[ KB ] Original Dual PIII Notes

Post Reply
Derek
Site Admin
Posts: 2489
Joined: Tue Jul 23, 2002 3:55 pm
Location: Canada
Contact:

[ KB ] Original Dual PIII Notes

Post by Derek »

Originally written in October of 2001 by Tim, for the latest Pentium3 modification see: Anton's BP6 Dual Pentium III Mod

Introduction:
Ever since the first SMP capable FC-PGA PIII's were introduced to the market, BP6 users have been trying to get them to work on the BP6 motherboard. For those of you that don't know the history about the BP6 let me give you a quick overview: The BP6 is the first ever motherboard that was designed to run dual Celeron PPGA CPU's. No other company even attempted to try this, except for Abit (famous for making highly overclockable motherboards). According to Intel, PPGA Celerons running dual is an unsupported feature. So from the get go, this mobo (slang for motherboard) was a unique product. Celerons were cheaper than PII's and the advantages of running them in SMP were a big welcome from the hardcore overclocker community. Abit took a risk by producing this "unsupported motherboard". Because of all the undocumented PIII core changes that result with development of new technology, and hardware, the BP6 ended up not having the future capability of upgrading to PIII's, (mainly because there was a different pin layout for the new PIII's and a Vcore change).

However, the BP6 community is vast and comprised of many hardcore computer users, many of which have electrical engineering degrees or were born with a soldering iron in their hand. Over the past few years, this board has literally been backward engineered to be more stable and to allow the use of PIII's in SMP mode. All credit goes towards the loyal BP6 users, manufacturers (Abit, Powerleap, Intel, Evergreen) for providing the info, tips, tricks and hardware to make this possible.

The main issues of why the PIII is not natively supported on the BP6.
1. Pin layout is different (PPGA vs. FC-PGA)
2. Voltage differences (Vcore, VTT, etc.)
3. Voltage stability on the BP6
4. BIOS Microcode (no PIII support in the OEM BIOS)

As a result of all of these known issues, the BP6 users out there have tweaked, soldered, flashed, and finally come up with a workable solution to hacking the BP6 for PIII SMP support.

Here are a few links to read up on modifying your BP6 for dual PIIIs:

Powerleap:
Creator of the NEO S370 adapter. This adapter saves some steps in allowing you to modify the adapter itself instead of re-engineering the BP6 Ziff sockets. You will need 2 of these (modified). You will want to check to see if they have the NEO S-370 plus.

VTT Modification:
Solves stability issues on the BP6. This is necessary on the BP6 to prevent lockups under Win2K.

Yoichiro's site:
PIII SMP Success How-To Guide - He is a Japanese BP6 user that created a good in-depth set of articles on modifying the BP6 for PIII SMP support.

Aryan:
Another BP6 user and his article on PIII's on the BP6

Powerleap NEO S370 mod:
Shows how they got PIII's working on th BP6 in SMP mode.

Evergreen Technologies:
Spectra III - Maker of computer upgrade CPUs. The Spectra III 850 CPUs are currently being tested here at BP6.com as a possible alternative to the NEO S370. It's basically a PIII CPU mounted on an FC-PGA to PPGA Socket 370 adapter. The good thing about these Spectra IIIs that the adapter has a built in multiplier adapter and may not need further modifications.
Last edited by Derek on Sun Sep 02, 2012 4:01 pm, edited 1 time in total.
tjcarst
Posts: 14
Joined: Thu Apr 10, 2003 7:54 pm
Location: Lincoln, Nebraska, USA

Status Spectra III 850 of testing?

Post by tjcarst »

Were the Spectra III 850 successful?
Derek
Site Admin
Posts: 2489
Joined: Tue Jul 23, 2002 3:55 pm
Location: Canada
Contact:

Re: Status Spectra III 850 of testing?

Post by Derek »

tjcarst wrote:Were the Spectra III 850 successful?
Not yet.
kuun
Post-O-Matic
Posts: 1566
Joined: Sat Nov 02, 2002 5:26 pm
Location: Tennessee
Contact:

Post by kuun »

whats a spectra III 850?
!!! WARNING !!!
The following forums: www.bp6.com
are infected with the following VIRUS(s): Kuun.infected.all.posts.Win2K.user

The following IRC servers has been exploited: irc.bp6.com
with the Following Exploit: Kuun.lurks.using.mIRC.v5.82.exploit
hyperspace
Board Admin
Posts: 1395
Joined: Wed Jul 24, 2002 7:39 am
Location: Lincoln, NE USA
Contact:

Post by hyperspace »

kuun wrote:whats a spectra III 850?
http://www.evertech.com/category.cfm?Category=57
Quantum WormHole

Image
lost in hypertime...
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

Hey all. I'm new around here. I just did the PIII mod using Neo S370s, following Yoichiro's guide. I've got a version 1.0 BP6. I cut the trace on the S370s and added the wire from AN15 to N33. The dual proc works fine.

I tested her with just this mod and the board was stable to a maximum of 83MHz FSB.

I then changed out EC10 for a 1500MFD cap (the previous owner modified the board using the $2 VTT fix, so my Q6 is a LM317T soldered right onto the old regulator), and the board is stable at 100MHz FSB, so I'm underclocking (yes, it rankles, I'm not much for overclocking---I like stability---but underclocking makes me feel gipped :?) my 866MHz PIIIs to 650MHz right now. This seems to be the maximum I can boot at.

Something funny happens when I try 110, 124 or 133 (the ultimate goal). The board powers up, the kbd leds flash, the CD checks for media, and I get a whole lot of nothing on my monitor, it's just blank, nothing is displaying.

The rest of my hardware setup looks like this:

BP6 RV BIOS w/o HPT366 (I don't use them)

384MB PC133 SDRAM (256+128)

Western Digital 20GB drive
iOmega ZIP 100 (IDE)
LG DVDROM

NE2k PCI ethernet

SB PCI128 soundcard

GeForce4 Ti4200 -or- 3dfx Voodoo3 2000

Slackware Linux v10.0, Kernel v2.6.7

I get the same problem with both video cards, ie, neither display anything above 100MHz FSB. I downloaded and installed the RV BIOS to get the PIII microcode and hoping that the problem would clear up with the new BIOS (admittedly, a long shot, and I realized this at the time), the RU BIOS worked for booting to 650/100MHz.

Anyone have any idea why 110, 124 and 133MHz FSB would cause nothing to display? I've been thinking about trying to rip into BP6FSB and port it to Linux, to see if I can boost my clock on the fly, but I don't have time for that today.

Thanks all, this site is awesome, I wouldn't have gotten as far as I have without you guys. My system already feels much faster (but I still only get a maximum of 20fps in tenebrae and 3fps in hard to render areas... It's unplayable... (looks like they're trying to multithread it, maybe it's time to help them out a little... :thumbs:).

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
hyperspace
Board Admin
Posts: 1395
Joined: Wed Jul 24, 2002 7:39 am
Location: Lincoln, NE USA
Contact:

Post by hyperspace »

Welcome to BP6.Com, phaedrus !

Dual PIII's and Linux! Nice work! Not sure if you have read all of Yoichiro's modifications he did. From what I have read on his site, even if you do all the modifications, you still may not be able to get 110 or higher.

When you are running at 100 or 83, are you experiencing any random lockups or reboots?
Quantum WormHole

Image
lost in hypertime...
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

hyperspace wrote:Welcome to BP6.Com, phaedrus !

Dual PIII's and Linux! Nice work! Not sure if you have read all of Yoichiro's modifications he did. From what I have read on his site, even if you do all the modifications, you still may not be able to get 110 or higher.

When you are running at 100 or 83, are you experiencing any random lockups or reboots?
I've gone through Yoichiro's page, and decided to take them one at a time.

Since I put in the 1500MFD cap, I've been running perfectly smoothly at 100MHz FSB, and I've been stress testing it under Quake1, Q3A and Tenebrae, and she doesn't lock up (if she doesn't lock up in the first hallway in e1m1 under tenebrae, methinks she's never going to lock up, for some reason, Quake & Tenebrae have a hard time rendering that hall with water vissed maps, something to do with the vising process).

But, she's a helluva lot faster than with the dual 366s (which I didn't oc, because I was never happy with their oc'd stability). I'm pretty happy.

As for the linux bit, I've only ever run Slackware Linux on my BP6, because, well, it's just the best.

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

Hey guys. I had a thought. I'm an electronics technician in a physics lab at the UW (this is why I got the mod going in a weekend, experience and access to the right tools).

Anyway, I keep worrying about the wires on the back of the sockets for the Neo 370S mod--solder is bad for making mechanical connections, and while the wire is light, and I'm not hauling my box around, I'd be worried about going to a LAN party with it.

So, my idea is this: glue the wires down with some kind of epoxy. I haven't tried it yet, but I think even just a 5 minute epoxy would make a very solid mechanical connection with the board, and shouldn't damage the PCB. Any thoughts?

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
hyperspace
Board Admin
Posts: 1395
Joined: Wed Jul 24, 2002 7:39 am
Location: Lincoln, NE USA
Contact:

Post by hyperspace »

Sounds like a plan.
Quantum WormHole

Image
lost in hypertime...
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

phaedrus wrote:So, my idea is this: glue the wires down with some kind of epoxy. I haven't tried it yet, but I think even just a 5 minute epoxy would make a very solid mechanical connection with the board, and shouldn't damage the PCB. Any thoughts?
I have been thinking about doing the same with the capacitors that I have replaced on the version 1.1 board that I have. They are larger than the original (both physically and electonically) and because I am using large coolers I have the new caps not sitting tight on the motherboard itself (they are also bent slightly to one side - away from the cooler).

I got a "plastic gun" that melts plastic sticks to glue things together. I have been thinking about using that to make the caps sit firmly in place. I am a bit weary of having them sitting loose on the board.

I'd say go for it. I seriously doubt it will damage anything.
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
Billl
Posts: 211
Joined: Fri Jul 25, 2003 1:52 pm
Location: USA

Post by Billl »

I work in what used to be a card manufacturing shop. Now all we do is customer returns, but I do know a thing or two about card manufacturing. All we ever used was either two part epoxy or RTV to hold things in place. So you should be just fine. I'm not sure if I got his meaning about the attaching the wires though? If it were me I would solder them and then epoxy them to stop any movement at some nearby convient spot.

just my .02


Billl
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

This is linked to in the article, but here is a good picture of what I'm talking about. To get the PIII Neo mod going, you have to add the wires in the picture, and there's no good place to make a mechanical connection.

Most of the work I do in the lab is point to point turret construction. We do lots of radio frequency and cryogenic work, so sturdy construction with considerations of crosstalk and shielding is usually the name of the game. Since we do tons of custom designed equipment and only have 5 people working in the lab itself, quick and easily done by hand is also the name of the game. Turrets on PCB usually ends up being the best construction technique for us.

So, I'm not as familiar with etched PCB construction techniques, but I do know solder is a bad mechanical connection, and that we often use 2 part epoxy to glue DIP sockets and certain capacitors to the back of single sided PCB. I just did this on a board I built the other day, which is what made me think of using the epoxy to make a solid, insulated mechanical connection. I would just be glueing them down to the back of the board, it being the only nearby, convenient spot.

Billl, thanks for the confirmation on it working fine on etched PCB parts. I felt a lot better trying this knowing it's common repair practice (I don't want to screw up my BP6, it's my baby). I've been fretting over those wires--I've been avoiding moving my box just to minimize the chance that the wires would break. Anyway, I went ahead and did it, and it's working just fine--I'm typing this post on her. The epoxy is mostly cured, and over the next day or so, it'll finish curing. This should probably get added in as standard proceedure for those wires (they'll be much more reliable, and while we haven't really heard reports of the modded boards breaking, I wouldn't take a board without the wires glued down to a LAN party or the like).

I took the opportunity to take some pictures, I'll post them later (girlfriend owns the digital camera, and has her computer set up to grab the pictures. Girlfriend is now in bed).

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

Phaedrus: Could you please outline what mods you have already done on your BP6? My dual 650MHz machine is giving me problems but I haven't had time to look into it yet. Just wondering what you have done so far.
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

Sure thing. With pictures even! Check out my gallery, I've got pretty clear shots of all the mods I've done.

So far, I've done the Neo 370S mod (cut the trace on the adaptor, on wire on the mobo) and epoxied the wire down.

The previous owner put an LM317 in for Q6. To stabilize the board with the PIIIs in it, I changed out EC10 (was 2200MFD) for a Radio Shack Special 1500MFD cap (1000MFD + 470MFD 25V caps tied in in parallel). Without the RSS, I was stable at 83MHz FSB, and with it, I was stable at 100MHz FSB.

To date, that's all I've done. My other hardware is an SB PCI128 sound card, 384 MB of PC133 RAM, a GF4/Ti4200 (also worked fine with Voodoo3 2000), a Western Digital 20GB hdd, an iOmega Zip Drive, an LG DVD-ROM drive and a 400W PSU. Standard keyboard and an optical mouse plugged into the PS/2 port via an adaptor.

I'm planning on recapping the whole board and shorting diodes D1 and D2 (or whichever are suggested by the Neo guys).

I got to get to class. Later.

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

Hey thanks! Good pics there! Tell me, how was the LM317 job done? Did he remove the 1085?

I got myself an LM350A but I am sort of sceptical. The 350 can regulate a continuous 3 amp current but so can the 1085. My problem is that I don't know the 1085 that well and documentation on it is not easy to find.
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

The 1085 is still there. I, personally, would have removed it (now that I think of it, LM317's are cheap, I might just redo the job). My Vtt voltage still fluctuates +/-10%, but it hasn't been affecting stability. I don't know that I've ever been able to find documentation on it [the 1085].

I went by Yoichiro's table for EC10 based on the Q6 regulator used (which I'm sure you've seen).

I've seen at least 3 different suggestions for what to do when re-capping your board. Yoichiro has his suggestions (under further mods), Friendtech has an update on the Neo 370s mod where they suggest cap value changes (as well as shorting a couple diodes and changing some resistor values) and finally, the www.motherboardrepair.com guy [Gary Headlee, who's been getting quite a bit of flack on these forums of late] recommends moving every cap above 1000MFD to 1200MFD (I emailed him to ask about this and got no responce, go figure).

I'm still trying to figure out which way to go. I'm going to replace all my caps, but I don't know which of those three suggestions to go with. I'm actually leaning toward the 1200MFD all around solution with EC10 kept at 1500MFD, and then if that doesn't get me to 133, I'm thinking about trying to short the diodes as per the Friendtech mod, and see what happens from there.

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

I can't say that I have seen the extreme Vtt fluctuation on either of my boards (ver 1.0 and 1.1) which is why I am also sceptical on replacing the Q6. I have also seen all of the mods sites that you mentioned and there is one thing that bothers me about the Friendtech's recommendations is that they don't mention why these changes should be implemented.

I can mention that the caps that I replaced were all of bigger size (in terms of microF and physical size) than the originals but I used the same voltage specifications. I replaced the 8 caps around the voltage regulators and the 12 caps around the ZIF sockets.

I am not so sure you will ever see 133 on the BP6 no matter how much you try. I think the adventurous over on this forum have reached 110 but thats it. I makes me wonder how the NEO 370S handles overclocked frequency as opposed to the frequency switches on the adapter itself?
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

purrkur wrote:I can't say that I have seen the extreme Vtt fluctuation on either of my boards (ver 1.0 and 1.1) which is why I am also sceptical on replacing the Q6. I have also seen all of the mods sites that you mentioned and there is one thing that bothers me about the Friendtech's recommendations is that they don't mention why these changes should be implemented.
If you never see the fluctuations, your Vtt might not be the problem then. My Vtt only ever droops, which means it droops into the range specified by Intel for the PIIIs. This might be why I'm ok for Vtt.

Yeah, Friendtech doesn't say why. I'm going to email them before I actually short the diodes, but looking at their mods, it's the only one that really looks like it might make a big difference. My reasoning on that is that the others are cap value changes and resistor value changes, which I think aren't going to make that much of a difference, but the diodes are right next to the CPUs, and my hunch says that's half of it (although, I want to figure out why they're there first).

However, the first step is to just recap the board (I'm pretty sure I've got some ugly leakage somewhere, and I just don't like having those shoddy caps in there).
I can mention that the caps that I replaced were all of bigger size (in terms of microF and physical size) than the originals but I used the same voltage specifications. I replaced the 8 caps around the voltage regulators and the 12 caps around the ZIF sockets.
All the decent replacement caps I've seen are bigger, even if of the same value. The cap suggestions all seem to point to keeping near to the same values as on the board (with the exception of EC10, which is teensy on some boards) or even reducing a bunch of them (the flat 1200MFD suggestion by motherboardrepair.com). This might be your stability problem.
I am not so sure you will ever see 133 on the BP6 no matter how much you try. I think the adventurous over on this forum have reached 110 but thats it. I makes me wonder how the NEO 370S handles overclocked frequency as opposed to the frequency switches on the adapter itself?
Damnit, I'm going to try. It may require Peltier cooling, but I'm going to try.

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

phaedrus wrote:Damnit, I'm going to try. It may require Peltier cooling, but I'm going to try.
Thats the spirit!!

Keep us informed of your findings. I'll look into my issues hopefully soon to see what is going on. It is not really stability that is my problem. The screen/video freezes on me but the machine is still running. I can log into it from another machine without issues. It isn't doing anything special either. The logs don't really reveal anything either :(
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
phaedrus
Posts: 75
Joined: Tue Jul 20, 2004 11:27 pm
Location: Seattle, WA
Contact:

Post by phaedrus »

I've had similar problems. What video card are you using? I've found that the 3dfx drivers are unstable (and unmaintained, so it's not likely to get fixed any time soon) as well as the closed source nVidia drivers. I've setup my system so I can run with the opensource nv driver when I don't need the 3d acceleration. That cleared up my problems.

Jeff
"If it ain't broke, mod it till it is"
They said... and now my BP6 needs new processors... D'oh
Slackware Linux v10.1
Image
purrkur
Linux Guru
Posts: 687
Joined: Fri Dec 12, 2003 5:57 pm
Location: Sweden
Contact:

Post by purrkur »

phaedrus wrote:I've had similar problems. What video card are you using? I've found that the 3dfx drivers are unstable (and unmaintained, so it's not likely to get fixed any time soon) as well as the closed source nVidia drivers. I've setup my system so I can run with the opensource nv driver when I don't need the 3d acceleration. That cleared up my problems.

Jeff
Interesting. I am using the closed source Nvidia drivers. I have never had any issues with them before. Time to fire up the old baby once more to see if things change if I change the driver...

Thanks for the tip.
2x533MHz@544MHz, 2.0V
640MB PC100 memory
Realtek RTL-8139 NIC
Maxtor 6Y080L0 80GB hdd
Debian Linux stable with 2.4.8 kernel
Post Reply