overclocking & undervolting tualatin at the same time?

dan · Post by **dan** » Thu May 20, 2004 2:58 pm

currently my celeron tualatin is 1100 running on a measely 100mhz FSB. it is UNDERVOLTED on my SOYO TISU to 1.07 volts, which SISandra reports at 14 watts.

so it is fanless. it is quiet.

i'm kinda adventurous, i would like to OVERCLOCK to 133mhz FSB (SOYO TISU has a weird bug that the onboard sigmatel audio won't work when overclocked. It must be either 100mhz or 133mhz fsb. also, you get weird checksum error defaults loaded when you OC)

but i still want cool and quiet, hence i want to UNDERVOLT. my calculations suggest i need to undervolt to 1.2-1.3 volts to keep it at 20 wattage range.

my calulcations suggest that at default, 1.475 volts, my overclocked cpu will put out around 30+ watts, which i probably cannot cool passively, i probably will need a fan, spoiling the silence of my machine.

anyone have experience with this? i don't need it prime95 stable, only stable to surf the web, and check email, and write stuff on word.

between the two choices of faster or more silent, i would prefer more silence, but i want it all baby!

sthayashi · Post by **sthayashi** » Thu May 20, 2004 3:43 pm

dan wrote:anyone have experience with this? i don't need it prime95 stable, only stable to surf the web, and check email, and write stuff on word.

If it's not Prime95 stable, then it's not stable.

dan · Post by **dan** » Thu May 20, 2004 4:23 pm

LOL
well i've not run prime95, but i've never had it lock up on me, it never crashed or never froze.

my current set up being celeron tualatin 1100 @ 1.07 volts on a SOYO TISU, FANLESS

sthayashi · Post by **sthayashi** » Thu May 20, 2004 7:32 pm

The prime95 torture test performs intense calculations that should result in predetermined values. If your computer generates an answer that is NOT in line with the predetermined results, then it cannot be considered stable.

Most let the torture test run for 24 hours in order to confirm stability.

Edward Ng · Post by **Edward Ng** » Thu May 20, 2004 7:38 pm

Erhm; if U add a very low speed quiet fan like an undervolted AF80CT, you can get much, much more performance and still not hear it from anything over a foot away like I did with Sigma One, which is a sticky in the gallery. I mean, isn't it worth it to run at 2G? Oh well, guess I'm the only one who thinks virtually silent it as good as truly silent...

-Ed

PS I Prime95 all my machines for minimum 48 hours straight before declaring, "it's stable."

Putz · Post by **Putz** » Thu May 20, 2004 7:57 pm

dan does have a point, though. It can still be stable at lower temperatures without being "stable" as you've described it, Edward and sthayashi. It might not meet your criteria for what you're comfortable with, but it's his choice.

That said, I also agree with you that I wouldn't be comfortable with that situation. For instance, Word has sometimes decided to consume 100% CPU on me under certain circumstances... if that happened when I hadn't saved in a long time, and my CPU overheated to the point of hanging the machine, I'd be pretty upset. Nevermind that I'd have trouble sleeping at night knowing that at any minute, my system could contract a virus or run into an infinite loop that would cause the CPU to overheat and potentially cause a fire (that risk is very low, but I'd worry about it anyway). Anyway, my point is that I agree that if you could even just add a 5-volted Panaflo L1A in the machine somewhere, it would still be essentially completely inaudible, and have much, much greater cooling headroom.

sthayashi · Post by **sthayashi** » Thu May 20, 2004 9:47 pm

If it fails a Prime95 torture test, then there is proof that there is at least one series of instructions that will cause it to spew an error when it shouldn't. And at that level, you can't predict what would or could happen to your machine. Everytime a program or even windows crashed, you could never be sure whether it was software-related or not.

You wouldn't choose a file at random on your machine and just change one byte, would you? But if it fails a decent prime95 test, then you've opened yourself up to that possibility.

nutball · Post by **nutball** » Thu May 20, 2004 11:16 pm

sthayashi wrote:If it fails a Prime95 torture test, then there is proof that there is at least one series of instructions that will cause it to spew an error when it shouldn't. And at that level, you can't predict what would or could happen to your machine. Everytime a program or even windows crashed, you could never be sure whether it was software-related or not.

You wouldn't choose a file at random on your machine and just change one byte, would you? But if it fails a decent prime95 test, then you've opened yourself up to that possibility.

Yeah, but it's his machine. "Stable" isn't an absolute term, it's a relative term. Different people mean different things by "stable". As we're talking about his machine, we should use his definition of "stable".

Post by **Rusty075** » Fri May 21, 2004 5:38 am

Dan, my advice would be to just try it and find out!

In the time spent waiting for someone who doesn't really know your hardware to guess at whether it would work....you could have known for sure.

Just keep an eye on temps, and if it gets scary hot, or crashes, then shut it down.

I don't think Prime95 is the end-all-be-all for stability, personally. According to that logic your car shouldn't be considered "drive-able" until you've driven it as fast as it can go for 24 continuous hours. If I was selling/giving the system to someone else I would require that kind of stability, but if it doesn't crash during your typical use, I'd call it good. Data integrity is probably about the same as any other Windows-running machine.

Post by **Ralf Hutter** » Fri May 21, 2004 5:50 am

I'd certainly try it at 133MHz. I've had a lot of experience with Tualerons and PIII Tualatins and have found that they will always run at much lower than default Vcore at default speed, and even run quite well with smaller Vcore degreases when they're being OCed. I've never seen a 1.1A that won't run at 133MHz at default Vcore, and most will run it at 1.35-1.40V.

I'd also mirror Edward Ng's advice about installing a very slow speed fan on your existing heatsink. A very slight amount of air can make a huge difference in your CPU load temps without being able to be heard outside the case. My normal PIII/Tualeron cooling setup was a Thermalright SLK800 with a 80mm or 92mm 5V Panaflo L1A as a cooling fan. This sort of thing would reduce the load temps by 15-20°C! And I'll gaurantee that you won't be able to hear a 5V 80mm L1A from outside the case.

sthayashi · Post by **sthayashi** » Fri May 21, 2004 6:37 am

Rusty075 wrote:I don't think Prime95 is the end-all-be-all for stability, personally. According to that logic your car shouldn't be considered "drive-able" until you've driven it as fast as it can go for 24 continuous hours. If I was selling/giving the system to someone else I would require that kind of stability, but if it doesn't crash during your typical use, I'd call it good. Data integrity is probably about the same as any other Windows-running machine.

Since I've owned cars that have stalled when sitting in traffic, not had power up hills, and couldn't break 40MPH, I would say that a 24 hour drive is a damn good way of making sure your car is stable

But not the bit about driving at max speed.

But I do agree with your statement to Dan.

Rusty075 wrote:In the time spent waiting for someone who doesn't really know your hardware to guess at whether it would work....you could have known for sure.

dan · Post by **dan** » Fri May 21, 2004 1:38 pm

hi,
well let's suppose i take the trouble to download prime95 and run it and it fails. so what? does this mean i shouldn't be leaving my computer undervolted to 1.075 volts? i use my computer for email, word processing, and web surfing, and photo/music stuff/game stuff about 1-2 hours daily. it never crashes on me.

anyhow, SISandra reports that my cpu's power consumption is 14 watts.
If i increase core voltage, with my heatsink fanless/passive, it would increase power consumption, which may make it unstable due to excessive heat!

anyhow i bought my SOYO TISU b/c www.anandtech.com called it the "overclocker's motherboard" but actually it doesn't overclock well.
it has a bug where you get a "cmos checksum error defaults loaded" from what i've read you must disable the BSEL1 pin. also, overclocking disables the onboard sigmatel audio.

from my experiments, going from 1100*100 to 1450*132mhz i need to increase vcore from 1.075 to 1.475, default, but then it freezes, presumably due to heat problems.

Edward Ng · Post by **Edward Ng** » Fri May 21, 2004 4:44 pm

Dan, this is the reason why I recommended trying a very low speed fan; it will be virtually unnoticeable, even in the quietest of conditions (I'm assuming you don't work with your computer within 12" from your head, but then again, maybe I'm wrong on that front), and it will go an extremely long way to cooling your system. Most likely you do not need to raise your voltages; just the extra cooling will finally stabilize your machine that much. And even if it's only 99.9% stable after mounting the fan (let's just say you're about 97.5% stable, right now), raising your voltage maybe 0.025 or 0.05 will round out that stability figure to a solid 100%, without a perceptible change in noise (again, assuming the computer isn't like, in your face or right next to your ear, AND your room has to be dead silent, anyway, at that rate!), and probably still run cooler than it is now.

I understand the novelty of a 100% passively cooled system, but virtually silent active cooling is truly, significantly easier to achieve, and will not change your, "user experience," one bit. Take Sigma One as an example...

Honestly, would it kill you to give it a try?

-Ed

Post by **Ralf Hutter** » Sat May 22, 2004 6:00 am

Edward Ng wrote:Dan, this is the reason why I recommended trying a very low speed fan; it will be virtually unnoticeable, even in the quietest of conditions (I'm assuming you don't work with your computer within 12" from your head, but then again, maybe I'm wrong on that front), and it will go an extremely long way to cooling your system. Most likely you do not need to raise your voltages; just the extra cooling will finally stabilize your machine that much. And even if it's only 99.9% stable after mounting the fan (let's just say you're about 97.5% stable, right now), raising your voltage maybe 0.025 or 0.05 will round out that stability figure to a solid 100%, without a perceptible change in noise (again, assuming the computer isn't like, in your face or right next to your ear, AND your room has to be dead silent, anyway, at that rate!), and probably still run cooler than it is now.

I understand the novelty of a 100% passively cooled system, but virtually silent active cooling is truly, significantly easier to achieve, and will not change your, "user experience," one bit. Take Sigma One as an example...

Honestly, would it kill you to give it a try?

-Ed

Dan - he (Edward Ng) really is right about this. I've played around a lot with passively/quietly cooling PIII Tualatins and have independently arrived at the exact same conclusion.

My suggestion is to throw a 5V L1A on your heatsink and play with it/listen to it for a while. I think you'll be pleasantly surprised.

dan · Post by **dan** » Sat May 22, 2004 1:58 pm

well toms hardware at
www.tomshardware.com/cpu/20020103/celeron1300-08.html -

states that FSB 100mhz is a severe limit on the tualatin's performance.
but since i only use my computer for email, web, and word, and occassional games, i'm unsure whether i want to break the BSEL1 pin, which is presumably a permanent modification, since coating it with nail polish didn't do the trick.

i guess i can add a fan. is there a way for me to ad pics to this?

Post by **Ralf Hutter** » Sun May 23, 2004 6:05 am

dan wrote:well toms hardware at
www.tomshardware.com/cpu/20020103/celeron1300-08.html -

states that FSB 100mhz is a severe limit on the tualatin's performance.
but since i only use my computer for email, web, and word, and occassional games, i'm unsure whether i want to break the BSEL1 pin, which is presumably a permanent modification, since coating it with nail polish didn't do the trick.

Going from 100MHz to 133MHz results in a very nice performance increase in the Tualerons. I think a 1.1A running at 133MHz is a great way to have a fast, quiet system. Get yourself an SLK800 and an 80mm L1A, run the fan at 5V and the CPU at 133MHz and you'll be in heaven.

dan wrote:i guess i can add a fan. is there a way for me to ad pics to this?

Sure, but you need to have your pics hosted somewhere first. Then you just enclose the direct link to your picture in a set of these brackets: .

So it would look like this:

[img]www.yourhostingsite.com/yourpic.jpg[/img ] (w/o the space in the last tag)

Techno Pride · Post by **Techno Pride** » Sun May 23, 2004 6:15 am

I think burnmx.exe is a better stress test for your processor.

My Cel 1.2 passes Prime95 and burnk6 @ 1.6GHz, but fails burnmx.

dan · Post by **dan** » Sun May 23, 2004 11:47 am

thanks Ralph,

i know Tomshardware and anandtech recommend the athlon xp and duron, but both have 2x the power consumption @ same MHZ as the
Tualatin.

"and you'll be in heaven"

LOL

Post by **Ralf Hutter** » Sun May 23, 2004 12:02 pm

dan wrote:thanks Ralph,

i know Tomshardware and anandtech recommend the athlon xp and duron, but both have 2x the power consumption @ same MHZ as the
Tualatin.

"and you'll be in heaven"

LOL

Yes, that's true but you've already got the board and the CPU. Why switch horses in the middle of the stream?

dan · Post by **dan** » Sun May 23, 2004 9:04 pm

oh i'm not switching horses midstream, i'm sticking with TISU/tualatin.

Originally, i tried fingernail polish but that didn't disable the BSEL1 pin!

so how do you break the pin without damaging adjacent pin?
well the OVERCLOCKERS.COm forum, one guy suggested using super thin pliers. i went to Radio Shack, and their thinnest pliers were not thin enough. another guy suggested using an eye of a sewing needle.

i eventually plan to do this. Once I break the BSEL1 pin i cannot go back to 100mhz FSB.

i plan to post on a new thread, but my newest project is how to solder two heat sinks together.

Michael Dooley · Post by **Michael Dooley** » Mon May 24, 2004 8:03 am

Originally, i tried fingernail polish but that didn't disable the BSEL1 pin!

Perhaps using the fingernail polish on the socket would work. Alternatively, have you thought about using heat shrink on the pin? Something extremely thin mind you. Good luck dan.

dan · Post by **dan** » Mon May 24, 2004 4:55 pm

um, using it on the socket was something i thought of but i would think that is as irreversible a change as breaking the pin.

i...uh...returned the fingernail polish at walmart LOL.

what's heat shrink? i'll post my results.

Michael Dooley · Post by **Michael Dooley** » Mon May 24, 2004 5:55 pm

Fingernail polish can be reduced or cut or weakened or even removed by using acetone - usually. The label might tell you which type of solvent the nail polish or pigment was "floating" in. If I'd face4d the same challenge, I'd have coated that teeninesy pin with several coats of the nailpolish. Then when inserting the chip I might try a very small drop of lubricant on that one pin socket so that the polish was not scraped off as you shoved the chip into the socket. I'm assuming that others have done this trick and succeeded.

Heatshrink is tubular stuff that is usually used to insulate an electronic joint of one kind or other. You solder the wires together, but in preparation, you have previously positioned a 3/4" or 1" long tube of heat shrink on one of the wires. After the solder has set and you are satisfied that the connection is okay, you slide the heatshrink over the connection and apply a source of heat. The heatshrink then shrinks to fit. Its nifty stuff.

Perhaps it's available in a small enough diameter at a place like Radio Shack? I've never done this particular thing but offered it as a possibility.

Good luck dan.

dan · Post by **dan** » Tue May 25, 2004 9:09 am

but is it easy to remove the heatshrink?

if not, then i might as well break the pin. there are no pliers at radio shack slender enough, but on overclockers one person recommended using the eye of a sewying needle

nmuntz · Post by **nmuntz** » Tue May 25, 2004 11:36 am

I thought I'd throw in my 2 cents regarding my experience with Tualerons. I have a 1.3GHz part (default is 1.45v). At the rated speed it would not run reliably at anything below 1.40v (weird errors when doing heavy compiling). When clocked at 866 it would run at 1.30v. This was using a BH6 and Upgradeware Slot-T adapter.

dan · Post by **dan** » Tue May 25, 2004 8:45 pm

overclockers.com does have a list of overclocks and none of them are lower than default voltage, but the ASUS TULC does not allow undervolting from what i heard, the SOYO TISU does. also the TISU, unlike ASUS tends to undervolt, as anandtech points out.

also i'm only interested in internet/web/email/word processing/game stability, not prime95 stability.

it takes me over an hour to take apart the pc so i plan to break the BSEL1 pin AND solder/heat glue two (or more) heatsinks together at the same time.

dan

Edward Ng · Post by **Edward Ng** » Tue May 25, 2004 8:53 pm

I'm assuming you're soldering two sinks together to save money, right? Because it's still possible to find Slot One coolers on the market. I personally recommend thermal epoxy over solder, however.

-Ed

EDIT: Here's one place selling Slot 1/A coolers, and this vendor gets extremely good marks at ResellerRatings.Com.

dan · Post by **dan** » Wed May 26, 2004 8:06 am

Hi edward NG

i'm surprised you recommend heat epoxy over solder, as solder is metal.

oh, while my aopen 440bx used slot one, the soyo tisu is 815ep step B, and uses socket 370 so i don't see how a slot one cooler will help

Edward Ng · Post by **Edward Ng** » Wed May 26, 2004 8:14 am

If it's a socket, rather than slot chip, why are you soldering two sinks together?

I recommended epoxy because you can apply epoxy to the entire mating surface; I am not sure how you're attaching the sinks together (surface to surface, flat, or edge to edge?), but if you're going surface to surface, with solder, you can probably only seal them at the edges, because of how solder works (you can only solder in a line, much like welding); with the epoxy, you can apply a layer to the entire contact surface area, not just the outside edges; this would ensure superior transfer from sink to sink.

I mean, if you're only going edge to edge, then by all means, good old hot solder or welding should be at least as effective. Of course, I've burned myself quite nicely while soldering on past occasions (not that I'm very good at it...), but never caused any self-harm with epoxy. It's also probably a good thing I've never attempted welding.

-Ed

dan · Post by **dan** » Wed May 26, 2004 8:57 am

hi ed,
thanks for the super-fast response. my lunch time is almost over but i'll check SPC tonight.

oh well 1- SOYO TISU has a bug that when you overclock a celeron TUALATIN you get "cmos checksum error default loaded" so i don't know if it can overclock since the bios resets it when i try to overclock.
ironically, ANANDTECH called this the overclocker's motherboard LOL.
actually it's a terrific UNDERVOLTER"s motherboard.

2- over at overclockers and soyo message board the solution is to disable the BSEL1 pin. i wanted to disable it REVERSIBLY using a thin plastic tube but none i've found is thin enough as the i370 pin holes are very thin, and red fingernail polish didn't work, so when i do break the pin, power consumption will go up

3- currently i have it undervolted to 1.07 volts (bios set at 1.1 volts but TISU undervolts). It is FANLESS. it is passively cooled and "stable".

well when I overclock, POWER CONSUMPTION WILL GO UP.
I STILL WANT FANLESS PASSIVELY COOLED. So, I plan to glue heat sinks together. More heat sinks = more cooling.
1- due to more surface area 2- due to the fact that it takes more thermal energy to heat a given amount of mass, m.

if it is unstable, as i define stable, then i'll be forced to add case fans, as breaking the BSEL1 pin is a permanent overclock.

overclocking & undervolting tualatin at the same time?

overclocking & undervolting tualatin at the same time?

Re: overclocking & undervolting tualatin at the same tim