Building a PC, Part III - Overclocking

This is the article series that make me love this blog!
It’s exactly as I am: a developer with HW passion.

Althought I was a PC builder before becoming a professional developer, I still love assembling and tweaking with HW, Router and all those things that are not-strictly-development stuff.

Please, don’t hesitate to continue with this kind of post…

Coding horror is THE BEST Blog I’ve ever read…

Igor A.!

Core Temp only appears to report Core Temperatures (obviously).

You might want to also take a look at SpeedFan:
which will report and graph other readings (such as case and NorthBridge temperatures), fan speeds, and hard drive SMART statistics.

More accurate than your “built-in Mark I finger” and (as it’s freeware) it’s a bit cheaper than a fancy-shmancy laser temperature gun (which I don’t think can give accurate readings anyway, as you have to remove the case side panel to use it, thus possibly significantly altering the air flow).

I’m about to build a new developer rig in an month or two so these posts are really useful - good balanced as well, not too techy and not too trivial.

I’ve really enjoyed this series of articles – it’s nice to know that I’m not the only developer around who likes to (a)understand what’s going on under the hood, and (b)have as much control over the hardware as possible.

The over-clocking bit is interesting; I’ve done it before, but not on any modern system, and the tools have certainly advanced. I’m curious, though, about how much real performance gain you get from the extra 400MHz you’ve tweaked from the chip – is it worth the extra heat and noise?

I’m personally in the “under-clockers” category: the bottleneck in my development and testing is almost always I/O (RAM, disk, or network). I’ve been running an Athlon 2400+ (1.9GHz, IIRC) under-clocked for a few years, and I still only rarely peg the processor during dev activities.

By under-clocking, I was able to reduce the heat output by about 15 deg F: just enough to push the requirements for the CPU fan down. The system is so quiet that one can barely hear it running: the noise from the HDD seeks is as bad as it gets (well, except for the occasional use of the optical drive – those are loud bastards).

In terms of performance, I’ve gotten the best gains from using proper pairs of high-speed RAM, using the RAID-1 feature of my SATA controller, and using quality networking components (esp. good switches).

Hey Dan,
ive got a P5B-e ( :stuck_out_tongue: ) i never tried past 333mhz, what kind of temps do you get reported in the bios, and on what cooling.
I have an artic freezer 7 (not pro) and i max about 53c(reported in bios) with the fan at 800rpm and a couple of case fans.

Also, is your Northbridge REALLY hot?

One thing with overclocking, is that its always really interesting to see what people with similar/same hardware managed to reach, as i now feel small and inferior :frowning:

As a software developer, the term ‘irresponsible’ comes to mind when reading this post.

There are many cases where you just might get away with overclocking, for certain loads. But you can’t conceivably cover every permutation out there. - some real life information on the impact of overclocking that developers face as a result of doing it. As Raymond says at the end of the article, Moral of the story: There’s a lot of overclocking out there, and it makes Windows look bad.


the reason jeff was running Prime 95 (on each core) was to check that his oveclock was STABLE, the guilty overclockers in that article were the mom and pop stores that were overclocking, then obviously not checking them under load/after a period of time.

Meccano, nice board :slight_smile: You have to avoid “auto” in your voltages, is the reason of the northbridge going hot.
Fix it to one of the lower voltages and you’ll see it works great.

The same with memory and cpu voltages, I have CPU - 1.1625V, memory - 1.9V and chipset ( there is only one option in my board ) to 1.8V if I remember well.

And when going over 333Mhz, you have to fix the memory latencies. I have 4-4-4-12, but I want to go over 400Mhz I have to use 5-5-5-15 and the temperature go out of hand.

Anyway, summer it’s not the time to experiment with overclocking :))


Like i mentioned i do undervolt my CPU, but i will definitly try setting a specific Northbridge voltage :smiley:
Im from scotland, so excess heat really isnt an issue


If you want to quiet down the HDD seek then I recommend a Samsung SpinPoint drive.
I’ve got a P120 and it’s pretty quiet.
Though strangely SilentSeek was turned off by default on my drive and had to be turned on using the HUTIL tool:

SilentPCReview like them too:

Jeff, perhaps you could run a programmer type of benchmark to see how valid overclocking is. Something real world like compiling a large project using Visual Studio.

It would be interesting to see what differences, if any would be seen.

I have a sneaking suspicion that the bottlenecks in the system are all I/O related.

But on the positive side I am sure it would give another FPS or two to Quake or Doom, or whatever the current shooter of the day is.

Great 3 part series. It made me want to dabble in the diagnostics and overclocking aspect of building a PC from scratch. I’ve shied away from it in the past.

And I agree with Davide. For the sake of the development theme of the PC and your Blog, it would be cool to see compile times on a large VS2005 project differ between the overclocked CPU and non-overclocked.

That machine is a BEAST. I’m jealous!

Great articles! Been a fan of your articles for a while now and that intel chip is impressive. I have only one problem with the rig, using a secondary manufacturer for the motherboard. I’ve always had those things toast very quickly, whereas straight intel on intel motherboards run forever. Is there a reason you chose to go with a secondary motherboard manufacturer?

Anyone heard how to increase the memory rating on the windows experience index?

I have 5.9 for all but memory which is at 5.1.

Jeff, have you done any over clocking on the video boards and using the latest drivers, don’t think they are certified yet. You should easily get a 5.9 on the gaming graphics.

In line with Robert, I’ve found that GCC and any given multithreaded video encoder are much better torture tests than prime95. It’s a great starting point, don’t get me wrong! But gcc reveals hidden instabilities, because it’s memory heavy and works so many different bits of your cpu at once - MSVC might be good for this too, I haven’t used it as much. Since gcc’s a pain to set up on windows initially, another option is Handbrake or DVD-RB, free dvd encoders, which will exercise the vector and memory/cache units excessively. So many doom9 threads start with “I overclocked but it passed every test” that end with “I lowered the fsb and everything works now!” (But everyone OCs for video encoding, cutting 2-8 hours off is always worth it. Except us poor laptop savages.)

Jeff, this series of articles has been very helpful. I’ve got another build coming up and appreciate browsing your component choices and picking up a couple new tricks.

I second the question about revisiting the “suspicious” memory score. Please. :slight_smile:

Thanks! -bill

I whole-heartedly agree this was a very informative article right up there with the stuff from Tom’s H/W! I thoroughly enjoyed the read! Didn’t know if you were going to play with the memory stepping of the RAM and see if that could be tweaked back up to where it belongs? Definitely, waiting on the silencing part to come.

As some one who sold hardware . . .


By the way, Core 2 Extreme processors leave the multiplier unlocked. Great utility for overclocking.

I’m suprised the computer isn’t quiet enough for you using the Antec case?
I’m running a similar build using an older Antec P180 case and it’s so quiet the only way you can tell it’s on is when you insert a disk into the dvd drive since spinning up a disc is the loudest noise it makes!