Take a look at our
ThinkPads.com HOME PAGE
For those who might want to contribute to the blog, start here: Editors Alley Topic
Then contact Bill with a Private Message

PAE/NX CPU swap in T42P (2373-HTU)

T4x series specific matters only
Post Reply
Message
Author
surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

PAE/NX CPU swap in T42P (2373-HTU)

#1 Post by surrealmode » Fri Oct 19, 2018 5:06 pm

My trusty old T42P has just received a partial restoration - motherboard swap to the original (mobo was replaced in '09 due to physically damaged USB port/password, but new mobo ended up having broken USB), replaced USB port, keyboard, keyboard bezel, CMOS battery, and CardBus slot (bent pin) along with a deep cleaning. An mSATA SSD is on the way, and I've got an mSATA<->IDE converter with the better behaving Marvell chipset.

It's working well aside from one thing - the CPU lacks the PAE/NX capabilities required for Windows 10.

My goal (for purposes of fun and nostalgia, mostly) is to get 4 different versions of Windows booting. Windows 98 SE, Windows XP Pro, Windows 7 Pro, and Windows 10 Pro. However, I've run into an error with the Windows 10 install due to aforementioned issues with the CPU (currently a Pentium M 745 @ 1.8GHz).

I've scoured forums far and wide, both here and elsewhere, even some foreign language forums via Google Translate.

I see a couple possible options:
1.) Swap with a Pentium M 745A, which is 400MHz FSB (SL8QZ)
2.) Swap with a Pentium M 780 (or similar), which is 533MHz FSB (SL7VB)

From what I've seen, the 745 A is a bit of an oddball. Checking the Linux driver (drivers/cpufreq/speedstep-centrino.c), there are *still* no references to the C0 stepping CPUs, meaning that the patch described several years ago would still be necessary. NHC was suggested for XP (and Windows 7?) to properly scale the CPU frequency. The BIOS would not contain the proper microcode either. It seems it might be possible, but modding the BIOS to replace the microcode blobs in it seems risky at best.

The 780 would require some sort of pin mod to have it run at the correct 533MHz bus speed. I imagine it'd also suffer from the same issues with missing microcode. I've seen a guide to overclock the older Dothans from 400MHz -> 533MHz, but am unsure if the same mod would be sufficient to have an 780 run at 533MHz in a T42P.

Has anyone out there completed such a swap successfully and got Windows 10 (or unmodded Windows 8(.1)) running on a T42P or equivalent? If so, which CPU and steps did you take to get things running smoothly? I do not want to swap the motherboard with e.g. one from a T43.

My ideal solution would be to replace microcode in the BIOS with that of the swapped CPU, and to have frequency scaling being done by the OS without a userland utility (e.g. NHC).

I understand that there's no real point to any of this (the laptop is old and slow, even if it's now shiny). The laptop does have significance to me, though, and with the Thinkpad Dock II I just got, it could be useful for running certain older programs/utilities that don't behave well on UEFI/64-bit systems.

Thanks!

ajkula66
SuperUserGeorge
SuperUserGeorge
Posts: 16335
Joined: Sun Feb 25, 2007 11:28 am
Location: Brodheadsville, Pennsylvania, USA

Re: PAE/NX CPU swap in T42P (2373-HTU)

#2 Post by ajkula66 » Fri Oct 19, 2018 5:26 pm

Welcome to the forum!

PM 780 will work, with or without pinmod (only required for it running at 533Mhz) and you should be able to install W10 without issues.

You can pick up a PM750 for $5 and give the concept a test drive.

I don't run W10 but would imagine that NHC would work, I have it running in W7 on my T43pSF.

Happy modding.
...Knowledge is a deadly friend when no one sets the rules...(King Crimson)

Cheers,

George (your grouchy retired FlexView farmer)

AARP club members:A31p,T43pSF

Abused daily: T61p

Work in progress: X60T

PMs requesting personal tech support will be ignored.

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#3 Post by surrealmode » Fri Oct 19, 2018 6:53 pm

ajkula66 wrote:
Fri Oct 19, 2018 5:26 pm
Welcome to the forum!

PM 780 will work, with or without pinmod (only required for it running at 533Mhz) and you should be able to install W10 without issues.
Thanks!

Is the pinmod for overclocking a 400MHz Dothan to 533MHz the same that's needed for swapping in a 533MHz CPU into a 400MHz T42 motherboard? I found The Ultimate Pentium-m Pinmod/Voltmod Guide from overclockers.com:

Code: Select all

Q: Can I use my 533 mhz processor to do the Pinmod?
A: No you need a 400mhz processor for it to work.
That makes me wonder if there's a separate pinmod for swapping in native 533MHz CPUs.

I was also wondering - will the CPU power utilization be the same if it's running with the same multiplier at 400MHz vs 533MHz? Or will the max TDP be 75% (400 ÷ 533)? I'm thinking it would use less power but am not certain. If the CPU selects voltages based on the multiplier irregardless of which FSB it's running at, then it seems it would use significantly more power MHz for MHz.

Hopefully someone can chime in about modding the BIOS with correct microcode. It'd make sense for the microcode to be loaded based on CPU ID metadata attached to each blob, but they might not have done it that way. In that case simply replacing the blob wouldn't be enough I think.

ajkula66
SuperUserGeorge
SuperUserGeorge
Posts: 16335
Joined: Sun Feb 25, 2007 11:28 am
Location: Brodheadsville, Pennsylvania, USA

Re: PAE/NX CPU swap in T42P (2373-HTU)

#4 Post by ajkula66 » Fri Oct 19, 2018 7:28 pm

surrealmode wrote:
Fri Oct 19, 2018 6:53 pm

Is the pinmod for overclocking a 400MHz Dothan to 533MHz the same that's needed for swapping in a 533MHz CPU into a 400MHz T42 motherboard? I found
That makes me wonder if there's a separate pinmod for swapping in native 533MHz CPUs.

I was also wondering - will the CPU power utilization be the same if it's running with the same multiplier at 400MHz vs 533MHz? Or will the max TDP be 75% (400 ÷ 533)? I'm thinking it would use less power but am not certain. If the CPU selects voltages based on the multiplier irregardless of which FSB it's running at, then it seems it would use significantly more power MHz for MHz.
I'm afraid that I've got no wisdom to offer here since I've never done either of the pinmods myself.

As I've said in the previous post, spend $5 on a PM750 and you'll know where you stand in general terms.

The only BIOS mod that I'm aware of when it comes to T42 (and previous T4x models) deals with whitelisting the wireless cards and SLIC 2.1, not with any added microcodes and stuff like that.
...Knowledge is a deadly friend when no one sets the rules...(King Crimson)

Cheers,

George (your grouchy retired FlexView farmer)

AARP club members:A31p,T43pSF

Abused daily: T61p

Work in progress: X60T

PMs requesting personal tech support will be ignored.

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#5 Post by surrealmode » Fri Oct 19, 2018 7:46 pm

ajkula66 wrote:
Fri Oct 19, 2018 7:28 pm
I'm afraid that I've got no wisdom to offer here since I've never done either of the pinmods myself.

As I've said in the previous post, spend $5 on a PM750 and you'll know where you stand in general terms.

The only BIOS mod that I'm aware of when it comes to T42 (and previous T4x models) deals with whitelisting the wireless cards and SLIC 2.1, not with any added microcodes and stuff like that.
Yep, I saw discussion of a microcode mod and perhaps someone bricking their laptop by attempting it, but no reports of success. There are existing utilities for Phoenix BIOS (e.g. PhoenixTool) to do things like SLIC mod - I don't know about microcode though. I've had success doing that on newer UEFI firmwares.

I did find the no-1802/SLIC BIOS several days ago after much searching. Someone was kind enough to re-post it on these forums here. It'd fallen off the net - those fly-by-night file sharing sites aren't reliable long-term storage. It's working fine.

I have a hardware flasher setup. If I do end up attempting to mod the BIOS myself, I'll make sure to keep a dump of the flash beforehand.

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#6 Post by Screamer » Sat Oct 20, 2018 5:45 am

surrealmode wrote:
Fri Oct 19, 2018 5:06 pm
Has anyone out there completed such a swap successfully and got Windows 10 (or unmodded Windows 8(.1)) running on a T42P or equivalent? If so, which CPU and steps did you take to get things running smoothly? I do not want to swap the motherboard with e.g. one from a T43.
I doubt anyone has, unfortunately. Although I did have an R50e that was using a Pentium M 725A to run Windows 8.1, but that was ages ago and I can not exactly recall the details well.
surrealmode wrote:
Fri Oct 19, 2018 6:53 pm
Is the pinmod for overclocking a 400MHz Dothan to 533MHz the same that's needed for swapping in a 533MHz CPU into a 400MHz T42 motherboard? I found The Ultimate Pentium-m Pinmod/Voltmod Guide from overclockers.com:

Code: Select all

Q: Can I use my 533 mhz processor to do the Pinmod?
A: No you need a 400mhz processor for it to work.
That makes me wonder if there's a separate pinmod for swapping in native 533MHz CPUs.
Yes, but that guide was only meant for laptops equipped with an Intel 82915PM or an Intel 82915GM memory controller hub. The T42(p) and before were equipped with an Intel 82855PM memory controller hub, which does not officially support the 533MT/s front-side bus speed. There isn't a separate pinmod meant for swapping in native 533MT/s processors either; the BSEL pin mod is the only way to do that.

Another way to force it to run on the unsupported 533MT/s bus, is to bridge a connection between pin 55 and pin 54 on the phase-locked loop chip. However this will overclock your RAM, so it was not advised as much as a pin-mod would have been.
surrealmode wrote:
Fri Oct 19, 2018 6:53 pm
I was also wondering - will the CPU power utilization be the same if it's running with the same multiplier at 400MHz vs 533MHz? Or will the max TDP be 75% (400 ÷ 533)? I'm thinking it would use less power but am not certain. If the CPU selects voltages based on the multiplier irregardless of which FSB it's running at, then it seems it would use significantly more power MHz for MHz.
No, it will not be the same because of the increased clock speed. The multipliers themselves do not change the power consumption, but the clock frequency and voltage does. The theoretical maximum TDP is not a way to measure power consumption, it was meant to measure the heat output (in watts) from the processor. If you want to measure power consumption, download the 533MT/s Pentium M datasheet and get the IccMax values multiplied by voltage. This is the correct way of measuring the theoretical maximum power consumption in watts.

The processor does select voltage values based on the multiplier setting, but it is only meant for a few and not all multiplier settings if I recall. The 1.6GHz Banias that I have in my R40, seems to do that.

Slightly more power though, the clock speed does not cause the power consumption to rocket through the roof. The voltage is the major factor that does that.
surrealmode wrote:
Fri Oct 19, 2018 6:53 pm
Hopefully someone can chime in about modding the BIOS with correct microcode. It'd make sense for the microcode to be loaded based on CPU ID metadata attached to each blob, but they might not have done it that way. In that case simply replacing the blob wouldn't be enough I think.
I took the R50e's microcodes out and placed them in the T42's ROM, here is the modified ROM. You might be right about it not being enough, but I believe it should work.

Remember though, I can't be held responsible for a brick. If it ends up as a brick, you have my apologies.

EDIT: Uh, I think it is recommended that you don't flash the BIOS with the ROM that I have provided. Recently I have noticed some posts saying that the C0 stepping/6D8 Dothan processors, mainly the 725A, 735A, and the 745A, seems to work. The only problem is a non-functional SpeedStep, which means the processor will always be running at its lowest multiplier (6x100) or 600MHz.

I suggest buying the 745A as you mentioned earlier, or go with a Celeron M 370/380/390. The Celeron M 370/380/390 processors always have NX and PAE enabled, and should be mostly cheaper than the Pentium M 'A' revisions.
Last edited by Screamer on Sat Oct 20, 2018 1:30 pm, edited 1 time in total.

kfzhu1229
ThinkPadder
ThinkPadder
Posts: 1164
Joined: Tue Sep 22, 2015 10:59 pm
Location: Toronto, Canada

Re: PAE/NX CPU swap in T42P (2373-HTU)

#7 Post by kfzhu1229 » Sat Oct 20, 2018 11:44 am

At this point I think replacing the motherboard with a T43p one makes more sense if you plan to do the BIOS mods anyways.
But a T43p would only run Windows 10 1607 well. Older versions aren't as matured and newer versions have problems with the ATI Radeon drivers and DirectX 9.0, while the latest 2 versions give you a BSOD for some odd motherboard compatibility issues.
If you plan to install Windows 10 1607, make sure you install it without any internet connected and then disable windows update right away.
ajkula66 wrote:
Fri Oct 19, 2018 5:26 pm
Welcome to the forum!

PM 780 will work, with or without pinmod (only required for it running at 533Mhz) and you should be able to install W10 without issues.

You can pick up a PM750 for $5 and give the concept a test drive.

I don't run W10 but would imagine that NHC would work, I have it running in W7 on my T43pSF.

Happy modding.
I don't think you can get NX and PAE working on a T42 motherboard as the motherboard itself doesn't support it.
_________________________________
T23 PIII 1.13ghz 1gb W7
2xT43 14.1" 2.26 SXGA+ 2gb 1*fp W10
T530i 15.6" i7 16gb fp W10
Flexview UXGA:
A30p PIII 1.2 1gb W7 (IDTech)
T43p 2.26 2gb fp W10

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#8 Post by surrealmode » Sun Oct 21, 2018 5:53 pm

Screamer wrote:
Sat Oct 20, 2018 5:45 am
EDIT: Uh, I think it is recommended that you don't flash the BIOS with the ROM that I have provided. Recently I have noticed some posts saying that the C0 stepping/6D8 Dothan processors, mainly the 725A, 735A, and the 745A, seems to work. The only problem is a non-functional SpeedStep, which means the processor will always be running at its lowest multiplier (6x100) or 600MHz.

I suggest buying the 745A as you mentioned earlier, or go with a Celeron M 370/380/390. The Celeron M 370/380/390 processors always have NX and PAE enabled, and should be mostly cheaper than the Pentium M 'A' revisions.
Thanks a lot for the detailed reply, this is extremely helpful information.

I see some discussion of the 133MHz PLL mod here here, but unfortunately the linked images have disappeared. Searching for the specific PLL chip (ICS950810) brings up this post (in Chinese) at thinkpad.com. Thankfully, there is an image of the PLL chip there.

So the steps for swapping in a 533MHz CPU and running it at the intended speed seem to be:
1.) Bridge pins 54 (FS0) + 55 (FS1) on the ICS950810. This corresponds to the 133MHz on the table from the ICS950810 datasheet (see below).
2.) Optionally flash SPD of the RAM so it runs at a lower speed (which ends up being the correct speed when the PLL is modified).
3.) Possibly mod BIOS to include correct microcode.

Image

This makes the 745A more appealing as it's not just a matter of putting a bit of wire jumper in the CPU socket as I'd hoped it would be.
kfzhu1229 wrote:
Sat Oct 20, 2018 11:44 am
At this point I think replacing the motherboard with a T43p one makes more sense if you plan to do the BIOS mods anyways.
But a T43p would only run Windows 10 1607 well. Older versions aren't as matured and newer versions have problems with the ATI Radeon drivers and DirectX 9.0, while the latest 2 versions give you a BSOD for some odd motherboard compatibility issues.
I would prefer *not* to swap out the motherboard. I just completed a tedious swap and USB repair of the laptop's original motherboard and was hoping to keep at least that much intact.

Regarding ATI drivers (for the FireGL Mobility T2), I had some fun :roll: modding the DriverHeaven Mobility Mod tool to properly alter the last Vista legacy release so they'd install on Windows 7. Here are the modded files (to be placed in Mobility Modder's Program Files directory, replacing the original files by that name). Maybe it'll end up being useful to someone down the road. I realized after the fact that I didn't do it quite right, but maybe once I get to reinstalling Windows 7 on the SSD I'll make a cleaner bundle.

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#9 Post by Screamer » Mon Oct 22, 2018 11:13 am

No problem.

Before you do go along with the phase-locked loop chip modification, be aware that this is not the only phase-locked loop used in the T4x series. Other than the ICS950810 chip which I do find a little common, the CY28346ZI-2 chip is another potential phase-locked loop that may show up in a T42. The pin-outs for the CY28346ZI-2 chip, looks identical to the ICS950810 chip's pin-outs. In other words, don't worry about the pin-outs being different, they are the same.

The only difference between these 2 chips, is that the CY28346ZI-2 is capable of being overclocked by using SetFSB, whereas the ICS950810 can not be overclocked with SetFSB or with any other program that modifies the phase-locked loop's frequency divider.

Step 1 seems right, but step 2 lacks a little information on the lower speed part. According to my previous experiences with a friend of mine that forced his R51 to use the 533MT/s bus speed, he always made sure that the RAM's speed is set to DDR266 using SPDTool. I believe this results in a DDR speed of approximately DDR355 (266.66 x 1.33), in comparison to the original untouched DDR speed of approximately DDR443 (333.33 x 1.33).

DDR355 isn't exactly the correct speed; just a stable speed that is far less likely to induce instability. Although I would strongly recommend to get a pair of DDR400 modules in this case, because DDR355 is still in the overclocked range for DDR333 modules.

As for step 3, give me some time to let my R50e be a guinea pig for microcode swapping. That is if you don't mind, because I don't think it is right giving you the extra work required to revive your T42p if the modified BIOS fails to function. Even with a hardware flasher, I would imagine that is still tedious to do.

EDIT: Should work, here is the modified ROM with support for the C0 stepping Dothan. There is only one drawback however; this BIOS can no longer accept an A0 stepping Banias, but the B1 stepping Banias is still accepted. Good luck with the BIOS flash, I hope it works.
Last edited by Screamer on Tue Oct 23, 2018 11:49 am, edited 1 time in total.

kfzhu1229
ThinkPadder
ThinkPadder
Posts: 1164
Joined: Tue Sep 22, 2015 10:59 pm
Location: Toronto, Canada

Re: PAE/NX CPU swap in T42P (2373-HTU)

#10 Post by kfzhu1229 » Mon Oct 22, 2018 10:30 pm

surrealmode wrote:
Sun Oct 21, 2018 5:53 pm
So the steps for swapping in a 533MHz CPU and running it at the intended speed seem to be:
1.) Bridge pins 54 (FS0) + 55 (FS1) on the ICS950810. This corresponds to the 133MHz on the table from the ICS950810 datasheet (see below).
2.) Optionally flash SPD of the RAM so it runs at a lower speed (which ends up being the correct speed when the PLL is modified).
3.) Possibly mod BIOS to include correct microcode.

This makes the 745A more appealing as it's not just a matter of putting a bit of wire jumper in the CPU socket as I'd hoped it would be.
How well in the end does that work though? I didn't bother with it because I have the Radeon 7500 inside my T42 and R51. Does that allow the older versions of Windows 10 setup to at least boot? And also I heard that if you are still considering Windows 10, the older AGP interface has more bugs in Windows 10 than the newer PCIe interface used in T43p motherboards.
surrealmode wrote:
Sun Oct 21, 2018 5:53 pm
Regarding ATI drivers (for the FireGL Mobility T2), I had some fun :roll: modding the DriverHeaven Mobility Mod tool to properly alter the last Vista legacy release so they'd install on Windows 7. Here are the modded files (to be placed in Mobility Modder's Program Files directory, replacing the original files by that name). Maybe it'll end up being useful to someone down the road. I realized after the fact that I didn't do it quite right, but maybe once I get to reinstalling Windows 7 on the SSD I'll make a cleaner bundle.
At least for the drivers for Mobility Radeon X300 and FireGL V3200 found on T43/p boards, the driver that Lenovo China provides works right away even in Windows 10 and I could not find any version that works better than that. Maybe compare your experience with your drivers to the ones on the driver archive of this forum as well as Lenovo China at think.lenovo.com.cn (Or I can find it if you find translating Chinese too difficult)
_________________________________
T23 PIII 1.13ghz 1gb W7
2xT43 14.1" 2.26 SXGA+ 2gb 1*fp W10
T530i 15.6" i7 16gb fp W10
Flexview UXGA:
A30p PIII 1.2 1gb W7 (IDTech)
T43p 2.26 2gb fp W10

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#11 Post by surrealmode » Wed Oct 24, 2018 9:40 pm

Screamer wrote:
Mon Oct 22, 2018 11:13 am
Before you do go along with the phase-locked loop chip modification, be aware that this is not the only phase-locked loop used in the T4x series. Other than the ICS950810 chip which I do find a little common, the CY28346ZI-2 chip is another potential phase-locked loop that may show up in a T42. The pin-outs for the CY28346ZI-2 chip, looks identical to the ICS950810 chip's pin-outs. In other words, don't worry about the pin-outs being different, they are the same.

The only difference between these 2 chips, is that the CY28346ZI-2 is capable of being overclocked by using SetFSB, whereas the ICS950810 can not be overclocked with SetFSB or with any other program that modifies the phase-locked loop's frequency divider.
Got it, thanks. I'll probably go with the bridged pins mod instead of using something like SetFSB, but good to know. I took a couple pics of the motherboard before I put it back together but I can't positively ID the PLL. Here is a gallery with a couple pictures of the board. You can zoom the images with a mouse wheel. I think I see an ICS chip to the left of the Intel chip, below the FireGL chip. I'll make sure to get better pictures when I open it back up.
Screamer wrote:
Mon Oct 22, 2018 11:13 am
Step 1 seems right, but step 2 lacks a little information on the lower speed part. According to my previous experiences with a friend of mine that forced his R51 to use the 533MT/s bus speed, he always made sure that the RAM's speed is set to DDR266 using SPDTool. I believe this results in a DDR speed of approximately DDR355 (266.66 x 1.33), in comparison to the original untouched DDR speed of approximately DDR443 (333.33 x 1.33).

DDR355 isn't exactly the correct speed; just a stable speed that is far less likely to induce instability. Although I would strongly recommend to get a pair of DDR400 modules in this case, because DDR355 is still in the overclocked range for DDR333 modules.
If I can find a cheap pair of DDR400 I'll go with that, otherwise I would at least make sure that it's stable.
Screamer wrote:
Mon Oct 22, 2018 11:13 am
As for step 3, give me some time to let my R50e be a guinea pig for microcode swapping. That is if you don't mind, because I don't think it is right giving you the extra work required to revive your T42p if the modified BIOS fails to function. Even with a hardware flasher, I would imagine that is still tedious to do.

EDIT: Should work, here is the modified ROM with support for the C0 stepping Dothan. There is only one drawback however; this BIOS can no longer accept an A0 stepping Banias, but the B1 stepping Banias is still accepted. Good luck with the BIOS flash, I hope it works.
Thanks for testing that. I'm taking a look at the differences between a hex dump of a backup made with the WINPHLASHER and the latest T42.ROM that you provided. There seems to be large differences between the two. I'm thinking some of it might be the CMOS configuration, some must the microcode, some appears to be the SLIC tables, and I'm not sure if yours had the no-1802 built into it either.

Ideally I'd like to merge those differences (at least the ACPI tables and no-1802) into the final image I flash.

I see a couple tutorials on how to update microcode in Phoenix images. There appears to be a utility cbrom195.exe detailed here, and an official one called Phoenix Bios Editor that is apparently a manufacturer utility. Did you use either of those or do something a bit lower level?

If needed I can dump current copy with WINPHLASH and upload it somewhere, or use the steps you took to insert the microcode into the dump. The version of 1RETDRWW that I flashed and have been using is located on these forums here.
kfzhu1229 wrote:
Mon Oct 22, 2018 10:30 pm
How well in the end does that work though? I didn't bother with it because I have the Radeon 7500 inside my T42 and R51. Does that allow the older versions of Windows 10 setup to at least boot? And also I heard that if you are still considering Windows 10, the older AGP interface has more bugs in Windows 10 than the newer PCIe interface used in T43p motherboards.
I got the legacy Vista drivers working OK in Windows 7 after using a modded Mobility Modder. From what I understand those should also work in Windows 10. It wouldn't be the end of the world if I wound up needing to use a generic driver, though.
kfzhu1229 wrote:
Mon Oct 22, 2018 10:30 pm
At least for the drivers for Mobility Radeon X300 and FireGL V3200 found on T43/p boards, the driver that Lenovo China provides works right away even in Windows 10 and I could not find any version that works better than that. Maybe compare your experience with your drivers to the ones on the driver archive of this forum as well as Lenovo China at think.lenovo.com.cn (Or I can find it if you find translating Chinese too difficult)
The last drivers released by IBM/Lenovo for this FireGL Mobility T2 were some ancient XP drivers. I checked the Lenovo China site and found the same there - nothing for Windows 7 or newer. As I'd mentioned I used modded "official" drivers from amd.com - those are the newest in existence AFAIK, and mostly identical except for the included .inf files.


In other news, I've successfully completed the first few steps of getting Win98SE and XP Pro dual-booting. I found the correct version of XP SP3 OEM and was able to activate it using the COA sticker on the bottom of the laptop. I did run into a snag with Windows XP seeing its own drive as D:\ instead of C:\ due to the FAT32 Win98SE partition being visible to it. I worked around that today using a bootloader called Plop Boot Manager which very conveniently will re-write the partition table depending on which OS you've selected from the menu.

Win98SE sees a single, bootable FAT32 primary partition located between 12-16384MB. WinXP sees a single, bootable NTFS primary partition located between 16416-49184MB. The odd gaps are to keep things aligned to 12MB chunks, which has been considered "safest" for the flash layout (erase block etc.) on the Samsung Evo SATA drives.

From what I understand Windows 7 and Windows 10 are a bit smarter than XP/98 in terms of recognizing which is their boot drive - but I will probably use Plop to isolate those from each other as well.

kfzhu1229
ThinkPadder
ThinkPadder
Posts: 1164
Joined: Tue Sep 22, 2015 10:59 pm
Location: Toronto, Canada

Re: PAE/NX CPU swap in T42P (2373-HTU)

#12 Post by kfzhu1229 » Wed Oct 24, 2018 9:57 pm

surrealmode wrote:
Wed Oct 24, 2018 9:40 pm
I got the legacy Vista drivers working OK in Windows 7 after using a modded Mobility Modder. From what I understand those should also work in Windows 10. It wouldn't be the end of the world if I wound up needing to use a generic driver, though.
Does your computer boot Windows 10 1607? You said you modded the motherboard to enable PAE and NX, and that is a good way to test whether your mod worked.
Well the downside with these newer Sonoma CPUs is that they run at a higher TDP and they give much shorter battery life. I remember my T42 lasts about twice as long as my T43p under light tasks such as PDF reading. You can notice the higher TDP when you are running modern applications that give the CPUs a good workout. So I guess in that case, it's your choice to either undervolt or overclock with SetFSB if you want more performance. I chose to go with undervolting it and it makes a pretty decent difference in terms of battery life both when idling and under full load and so that my V3200 graphics chip doesn't fry.
_________________________________
T23 PIII 1.13ghz 1gb W7
2xT43 14.1" 2.26 SXGA+ 2gb 1*fp W10
T530i 15.6" i7 16gb fp W10
Flexview UXGA:
A30p PIII 1.2 1gb W7 (IDTech)
T43p 2.26 2gb fp W10

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#13 Post by surrealmode » Wed Oct 24, 2018 11:23 pm

kfzhu1229 wrote:
Wed Oct 24, 2018 9:57 pm
Does your computer boot Windows 10 1607? You said you modded the motherboard to enable PAE and NX, and that is a good way to test whether your mod worked.
Well the downside with these newer Sonoma CPUs is that they run at a higher TDP and they give much shorter battery life. I remember my T42 lasts about twice as long as my T43p under light tasks such as PDF reading. You can notice the higher TDP when you are running modern applications that give the CPUs a good workout. So I guess in that case, it's your choice to either undervolt or overclock with SetFSB if you want more performance. I chose to go with undervolting it and it makes a pretty decent difference in terms of battery life both when idling and under full load and so that my V3200 graphics chip doesn't fry.
I haven't tried 1607, but 1809 boots to an error screen after a fairly long timeout. IIRC it specifically mentioned PAE, and the issue goes as far back as Windows 8.x. No mods to the motherboard have been done yet aside from replacing the USB port. The CPU is the real thing holding me back, though.

It'll be a C0 - not sure which one yet.

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#14 Post by Screamer » Thu Oct 25, 2018 8:14 am

surrealmode wrote:
Wed Oct 24, 2018 9:40 pm
I took a couple pics of the motherboard before I put it back together but I can't positively ID the PLL.
Looks like an ICS950810 to me, even with the poor image quality.
surrealmode wrote:
Wed Oct 24, 2018 9:40 pm
If I can find a cheap pair of DDR400 I'll go with that, otherwise I would at least make sure that it's stable.
I found these 2, below $11 offers for a pair of DDR400 modules: [1] [2]

If you were to use the same DDR333 modules, you would need to loosen the timings using SPDTool. I can't exactly remember how loose the timings had to be to bring it to a stable state at DDR355 though, but it was definitely dependent on the DDR333 module for sure.
surrealmode wrote:
Wed Oct 24, 2018 9:40 pm
Thanks for testing that. I'm taking a look at the differences between a hex dump of a backup made with the WINPHLASHER and the latest T42.ROM that you provided. There seems to be large differences between the two. I'm thinking some of it might be the CMOS configuration, some must the microcode, some appears to be the SLIC tables, and I'm not sure if yours had the no-1802 built into it either.

Ideally I'd like to merge those differences (at least the ACPI tables and no-1802) into the final image I flash.

I see a couple tutorials on how to update microcode in Phoenix images. There appears to be a utility cbrom195.exe detailed here, and an official one called Phoenix Bios Editor that is apparently a manufacturer utility. Did you use either of those or do something a bit lower level?

If needed I can dump current copy with WINPHLASH and upload it somewhere, or use the steps you took to insert the microcode into the dump. The version of 1RETDRWW that I flashed and have been using is located on these forums here.
No problem, but I can't assure you that it will allow the C0 stepping Dothan processors to utilize their NX capability. Part of the NX capability has to be handled by the BIOS, if I am not wrong. It can also be forcibly enabled by manually setting bit 11 in the processor's model specific register named 0xC0000080, but I don't think that might be possible.

The winphlasher backup ROM might be a little different from the stock flashable ROM, at least that is what I noticed on my end. It is no CMOS configuration difference as far as I know, but it is more of how the backup ROM doesn't contain the flashable binary that allows the ROM to be flashed using a BIOS update utility. I don't think the backup ROM thing applies to every BIOS in existence though, because some backup ROMs look identical to their stock flashable ROMs.

The BIOS that I have provided is actually a minor edit based on the no-1802 + SLIC 2.1 BIOS that you have currently on your T42p, I forgot to mention that. I did not use cbrom195.exe or Phoenix BIOS Editor, but yes, I did use something that is very low-level. I used HxD to modify the ROM and to get the microcodes swapped, it actually works very well for that purpose. On the other hand, cbrom195.exe is a hit or a miss, because some BIOS images do not support it for some reason. Phoenix BIOS Editor, generally only works for editing BIOS images made before 2006. I think it has something to do with the BIOS' structure or the compression method used to compress the BIOS, at least for the newer BIOS images. It is a tad bit annoying to swap or add microcodes with it though, which is why I typically use HxD for BIOS editing.
surrealmode wrote:
Wed Oct 24, 2018 11:23 pm
It'll be a C0 - not sure which one yet.
If you are still not sure, and you are looking for some possible options, have a look at this list:

Code: Select all

Celeron M 350J - SL86L/SL8MK
Celeron M 360J - SL86K/SL8ML
Celeron M 370 - SL86J/SL8MM/SL8K6
Celeron M 380 - SL86H/SL8MN
Celeron M 390 - SL8MP
Pentium M 715A - SL89U
Pentium M 725A - SL89T
Pentium M 735A - SL8BA
Pentium M 745A - SL8QZ
Pentium M 745 - SL8U6

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#15 Post by surrealmode » Fri Oct 26, 2018 1:04 am

Screamer wrote:
Thu Oct 25, 2018 8:14 am
Looks like an ICS950810 to me, even with the poor image quality.
Cool, so no SetFSB.
Screamer wrote:
Thu Oct 25, 2018 8:14 am
I found these 2, below $11 offers for a pair of DDR400 modules: [1] [2]
Thanks, I've got the cheaper of the 2 in my cart - both Singapore Micron chips and the cheaper one actually advertises a warranty. ePacket on both.
Screamer wrote:
Thu Oct 25, 2018 8:14 am
If you were to use the same DDR333 modules, you would need to loosen the timings using SPDTool. I can't exactly remember how loose the timings had to be to bring it to a stable state at DDR355 though, but it was definitely dependent on the DDR333 module for sure.
I'll probably get those Crucial DIMMs and leave the existing ones untouched so I can fall back if I run into unsolvable issues with the 533MHz or overclocking mods.
Screamer wrote:
Thu Oct 25, 2018 8:14 am
The BIOS that I have provided is actually a minor edit based on the no-1802 + SLIC 2.1 BIOS that you have currently on your T42p, I forgot to mention that. I did not use cbrom195.exe or Phoenix BIOS Editor, but yes, I did use something that is very low-level. I used HxD to modify the ROM and to get the microcodes swapped, it actually works very well for that purpose. On the other hand, cbrom195.exe is a hit or a miss, because some BIOS images do not support it for some reason. Phoenix BIOS Editor, generally only works for editing BIOS images made before 2006. I think it has something to do with the BIOS' structure or the compression method used to compress the BIOS, at least for the newer BIOS images. It is a tad bit annoying to swap or add microcodes with it though, which is why I typically use HxD for BIOS editing.
I actually realized that the Winphlash dump was a backup I made before flashing the modded BIOS, and I got the two hexdumps reversed when I diffed them. Sorry.
Screamer wrote:
Thu Oct 25, 2018 8:14 am
If you are still not sure, and you are looking for some possible options, have a look at this list:

Code: Select all

Celeron M 350J - SL86L/SL8MK
Celeron M 360J - SL86K/SL8ML
Celeron M 370 - SL86J/SL8MM/SL8K6
Celeron M 380 - SL86H/SL8MN
Celeron M 390 - SL8MP
Pentium M 715A - SL89U
Pentium M 725A - SL89T
Pentium M 735A - SL8BA
Pentium M 745A - SL8QZ
Pentium M 745 - SL8U6
I'm strongly considering getting the M 745A, but for just a few bucks more I can pick up an M 770, so I'll probably get both. I think those and the RAM should give me some flexibility to find a stable configuration.

eBay has a 10% sitewide going on in the Americas until 8AM (6 hours from now) which makes it even more enticing.

Edit: 745A, 770, 2x1GB 400MHz Crucial Ram, and a replacement MCF-205AM05 fan are on the way (should have the right hole/tab layout, "Made in Japan" etc. - most of them are for the newer T43 cooler it seems). It'll be a few weeks before most of it arrives, but the 745A should be here in a week or so. I might wait until the other parts get here before I pull it apart again.

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#16 Post by Screamer » Fri Oct 26, 2018 8:55 am

surrealmode wrote:
Fri Oct 26, 2018 1:04 am
I actually realized that the Winphlash dump was a backup I made before flashing the modded BIOS, and I got the two hexdumps reversed when I diffed them. Sorry.
I see, that is ok. Was the modded BIOS, the BIOS with the Dothan C0 microcode I gave you in the previous post, or was it the BIOS without the Dothan C0 microcode? I was wondering if it worked.
surrealmode wrote:
Fri Oct 26, 2018 1:04 am
I'm strongly considering getting the M 745A, but for just a few bucks more I can pick up an M 770, so I'll probably get both. I think those and the RAM should give me some flexibility to find a stable configuration.

eBay has a 10% sitewide going on in the Americas until 8AM (6 hours from now) which makes it even more enticing.

Edit: 745A, 770, 2x1GB 400MHz Crucial Ram, and a replacement MCF-205AM05 fan are on the way (should have the right hole/tab layout, "Made in Japan" etc. - most of them are for the newer T43 cooler it seems). It'll be a few weeks before most of it arrives, but the 745A should be here in a week or so. I might wait until the other parts get here before I pull it apart again.
Good idea, but why pick the fan from the T43? The only advantage it has in comparison to the T42's fan, is a higher maximum fan speed and a higher cubic-feet-per-minute rating.

kfzhu1229
ThinkPadder
ThinkPadder
Posts: 1164
Joined: Tue Sep 22, 2015 10:59 pm
Location: Toronto, Canada

Re: PAE/NX CPU swap in T42P (2373-HTU)

#17 Post by kfzhu1229 » Fri Oct 26, 2018 10:14 am

surrealmode wrote:
Wed Oct 24, 2018 11:23 pm
I haven't tried 1607, but 1809 boots to an error screen after a fairly long timeout. IIRC it specifically mentioned PAE, and the issue goes as far back as Windows 8.x. No mods to the motherboard have been done yet aside from replacing the USB port. The CPU is the real thing holding me back, though.

It'll be a C0 - not sure which one yet.
That may be a sign that your motherboard is still not supporting PAE and NX. If 1607 and Windows 8.x doesn't boot and give you BSOD for PAE, that means your motherboard is definitely not supporting PAE and NX. I guess that's fine really as long as you aren't using 8.x or 10 as your OS of coice. (1809 doesn't run on a T43p either due to a driver glitch when booting up from the setup, 1607 works fine)
_________________________________
T23 PIII 1.13ghz 1gb W7
2xT43 14.1" 2.26 SXGA+ 2gb 1*fp W10
T530i 15.6" i7 16gb fp W10
Flexview UXGA:
A30p PIII 1.2 1gb W7 (IDTech)
T43p 2.26 2gb fp W10

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#18 Post by Screamer » Fri Oct 26, 2018 12:08 pm

I am sure that it isn't the motherboard alone that decides or limits whether if PAE or NX can be used, it should be reliant on the BIOS and the chipset for it to properly support that. Assuming that the processor's registers can be edited before the installation of Windows 8.1/10, it should be possible to run it on a T42. Or, even on older platforms like the T41 and the T40.

As for NX, that should mostly depend on the BIOS if I am not mistaken. Intel did mention that NX is recommended 'only' for the 915/910 chipset derivatives, but I believe that is just a recommendation and not a strict requirement. In any way, even if the C0 stepping Dothan processors don't display their NX capabilities or just have it disabled, it should be theoretically capable of being enabled by modifying the register 0xC0000080 on the said processor.

I used that method of enabling NX on my G1 stepping Prescott ages ago, and that if I can recall right, worked. Though, I only tested that in Windows XP (with data execution prevention) and I don't remember using 8.1 or 10 for that matter.

Just a heads-up if anyone is interested in this PGA478 Prescott; it was a limited-production run unit that had EM64T and NX capabilities enabled on it for some reason. The S-Spec for that particular Pentium 4 was SL8JX.

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#19 Post by surrealmode » Fri Oct 26, 2018 8:25 pm

Screamer wrote:
Fri Oct 26, 2018 8:55 am
Good idea, but why pick the fan from the T43? The only advantage it has in comparison to the T42's fan, is a higher maximum fan speed and a higher cubic-feet-per-minute rating.
Actually, the majority of replacement "Long Fan" heatsink+fan combos advertised as being 13R2657 appear to be a newer T43 variant with a different heatpipe layout (even if the mounts/size overall is same/similar). The vast majority of replacement fans seem to target this model and have a different footprint making it difficult/impossible to mount them on the actual T42P heatsink. This can be seen in various comments on AliExpress e.g. the below picture (T42 model on left)
t42_t43_sidebyside.jpg

I meant to say I picked up the harder to find one on the left, which should actually fit my existing heatsink.

Edit: convert inline image to external link.
Last edited by surrealmode on Sat Oct 27, 2018 12:01 pm, edited 1 time in total.

Screamer
Sophomore Member
Posts: 233
Joined: Sun Apr 15, 2018 7:15 am
Location: Singapore

Re: PAE/NX CPU swap in T42P (2373-HTU)

#20 Post by Screamer » Sat Oct 27, 2018 5:56 am

Ah, no wonder. That outright confused me, I wasn't aware of the strange fan replacements that had incompatible mounts. Thanks for the clarification.

Also, you might want to remove the image and convert it into a link. The forum rules state that any images used in a post, should be below the 50KB limit.

surrealmode
Posts: 9
Joined: Fri Oct 19, 2018 4:17 pm
Location: Asheville, NC

Re: PAE/NX CPU swap in T42P (2373-HTU)

#21 Post by surrealmode » Sat Oct 27, 2018 11:58 am

oops, ignore - meant to edit not reply to myself :eek:

kfzhu1229
ThinkPadder
ThinkPadder
Posts: 1164
Joined: Tue Sep 22, 2015 10:59 pm
Location: Toronto, Canada

Re: PAE/NX CPU swap in T42P (2373-HTU)

#22 Post by kfzhu1229 » Sat Oct 27, 2018 11:01 pm

Screamer wrote:
Fri Oct 26, 2018 8:55 am
Good idea, but why pick the fan from the T43? The only advantage it has in comparison to the T42's fan, is a higher maximum fan speed and a higher cubic-feet-per-minute rating.
Well and it also have the downside of being a lot noisier because it spins faster. Since the T4x series have their fans turned to max speed during BIOS (Fan 64 mode), a T43p sounds like an aeroplane when it is posting. But if you wanna use a Pentium M 7x0 series CPU and a FireGL T2 on the T42p at the same time, the only way to keep it from throttling is to use a T43/T43p long fan or undervolting the CPU and use high quality thermal compound.
_________________________________
T23 PIII 1.13ghz 1gb W7
2xT43 14.1" 2.26 SXGA+ 2gb 1*fp W10
T530i 15.6" i7 16gb fp W10
Flexview UXGA:
A30p PIII 1.2 1gb W7 (IDTech)
T43p 2.26 2gb fp W10

Post Reply
  • Similar Topics
    Replies
    Views
    Last post

Return to “ThinkPad T4x Series”

Who is online

Users browsing this forum: No registered users and 20 guests