Same idea as having your car race tuned or as this is done today, a new engine managment system.

For PCs the same risks as the car exist. If you put a 1.6GHz PIV in a motherboard that supports overclocking, it might break if you take it beyond 2.4GHz. Ideally you would set the chipset, RAM, PCI and AGP rates to as close as their normal capabilities as possible. Some motherboard manufacturers make this easier than others.

The warranty on any part of doing this can be invalidated. Fortunately many components have protective circuits that kick in before anything fries. However you can lose hard disc consistency when the failure occurs during disc activity. This is more likely with videography as the CPU can be the the most busy and therefore hottest when processing disc based media/timelines.

The rub, well it seems worth overclocking when the manufacturer has already made a CPU with a variable clock multiplier that allows the chipset/motherboard maker to try all the frequencies on that generation of CPU between say 2.6GHz and 3.4GHz. Then for the retailed version the clock multiplier is fixed (by the marketing department more than the engineering department of Intel/AMD). You don't have to change any clocks by soldering or anything. You change the variable MHz clock that the specialist motherboard manufacturer has provided. This for PIV is usually a change in the front-side-bus speed with optionally some higher voltages applied to it and/or the chipset. Also optionally the PCI bus fixed (for 32bit systems) to 33MHz and AGP to 66MHz.

With beyond spec cooling it is possible to run every solid state component in a PC even up to 4GHz with mammoth memory bus transfers and rocket fueled graphics systems. However you sometime wonder why you did it when the equivalent stock machine will achieve the same figures if you just wait between 6 and 18months (depending on the economy mostly).

If you don't understand a word I've just said, overclocking may not be for you. Windows can be unreliable at times, as can Vegas. You might not want any other uncertainties to be introduced (which supercooling/refrigeration can muster if you start chilling the board so that water frosts/melts on it etc).

Apparently some Intel mobos now support overclocking. With the friends this gains, I'm sure Intel lose nothing by doing so. Probably they get more sales this way. However I'm sure they will always state that overclocking is in contravention of their warranty agreement for retail components.

I overclock my out of warranty PIV 1.6 512k cache CPU to 2.0GHz with i850 chipset burning fine at 125MHz and RDRAM at PC1000, 33MHz PCI bus, 66MHz AGP. I won't go faster as I'm using no additional cooling and my DRAM clock generators are only rated to PC900. Having the PC running for a year seems to do what folklore predicted which is to let me push the machine further now that it has been run in. The Abit TH7-II when launched couldn't do PC1000 from the reports I read, not with PC800 units with any amount of liklely overclockability. How do I explain why I can overclock more today than a year ago? I see that it is a bit like the migration on the Serengeti plains. All the stragglers have now been eaten by the predators so that my herd is now faster on average. Or some other physical analogy to do with electron migration going in favour instead of against me.

I wouldn't overclock in warranty or a main machine if I made money on it. Some people do though.

If you don't know what overclocking is, you probably should avoid trying it.

Or at least get more knowledgeable before you do.

Lot's of computer parts/computer resellers have turned off any features you could use to overclock anyway, so this may not even work on your machine. I assume you didn't build your computer.

http://www.hyperformance-pc.com/overclocking_faq.htm


Have a spare board & cpu layin around.

I'd get a little nervous doing this to a machine that renders and encodes. Both of these processes are CPU bound and run for a very long time.
The CPU may be able to get rid of the extra heat into its thermal mass for a short burst but hours of encoding means a lot of heat to dissipate and the core temperature may get too high. A few people here have traced crashes during encodes to cooling problem on stock machine.

To give you an idea, my office gets noticeably warmer when the PCs encoding.

I have a little 1.2GHz AMD T'Bird that I overclock to 1.3GHz (10%). Because of the way I do it, everything on my system is running 10% faster, too--the video card, the IDE buses, and so forth. System has been extremely stable for the past two years, and right now my CPU temp as measured by Motherboard Monitor shows 44C. When doing a major render the temperature rises to about 50C, which is still WAY below what AMD says is too hot.

By way of example, www.tomshardware.com just did a bit of an extreme overclocking experiment--they boosted a 3.2GHz Pentium4 to 5GHz (!) by cooling it with liquid nitrogen! (http://www20.tomshardware.com/cpu/20031230/index.html) Based on the size of the CPU die and the amount of heat it gave off, they calculated that it was equivalent to 1.6 megawatts per square meter! Without the cooling, the CPU chip would turn into a Crispy Critter is a fraction of a second.

Today overclocking is safe, fast and simple. IF your motherboard and processor support it.

Just to give the newbies an idea how "complicated" it is.

I recently (several months ago now) build a ASUS P4P800. To overclock it, all I had to do was reboot, enter setup before the PC went into Windows, go to the advanced screen, select the overclock option I wanted as a percentage, (several presets) reboot. Total time 30 seconds. Yes today it is that simple.

Risk? About the same as crossing the street. In 99.99% of the times the only thing that can go wrong is your PC won't boot or hang entering Windows. If that happens, shut down, (push main power button) reboot, go back into setup try a lower setting.

The danger from overclocking is in the past you did have to be a guru, or just say what the heck and hardwire some pins on the CPU/motherboard. Today you don't need to to that.

So the downside is very miminal. The upside is your PC will be anywhere from 5-30% faster. Realisticly, expect maybe 10-15% tops unless you have a very stable MB and some very fast memory.

I got as high at 25% overclocking but it wasn't stable that high so like I said, I just dropped it back and its been rock solid stable.

Yeah, BB's right about the 10-15% thing. I have an ASUS A7V8X-X MB and an AMD XP 1800 (~1.5ghz). In the bios these is overclock settings (for both multiplier and bus). I oc's the buss from 133 to 145. Anything above 145 was giving me unstable results (mem is only rated 133). If my computer didn't boot, i just hit reset and it defaulted to a 100mhz bus. :)

Anyway, you only need to be a guru if you want to OC an AMD wiht the multiplier too. At the factory AMD cuts some connections on the chip. These are the multiplier locker. Re-connect these and multiplier OC's is possible.

I've been OCing since I built my first 486. As BB and others say, there's little risk in actually doing the procedure nowadays. The real risk is getting too greedy and OCing too high and leaving it there. Then the system is unstable and crashes (usually with you & a client right in front of it!). But if I tested a setup well (say, a 5 hour overnight render), it's never NOT remained stable.

My 2500+ Athlon came off the same assembly line as the 3200+. It just wasn't tested as high. But since many 2500's are labelled as such to fill orders (cheap processors are always the biggest sellers), many WILL do 3200+, with no heat or stability difference from a "true" 3200. Mine does 3100+ stably. It'll do 3200, but was fudgy, so I backed it down 3%. Now it swings.

I remember BillyBoy saying that all the processors in a family come off the same line and they clock them based on their stability. So if my P4 3.0Ghz is the same as a P4 3.2Ghz, I assume it failed to stabilize at 3.2Ghz but passed at 3.0Ghz and that’s why they labeled it as such. Wouldn’t it be risky to try and run it a 3.2 since Intel didn’t think it could run at 3.2?

BTW, BillyBoy, you are the main reason I built my own PC this time (it was my first!). There was a post you made a while back that convinced me to do it and I just wanted to thank you. This puppy screams! (…even my wife is impressed, and it ain’t easy to impress the wife anymore) ;-)

~jr

I'd love to see how Intel does do the tests to determine how they rate one chip 'better' than the next one speed wise. The test may be on their web site, but the Intel site is so big its hard to find stuff and there is no real reason for Intel to make it public since they in effect would be shooting themselves in the foot if they spill the beans. The higher they rate a CPU the more money they make.

What a lot of people still don't realize is with just a little effort and the right MB you can get the same or better performance from a lower rated chip (like a P4 2.4 or 2.6 by just overclocking it and keeping several hundred dollars in your pocket compared to what a 2.8 or 3.0 chip would cost, that you didn't overclock.

By no means am I expect on this topic. There is a relationship of the memory used your little RAM sticks (their clocking) which determines how high you can push the FSB (front side bus) which more or less determines how fast the main CPU can access memory. In theory, the faster the CPU can make the round trip to and back from RAM, the faster your PC can function because the wait state is reduced.

If you want to go more extreme with your overclocking (not recommended) check out some of the web pages that review specific motherboards. These guys play around all day seeing how far you can push things. So just because your MB says it just supports 3200 RAM, doesn't mean it can run with 3600 or 3800 RAM, (if you can find it) because that will alllow the FSB to have a still higher multiplier factor. Again, that's similar to turbocharging a engine. You're just making things run faster.

If you get too extreme then heat can be a problem and you start to risk burning up the CPU since to get that super extra speed out of it you're going to have to increase the voltage to the CPU and they are very sensentive to just fractions of a volt more.

Again, each chip is different, so if you're really lucky your particular CPU chip may be able to handle a extra quarter volt or so and if combined with a super fast CPU to start with and ultra fast memory, it could really fly. You may be able to approach a 4.0 GHz, maybe more. I doubt it would be stable though.

Wow that was a quick response. ;-) I purchased Kingston HyperX PC3200 DDR400 but they claim they tested it up to DDR433. I’ll have to try OC’ing it a bit to see what I can get. I could have gotten PC3500 memory for $6 more but it was CAS 3 and the PC3200 I got was CAS 2 (2-2-2-6 to be exact) so I went with the CAS 2 memory. I tried the Sundance rendertest and got a time of 1:35 so I’m happy with the system.

~jr