1U type coolers (Vantek) were the choice of the authors of this article:
http://tipperlinne.com/p2bmod.htm
scroll down for photo + hardware details

PARKE wrote on 2020-09-08, 11:03:

1U type coolers (Vantek) were the choice of the authors of this article:
http://tipperlinne.com/p2bmod.htm
scroll down for photo + hardware details

Thanks for your input, so the stacked hight of that heatsink & Fan combo according to:

https://www.overclockers.com/vantec-1u-cck-6012/

Seems to be 33mm and from that photo on the link you sent, it seems there is room to spare by some margin, clearly, you don't want to obstruct too much to reduce airflow through the fan by getting too close to the back of the other slotket, also need to be mindful of the row of capacitors sat there as well. At a guess, it seems there is around 40 - 45mm to play with.

Cheers

limsolo wrote on 2020-09-08, 11:24:

clearly, you don't want to obstruct too much to reduce airflow through the fan by getting too close to the back of the other slotket,

Yes, exactly.

This is how I've currently got my P2B-DS kitted out

Two MSI MS-6905 Mater slockets & two Coolermaster DP5-6G11 (known in some markets as the CM DRACO Slimbird GP) but with uprated fans (Gelid Silent 6).

The gap from the base of the rear heatsink to the back of the front slocket PCB (for this to work had to remove the plastic backplate from front slocket) is 50mm, and the heatsink / fan combo is 43mm. I'm using bare die processors so with IHS would be probably 1mm difference.

PC Hoarder Patrol wrote on 2020-09-08, 20:05:

This is how I've currently got my P2B-DS kitted out […]
Show full quote

This is how I've currently got my P2B-DS kitted out

Asus P2B-DS Slockets.JPG

Two MSI MS-6905 Mater slockets & two Coolermaster DP5-6G11 (known in some markets as the CM DRACO Slimbird GP) but with uprated fans (Gelid Silent 6).

The gap from the base of the rear heatsink to the back of the front slocket PCB (for this to work had to remove the plastic backplate from front slocket) is 50mm, and the heatsink / fan combo is 43mm. I'm using bare die processors so with IHS would be probably 1mm difference.

Thank you for such detailed information this will help me a lot!, if I may ask you a couple of questions as your setup is practially what I am aiming for.

What version of the motherboard is it 1.06 d02 or d01, I don't see the yellow sticker near the AGP slot so I don't think is d03. What revision or the MSI MS-6905 I can see the TVC IC's on the left as I am guessing 2.0 or later?. What did you do to run SMP did you have to solder a wire from jumper J3 pin 3 to socket pin N33 (or are they Celerons) if you are using PIII's's are they Coppermines or Tullys-S (with IHS removed)

Again big thanks photos and info provided so far are a big help

My motherboard is v1.06 D02, but is factory-fitted with a pair of the improved Harris VRMs (HIP6004BCB) , capable of voltages down to 1.3V.

The MSI MS-6095 Masters are nominally v2.3, but unlike the versions with the 4-pin J3 jumper block, mine only have 3 pins so yes, I've soldered a wire between the pin on J3 and the N33 pin on the socket.

I'm using a pair of Coppermines, 1000/256/100/1.75V (SL5QV).

PC Hoarder Patrol wrote on 2020-09-09, 20:55:

Asus P2B-D CPU heatsink clearance for Slotkets My motherboard is v1.06 D02, but is factory-fitted with a pair of the improved Ha […]
Show full quote

Asus P2B-D CPU heatsink clearance for Slotkets
My motherboard is v1.06 D02, but is factory-fitted with a pair of the improved Harris VRMs (HIP6004BCB) , capable of voltages down to 1.3V.

The MSI MS-6095 Masters are nominally v2.3, but unlike the versions with the 4-pin J3 jumper block, mine only have 3 pins so yes, I've soldered a wire between the pin on J3 and the N33 pin on the socket.

I'm using a pair of Coppermines, 1000/256/100/1.75V (SL5QV).

Thanks again, on J3 looking at the instruction manual for the v2.0 MS-6905 (the only one I am aware of with TVC chip & 3 pin J3, though as you said have a different version which is not something I have seen mentioned any place 🤣), it says SMP is pin 2&3 shorted, Coppermine is pin 1&2 shorted. I know the board only officially supports SMP on Celerons in SMP. Therefore once the wire is soldered on (do you have to do this on both boards or just 1?), which jumper position on J3 do you select, I know you have to short pins 2&3 otherwise it would be pointless soldering the wire on but is another jumper on pin 1&2 of J3 required for Coppermine CPU (logically the answer is yes as you have Coppermine fitted in your boards) meaning that 1&2 and 2&3 are jumpered at the same time in your case?

Out of curiosity do you know what the difference is between yours and a d03, what clock generator IC is on yours?

Thank you

MSIs version numbering on these seems slightly screwy at times - as I said, mine are nominally v2.3 - one is marked as 2.3 (sticky label) & one as 2 (silkscreened on PCB) but they're identical down to age, layout & individual component numbers (maybe from different fab houses?). Not sure I've actually seen a pic of a v2.3 with the 4-pin J3 but I've seen them discussed on other forums - apparently the 4-pin version does away with need for the J3 / N33 wire mod.

To make it work properly, I've wire-modded both mine and they're also both single jumpered on pins 1-2 on J3 (Coppermine 256k, as 2-3 is reserved for dual Celeron only). What you're describing is actually how the 4-pin version works - dual Coppermines are jumpered on 1-2 and 3-4.

Differences on these boards can be hard to pin down without actually seeing them, as quoted on that web page I linked you to in an earlier thread...

Q: Are there differences between boards even if they have the same revision (or even same pcba number)?
A: Yes there are. Most notable are the P2B rev. 1.10, the P2B-D/-DS rev. 1.06 pcba D01/D02 (and even some 1.05 boards) and the P2B-F (those with the not supported pcba numbers for coppermines) boards, some of these boards indeed have the newer voltage regulator chips and thus can run slot1 coppermines (and fcpga coppermines with slotkets without voltage adjustment jumpers) just fine.

The clock generator on my board is ICS 9150AF-08, rather than the newer / better spec ICS 9250CF-08.

The 9150-08 is buggy according to this old thread:
https://www.motherboardpoint.com/threads/need … mod-help.22785/
quoting:
[[...Yeah, upgrading older P2B-series boards would be a lot simpler if the
9150-08 133Mhz setting actually worked, instead of producing an ugly
signal the BX interprets as 109Mhz.
P2B...]]

Good find thanks, ICS 9250CF-08 is available on eBay, I have the necessary skills and equipment to fit the chip. All I need then is 2x MSI MS-6905 Rev 2.x slot converters.

My board is a ver 1.06 d02 with the correct VRM's. This guy in Korea is selling 1.4S Tullys with an integrated baseplate/adaptor which seems to pin mod the chip: Here so that saves a bit of hassle.

Thanks for the info everyone who replied 😀

First of all, the ASUS 440BX boards are probably the best documented of all, there is a huge amount of knowhow still available on the web.
Then, I don't think that the MS-6905 slotkets are the only ones that are fit for this job; Iwill and Soltek come to mind, besides others, plus of course the ASUS mentioned in the Tipperlinne article.
Take also into consideration that the korean modded chips are thicker than unmodded ones - this re distance between the slotkets.

PARKE wrote on 2020-09-10, 17:52:

First of all, the ASUS 440BX boards are probably the best documented of all, there is a huge amount of knowhow still available on the web.
Then, I don't think that the MS-6905 slotkets are the only ones that are fit for this job; Iwill and Soltek come to mind, besides others, plus of course the ASUS mentioned in the Tipperlinne article.
Take also into consideration that the korean modded chips are thicker than unmodded ones - this re distance between the slotkets.

Agreed on all points, I'd love to get Upgradeware slot T but I have spent the last week searching, the only one coming up with any hits is the MS-6905.

Korean guys says that adaptor adds about 1mm

Cheers

It will be interesting to follow the process and I hope you will be successful and post an update when you are done.
There is one aspect that also needs attention. When you install the 9250 chip you will have working 133Mhz but given the 3x2 pin layout of the jumperblock on the P2B-D the selection is limited. In this case the 133Mhz setting that is available (of the two that are supported by the chip) runs the PCI at 44.3Mhz (133/3) which may be less than desirable. The Tipperlinne P2B Modification guide deals with that issue under the [150Mhz FSB on the Asus P2B-DS Rev 1.06 D03] header.

OK so an update, my board turned out to be a rev 1.06 D02 which has the latest/necessary voltage regulator chips HIP 6019 BCB & HIP 6004BCB to drive lower voltage CPU's such as PIII-S 1.4Ghz. I have the old 9150 clock gen chip which I confirmed with SoftFSB wont produce a proper signal at 133Mhz it downclocks the CPU's (I forgot exactly what but it was lower than 112Mhz FSB). Anyway it will run reliably at 112Mhz which for now is good enough. So far it seems that the Slot 1 sockets on the motherboard might be a bit flaky.

2x PIII Coppermine (slot 1 100Mhz FSB spec) 700's @ 784Mhz 0r @ 700Mhzrun fine both are slot 1 CPU's and seem to be recognized fine as a pair unless disturbed then can be a bit of a hassle to get them to both be recognized at POST time but otherwise runs fine with Windows 2K, Haiku and BeOS, installed Lubuntu LTS and ran fine. 1gb of 133Mhz ram in this configuration runs fine. Its been running fine with these since I built it 10 days ago then last night I tried the Slot0T's and became aware of problems as below.

2x Upgradeware Slot-T with 2x Tualatin 1.4S, bit of a pain to be honest, very flaky getting both CPU's to be recognized at POST time. multiple remove and inserts (reseating) seems to work a little, blown slots out with airduster, cleaned Slot-T edge connector with rubber eraser, also polished with lint cloth and tiny bit of metal polish as they where tarnished. When system posts, BIOS POST screen sometimes corrupted with oddities, also hits and miss how much memory is reported seems to see 1.5 sticks sometimes???. Remove 2 sticks, leave 2 in, 512MB so posts a bit better then when running Windows 2K getting BSOD with some kernel error. Sorry to be a bit vague but it was going on last night and it was late. Could not install Lubuntu got a couple of kernel panics and also crashes.

I think the connection of the slot closest to the ram and the Slot-T in it is the problem, I have swapped Slot-T's, same, checked voltage jumpers 1.45v as it should be (confirmed with CPU-Z). But sometimes getting this slot to recognize something is plugged can be a little difficult, the proper Slot 1 CPU's I have being the easiest and most reliable.

POST screen reports them as Tualatin Celerons even though they are full fat 1.4-S??

I did try a couple of S370 PIII 1Ghz Coppermines (1Mhz FSB ones) I have and maybe they where better but would not run with 112FSB, I didn't test for long, I tried them when the Tualatins would not POST before I realized one socket was not making good contact.

I have the 1014d.003 Beta BIOS which I used BIOS Patcher 4.23 to fix as my DVD drive disconnected randomly before I did the BIOS patch.

Not sure what's going on here, anybody got any tips or observations?

Any ideas how to clean or improve the connector quality between the edge connector and the slot?

Thanks

limsolo wrote on 2020-11-08, 09:25:

2x Upgradeware Slot-T with 2x Tualatin 1.4S, bit of a pain to be honest, very flaky getting both CPU's to be recognized at POST […]
Show full quote

2x Upgradeware Slot-T with 2x Tualatin 1.4S, bit of a pain to be honest, very flaky getting both CPU's to be recognized at POST time. multiple remove and inserts (reseating) seems to work a little, blown slots out with airduster, cleaned Slot-T edge connector with rubber eraser, also polished with lint cloth and tiny bit of metal polish as they where tarnished. When system posts, BIOS POST screen sometimes corrupted with oddities, also hits and miss how much memory is reported seems to see 1.5 sticks sometimes???.
8><CUT
I did try a couple of S370 PIII 1Ghz Coppermines (1Mhz FSB ones) I have and maybe they where better but would not run with 112FSB, I didn't test for long, I tried them when the Tualatins would not POST before I realized one socket was not making good contact.
8><CUT
Not sure what's going on here, anybody got any tips or observations?

The P2B-D rev 1.06 D03 that I have here does not like 4x256 sticks when there are two 1GHz 1000/133 SECCs installed. Recognition is erratic, sometimes 2, sometimes 3. With fsb 100 Seccs there is no problem.

Re the 112 fsb, according to the specsheet of the ICS 9150 the performance is 'not guaranteed'. And at 112 fsb the PCI is overclocked wich may also cause instability?.
I guess that your system would improve with replacement of the ICS 9150 with the later/better model.

Another point of interest is that the voltages delivered by Upgradeware/Powerleap slotkets are not quite up to par for what a Tualatin on a 440BX actually requires (although I don't know if that can cause instability) . You can read details about that aspect in this discussion:
https://forums.overclockers.com.au/threads/sl … slotket.164450/

Hi Parke

Thanks for weighing in on this. OK had another go at reseating CPU 1 (the one closest to the ram) bit more metal polish and seem to be OK now. Rammed the CPU fast this time as well!! Lol.

I was thinking before I actually saw the board in the flesh to fit a new clock gen but I don’t think my skills are good enough I cannot say confidently I can do it I am not bad and have the gear but that is boarderline the legs are too tiny and my eyes are too old maybe 20 years ago when the board was new..... P2B the guy (Simon) in that thread is who I bought the Slot-T’s from and the PIII’s CPU’s he said he never got ram to work properly at even 124Mhz FSB without having to enter and exit the BIOS each time (hassle) so for what it’s worth another few MHz on the FSB does not seem to be worth the hassle swapping the clock gen IC I have it running 1175 MHz which is plenty fast.

I tried on Win2K to use the MPS HAL/Kernel instead of the ACPI SMP HAL but still BSOD at 100 and 112 FSB and I am using Blackwing Cats Kernel EX too. I seen a few reports that Tualatins require Windows to reinstalled so trying that and will report back. Memtest on a Linux distro passes OK. So made some progress.

I think there are two sides to this coin.
The board is probably ok when you run Coppermine fsb 100 cpu's at fsb 100 - which is what the board was intended for.
Eveything else that you describe that you did and did not work properly translates to running stuff out of spec, either fsb or voltage.
My idea is (having read various publications by Simon Edwards and Keith Hui aka the P2B team) that if you want to run this particular board out of spec you better go the extra mile and apply the modifications that worked for them. Some of the combinations you tried are seemingly not of the kind that respond dutifully to replugging just a couple of jumpers or re-installing software. So in my idea it seems to boil down to a simple choice, either run the board in spec -or- modify the hardware. Overclocking/overvolting comes at a price. Do you agree ?

Hi Parke

I don’t know enough at this stage to disagree that’s for sure. Still very much in the experimental stage with it all. P2B in an email said to me he’s run them for ages out of spec voltage wise and it had run stability at less than 124Mhz FSB with less then 1gb ram he sort of intimated that was the limit, I think what you say makes sense if you are at the bleeding edge which I am not as I think possibly the law of diminishing returns lies ahead. My old win2k install which sat on a SCSI Atlas 10kIV via a PCI 2940U2W would randomly BSOD shortly after hitting the desktop and it would die shortly after running Prime95. Just done a fresh install on a IDE - SD card and it was running fine for around 45 mins even running prime 95. Will test further tomorrow more experiments to be done. So far could be PCI SCSI card running out of spec or the fresh Windows 2k install. I still have the old Win2K on the SCSI drive might reinstall KernelEX see what happens the BSOD was in the kernel (ntoskrnl.exe)

Interesting topic.

I run here 1.06 D03 with ics 9250cf-08 and dual 1000/256/133 slot 1 cpus.
Two mods are applied to the board - one for higher ram voltage and another for additional fsb/pci frequency options.

I have it completely stable at:
- 140mhz FSB and 33.44mhz PCI with 3x256mb ram, it takes the highest quality NEC or Mosel Vitelic chips to get there. There is no way to get 4 dimms 1gb ram working in this configuration.
- 133mhz FSB and 44mhz PCI with 4x256mb ram. In this configuration most pc133 rated dimms will do.

I am not sure if ics 9250 is pin compatible with 9150.
I have vague memories that I looked at this before, while trying to upgrade another p2b-ds 1.04 board to 1.06 standard and it was not possible.