Overclocking my Athlon 3000--expert opinions needed!

Hi everyone,

I'm relatively new to the hardcore overclockers scene but a veteran of the hobbist overclocker scene--the most overclocking i've done is jump the mhz of a cpu by X amount--thats about it.

I recently got an athlon 3000 barton and i got a killer 84 cfm tornado fan and artic silver stuff, everything neat!

In anycase, I have an asus a7v333 and i can overclock the bus speed (166mhz) by increments of 1mhz.

Forgive some of my n00b statements if you see any--but I was under the understanding that the biggest restriction in overclocking was CPU temperature. With this great new heatsink and uber-fan, regardless of the voltage setting on the cpu and the mhz i brought it up to (brought it up to about 198x13, or 2574mhz) the temperature never got above 142 degrees F. (it freezes due to heat around 160-167). So..I was really shocked to find that around 2450mhz (stock is 2167) my computer would boot just fine, keep a safe temperature, but it would lock up the second I started doing something 3d wise, (my test was to start 3dMark2003)

I didn't know what the problem was, still not sure. I have pc 2700 memory which i think by default is rated at 166mhz--and putting that up to 175-200 seems logical that it might be the cause of the problem, but also i hear that i'm overclocking my pci and agp ports? If that is the case, that would explain why the second my 3d card started to run, it froze on me.

So here is what i've done to remedy the situation:

I purchased one stick of pc 3200 memory which can run at 200mhz+.(from corsair)--that is to make sure my ram isn't giving out. Meanwhile, I bought a pen with 'condusive ink' on it to unlock the multiplier so I wont have to jack the AGP/PCI bus speeds into crash-levels.

Am i on the right track here? Am I missing something that you guys see--overlooking something?

I would appreciate any input on it, I'm pretty new to this but hoping to make due with what i have.

Thanks!

E

More than likely your motherboard has a 1/5 pci divider (I am not positive so you will probably want to verify this). with a 1/5 divider that means that at a 166 fsb your pci slots will be running at 66mhz (166/5) and your agp will be at 66 (double pci). Running a 198 fsb with a divider of 1/5 will get you around 40mhz on your pci which is danger territory for alot of hard drive corruption and some video cards don't like that.

Also your motherboard chipset itself might not be able to run at those speeds, especially if you don't have sufficient cooling on the northbridge chip.

You are on the right track but if you are going to get into serious overclocking I would just recommend that you get an nforce2 motherboard (such as the Abit NF7-S). You won't need to mod your chip to use different multipliers(as you would have to with your current motherboard I believe), the rev 2 boards are speced to run fsb of 200 (but many can run a good chunk over that) and they have a locked agp/pci bus speed so you don't need to worry about nasty stuff like hard drive corruption.

Yes, PC2700 RAM could do about as much as your mobo can handle since it doesn't have a PCI lock. Between 166 and 180ish FSB.

Unfortunately you will not be able to take full advantage of PC3200 RAM with your motherboard.

190 FSB / 5 is 38 and just about everyone here agrees that higher than that would be running your hardware out of specs to much risking hard drive corruption.

Yes 3000+ XP is probably easier to overlcok but just about every T-Bred B and Barton average 2.3 GHz. Extra 100 MHz is possible with luck and other components being excellent. 2.5 GHz is rare.

Your options are:
1. 12.5 or 13 multiplier times 180ish FSB
2. Purchsase an nForce2 mobo which can handle well over 200 FSB.

Prime95 Torture Test your final overclock for 12 hours.
You'll find that 24xx MHz more often than not fails the stability tests after booting into Windows, although it may take 8 hours of Torture testing to fail.

Thanks for the info Death

Now--I'm not going to be able to purchase a new mobo--sticking with what I have for a good long while. I think you're definately right about the 1/5 pci divider, in the bios it looks like 166/33, etc and goes up.

Now am I right--and i've read this on toms hardware--that by unlocking the multiplier you don't have to increase the bus speeds of the agp/pci slots--thus its safer? I saw some benchmark there too which showed that the 166x15 was great--can you explain more throughally the purpose for unlocking the multiplier?

Thanks a ton!

E

ekidhardt said:

Thanks for the info Death

Now--I'm not going to be able to purchase a new mobo--sticking with what I have for a good long while. I think you're definately right about the 1/5 pci divider, in the bios it looks like 166/33, etc and goes up.

Now am I right--and i've read this on toms hardware--that by unlocking the multiplier you don't have to increase the bus speeds of the agp/pci slots--thus its safer? I saw some benchmark there too which showed that the 166x15 was great--can you explain more throughally the purpose for unlocking the multiplier?

Thanks a ton!

E

Click to expand...

Using the multiplier to change the CPU speed, is usefull when your chipset wont do anymore FSB. I dont think you need to unlock barton multi's do you?! (I still have a 1700+ )

Yes, the PCI lock depends on the FSB, not on the multiplier.

Therefore, since the expected stable overclock is between 2.3xx and 2.4xx MHz, you're probably looking at 12.5 or 13 multiplier which you have so no unlocking is needed and 18x FSB, where x is between 1 and 9.

"Using the multiplier to change the CPU speed, is usefull when your chipset wont do anymore FSB. I dont think you need to unlock barton multi's do you?! (I still have a 1700+ )"

" you're probably looking at 12.5 or 13 multiplier which you have so no unlocking is needed"

Its a 13x multiplier, however, the instability in my computer is due to the bus rates being increased too high by bringing it near 200mhz fsb--so by using a higher multiplier, I can keep those agp/pci rates at an operable range.

quote from toms hardware

"Our benchmarks showed that the Athlon XP 3000+ at standard clock speed (13 x 166 MHz = 2166 MHz) can't hold a candle to the P4 3.06 GHz together with the latest applications. The only exception is in UT 2003, where the AMD CPU clearly took the lead. Once the Athlon with the Barton core is overclocked to 2500 MHz (15 x 166 MHz), it can hold its own or even beat the P4."

So I want that 15x multiplier basically.

By unlocking multiplier and making it 15x, I'm looking at 166mhz bus and leaving the pci/agp settings to 33mhz (166/5).

and I wonder if I can go higher even

am i missing anything here?

E

1st point: don't trust Tom's. They spin results so it makes Intel look better.

Your multipliers are already unlocked and I believe you have access from 13-18x ish on the A7V333, which is a board I used to have (crappy revision I had, AGP controller fails at 166MHz). Stick with the 166 MHZ bus speed and increase the multiplier and voltage as necessary, but don't let temperatures get too high under CPU-intensive programs.

Oh!

I never checked, now i feel like a bonehead

lol..oh well, thanks for the info l337--you know what i noticed though, my temp never got high no matter what setting i had it at--even at max voltage it was approx 140 degrees--couldn't really get it higher--it freezes around 165-170, go figure.

Thanks again for the info, we'll see how it goes, i think i'm learning here!

E

Heh 170F? That's loads hotter than my Pally ever got - keep it under 120F just to be on the safe side, plus it may keep your CPU running for longer. Electron migration takes place faster at higher temperatures.

What heatsink are you running? You mentoned the fan and TIM but not the lump of metal =X

I'm using the thremaltake800a i believe, just got it--seen at the picture here

http://www.hardocp.com/image.html?image=MTAyNDY0NDg2M0pkQzJ2VkdmZ0ZfMl8yX2wuanBn

I believe its the best air heatsink available, weighs 3-4x what my last one did (stock fan)--so I think my set up is decent.

If you don't mind, i've got an airflow question too--I've got an 80mm front fan intake and a 120mm exit fan--along with the badass 84cfm tornado on my cpu--i'm sort of thinking of adding another fan for intake on top of my computer--useful?

I also noticed after putting the sides on the case that my temperature raised by about 4 degrees--not bad.

With stock fan/heatsink and radio-shack goop (i know i'm ashamed) i was runnning standard at 140 degrees and 135 on cooler days--the second i put the new one in, it went to 104--and the bios says 120 give or take 1-2.

Anywho, thanks a bunch for the advice!

E

if you put a fan on top of case, it should be exhaust, for heat rises, also, do the math on how much intake and how much exhaust you have, try and keep them even, some fans(especially power supply fans) aren't that powerful, so if the exhaust is too high, air flow could be restricted through the p/s, also, you don't want too much air being blown in, then you get pressure, more pressure = more heat

Whether positive or negative pressure actually causes temps to rise is debatable, just try to keep both intake and exit the same number of CFMs.

You should have plenty of airflow provided that your fan grilles are not too restrictive - like some you get on punched-out cases.