WIP: Karaktu's (and Zap's!) Unofficial Guide to mobile Celeron overclocking success!

[nickdigger]

If it only lasts 1 minute in Prime95, it will affect your daily use. You can find the temp limit from the datasheets at support.intel.com. Most of the p4-478's seem to say 100celsius, FWIW.

 

[buzzly]

Finally had a chance to try my 1.6 Mobile Celeron on my main company Asus P4G8X. With the bended bin mod, the board recongized the CPU and set it to 1.25v. The 133fsb mod did not have any effect on the board as it recongizes it as 100fsb 1.6 mobile Celeron.

However, the highest stable speed was around 175fsb=2.8ghz @1.65v. At first I thought it was the CPU, now I believed I had exceeded the limits of this board--- max. 533fsb.

 

[Zap]

YMMV when it comes to overclocking. Besides, what other CPU will let you get even that close to a full 2x overclock? The ones I have can't hit 3.2, but some can get pretty darn close. Oh yeah, one exception was a 2.4GHz mobile Celeron that can hit 3.2GHz with detected 1.6v and set to 133MHz FSB. It was a D1 core, IIRC.

 

[rogue1979]

I have messed with a dozen or so mobile celerons of various speeds. Only two were able to reach 3.2GHz. Both were 2.4's which defaulted to 1.6v on a P4S800-MX, and both hit over 3.4GHz.

 

[sandrogalli]

It doesn't make it through prime95 but I wasn't too sure about the generic ram i have in there. The chip made it fine through pcmark and a good few hours of prince of persia 2. I was worried about the temps, but its at 183fsb now anyway and that seems to have taken them down a few degrees. You're right, the extra 2hundred mhz isn't worth the hassle.

The funny thing is, it did do 200fsb at default voltage when I first got this chip, and ran for a few months without a hitch. Does anyone else know a chip to tire like that?

 

[GL Joe]

Hello all! I've been following this thread and picked up a mobile CPU to try out myself. Forigve me if I missed it earlier in the thread, but has anyone tried any of the boards that have a 478 socket but are based on the Intel mobile chipsets?

I have a mobile p4 (not p4-m) 3.2 HT I picked up. Also arriving tomorrow is an Asrock P4915GM-Dual, a micro atx board with the 915mobile chipset and a socket 478.

WHile I haven't been able to find where anyone has used this board for a Mobile P4, I am hopeful it will boot up and recognize the CPU since it IS a 915 mobile chipset.

It also has PCI-E and AGP support, though it looks questionable for supporting really high end cards. For me, if it support my X1600 Pro PCI-E then it will be a real boon for my HTPC.

 

[Zap]

I couldn't find info on that specified motherboard. The only socket 478 mATX "dual" series motherboard I could find at ASRock's site is this P4Dual-915GL. It is a mATX board for socket 478 that uses the Intel 915GL chipset and has "AGI" slots for PCIe and AGP. The 915GL chipset does not officially support AGP nor PCIe 16x. It does have one or multiple PCIe 1x lanes. Likely the PCIe 16x slot is wired up to a number of 1x lanes, possibly resulting in 4x operation like many unofficial SLI board's second 16x slot. Also, the physical AGP slot is likely wired to PCI. These are limitations for sure, but still it would be interesting to see what this board can do.

Keep us updated. :thumbsup:

 

[slugi]

Lets hope that P4Dual-915GL is better than Asrock P4Dual 880Pro, i got problems with pci-express cause it is works only @ 4x. Also OC was impossible. SuperPi 1M with M Celly @2.4GHz was 1,24 minutes with tweaked XP pro...

But just got new rig up with new 18$ mobo... prolly works with Cellys either.
Abit IS7-V P4-M 1.9GHz

And addy to coolers Arctic Cooling C S-Silent 4 ULTRA

 

[osage]

OK so I got a Mobile Celeron 1.8 here on the forums with the bent pin for desktop mobo, put it in a Soyo P4X400 Dragon Ultra mobo. cpu boots up as Celeron 1.8A, but the voltage defaults to 1.18v.
The mobo has a wide range of voltage adjustments, but any changes I make won't stick, board won't post, or just falls back to 1.18v.

I have dropped wire into the socket to try to get 1.6v, but it dosent work, so just get another mobo, or is there something else I can try ??

So any idea what course of action to take ?

 

[rogue1979]

Do you know if the cpu will post without the bent pin?

Sometimes the wire hoop in the socket can be quite frustrating, you have to try several times. Did you use the guide from this thread and make sure that you put the wire in the correct place?

Is there a bios update for your motherboard that addresses core voltage compatibility or issues?

 

[osage]

no idea if the board will post without the bent pin, I was going to try bending it back, but it broke off in the process.
yeah I tried the wires 4-5 times. bios has the options for voltage, but I think they just won't take without the AE1/VID4 pin

 

[Zap]

You can use a wire in the empty hole plus the VID hole next to it to simulate an unbent pin. 1.18v is probably the 1.2v of the bent pin with a bit of undervolting. Undervolting is pretty common on motherboards. You may need to look up information on the mobile CPUs as they may not match the pinout of the desktop chips, thus making the wire voltage trick not work properly.

 

[rogue1979]

The VID location is the same for the P4 desktop and the Mobile Celeron. However, the pin out values are different. Connect VID2 and VID3 for 1.60v on the Celeron.

 

[osage]

I will give the wire in the empty hole plus the VID hole next to it a try, got nothing to lose.

thanks

 

[osage]

"got nothing to lose."

OK, so that theory is full of holes.........I had a working system and now I don't.

put a wire in the AE1 hole plus the VID hole next to it and it wouldn't post... removed/reinstalled the cpu a few times but no go. pulled the wire out and was just going to be happy with a cheap 1.8 Cel combo, but now it won't post at all.

I gave it the usual chants and recited all the magic phrases to no avail. stuff happens.

the board was always a bit flacky every now and again when it would boot up the onboard nic would not initalize, and the temps always seemed very high for 1.18v 43C at idle in XP, I got the same readings in the bios, Winbond Hardware monitor and Speed Fan.

I knew it was a crap shoot when I got the board, but it was fun to play with for a few days.

 

[nickdigger]

Your 1.18 was probably 1.175 (VID = 11011). Grounding VID4 (01011) would give you 1.575, which shouldnt have killed your cpu. Connecting VID4 + VID3 would still give you 11011, since VID3 = "1".

 

[osage]

I'm not really thinking it's the cpu at this point, I lean more to the mobo being the problem.
Not sure however, may have to try the cpu in a known working mobo to find out.

 

[osage]

well I got it to post and boot again, left the jumper to clear cmos for a hr or so while I had a :beer:, pulled all the wires and it fired up.
As soon as I change voltage in the bios or by wire in the socket it won't post.

Lame mobo I guess, but at least the cpu and board work.

thanks for all your suggestions, and tips.

 

[GL Joe]

Originally posted by: GL Joe
Hello all! I've been following this thread and picked up a mobile CPU to try out myself. Forigve me if I missed it earlier in the thread, but has anyone tried any of the boards that have a 478 socket but are based on the Intel mobile chipsets?

I have a mobile p4 (not p4-m) 3.2 HT I picked up. Also arriving tomorrow is an Asrock P4915GM-Dual, a micro atx board with the 915mobile chipset and a socket 478.

WHile I haven't been able to find where anyone has used this board for a Mobile P4, I am hopeful it will boot up and recognize the CPU since it IS a 915 mobile chipset.

It also has PCI-E and AGP support, though it looks questionable for supporting really high end cards. For me, if it support my X1600 Pro PCI-E then it will be a real boon for my HTPC.

___________________________updated_________________________________________________
Results: not good.

Zap was right, it is a 915GL, not a 915GM as I had mistakenly supposed. And while it DOES support EIST, it only supports it for the Prescott. It does not recognize my Mobile P4 as supporting SpeedStep.

I may as well give a mini review on the board. It is quite stable and would let anyone run a socket 478 processor on a board with all the features of 915 chipset. The PCI-E is x4, but I was able to use an X850XT PCI-E in it with no trouble. I didn't benchmark, and while overall bandwidth would be limited by 4 lanes, I don't think it would be THAT limiting, as 16 lane bandwidth is seldom achieved/saturated/realized today.

THe board has NO overclocking functions. Even fsb changes seem not to take. I manually set fsb to 800, which on a P4P800-VM and Shuttle SB75G2 ran this processor at 2.4ghz, but after reboot, no good.

At any rate, the Mobile P4 3.2 booted with 12 multiplier x 133 fsb for 1.6Ghz.

I would like to let the group know that the above boards do support the mobile p4s, and with ddr500 on the shuttle i was able to run this chip at 3 ghz (250fsb) at 1.2 volts.

At any rate, the Asrock board would be great HTPC board for supporting stock speed CPUs but is not the enthusiast board I had hoped for.

It's for sale and so is the Mobile P4 3.2 (24x 533 w/ HT).

Cheers!


Edit: Forgot to mention...enabling Speedstep in Bios did not survive reboot. Also. I tried using RMCLock and EIST.exe in Winxp with no luck...all multiplier options were grayed out.

 

[Zap]

GL Joe
PCIe 4X is IIRC same bandwidth as AGP 8X so performance wouldn't be hindered too much, but there are possible/unknown issues. Theoretically it works and glad to see it did in your case. If your board does not let you select FSB but will detect pins, break off BSEL1 while leaving BSEL0 intact will tell your board to run it at 200MHz FSB (quad pumped to 800MHz).

nickdigger
You are on the right track, but the CPU defaults to 01010 and breaking AE1 would give it 11010 - well, according to one datasheet the mobile Celeron should default to 01001 but it doesn't "act" that way on a desktop board, thus I say 01010 is the way it "acts."

osage
My take on why the system wouldn't POST with the wire is that you made the board detect 00010, which is invalid. Likely the wire touched the remaining pad on the bottom of the CPU, thus shorting out VID3 (which defaults to open = 1). If it didn't touch the pad, wouldn't have changed anything. Why the board wouldn't power up after wires removed... well, the voodoo chants and clear CMOS for an hour fixed that I guess.

You can try using a wire to short out VID2 to VID3, which should give you 1.400v (but your board will probably report 1.380v). You could also try VID2 to VID1 for 1.250v or if you want to go for broke again and want to simulate the VID4 pin still being there, you can try putting a loop into itself hoping that it makes contact with the CPU, or you can wire to AF2, which is the closest VCC (power) pin. You can also get 1.450v with VID3 to VID2, and VID1 to VID0 all at the same time (2 wires).

Besides the VID4/AE1 bent pin I haven't tried the other voltages, but it should work in theory. I'll test out that theory soon on my remaining Keleron 1.6.

 

[Zap]

Originally posted by: GL Joe
At any rate, the Mobile P4 3.2 booted with 12 multiplier x 133 fsb for 1.6Ghz.

Edit: Forgot to mention...enabling Speedstep in Bios did not survive reboot. Also. I tried using RMCLock and EIST.exe in Winxp with no luck...all multiplier options were grayed out.

In the datasheet for your CPU there is mention of the "Intel Speedstep Applet." I wonder if that is a special software that can overrider BIOS? Also, I found out the GHI# pin (A6) is responsible for the change from 1.2v 12x multiplier to 1.55v 24x multiplier. I wonder if that pin can be used to put the CPU into the "high performance" mode as opposed to the "battery optimized" mode it defaults to? You can try bending it out to simulate no signal, or connecting it to VCC (nearest is pin B7) to simulate a signal. Let us know if this works! If it does, this will open up a whole new world for mobile P4 CPUs.

 

[jjmIII]

What should I expect from this Mobile Cel 2.0 in this system ??

 

[nickdigger]

Zap: according to one datasheet the mobile Celeron should default to 01001 but it doesn't "act" that way on a desktop board, thus I say 01010 is the way it "acts."


It acts as advertised on my GA-8IGX. The damn nail polish kept stripping off my pins, so ultimately i had to pop the s478 hatch & stick scotch tape into the AE3+4 holes.
1.300Vlap == 01001 == 1.625Vdesk, stock pins.
01111 == 1.475Vdesk, AE3+4 insulated

I assumed his chip was 1.200v default (01011), which made the best explanation, to me, for how he got 1.18 (1.175 == 11011).


Do SpeedStep chips typically need chipset & BIOS support to access GHI#, or can a simple app do this on an old 478 board? Also, the sheet says GHI# = Low --> high speed; GHI# = High --> low speed; is Low eq to Vss or Vcc? On the surface, it seems to be a bit tricky due to Sleep State changes, etc.

Google found a discussion here: http://www.howtofixcomputers.com/bb/ftopic92449.html

 

[stevty2889]

Originally posted by: nickdigger
Zap: according to one datasheet the mobile Celeron should default to 01001 but it doesn't "act" that way on a desktop board, thus I say 01010 is the way it "acts."


It acts as advertised on my GA-8IGX. The damn nail polish kept stripping off my pins, so ultimately i had to pop the s478 hatch & stick scotch tape into the AE3+4 holes.
1.300Vlap == 01001 == 1.625Vdesk, stock pins.
01111 == 1.475Vdesk, AE3+4 insulated

I assumed his chip was 1.200v default (01011), which made the best explanation, to me, for how he got 1.18 (1.175 == 11011).


Do SpeedStep chips typically need chipset & BIOS support to access GHI#, or can a simple app do this on an old 478 board? Also, the sheet says GHI# = Low --> high speed; GHI# = High --> low speed; is Low eq to Vss or Vcc? On the surface, it seems to be a bit tricky due to Sleep State changes, etc.

Google found a discussion here: http://www.howtofixcomputers.com/bb/ftopic92449.html

Low is connected to VSS

 

[Zap]

Originally posted by: jjmIII
What should I expect from this Mobile Cel 2.0 in this system ??

At least the same as your 1.6@2.66GHz. Likely you'd get 3GHz or more but YMMV. I have a similar chip (don't remember SPEC) in my HTPC, but I'm running it near silent at 1.2v default MHz, not a good candidate for 3GHz+ testing.

nickdigger
AFAIK all the mobile Northwood Celerons are 1.3v according to all the ones I've looked up at Intel's Processorfinder site. According to this datasheet which supposedly corresponds to these CPUs we're talking about, 1.300v is 01001. When I say how it "acts" on a desktop board, I do realize it doesn't mean it does the same thing on all desktop boards, indeed my Albatron board defaulted to something closer to 1.5 or 1.55v. Perhaps I should say commonly defaults to since most of the boards that have worked successfully will think the CPU asks for 1.600v and 1.200v with the bent pin, thus my assertion of 01010 default, 11010 bent pin.

I think also that we're reading the pins in opposite order from each other. I say 11010 and you say 01011 (brings to mind a certain song). You're robably naming the pins VID0-VID4 while I'm naming them as written in the datasheet, which is reverse order of VID4-VID0. But wait, your 01111 does match 1.475v. Which is it?