degrees that your component could tolerate, adding in a hefty safety margin still means you could run 10-20°c with barely any impact on the lifespan of your component. (If you’re after some further reading, have a look at this great article on component degradation by AnandTech.)
This is what makes overclocking more of an art, than a science. Even processors of the same model will have minor differences between them, one might be able to tolerate a couple more degrees than the other, another might handle heat well but not take so well to increased voltages. Therein lies the skill of overclocking, carefully testing the limits of your components without damaging them. Some people are fanatical about overclocking, always trying to eke out some extra performance no matter how long it takes, one of the more well known examples is of people attaching metal enclosures to the motherboard for holding liquid nitrogen. I am most definitely not of that league, before now I had never performed even a minor overclock. However when I was planning this computer I set my aim for the long-term, keeping a gaming computer competitive can be a difficult and expensive process, but by making the correct decisions you can minimise this by quite a large margin. Apart from replacing the cooler on my graphics card (which is just one of those things that happens some times) and the standard extra hard disk space I haven’t spent any money on my computer in almost two years, and it’s still capable of playing modern games on fairly high settings. But my aim was even more long term than this, I decided to purchase an OEM processor so that I could spend the savings on a better third-party fan. My choice of processor model was also carefully made, around the time there were two directions I could have gone, a faster dual core like an E8500 or slower quad, the most recommended at the time being the Q6600. Whether I made the right decision I will probably never know, but one of the reasons I went for the Q6600 was because of it’s almost legendary overclocking capabilities, but it was always my intention to wait as long as possible before attempting an overclock, definitely beyond my warranty. So here we are…
I won’t be going into the details of performing an overclock, I knew practically nothing before I started and still don’t know all that much now. I will instead point any curious readers to the same guide I used. It’s examples are specific to the Core2Duo and Core2Quad ranges, but it gives a huge amount of information that should help you whatever processor you have. I found it a brilliant guide not only giving me the information I needed to perform an overclock, but also providing lots of extra reading for those who wish to learn more.
When you are about to perform an overclock, it’s often a good idea to set your aim beforehand. This not only holds you back from getting carried away and making silly mistakes, it also gives you something to aim for. I had read that people have managed to overclock the Q6600 up to 3.6Ghz without any effort, this was a little ambitious for me so after reading through the above guide I settled upon 3Ghz. If I could achieve more then I would go for it, but this was a good starting goal.
What I hadn’t realised is how long it’s recommended you test your system after making any changes. A lot of guides suggest only making minor changes, and running a test after each change. That is for exploratory overclocks though, I was treading well worn territory and knew that my aim was easily in reach, it really pays to do research on your components as it could save you a lot of time. I made the various preparations in the BIOS then I was reading to start changing settings, to reach my goal of 3Ghz all I had to do was increase the FSB to 333. I restarted the computer, watching and waiting the whole while, fortunately it booted fine and I was soon running a stress test with Prime95. After several minutes no errors had appeared so I decide to start on the next section of the guide which was optimising the voltage levels. This is where I ran into problems, I had left the power settings on automatic, as the guide advised, but when I tried lowering the settings the computer failed to boot Windows. Looking at the basic settings in the guide I thought I would get away with only changing the Vcore, after all this wasn’t required for smooth running but merely to lower the heat levels. The original voltage level was 1.3250 and my motherboard had options in steps of 0.0250, I took the settings listed in the guide and found the next notch up which gave me 1.2750, after that failed to boot I tried 1.3000 which also failed to boot. In the end I decided to just leave the settings on automatic, the temperatures were well within reason and the guide did make it clear that settings could vary widely between different computers.
I did also attempted to increase the FSB to 400, however this caused my computer to fail to boot and it just kept restarting and cutting out. After clearing the CMOS and re-entering my previous settings the computer started fine. I am almost certain this problem was down to my RAM. There are two settings you can adjust to overclock a CPU, the Multiplier and the FSB, the multipliers on Core2’s are locked with a maximum of 9, which mine was already on so the only setting I could adjust was the FSB. The Front Side Bus is not a part of the CPU but the motherboard which means changing it also impacts on other parts of your system. As you increase the FSB the speed of the RAM also increases, to allow different settings you can adjust the RAM divider which I did as part of the preparations. At an FSB of 333 the RAM was running at 833, just 33 above it’s standard rating, however at 400 it was trying to run at 1000 and I didn’t have any lower divider options to step it down. It’s probable I’m missing something but as my understanding of overclocking is still fairly weak I just decided to take my 3Ghz and run.
According to most guides (including the one I listed) you need to run a full stress test on your CPU before it can be considered stable. This guide recommends running Prime95 for 24 hours, which seems a little extreme to me. This was further backed up by some extra reading I did, some opinions suggest that technically to guarantee stability you should be running the stress test for several days and that it’s unrealistic to require someone to run it for that long. If I were overclocking with unfamiliar settings it might have been a good idea to run it for 24 hours, however I was fairly certain I wouldn’t run into any problems given that I was using already tested settings, and what are the chances that I will ever run my computer at full capacity for 24 hours? In the end I just decided to run it during the day, since my computer is in my room I wouldn’t be able to run it all night anyway, and 12 hours seemed like a more realistic time frame.
Aside from errors you should also run a temperature monitoring program like Core Temps, which I have been using for all my testing. On the evening I performed the overclock I ran Prime95 for a few hours and it never exceeded 61°c, fairly hot but well within what I understand to be reasonable limits for my CPU. (This actually made me more worried about the temperatures I was operating at before, but I shall not mention that again…) The next day I started the test running as soon as I got up, and stopped it just before I was going to bed. Results in the next section.
Last week I mentioned the games I was using to test my new settings, but last week was only about temperature levels, this week I needed FPS measurements from the games which requires a more strict set of testing rules. In each of the games I decided upon my own routes for testing, there are no set tests you can perform in Oblivion or Wurm Online (Crysis does have a synthetic benchmark) but I quite like this fact. Opinions will be split here and I don’t think either side is ‘right’, one side will say that for a fair test you need to replicate the exact same conditions for every run, the other side will say that whilst you do need to keep the situation as similar as possible, having the player perform the test threes times then averaging the result gives a more realistic marker of what you will actually experience in the game.
This test takes part just after the mid-point of the game, there are two reasons I selected it. One, this is the most straining portion of the game, not only is there a lot of action, it’s well known that the ice effects can often be crippling for even high level computers and the aliens also add a lot of blurring effects to the mix. Two, it’s completely on rails.
I chose the second test mainly because it’s probably the most beautiful part of the entire game. This also has the benefit of providing a calmer test with no action, which I would assume is what stresses the CPU the most.
It should be noted that I was running quite a few mods (although no major graphics packs such as Quarls) and decided to just leave them instead of going through the messy process of turning them all off. I ran with exactly the same settings both before and after so it shouldn’t make any impact on the results, it may just mean you won’t get exactly the same FPS as me even on similar settings.
A nice walk in the country from Battlehorn Castle, includes both some close range sections and more expansive views.
Similar to the walk path, except riding on a horse at full speed before stopping to take in the view at the end. Likely to require more CPU power as it moves across the land, especially at the cell transition.
Just the one test here, a simple stroll from my house up and through part of the village, starts off in a large open area then enters a housing district which contains more CPU sapping objects. Due to the time between my first readings and overclocking I decided to run another test using water cooling but no overclock, the landscape of Wurm is constantly changing.
Please note the different tests that have been performed here. Since I performed a second set of temperature tests last week to discover base difference between the Fan and Water Cooler, I performed a further temperature test this week to reveal the change after overclocking. The FPS measurements however are from my original set of tests with the fan, this is not only because there shouldn’t be a difference between the water and fan tests since they used the same settings, but it also gives me an actual comparison of how much better off I am with the new cooler.
Also, I was originally going to include all of the FPS data I collected in some fancy line graphs, but it ended up being too messy so I just averaged out the three sets of results.
These temps were taken as soon as the computer logged into Windows and were run for approximately five minutes. Nice results, barely 3°c difference.
Well that’s a lot of difference… Of course this may be proving that my original theory was correct, that my graphics card is capping any improvement I might otherwise get from the overclock. Guess we’ll need to move on to find out.
Now these are more interesting results, in the Walk test the FPS increases bye as much as 5 frames per second, that’s not really very much but it is a difference I suppose. However the Ride test shows that the CPU increase hasn’t solved my loading issues, so it’s almost certainly down to my hard disk or possibly a mod… Moving on.
Now this is more interesting, firstly it shows that running a second test was a good idea since the complexity of the area has changed by quite a bit. Secondly there is a much more noticeable increase in FPS, as much as 15 frames more than at stock on water cooling.
Ultimately I’m slightly deflated by these results. I’m not sure what I was expecting, but it just seems like very little increase. Okay so I haven’t made that big of a change, only an increase of .6Ghz but…I don’t know I just thought it would make a larger impact.
On the whole I think my first expedition into overclocking went pretty well but, to continue the hiking theme, I did see another huge mountain ahead and if I could only scale it I’m sure I would see a bigger impact. My alterations were pretty simplistic, only dabbling with a small selection of options, so I intend to read up more on the subject and hopefully I’ll be able to better balance my system. I might even manage that 400 FSB I attempted.
The cooler has proved itself pretty well though, a 5°c increase overall seems perfectly reasonable to me, I just wish I was getting more for it. The biggest benefit for me has undoubtedly been the quieter noise levels, even when I was running Prime95 I didn’t notice any audible increase, although the temperatures may just not have been high enough to kick the fans into action. Given what I’ve read of other people’s temperatures I’m sure I must have poor case airflow, no doubt because of all the snaking wires that sit at the bottom… But I’m now happy that at least my CPU is safe.
For reference, here are some more details on my main components:
Processor: Intel Q6600
Motherboard: Gigabyte P35-DS3
RAM: 800Mhz DDR2 4GB