IBM Thinkpad 600E bios mod for processor update

[Dr. Morg]

Usually 600E floppy is robust as train's toilet. So I suggest do nothing until your new motherboard arrives. I you still after that keep having this floppy problem, let us know.

After mother board replacement FDD works great!
All troubles with USB, COM and IrDa also disapear!

Now, I have a Mother Board for various tests with it.
Also, I have two CPU MMC-2 PII 366Mhz, one modem card, one sub card, one dead battery with working controller and one keyboard with dead trackpoint, palm rest for keyboard and bottom side of laptop.

If anybody needs something, feel free to email me: Dr.Morg@gmail.com

If nobody needs something, I'll be glad to help all us with experiments with BIOS or others.

Just let me know!

Thanks!

[brokencase]

It's been a while since I have been at this thread. Looks like there is still plenty of 600e users who are performing this upgrade.

I looked through the posts and thought I would comment on a few items:

1) Heat. If you are considering the speed step mod then you should also consider switching to the 600x heatsink/fan assembly. If you are running at the non-speed step speed you should be fine with the standard heatsink but you should use a good heatsink compound when you install the new cpu. I am running the mmc-750 at 650mhz with the standard 600e heatsink with no trouble. I have been running this machine continuously 24-7 since I have performed the upgrade.

2) CD-ROM issue. I ran into trouble installing Arch linux OS from CD. This turned out to be caused by the 108 fsb mod. The CD works fine in all other respects. My solution was to wire a switch in line with the resistor so I can go back to normal fsb if I ever need to.

3) There is a cache enabling solution for Linux (see earlier posts). However EVEN WITH the module loaded, /proc/cpuinfo will still report no L2 cache. Do not be alarmed by this, the enabler has worked. It turns out that there is a a kernel command line option that will force /proc/cpuinfo to report the correct setting of L2. The occurs because the linux kernel expects the machines bios to enable the L2 cache.

thats all i can remember

btw, can the individual who posted the deep sleep state code also provide the source code?



[/list]

[beatoem]

i used arctic silver 3 with the stock 600e heatsink + 810mghz and it only handles 15mins max on stress tests,conversely 600x heatsinks dont seem to fare any better either.700mghz max is the most any 600 sereis heatsink can cope during extended 60 min + stress tests.
i am pondering on filing down the screw locator bungs around 0.010-0.015 of an inch to see if better clamp loads will help,one things for sure you cannot go too far with clamploads on mmc cpu because you can very well disloged the alloy plate (because of a pathedic design)
these cpus should have had reinforced rivets affixed on the top end of the heatsink not the lame fashion that the intel dumbster decided to hold the cpu cover plate (asides from the 2 and not 4 screws to hold the thermal mat correctly inshape(cheap and pathedic engineering) yuk

[Dr. Morg]

Confirmation: DeepSleep DOS version 1.2 works now perfect also from autoexec.bat

I'm try to start DeepSleep DOS version 1.2 from autoexec.bat in WinXP Pro SP2, but it looks like it's not start sometimes.

Why can it occur?

[wmarcusm]

Confirmation: DeepSleep DOS version 1.2 works now perfect also from autoexec.bat

I'm try to start DeepSleep DOS version 1.2 from autoexec.bat in WinXP Pro SP2, but it looks like it's not start sometimes.

Why can it occur?

Dr. Morg,

If you are running NT/2K/XP, then install the 2K/XP version. The DOS
version will only run correctly on an OS prior to winNT (DOS/3x/9x/ME).
The autoexec feature in NT/2K/XP only provides backward compatability
for non real mode applications (those that do not touch hardware).

[brokencase]

wmarcusm

How about posting the source so I can write a linux module?

[Dr. Morg]

Dr. Morg,

If you are running NT/2K/XP, then install the 2K/XP version. The DOS
version will only run correctly on an OS prior to winNT (DOS/3x/9x/ME).
The autoexec feature in NT/2K/XP only provides backward compatability
for non real mode applications (those that do not touch hardware).

Whether is there other way to start DOS version before starting Windows?
There are set FireWalls, which are started before OS kernel start, why not to start DOS version also, it's would allow to boot OS already on the increased frequency and consequently, - faster!

It would be healthy, if it was possible to enable L2 Cache also before loading of OS kernel.

What can you tell on it?

There is only one nuance, - if from Deep Sleep utility we have a source code and it can be rewrote at any moment somehow for enabling L2 Cache of source codes is not present!: (

wmarcusm:
Can you find how we can enable L2 Cahce?

In turn, I'll try to look as it is possible to start the application before OS kernel boot.

[Dr. Morg]

I have excellent news!

I have found how to start Deep Sleep Utility in zero ring OS before loading OS kernel!

I need a source code of Deep Sleep Utility, or only its part is necessary for this purpose on Assembler which carries out translation CPU in Deep Sleep Mode and its return to a normal condition.

wmarcusm, if you'll give me a source code, I'll write the driver for start Deep Sleep Utility before loading OS kernel!

[Sharedoc]

I will be off-internet for next 8 weeks on a expedition to wilderness. Back in August.

[wmarcusm]

wmarcusm

How about posting the source so I can write a linux module?

brokencase,

I've added the source files for the DOS/Win3x/Win9x versions to the
download page http://home.pacbell.net/wmarcusm/deepsleep/

Most of the code is either devoted to finding the correct PCI device
(pxii4x), of computing the CPU frequency. The code that sets up the
deepsleep transition is in dsleep.c near the bottom of the file. You will likely want to use the Linux PCI routines to do the bridge lookup.

[brokencase]

wmarcusm,

Thanks. I'll see what I can do on the linux side...

[Dr. Morg]

brokencase,

I've added the source files for the DOS/Win3x/Win9x versions to the
download page http://home.pacbell.net/wmarcusm/deepsleep/

Most of the code is either devoted to finding the correct PCI device
(pxii4x), of computing the CPU frequency. The code that sets up the
deepsleep transition is in dsleep.c near the bottom of the file. You will likely want to use the Linux PCI routines to do the bridge lookup.

wmarcusm, can you post source code for Win2k/WinXP version?
Is it differ to enable Deep Sleep on Win98 and WinXP?

[wmarcusm]

brokencase,

I've added the source files for the DOS/Win3x/Win9x versions to the
download page http://home.pacbell.net/wmarcusm/deepsleep/

Most of the code is either devoted to finding the correct PCI device
(pxii4x), of computing the CPU frequency. The code that sets up the
deepsleep transition is in dsleep.c near the bottom of the file. You will likely want to use the Linux PCI routines to do the bridge lookup.

wmarcusm, can you post source code for Win2k/WinXP version?
Is it differ to enable Deep Sleep on Win98 and WinXP?

I thought you wanted to do the deepsleep transition at processor ring 0? The 2K/XP application runs in protected mode, and the piix4.sys driver does exactly what the DOS version does. The code to do the deepsleep transition is the same in all the versions. The only difference is how the PIIX4 is located. In the DOS/Win3x/Win9x version it is done by directly reading the PCI config registers and in the Win2K/XP version it is
done by searching the registry. You should have all the information you need to write a ring 0 device driver.

I'm a bit confused by the utility of what you propose however? Is it simply to place the processor in the high performance state earlier in the boot process? If so, this seems like a significant effort for very little return (a few seconds saved during boot)?

I have a version I'm testing and will release in the next few days that returns the processor to the high performance state after a suspend or hibernate. I'm hooking into the Power Management vector, and doing a deepsleep on a resume from suspend or hibernate event.

[wmarcusm]

It would be healthy, if it was possible to enable L2 Cache also before loading of OS kernel.

What can you tell on it?

There is only one nuance, - if from Deep Sleep utility we have a source code and it can be rewrote at any moment somehow for enabling L2 Cache of source codes is not present!: (

wmarcusm:
Can you find how we can enable L2 Cahce?

Dr. Morg,

I looked at the Linux device driver module brokencase did for the L2 cache enable (see earlier posting - this thread). It should be pretty easy to port that code to the DOS and Windows drivers.

To be effective it would also need to be called after suspend and hibernate events (since the L2 cache apparently is disabled on both of these events - at least on my TP600 ).

[Dr. Morg]

I thought you wanted to do the deepsleep transition at processor ring 0? The 2K/XP application runs in protected mode, and the piix4.sys driver does exactly what the DOS version does. The code to do the deepsleep transition is the same in all the versions. The only difference is how the PIIX4 is located. In the DOS/Win3x/Win9x version it is done by directly reading the PCI config registers and in the Win2K/XP version it is
done by searching the registry. You should have all the information you need to write a ring 0 device driver.

Yes, I want to do the deepsleep transition at processor ring 0!
And I know how to do it!

Is piix4.sys driver yours?
If so, have you source code of it?
If no, where have you found it?

I'm a bit confused by the utility of what you propose however? Is it simply to place the processor in the high performance state earlier in the boot process? If so, this seems like a significant effort for very little return (a few seconds saved during boot)?

I have many services on my TP600E, that's why I need enabled SpeedStep and L2 Cache on Windows boot.
Now Windows boot takes too much time!

I have a version I'm testing and will release in the next few days that returns the processor to the high performance state after a suspend or hibernate. I'm hooking into the Power Management vector, and doing a deepsleep on a resume from suspend or hibernate event.

That's good!

Will be waiting for release!

Good Luck!