Bios upgrade, overclocking

[nexus123]

Hi I'm trying to overclock an intel celeron 566, but the bios won't allow me to change any of the settings. I was wandering if anyone new of a bios flash that would allow me to tweak the front side bus and voltage. The PC is a hp pavilion 6645c I can't find the mother board model but the order part no. is d994a ne help would be appreciated,


Thanks, Nexus

[Cobra]

Nexus123,

AIDA32 will identify your motherboard for you so you can get a BIOS update:

http://www.aida32.hu/aida32.php

I remember having a Dell machine, and that would not alow me to touch any of those settings.

I haven't been able to update the guide recently, but I did write one for AMD overclocking. The same rules should apply to your processor. Remember than Intel quad-pump FSBs, so a motherboard-side FSB of 133MHz will amount to 532MHz (~533MHz) chip-side. Here's the guide, see if it helps:

https://www.videohelp.com/forum/viewtopic.php?t=213341&highlight=overclocking+guide

Good luck with your overclocking!

Cobra

[nexus123]

Thanks much man the guide I'm reading applies specific to my processor straight out of the box. It say I should be able to get it up to 901mhz I believe. Nething hirer would freeze it at boot. Neways appreciate the n-4mation.

[glockjs]

is overclocking a celeron a really bad idea since they run hot already???

[Cobra]

The guide can say that your chip will get to a certain speed, but this can't be guaranteed. You may have a core that has few mistakes on it and so can clock much higher, or you may have one that has a lot of imperfections and will never be stable or even able to reach those speeds. Overclocking is very CPU-specific.

As glockjs says, you must watch your heat with Celerons. However, you won't make any more heat if you don't increase VCore to regain stability.

Cobra

[gmatov]

Back in the days when the 500 mhz CPUs were the thing, the Celeron, garbage though it was said to be, with low levels of cache, all the o'clocker sites gave it top marks for o'clockability.

500 at 800, or more, no heat problems, stable, in many cases you HAD to raise core voltage to stabilize, got too hot without.

Go to www.tomshardware.com , or www.anandtech.com , check the o'c guides, maybe go to the archives to see what was done with the older CPUs.

Some here are relative newcomers to o'c'ing. They're starting with 2 gig machines already, Soft BIOS, just punch Page Up-Down keys to do it. Your machine is probably mostly jumpered. Gat a good glass and a light and the pic of the board, find the pinout clusters you need to reconfigure, make sure you jumper the right ones.

By all means, see if you can get a BIOS flash. At the time, 566 might have been all the board was capable of. May have raised it to 800 or so through BIOS flash.

You may have to go to Compaq (or whoever you said) to ID the board, as one I had here locked up tight each time I tried to run Sisoft Sandra to ID MoBo and components. Not sure, but I think they write into the BIOS to oat allow you to run an ID program. Or else it was a hell of a coincidence. All the other modules worked.

Good luck,

George

[Cobra]

George,

in many cases you HAD to raise core voltage to stabilize, got too hot without.

Raising VCore is what produces more temperature. Increasing clock speed has very little bearing on temperature. By increasing VCore, you are forcing more current through a complex circuit with a high resistance, and this in turn will waste more energy as heat.

As for overclocking more modern machines, you're right. It is a fairly unskilled procedure. That's why so many more people are doing it, it's just that some of those settings/the horror stories that go around scare some people off.

Nothing ventured, nothing gained! 8)

Cobra

[mike1061]

Why would you overclock? I have heard of this, but I don't understand it.
Thanks Mike.

[Cobra]

This is the same link as in my signature, to an overclocking guide:

https://www.videohelp.com/forum/viewtopic.php?t=213341

Overclocking is the process of driving computer components above the standard specification. This yields extra speed, but can also reduce the operational lifespan of components. In most cases, a mild overclock will gain you extra speed and not cause any undue stress on your components.

Cobra

[flaninacupboard]

Raising VCore is what produces more temperature. Increasing clock speed has very little bearing on temperature. By increasing VCore, you are forcing more current through a complex circuit with a high resistance, and this in turn will waste more energy as heat.

You sure that's right? i mean when you say current do you mean as in amperage, because obviously that's not the same as voltage. i do vaguely remember somoene telling me after OCing their XP2500 they the voltage to lower temps. makes sense i suppose, when you increase clock speed it's going to cosume more wattage, you can either increase the voltage or the amperage.

[gmatov]

Cobra,

Sorry, no, lowering the voltage causes a current rise. Raising it causes the current to fall. Current rise causes higher heat. Exactly why a fuse or circuit breaker can protect a circuit.

Heat trips the breaker, and current is the cause of the heat.

For instance, a 14 guage wire can carry 15 amps safely, must be protected by a 15 amp breaker. It doesn't matter if the circuit carries 12 volts, or 12,000 volts.

Cross country electrical transmission lines carry, as of a few years ago, 765,000 volts, in a wire of a given size, safely, because the current was within the limit of what the given wire guage was capable of, without heating up, as resistance is what causes heat, a result of volts X amps. E over I divided by R, or any permutation thereof.

Those 50 + million transistors in the core of the CPU probably traverse thousands of feet, if not several miles, of "wire". The longer a "wire" is, the more resistance there is. Resistance is measured per foot.

Just for the hell of it, you are aware, are you not, that if you did not buy the very top of the line CPU when it first is released, you have bought a "defect", a "second"?

If a 3.2 ghz is the very best, and you buy a brand spanking new 3.1, you have bought a CPU that has failed its test at 3.2? And if you buy a 2 ghz, with the same date of manufacture, you have bought a CPU that has failed at a dozen or more steps.

Should give you pause.They do NOT want to make 2 ghz CPUs on a 3.2 design. 500 buck CPU has to be sold for 80 bucks, or thrown in the trash.

It has been this way since the first CPU was made, whatever, 8086, 8088, 286, 386, 486, and AMD does the same. Better 7 bucks than the garbage can.

So, if it failed at, say, 2 steps up, and you O'C 3 steps up, you have an unstable system. The spec for that particular processor is written AFTER it has failed the testing stage.They do not tell you nor the CPU to raise the voltage, nor lower it, by 2 tenths of a volt to get it to run reliably.

They simply say it WILL run, if there are no glitches with the rest of your system, at the voltage written to the CPU..

"Speed for Free" is a great rallying cry, but most of us shoot for a stable system.

As for the original poster, he HAS to go to one of the O'C sites, see what others have done with the SAME board and other parts and CPU.

Again, the Celery CPU is a good overclocking candidate. Just don't expect miracles.

Flan, glad you see the point, some "electronic engineers" gave me hell a few months ago, saying there is "no resistance" in a CPU.

Cheers,

George

[indolikaa]

EDIT: Never mind.

[gmatov]

Indo,

Haven't the foggiest why you should post as you did. Agree, or no, the only options are yes or no.

Cheers,

George

[Cobra]

OK, I've had some sleep now and I'm looking at it a bit more fresh.

when you increase clock speed it's going to cosume more wattage

Got it. I took a look at the thermal specifications for the AMD chips and it is true. I should have realised this - the chip is doing more "work" and so must consume proportionately more energy in order to carry it out.

George, I'm not a physicist or an electrical engineer. You obviously have a grasp of what's going on here - fair enough. I think I was confusing raising VCore and frequency (looking back at my references) producing more heat and attributing it to the voltage increase.

Just out of interest, probably later today when I've moved back to uni, I'll take my XP2500 back to stock speeds and measure the temperature at 1.65 --> 1.85V VCore and see what difference it makes. Should be interesting.

Just for the hell of it, you are aware, are you not, that if you did not buy the very top of the line CPU when it first is released, you have bought a "defect", a "second"?

I learned about this years ago when I was thinking about overclocking a PPC603ev processor from 180MHz to 200MHz. I didn't do it in the end, knew very little about messing with computers and I didn't want to solder the wrong thing. I was aware, though, that the companies aim for high-speed chips and the ones that have less-than-perfect cores aren't stable at the highest speeds, so they're clocked back and back until they are. However, we're seeing a bit of a lag with AMD and their chips are actually very capable of overclocking to higher speeds with no trouble.

"Speed for Free" is a great rallying cry, but most of us shoot for a stable system.

My system is just as stable as a standard one. Reckless overclocking and poor testing will result in a system that craps out on you when you ask it to open Internet Exploder. It is a dodgy business, but a fun one to dabble in!

Anyway, that's that resolved. I'll need to edit my guide now, and it's your fault! :P

Cobra

(Edited numerous times because I can't spell and keep forgetting stuff)

[gmatov]

Cobra,

I'd advise against boosting the voltage just to see what happens. You actually may burn out your CPU.

Electromigration, or shorting to another trace, either of which will cook your CPU. Electromigration is the actual flow of the trace material, aluminum in a CPU, to junctions or transistors, building up to create a short circuit.

The main reason they are using lower and lower voltages is that the circuitry is getting packed so closely that even 1.8 volts or so will jump from line to line, just like the static spark that you get when you get out of the car. The silicon between traces is so thin that these voltages can jump from one to the other.

I am not an engineer nor a physicist, either. I have to assume that the reason the increased voltage at increased frequency is safe is because OF the higher freq, falls so quickly, as say, a 2800 over a 2500.

When you consider the distances we're talking about, it's almost unbelievable they work at all. 90 nanometer, 90 billionths of a meter in width, and the space between, the "insulation", probably the same or less.

When you consider that a transformer, such as the coil in your car, has copper windings with an enamel insulation of about a thousandth of an inch, and does not short from wire to wire, at say 27000 volts, it gives you some idea of how thin the silicon between traces is, to not be proof against less than 2 volts jiumping across conductors.

I, personally haven't been o'clocking sinse the 266 Pentium, and the 450 AMD. Might have been cheaper to have done so, but I just keep on moving up, 3 steps or so from something like the newest, maybe less, as I am at 2400XP at the mo', but there is a show this weekend, might be closing in on 3000 soon.

DVD Shrink and Autoshrink seem to go so damn fast, however, there's no real need for more.

Cheers,

George

I really am sorry about the guide. But there really are some good sites for o'clocking. www.overclockersworkbench.com is 1 if I don't have the url wrong.

[Cobra]

I really am sorry about the guide.

I was just joking! It's A Good Thing that I know this now, because I can update it and make it correct. Thanks for taking the time to explain my mistake - I do appreciate it.

My CPU is already running at 1.825V because it's not a good core and isn't stable at higher speeds without more voltage. It's perfect at 2.07GHz @ VCore 1.825V.

I did find out about electromigration, and mentioned it in the guide. Have a look, and follow the link - there's a scanning electron micrograph of a circuit that's been broken by the migration of metal ions. It's a good example. I don't mind testing stuff out on my CPU - I want to replace it soon with an XP Mobile CPU so I don't care what happens to it.

Thanks again,

Cobra

[mike1061]

gmatov
Is right about the voltage to amperage rule. Simplified, OHMs law states that voltage and amperage are INVERSLY proportional. That means, if all other factors remain the same, a change in one creates a reverse change in the other.
I do not know anything at all about overclocking, but I am an electrician. I also am leary of increasing the voltage to something above its rated voltage. I have, mostly by mistake, connected an appliance to the wrong voltage. Almost always with bad results. Something fried (burned). If you connect an appliance to an under voltage condition the results are very similer, again something will fry (burn) due to the increase in amperage.
Thanks Mike

[Cobra]

Mike,

We're only talking about small increases in VCore here, nothing very drastic. The chips can take it, and if something does go wrong most motherboards have a thermal trip that cuts the power to the CPU when it goes above a certain temperature (mine's is 85 deg C). You're right though, you can't be reckless.

George,

I did that test to see how VCore affect CPU temperature. You were almost right, it caused very little rise in temperature. Here are my results:

My CPU was not overclocked - it was a standard XP2500+ Barton for this test. All measurements are in DEGREES CENTIGRADE and I used CPU Burn-In on max heat generation to test both voltages for 1 hour.

VCore 1.650V (STANDARD)

Room = 19.5
Motherboard max = 28
CPU max = 52

VCore 1.850V (+200mV)

Room = 19.4
Motherboard max = 28
CPU max = 55

This gives the result of +3 deg C for 200mV increase in VCore. Therefore, for my CPU, you get a 0.015 deg C per mV rise in temperature. That's negligable.

Would this be specific to my processor, or common to all Bartons do you think?

Hope that was interesting,

Cobra

[flaninacupboard]

i would guess different scores for different cores. try again after lowering it to 1.3V and see what happens!

[Cobra]

I'd probably crash at VCore 1.3V - I have proven that I have a quite imperfect Barton core!

Cobra

[gmatov]

Cobra,

Go vice versa, raise the freq, leave the voltage alone, you might see a larger jump in temp, just shut down quick.

You are not doing more "work" with a voltage increase. You are running the CPU at overvoltage, which is the same as running a 120 volt lightbulb at 220 volts, or 240, if you want to be technical. You will blow the bulb.

Go to AMD's site, I really dont know if they give you a core voltage + the Amperage, but if you do, and they do, calculate it.. Or, vice versa, volts, watts, amps. If it draws 70 watts at 1.5 volts, or whatever, check the next higher CPU core speed. If it draws 70 watts at 1.52 volts, then the current draw will be the same. Etc. Keep your current within their limits and you should, not guaranteed, mind you, be OK.

They know what the traces will bear.. Hell, they're trying with all their might to get them thinner and thinner so they can get to the next stepdown in trace size, which is what all the hullabaloo is about, 135 nanometer to 90 nanometer, that's billionths of a meter in width for the aluminum "wiring" in that tiny piece of silicon.

Cooling is paramount in all electrical wiring when current gets high. Most wires are rated "in open air", meaning that there is air circulating around them. In conduit it is less. In a CPU, the heatsink has to do it all. And a very good fan.

Cheers,

George

[Cobra]

You are not doing more "work" with a voltage increase.

I know, I said we're doing more work with a rise in CPU clock speed. I don't know if "work" is the correct word for this, but you know what I mean. It's doing more processing, therefore it requires more power in order to do this. This generates extra heat... I think!

The first link in my guide covers all the thermal and electrical specifications of AMD processors. Here's the link (it's a .PDF):

http://www.amd.com/gb-uk/assets/content_type/DownloadableAssets/K7_Electrical_Specific...on_Rev_ENG.pdf

Here's what we're interested in:



Why is it that the processors with a higher clock speed have the same thermal output, until you get to the XP3000+ (FSB333), then it drops again with the XP3000+ (FSB400)?

Cobra

[gmatov]

Cobra,

I think that is the most they can get out of that D suffix CPU, and pushing it at that.

If you go back to the site, check to see if the new 3000 and 3200 are not the new interim sockets. They were supposed to/did go to a 700+ pin socket, and have a 900+ pin socket in their future. More pins, less "wiring" in "length" per pin pair or cluster, or however the circuits are arrayed, 32 bit processor, I guess 32 pins, and maybe grounds, 64 bit, same, more "circuits".

If you look at the last column, Amps, you see what I was speaking about, also. Those traces are microscopic, yet the current is stupendous. For that total current, a socket 462 is probably 231 circuits, wire pairs, each managing not to vaporize under load of a 1/3 Amp.

Also why they tell you you WILL cook a CPU in seconds without cooling attached. Even before any built in heat shutdown can work.

Work is a good word. At higher freq, we're doing the same amount of work in less time ( for instance, encoding a video ) or more work in the same amount of time.

I'll have to go take a look, myself, now, you've got me checking. Aroused my curiousity. Hell, I thought I got over this stuff years ago.

Cheers,

George

[Cobra]

George,

Overall, am I right in thinking the following?

- A rise in clock speed will cause a rise in temperature due to the increased amount of "work" being achieved in a set time, consuming more power in this set time as compared to standard

- A rise in VCore gives stability, does not raise temperature (well, it does but it's so tiny we can ignore it) but can contribute to electromigration when excessive voltages are applied

- Temperature is the main factor behind electromigration

The guide I'm writing doesn't have to pack all this information, just the essential facts laid down nice and simple. However, I find this very interesting so your input is most welcome.

Thanks,

Cobra

[mujahid7ia]

Don't they have programs in which you can change your fsb etc. within Windows, no BIOS needed. I think SpeedFan...

I tried it on ine, but my OEM Compaq mobo was not listed, and even if it was, what clock generator do I have, i don't know....

[Cobra]

There is an nVidia control for their nForce motherboards that allows you to do this. However, I have heard a couple of horror stories. I'm sure it's quite safe to use though, horror stories and bad reputations seem to spread very quickly.

Did you know that some MSI motherboards have dynamic overclocking built in? When you ask your PC to do something particularly stressful, the motherboard actually overclocked the CPU slightly on-the-fly to get it done quicker. I can't remember which model this was on, but Tom's Hardware Guide found out about it and MSI had a bit of a fit!

Cobra

(Edited once because I'm half asleep)

[mujahid7ia]

http://www.nvidia.com/object/sysutility_1.08.5.html

nVidia System Utility, but my Compaq is not nforce ... maybe i'll try it anyway.

I have used it on my nforce2 ultra 400 Msi k7n2.... works fine, but I prefer BIOS, more options, more advanced...

EDIT: anyone know about speedfan soft OC-ing?

[mujahid7ia]

not too many options:

[gmatov]

Cobra,


- A rise in clock speed will cause a rise in temperature due to the increased amount of "work" being achieved in a set time, consuming more power in this set time as compared to standard

Actually, it is doing more "work", drawing more current, but the voltage is too low, which is what causes the current to increase. That given CPU is going to draw XX watts, which is a product of Amps X Volts. If you do more work without increasing the voltage, the Amperage will rise, the heat output will rise.

A rise in VCore gives stability, does not raise temperature (well, it does but it's so tiny we can ignore it) but can contribute to electromigration when excessive voltages are applied

Actually, raising the voltage will not raise the temp, and that is not so much what causes electromigration, it's heat FROM excessive current, which is from insufficient voltage.

I just bought an Asus board today, to rebuild one of my 2400XPs. Think I will try some o'cing, just to keep from giving you any BS in my replies. Probably should go with the 2000XP, but think a 2400 might just be a better core. Might check at O'C board to see what the rating is on my stepping, or steppings.

Cheers,

George

[Cobra]

Actually, it is doing more "work", drawing more current, but the voltage is too low, which is what causes the current to increase. That given CPU is going to draw XX watts, which is a product of Amps X Volts. If you do more work without increasing the voltage, the Amperage will rise, the heat output will rise.

OK, got it.

Actually, raising the voltage will not raise the temp

I prove earlier in this thread that it did, but only by a tiny amount (so much so it's negligable). Don't know why, but that's what happened.

I just bought an Asus board today

Good man - Asus have always been my choice for all the PCs I've built for myself and others. Solid, reliable and good performing. A good choice. Which one did you get?

Thanks for the explanations, I'm glad we've got it cleared up. Do you mind if I quote some of your explanaions in the guide?

Thanks, and good luck with your overclocking!

Cobra

EDIT - I don't know if these are the "steppings" you're after, but if you click through on the link in my signature and look at the bottom for the first link it'll open a .PDF from AMD telling you the exact specifications of every K7 CPU.