Squeeze some extra Mhz out of your TB800

  • Thread starter Thread starter Guest
  • Start date Start date
  • Tags Tags
    CPUs
Toggle sidebar Toggle sidebar
G

Guest

Guest
Now I know that all of you are going to think I'm crazy, but I've done this with TWO processors and it's worked great both times. I've got a standard setup: Thunderbird 800, Tyan S2390 mobo (yes there are overclockers without Asus and Abit motherboards!) Crucial PC133 CAS2 memory, etc. Here's what I did to push my TB800 to 912Mhz (114Mhz bus)

1. Use a big heatsink (duh) but hook the fan up to some sort of a toggle switch you can access while sitting at the keyboard.
2. Un-overclock the chip back to factory settings.
3. Jump to the temperature gauge (I would not recommend doing this in any OS, but only in the BIOS)
4. Turn off the fan (!!!)
5. Watch the temperature until it reaches about 170F. Hold the temperature there (this will require turning the fan on and off in bursts) for as long as you can stand. Note that the chip *will* function after being this hot! I talked to an AMD tech personally and he said that the chips all leave the factory with 190F as max temperature.
6. The longer you hold temperature at 170, the faster your chip will go later on. I held mine for 30 minutes.
7. After you've "burned in" your chip (literally!) I would not recommend just turning the system off. Turn the fan on and let the chip come down to normal temperatures gradually. After being in such extreme conditions you wouldn't want to crack anything!
8. After getting all of your crazy peltier, water cooler, or whatever (I just have a 3dfxcool SocketAHO) back together and your chip running at -29C or whatever you like (mine runs at 111F) start bumping your bus clock up. Those of you with "overclocking" mobos will probably do better than me. I can almost guarantee at least 112Mhz bus stable.

Now I can see why you will think me crazy...I mean, overclocking means cold, right? Well I have no idea what happens to the chip when it gets hot like that, but it took an un-overclockable TB800 (I couldn't even POST at 102 bus) all the way to 912Mhz.

Known side effects:
1. PCI bus clock is unusually high...no HDD errors, but my first SB Live started popping and crackling. Had to send it back.
2. NIC quit working while overclocked. Probably due to bus speed.

Known remedies to side-effects:
1. Heat sink on SB Live chipset.
2. Heat sink on NIC chipset.

Well, I hope that this tip helps some of my fellow low-budget overclockers out there! Not all of us can afford crazy peltier and water coolers. I couldn't even afford a GlobalWin FOP-32. But you can see that we, too, can overclock.

-Skaven
Webmaster, Lunar Power
http://www.lunarpower.org
AOLIM: SkavenK
 
Very interesting! Thanks for the info!

I am just wondering, though..... How many times have you tried this?

----------------------
I don't hate Intel............ Do I?
 
Duplicate...I've done it twice! I burned up my first TBird (NOT doing this...I was a freakin' moron and accidentally turned my system on without a heat sink) So I bought another one (the one that wouldn't POST at 102Mhz bus) and burned it in, too. The results were spectacular. After burning in, I popped the chip out...no discoloration, burn marks, peeling, or any other signs of damage. Like I said, the cores are designed to take up to 190F before they start frying.

-Skaven
Webmaster, Lunar Power
http://www.lunarpower.org
AOLIM: SkavenK
 
Sounds like "snake oil" to me.

Where did you learn of this technique? Is there any scientific reasoning behind it?

It's interesting, though. It's also dangerous. Due to the differences between motherboard temperature sensor placements, sensor inaccuracies, and sampling rates in monitoring software getting a spot-on reading is difficult. Your 170 degrees (76 degrees C) is getting awfully close to the AMD spec'd maximum of 90 degrees C.

<P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 02/09/01 05:24 PM.</EM></FONT></P>
 
Well, as for "scientific reasoning" the only thing I can come up with is that the intense heat somehow "opens" the core's etching up a bit or something. I'm not an electrical engineer...however I do know the scientific method. I believe that my results would prove consistent all things considered. The machine stayed constant between the two settings (i.e. no new RAM to go with the overclocking) and the fact is that only after the "burn in" was I able to overclock beyond 102Mhz bus. Even more telling is that I went from terribly unstable at 102 to rock solid at 114. The only thing that changed between these was the burn-in on the processor. I do admit that it's close to the borderline. I chose 170F to account for any error in the temperature sensor and such. I'm also not saying that 170F is the "magic" temperature. That just happens to be the one I used. The key behind the tip was that running your chip extra-hot for a few minutes apparently does somthing to improve overclocking performance. Also note that I don't suggest running the processor without a heat sink at all...simply with the fan turned off. Because the heat sink is still attached and radiating heat, the temperature increase is very gradual; it took mine about 5 minutes to go from 98F to 170F. Once it reached that temperature, just bumping the fan on and off threw the temperature back to 160.

I'm also not sure what you mean by "snake oil" as I've never heard that phrase before, but if you're suggesting that I'm making all this up as some sort of a prank to get people to fry their chips...you'd be wrong. It's just that I know there's a lot of people out there who don't have the advantages of an Abit or Asus motherboard and a water cooler, and thus need to approach overclocking from the bus rather than the multiplier. My technique appears to improve the viability of bus overclocking on thunderbird 800's.

If you feel that 170F is too hot, try it in gradual steps. Try burning in at 140F for 30 minutes and see if you get an improvement. You just might...I really don't know. Anyway, I've got TestCPU and WCPUID screenshots of my system at 912Mhz at the following links:

http://calypso.spaceports.com/~power2k1/912.zip
http://calypso.spaceports.com/~power2k1/wcpuid.zip

The temperatures on TestCPU were before I got Arctic Silver paste and when I had my CPU fan on a 7V supply. The WCPUID screenshots were taken just now, and had the processor at 1.9V, the CPU fan on 12V, and arctic silver paste in between.

-Skaven
Webmaster, Lunar Power
http://www.lunarpower.org
AOLIM: SkavenK<P ID="edit"><FONT SIZE=-1><EM>Edited by skaven on 02/09/01 09:33 PM.</EM></FONT></P>
 
This is all very intersting but one thing is buggin me most of the 800 that I have seen and heard of will do 950 so you may not be changing the cpu but affecting the mobo with your proces, though I have had better overclockin sucses after letting the cpu run a default for a couple of days first.

SANDMAN
 
Sorry, "snake oil" was a bad choice. I just realized that had you understood the reference it could also be misinterpreted.

Snake Oil is a catch-all term for false medications sold by dishonest traveling salemen. It was claimed to have magical properties and would cure anything but in reality was little more than castor oil and foul smelling additives.

Incidentally, in this analogy I was not implying you were the seller of the "snake oil" but rather the unwitting buyer.

I apologize to you and all who misunderstood the term or my intent.

I still am not convinced that there is any validity to your procedure but as I said, it is interesting and thanks for posting it.
 
Jesse, I agree, but this could prove a lot more helpful in overclocking. Imaging using a combination of multiplier/fsb overlocking to get to a real high FSB and overall clock speed? PERFORMANCE, BABY!

----------------------
I don't hate Intel............ Do I?
 
You got to admit, though, it does sound nice to be able to OC the AMDs from both sides.
1: by this supposed "conditioning" to bumb the FSB
2: to use the Abits and Asus's to finess the multiplier

I'll give it a try on my 750 this weekend since it isn't doing much now.

This new forum sucks
 
I've seen this technique described at other forums as well at there seems to be a basis for it. I just got my 1K Tbird in the other day and ran it for 48 hours with "CPUBurn" and no case fans to drive the temp up. I did this at max voltage
at stock speeds. Right now on my first overclocking attempt I'm runnin 1210(11x110) while runnin Prime95 & a DVD at the same time to test stability. Tommorrow I'll try higher FSB and see what happens. All in all, I think this process generally "gets the proc used to" running beyond spec.
ASUS A7V 1005d bios
Tbird 1000@1210
Crucial 256meg PC133 2-2-2
ASUS V7700 Deluxe
SB Live! 1024
Netgear FA310TX
Don't Knock it till you try it 😉, Laterz.

"I yam what I yam!"
 
I believe this has validity.. I NEVER really believed in burn in before but a couple weeks ago I bought some new ADDITIONAL crucial ram and installed it and my system kept locking up .. off and on.. so once i got it on.. I just left the computer on all night with a couple minor things running.. just enough to flex the muscles a little but not over do it.. after this "burn in" .. my computer doesn't lock up anymore from the new ram..

I believe that by taxing the cpu.. you can achieve the same added stability.. Its electronics.. they work like that.. sorry i don't have more scientific numbers to prove it!
 
yeah, and when I want my computer to run faster I kick it. The harder I kick the faster it goes

Update: I just found out that using special connectors to the cpu will boost performance about 30% at the same clock. No need to overclock anymore! Additionally, using the same method I have my ram running with 35% increase performance (same method works with video card as well). I call the method foot-to-chip. As of now, I can only produce this effect once per unit. I am working on a way to use foot-to-chip method on cpu,mem and vid card at the same time. This requires case modifications since you need to get a full swing in there.
<P ID="edit"><FONT SIZE=-1><EM>Edited by m_kelder on 02/27/01 02:22 AM.</EM></FONT></P>
 
You shouldn't make someone laugh with their finger on the mouse button, the screen goes ape sh*t!
 
When it comes to OC, the CPU FSB is not the limiting factor but the chipset itself. The burn-in process is not new for me but I’ve never really believed it. Well, maybe it works but as I said the CPU FSB is not the limiting factor. Any ATHLON/DURON CPU can take a 133Mhz DDR (effective 266) FSB as long as you lower the multiplier and as long as the chipset is prepared to it (boards like the MSI K7TPro Turbo, the Asus A7V133, ...).
An ATHLON CPU that only OC from 100x8 to 102x8 is very strange. The worst I’ve ever seen was a CPU that could only OC 50Mhz (a lot more than 16Mhz!)
There must be something else. I don't now what!


OC is nice but risky!
 
I wouldn't recommend trying this. There is absolutely no validity to this post. Heating up the core to extreme temperatures will not cuase any physical changes other than tiny fractures in the traces. The term "burn in" is used to describe a test period before shipping a unit. It doesn't actually have anything to do with the temp of whatever it is you're testing. I am betting you had something loose or misconfigured in your initial setup, which was fixed when you installed that switch onto your CPU fan. The heat of the core has NOTHING to do with this. I worked for 3 years doing line and technician work at 2 elecronics manufacturing companies, one a power supply manufacturer (which can be touchy when they have to run within tight specs) and nothing good comes from cooking integrated circuits.

/Athlon-1.2GHz@1370MHz(137MHz*10)/Asus_A7V133/
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom