Install a 1066MHz FSB CPU on T61 (including Core 2 Quad)

[RMSMajestic]

@el-sahef
Congratulations on your Q9000 success!!!!!
I'll read your threads and investigate when I'm free (I'm a student of a university which has the lowest student satisfaction rate in UK :/. It's my final year now and I really need to get a decent mark). This is indeed very very very tempting!!! a QX9300, with only 2 cores running, is the same as a 200 EURO E0 stepping X9100.

I'll also contact leokim, or the person who modified this BIOS, or try to find some one in my uni who can modify BIOS.

This website might help
http://www.endeer.cz/bios.tools/bios.html

Would there exist a probablity that T61 could not supply sufficient current to start up the CPU, As for most C2Q laptops there are 3 CPU chokes?
Also, instead of short-circuiting BSEL[1] and VCC, I'm thinking of tweaking R1025, pull up on the CPU side and pull down on the NB side, so that in BIOS it will display the original frequency. But I have neither the time nor the facility to do this. :/

P.S. I've got 3 more questions:
1. Will the dual Core CPU work after your "quad core modification"?
2. could you please post an image of all the soldering/insulating/wiring?
3. Were you able to get the other 3MB cache of your Q9000 work?

P.P.S. I like your desktop background

[RMSMajestic]

@FryPpy
He downclocked the external clock. so that the top frequency is 200Mhz*7.5= 1500Mhz. multiplier is not capped. If the external clock is not downclocked, the RAM can't hold.

[el-sahef]

I got the Q9000 to work with all four cores active today, so no further BIOS modding seems to be needed. To achieve this, I loaded an edited APIC table via GRUB2. This will not work on T500 or W500.

Unfortunately it works not primestable yet and cooling is a huge problem. But currently I do not have a T61-cooler or T500-cooler available to test, so it might work with these and some undervolting.
I am trying to sort out the stability problems.

You can read all about it here (just use qoogle translate):
http://thinkpad-forum.de/threads/172581 ... en-bringen

P.P.S. I like your desktop background

Get it here: http://www3.zippyshare.com/v/51125405/file.html

[PowerPC]

I got the Q9000 to work with all four cores active today, so no further BIOS modding seems to be needed. To achieve this, I loaded an edited APIC table via GRUB2. This will not work on T500 or W500.

Unfortunately it works not primestable yet and cooling is a huge problem.

This is very important, you should make a new thread about it.

[RMSMajestic]

@el-sahef
This is indeed the greatest news of the year!!!! absolutely the greatest!!! 7 years after the release of T61, us thinkpad modders finally got there!!!!! And I think you should indeed start a new thread!!!! For entire 7 years there has been no successes on 965/GM45 quad core mod at all, and you are the first one who made it!!!!! congratulations!! You have successfully slapped lenovo right in the face!!!!

Regarding the heat issue, I think there might be two possible causes:
1. I can't see clearly but I suspect that you haven't polished off the screw holes of the modem, these two screw holes prevents the motherboard to fit perfectly with the roll cage, and hence can affect the contact between heatsink and CPU
2. I agree that T60p heatsink is too weak. Try those W500 heatsink. I'll ask my chinese friends to bring a lot of W500's heatsink pressure pieces this Easter


Regarding stability issue:
1. I can see that your Frankenpad has intel video card----Are you using a 65W power supply? When I was running my T61 with T9900 on 65W power supply it wasn't stable at all (CPU clock frequency alters a lot). Try it with 90W or 135W PSUs and see.
2. I have just disassembled my friend's Samsung R710, which came natively with Q9000, and there was ONLY two CPU chokes. So perhaps it's no tthe problem with these. Maybe it will become a problem when overclocking QX9300, but shouldn't matter when stressing a Q9000, which has a peak current about 11A(20%-25%) greater than X9100.

And I still have one question:
Will the Dual Core CPUs work after the quad core mod?

P.S. Undervolting examples:
http://forum.notebookreview.com/hardwar ... s-121.html

[el-sahef]

1. I can see that your Frankenpad has intel video card----Are you using a 65W power supply?

No, I use a 90W power supply and a charged battery is also attached.

I can't see clearly but I suspect that you haven't polished off the screw holes of the modem,

The contact to the heatsink seems to be okay. I will get a heatsink entirely made of copper from a T61 in a few days.

Will the Dual Core CPUs work after the quad core mod?

Yes, they do still work because pin D22 is "reserved" on these.

Some other news:
It seems to be that the FSB mod to 266 MHz is not stable with my hardware yet because a test with the E8335 SSpec SLGEB (same as T9800) I have was also not stable. Since FSB-1066-CPUs are stable in other people's T61 units, it is known that the chipset should be stable on FSB 1066.

After discovering this, I lowered the FSB to 200 MHz again, used some FSB-800-RAM where I did not edit the SPD and prime95 "in place lager FFTs" ran for 45 minutes already on all cores without errors. I will do some further testing, but at the moment, I highly suspect that I did something wrong when editing the SPD timing table for the RAM, so that it will produce the errors when running on FSB 1066. Will give an edit / update when this is verified.

This would also explain why "small FFT" test ran without errors for over two hours since the RAM is not tested much in this test but with "in place lage FFTs" and "blend", RAM is tested more.

So the probability that a QX9300 with FSB 800 and the multiplier changed to match the original speed of 2,5 GHz will work is high. But I do not have one to test.

EDIT:

Seemed to be the RAM or the chipset. I changed the SPD of the RAM to DDR2-677 with latencies from the DDR2-800 profile and the Q9000 works stable for two hours already with the "in place lage FFT" test on 2 GHz with FSB 1066. Before, I had the SPD flashed to DDR2-533 with latencies from the DDR2-800 profile. So either I did a fault when editing the SPD for DDR2-533 or the FSB:RAM-divider 3:4 of the chipset is not stable (but the 3:5-divider is):

http://abload.de/img/t61-q9000-02rejy5.jpg (500kB external image hoster link)

I will let the test run for some more time to make sure, but I think that the stability problems are solved now. Cooling might still be an issue.

[RMSMajestic]

I think you 've probably forget to change the latency for memory write.... And I'll ask my friend to bring 80 GBP-Qx9300s from China this easter, do you fancy one?

[PowerPC]

Would regular DDR2-1066 memory work?

[FryPpy]

Would regular DDR2-1066 memory work?

Where did you find it (in SODIMM not DIMM form factor)?

And I'll ask my friend to bring 80 GBP-Qx9300s from China this easter, do you fancy one?

Today there was one QX9300 for $152 on ebay... but gone:(

In one computer magazine of those days was an article comparing mobile quad core processors. And it shows that QX have better thermal design than Q9000 on it's base frequency. I can't find link for that article

[PowerPC]

Where did you find it (in SODIMM not DIMM form factor)?

I read a mention of a Finnish online store, "Jimm's PC", that once had a kit called "A-DATA SODIMM DDR2 1066 2X2GB 128MX8 CL5" (source here). This probably used hand-picked chips.

I also found this answer elsewhere:
"The underpinning for why PC2-8500 DDR2 memory is not available as 200pin SODIMM memory modules rests in its technical specification. The original JEDEC (the institution that develops memory standards used in computers) specification called for an operating voltage of 1.8V. At the time that PC2-8500 DDR2 desktop memory was launched, the (then) present manufacturing capabilities required an increase in voltage to 2.2V in order to grant stability. The by-product of this was not only an increase in power consumption but also produced heat. These two traits voided use of PC2-8500 DDR2 memory in laptops where low power consumption and heat output is the norm. By the time that the manufacturing process improved (shrunk) laptops had already started the shift to DDR3 memory, which uses less voltage at a JEDEC specification of 1.5V."

So... increase voltage a little? *BOOOOOM*

I was sure I read about it somewhere, I didn't remember it was so rare. I got the ThinkFever only around last year, I never dealt with laptop hardware before that, only with desktops.

[RMSMajestic]

Today there was one QX9300 for $152 on ebay... but gone:(

In one computer magazine of those days was an article comparing mobile quad core processors. And it shows that QX have better thermal design than Q9000 on it's base frequency. I can't find link for that article

I saw it, was about to buy but realised that the UK gov rips me off with 25 GBP import tax.....

And yeah you can probably run a QX9300 @3.06Ghz on 1.07V and it will be better than Q9000 in both performance and temperature
But QX9300 runs 2.5Ghz @ 1.225V I think..... With 6MB more cache it will be NO WAY cooler than Q9000 on stock settings

[RMSMajestic]

Where did you find it (in SODIMM not DIMM form factor)?

I read a mention of a Finnish online store, "Jimm's PC", that once had a kit called "A-DATA SODIMM DDR2 1066 2X2GB 128MX8 CL5" (source here). This probably used hand-picked chips.

I also found this answer elsewhere:
"The underpinning for why PC2-8500 DDR2 memory is not available as 200pin SODIMM memory modules rests in its technical specification. The original JEDEC (the institution that develops memory standards used in computers) specification called for an operating voltage of 1.8V. At the time that PC2-8500 DDR2 desktop memory was launched, the (then) present manufacturing capabilities required an increase in voltage to 2.2V in order to grant stability. The by-product of this was not only an increase in power consumption but also produced heat. These two traits voided use of PC2-8500 DDR2 memory in laptops where low power consumption and heat output is the norm. By the time that the manufacturing process improved (shrunk) laptops had already started the shift to DDR3 memory, which uses less voltage at a JEDEC specification of 1.5V."

So... increase voltage a little? *BOOOOOM*

I was sure I read about it somewhere, I didn't remember it was so rare. I got the ThinkFever only around last year, I never dealt with laptop hardware before that, only with desktops.

Probably not. The DDR2 1066 for laptops is very rare, I think only Kingston and A-DATA made some. And also, at 667Mhz it's latency is 4-4-4-13/14 instead of ordinary 5-5-5-15 and it will be overclocked to 888Mhz under that latency..... So perhaps you still need to flash spd.

[PowerPC]

(...) the UK gov rips me off with 25 GBP import tax.....

It could be worse, you know? At least it is not like our glorious Brazilian government, with its 100% import taxes, and if the value exceeds 500 USD, extra paperwork and a ~50 USD extra charge. Importing items with a price over $3000 demands... eh, better just forget it.

[el-sahef]

With the RAM now at DDR2-667, the "in place large FFT" test ran for 4 hours and the "blend" test for 6 hours without errors, so I think that it is stable now. Screens (these were with a PC fan blowing onto the heatsink, so ignore the temperatures):

http://abload.de/img/4std.inplacelargefftajdcu.jpg (500kB external image hoster link)
http://abload.de/img/65std.blendcwcoq.jpg (500kB external image hoster link)

Today, the T61 cooler arrived (all copper and goes up to 4500 rpm):

http://abload.de/img/khlereingebautyleyp.jpg (250kB external image hoster link)

Its performance is notably superior compared to the R500-cooler and the T60-cooler. But it seems that many (if not all) Lenovo coolers have a concave surface, so that contact is only good at the outer corners of the CPU. I also broke off all the outer corners from the CPU DIEs (but it still works). These splitters also made some serious scrates in the cooler surface, which I had to sand of with sandpaper of 1200 grain size.

http://abload.de/img/quadabgebrocheneeckenmbeze.jpg (225kB external image hoster link)
http://abload.de/img/khlerbodenausbesserunykdaa.jpg (190kB external image hoster link)

The temperature is between 80 and 84 °C at 2 GHz and stock voltages when running "in place large FFT" test at fan speed 64.

http://abload.de/img/temperaturkni8x.jpg (500kB external image hoster link)

When undervolted to 1,050 V at 2 GHz, temperature seems to be around 75 °C, but I did not run it that long with this setting yet. There might still be some more room for improvement.

I think that planning the cooler surface could make a huge difference but this "ring" makes it difficult. I do not know if the material is thick enough, so that one can sand it of completely without weakening the stability of the cooler. Using better thermal paste and finding a way so that the cooler gets more fresh air from the outside and not hot air from the inside could also help.

Since heat seems to be the only issue now, I think that if the cooling gets all improvements that are doable, someone could also test the higher clocked quad cores.

I'll ask my friend to bring 80 GBP-Qx9300s from China this easter, do you fancy one?

Will think about it . Depends also on the price. Maybe you could send me a PM when you are absolutely sure that you can get some. Cooling this down at full load will need some work and even then, it is not sure if it will be enough until someone tests it.

[FryPpy]

Today, the T61 cooler arrived (all copper and goes up to 4500 rpm):

But it seems that many (if not all) Lenovo coolers have a concave surface, so that contact is only good at the outer corners of the CPU. I also broke off all the outer corners from the CPU DIEs (but it still works).

Spring has come! And not only the flowers bloom but great ideas!


Some notes on cooling:
When i tested Q9000 i have used T61 HSF from discrete graphics. And it seems that it not injure CPU.
Take a look at this pictures:

Both T61 integrated an discrete HSF. Integrated HSF has better fan, but discrete has 2 heatpipes!
http://www.imagesup.net/?di=11139421782910

CPU Contact surface:

Integrated:
http://www.imagesup.net/?di=13139421779912

Discrete:
http://www.imagesup.net/?di=013942174308
YES - it have 1+mm more in diameter and smaller cylinder boundaries.

I tested once more - Q9000 can move inside cylinder 1 or 2mm without touching cylinder boundaries.

MHO it is better make hybrid HSF for C2Q. Use T61 discrete radiator (heatsink) (T(W)500 worse - it have smaller left radiator) and T500 or T61 int. fan.

do you fancy one

And YES - if you really can get QX9300 you definetly must PM me.

[el-sahef]

Since this whole thing seems to be in a state now where you can call it stable, I will try to explain the essential contents from this thread here in english.

FSB 1066 CPUs including Core 2 Quad on T61 mainboard:

1) Flash RAM SPD and modified BIOS:

a) Use DDR2-800 RAM, because after the FSB change, the RAM will run at 440 MHz (DDR2-880). Maybe you will need to edit the SPD of your RAM, so that it uses the timings for the DDR2-800-profile also in the DDR2-667-profile (which is the one that is used by the BIOS) so that it works stable at 440 MHz.

Edit (more detailed):
After the hardware modification to enable the possibility of using FSB 1066, the RAM will work at nonstandard 440 MHz instead of 333 MHz (if a FSB 1066 CPU is used - otherwise, it will still work with the normal speeds of 333 MHz (assuming DDR2 667 or DDR2 800 sticks installed). Without editing of the RAM SPD, most RAMs will not run stable with the higher frequency. Also, if the RAMs are DDR2-800, the BIOS will still use the timings from the DDR2 667 profile. Therfore, the SPD has to be edited in advance to force the BIOS to use higher timings and a lower FSB-to-RAM frequency-divider.
For flashing, there are edited profiles for 2 GB sticks and 4 GB sticks with 16 chips available here. Use the instructions in this post to flash them to your RAM. DDR2-533-profile is strongly recommended for boards with the Intel X3100 GPU, because the FSB-to-DRAM divider also affects the IGP clock. With FSB 1066 and the DDR2 667 profile, the IGP is overclocked by 166 MHz (render clock) and 110 MHz (display clock) which can sometimes lead to artifacts on the screen when running windows aero theme or graphics intensive stuff. With FSB 1066 and the DDR2 533 profile, the IGP is overclocked only by 33 MHz (render clock) and 92 MHz (display clock) which should never be a problem.
The DDR2 667 profiles will force the BIOS to run the RAM at 440 MHz and CL6 timings. This might work stable for DDR2 800 RAMs. The DDR2 533 profiles forces the BIOS to run the RAM at 354 MHz and CL6 timings, which should almost always work stable for DDR2 800 and DDR2 667 memory sticks.

b) Create a bootable USB pen drive that boots to bare DOS. Download the modified xiaofei290-BIOS from here or here, extract the archive and copy the "7luj26us"-folder on the USB pen drive. Be aware that the modded BIOS does not contain the microcodes for all the CPUs that were originally used in T61, so you should already have your new FSB-1066-CPU, because the old CPU will most likely not work anymore after the BIOS flash.

Edit: With the initial mod BIOS from xiaofei290 @ 51nb.com, Merom CPUs did not work any more in the T61 (only FSB 800 and FSB 1066 Penryn CPUs + T7100) because it lacked their microcodes. There are better mod-BIOS-files available now from highsun @ 51nb.com that do not have this limitation and have further advantages. They are also mentioned in 4). With those BIOS files, all socket-P-CPUs should work. It also enables the use of some FSB 800 CPUs (e. g. T6600) that did not work with the original Lenovo BIOS, without any hardware modification.
Some of them can also undervolt the nvidia GPU and activate PCI express power mangement functions (ASPM). All versions are available with a modified quadcore APIC table (for use with quad cores) and without it (for use with dual cores).

Download one of the BIOS files (which are linked here in the first table) acording to your needs regarding quadcore APIC, nvidia GPU voltage and ASPM and the DOS flasher program.

Extract the flasher and the downloaded BIOS file. Create a new folder with the name "DIR" and put in the extracted files (besides "liesmich.txt"). Rename the extracted BIOS file to "BIOS.ROM". The folder should now contain three files:

Code: Select all

go.bat
phlash16.exe
BIOS.ROM

[/size]
Copy this folder to the bootable USB pen drive you created. Boot from the USB pen drive into DOS, cd to the "7luj26us"-folder DIR folder

Code: Select all

cd DIR

[/size]
with the "BIOS.ROM" file and the flasher in it. Then type:

Code: Select all

phlash16.exe BIOS.ROM /S /X /C /MODE=3 /BO=BACK.ROM

[/size]
which will backup your current BIOS to BACK.ROM and then flash the modded BIOS.ROM. I did not try it, but you should be able to revert to the old BIOS with

Code: Select all

phlash16.exe BACK.ROM /S /X /C /MODE=3

[/size]
Alternatively, you can just type

Code: Select all

go.bat

[/size]
to immediately start flashing the mod BIOS without backing up the old BIOS.
The flash tool told me something about a wrong checksum before flashing, but I proceeded anyway and it worked, so this seems to be related to the modifications done to the BIOS.
Power off after the flash and...

2) Modify the mainboard:

http://abload.de/img/266-fsbe2cow.jpg (240 kB image hoster link)

Cut the yellow marked trace, then connect the marked side of the resistor with the marked side of the capacitor. This ensures that the chipset will always think that it operates at 200 MHz, which is needed. There is no need to revert this changes if you want to use a CPU made for FSB-800 when doing the modification this way (but the BIOS has to contain the microcode for the CPU).

Now you should be able to use FSB-1066-CPUs. If you have the hardware, you should test stability now with a normal FSB-1066-Core2, so that you can be sure that your RAM and chipset work stable at the higher speeds before proceeding with the next steps required for quad core support.

3) (only for Core2 Quad) Modify the mainboard further and insulate CPU pins:

Core2 Quad needs GTLREF (which is 0.63*VCC1R05B) at pin D22, which is reserved on dual cores. We can access the socket contact for this pin at the via shown in this picture:

http://abload.de/img/t61-mainboard-quadcor4qep0.jpg (210 kB image hoster link)

Connect a 2k resistor to this via and GND and another resistor with value of 1k to this via and VCC1R05B. This is how Lenovo does it in the W700.

Edit: Here is a much simpler solution suggested by FryPpy, which is confirmed to work.

To boot up with a quad, this is not even needed (only pin isolation), but i did not test if this is also the case when all four cores are activated (i do not think so).

Besides this, a core 2 quad has some more signals that seem to be used for testing when designing motherboards for this CPU and are not needed for normal operation. But some of the pins with these signals on a quad are connected to VCC or VSS on dual cores, so we have to insulate them on our quad core when inserting it in a mainboard made only for dual cores.
These are D8, AA7, AA8, AC8 and AE8.

I used the insulation of thin wires, cutted it into pices, so that I got small tubes which I could stick onto the pins with some tweezers.

http://abload.de/img/quadcore-pinsy1dm5.jpg (240 kB image hoster link)

Now the corresponding holes in the socket have to be drilled up. I opened the four clips at the side of the socket that hold down the top with the holes in it and placed a thin sheet of wood between the two parts of the socket, so that no chips will get into the socket. Do NOT try to drill the holes up without hanging out the top of the socket!

http://abload.de/img/bild3mns98.jpg (150 kB image hoster link)
(this image shows T500 board, but it is done the same way on T61 board)

Put the socket together again:
http://abload.de/img/sockelt6126e71.jpg (225 kB image hoster link)

For instructions how to hang out the top, see this post (contains also images). Do NOT drill the holes without hanging out the top, this will damage the socket contacts!

Another possibility is to cut off the mentioned pins from the CPU.

You should now be able to insert the quad core with the insulation tubes on the pins (or with the pins cut off) into the socket and lock it. After installing the cooler, booting with the quad core should work.

4) (only for Core 2 Quad) Activate all four cores:

Edit: With the original mod-BIOS from xiafei290 @ 51nb.com, it was needed to load a modified APIC table with the acpi command from the GRUB2 bootloader, because otherwise core 2 and 3 would not be activated (only core 0 and 1).

Meanwhile there are new mod-BIOS-files available @ 51nb.com that contain this modified APIC table, so the grub2-stuff is no longer needed to activate the two additional cores. Flashing the modified BIOS-files with the quadcore APIC table is sufficient. Credit to highsun @ 51nb.com who made these mod-BIOS-files.
They are linked here: http://www.thinkpad-forum.de/threads/172581
Some of them can also undervolt the nvidia gpu and activate PCI express power mangement functions (ASPM). All of them are also available without the quad core APIC table (for use with dual cores).
There will be an error message about wrong checksum while flashing them. This is normal. Ignore it and proceed with the flash.

Old quadcore activation method with the use of the original mod-BIOS from xiaofei290 left here for reference, but now obsolete:

The BIOS does contain the microcode for the quad cores, but to the operating system it will only notify two CPU cores. To make the operating system aware of the two additional CPUs, one needs to load a modified APIC table with the "acpi" command of the grub2 bootloader. This way, every operating system that will be booted by GRUB2 should be able to recognize all four CPUs (tested: linux and windows 7).

I will not go into the details here, if you want to know more, have a look at this thread: http://thinkpad-forum.de/threads/172581 ... en-bringen

Just download and extract this zip file: download

It contains the modified APIC table that we need to load with GRUB2. I assume that you have installed a working linux with GRUB2 and maybe also Windows 7, which you boot via GRUB2.
I use opensuse 13.1, for other linux distributions, it might be necessary to use other paths and commands.

Copy the skript "01-apic" from the zip file to "/etc/grub.d" and make it executable. For other linux distributions, it migth be necessary to change the path to "grub-mkconfig_lib" in the skript.
In the directory "/boot" create a directory "ACPI" and place the file "APIC.aml" in it. Then generate an new GRUB2-config by executing (as root)

Code: Select all

/usr/sbin/grub2-mkconfig --output=/boot/grub2/grub.cfg

[/size]
that will load the APIC table. You should see some output like "found custom APIC table" when creating the new config.

After a reboot, all four cores should be active.

EDIT:
5) (only for Core2 Quad) enable speedstep and maximum clock on cores 3 and 4
Because the modified BIOS does not activate speedstep and expose the different power/clock-states to the OS (e. g. in windows device manager, there are only two CPU cores instead of 4), this has to be enabled manually via 3rd party software. Otherwise, the cores 3 and 4 always operate at the lowest frequency. Credit to RMSMajestic for discovering this and for the method to circumvent the problem in windows.

For windows you can use the software throttlestop for this.
For linux, see the instructions posted here (contains also images)

Why it does not work on T500 and W500:

This information is based on tests with a T500 and an R500 mainboard. But since T500 and W500 use the same mainboard (and the same BIOS), I think that this information is also valid for W500 (and probably also T400 and R400, but this is not verified)

Generally, you can also get into the BIOS with a quad core installed in a T500. Only pin insulation is required (and if it would be able to activate all four cores, GTLREF modification would most likely also be needed). The BIOS seems to already contain the microcodes (at least the one for Q9000), since the processor was recognized as Q9000 in the BIOS.

http://abload.de/img/bild8yss96.jpg (135 kB image hoster link)

But to be able to boot an operation system, you will have to disable "core multi processing" in the BIOS because otherwise, the BIOS will hang while trying to continue the boot process to a bootloader like GRUB2.

http://abload.de/img/bild95hstz.jpg (145 kB image hoster link)

But with this option disabled, only one core is usable.

http://abload.de/img/bild7h7sy9.jpg (140 kB image hoster link)

Trying to load the modified APIC table with GRUB2 will fail with a system freeze (this also happens with a supported dual core installed and "core multi processing" set to "enabled" when trying to load any ACPI table). In the GRUB2 documentation, it says that the "acpi" command that loads modified tables does not work with some BIOSes, which seems to be the case here. When using the command with the option "-e", table loading works but this will make only GRUB2 itself use the modified table. Operating systems that are booted with GRUB2 will not be able to see it with this option, rendering the command useless.

To make things worse, the probability is high that even if the table loading could be done (without the "-e" option), it would still not work because in advance of all this I did a test on an x61t where I set "core multi processing" to "disabled" and then tried to re-activate the second core of the Core2 L7500 via APIC table loading, which was unsuccessful (table was loaded, linux kernel tried to activate the second core but said "CPU1 not responding" and continued to boot with CPU0 active only). So it seems to be that "core multi processing" has to be enabled if activating additional cores via APIC table should work. But on T500, this setting has to be disabled to be able to boot any operating system at all with the quad core installed.

So I think that for T500, W500 (and probably also T400 and R400) a modified BIOS with full quad core support is needed. The R500 mainboard even lacks the via needed to access the socket contact for pin D22.

[FryPpy]

I will try to explain the essential contents from this thread here in english.

Thank you for your translation!
All is very clear and simple.

But some questions remains:
- you have insulated 5 pins, other pins unique for C2Q (from Table 15 from Intel's Quad Core Ref) are reserved on C2D, BUT F8 (GTLREF_CONTROL). On C2D it is mapped to VSS. If you can start C2Q without insulating F8 - it means C2Q doesn't matter about F8. May be insulating F8 too is good thought?
- GTLREF2. You recreated W700 GTLREF_2 "powersupply", but W700 make second powersupply for GTLREF_2 and make it switchable thru GTLREF_CONTROL pin. May be better solution (for those, who can't solder SMD) connect GTLREF and GTLREF_2 all together? GTLREF_2 resistors are shown on your picture http://abload.de/img/t61-mainboard-quadcor4qep0.jpg at bottom right corner. Or GTLREF's can't be connected?

I'm going in;)
Hedgehog is shaved;)
http://www.imagesup.net/?di=113942735411
For me it is a simplier way than socket drilling.

PS
Someday i'll ask you many questions about you 60/500 frankenpad. I have read (translated) forum thread and it is good work!

[el-sahef]

Pin F8 (GTLREF_CONTROL) is connected to ground on dual cores and left open on quads, so it does not matter that it is connected to ground in the socket. On mainboards with support for dual cores and quad cores like W700, as you mentioned, there is a circuit that supplies pin D22 with GTLREF voltage when pin F8 is not connected to ground on the processor substrate (which means that a quad core is installed) and connects D22 to ground otherwise. So I do not think that there is a need to insulate F8.

Also, it does not seem to be necessary to set pin D22 to ground on dual cores ("reserved" on these). Dual cores also worked for me with GTREF_2 "power supply" connected to D22.

I did not try connecting GTLREF (pin AD26) and GTLREF_2 (pin D22) and my thinkpad is assembled again but this is a very good idea! If it works, this would tremendously simplify the modification to:
- cutting one trace
- soldering two cable connections
- cutting five pins off

So more people would be able to do it. Would be great if you can try it out and tell me if it works, so that I can update the how-to. I suspect that the board manufacturers did it the complicated way only because they did not want to have pin D22 connected to GTLREF on dual cores.

The two resistors on my previous picture are not the right ones. GTLREF resistors are R160 and R215 on T61 mainboard. I made a new picture how it should be done:
http://abload.de/img/gtlref-fryppy-neu45avr.jpg (340kB image hoster link)
Try to keep the cable as short as possible, because in the W700 schematic it says for the trace between socket contact and resistors: "trace length should be less than 0.5 inch", which is 2,54 cm. A straight connection between the vias would be 2,7 cm, but I do not think that this constraint is so strict that it would not work if the connection is some mm longer.

Edit: Confirmed to work in the german thinkpad forum.

Cutting the five pins off instead of insulating them is also a less complex solution. I did not want to do it like this, because I wanted to be able to sell the C2Q again if I could not get it to work.

PS: Just to make sure: I do not know if the flash of the modified BIOS also updates EC firmware, so you should probably check first if you currently have a BIOS with the latest EC firmware installed.

[Totoro-kun]

Hello,
This is a very interesting mod and I have spare T9550 and thanks to this amazing translation, my hands are itchy to go ahead

Howerver I have one concern about this modded bios. Now I have middleton bios in my T61, which enables sata 2 for my ssd and removes whitelist for my WLAN and WWAN cards. Can anyone confirm, that this moded bios also provides sata2 and removes WLAN/WWAN whitelists?

[el-sahef]

The whitelist is removed, an Intel 5100 wireless card works in my T61. Regarding the SATA 2 limitation, I do not know how to check this reliably. Is there a program that can read out the current transfer mode of the SATA link?

I ran the AS SSD benchmark on my SSD and the sequential read was 195 MB/s which is above the theoretical maximum of SATA 1 .

Alternative download link for the BIOS: http://www15.zippyshare.com/v/33223109/file.html

[Totoro-kun]

Thank you for confirmation, I've went ahead with this mod and can report a success!

I felt brave and decided to skip ram SPD flashing, because I am using Kingston 800Mhz modules (but that was not very wise thing to do). At first, I've flashed moded bios, but used GO.BAT instead of phlash16.exe command above (because it didn't work for me):

Code: Select all

GO.BAT

As some others here, I've also got checksum warning, however, ignored it and bios flashed fine. My T61 was already equipped with T9300 cpu, so laptop booted fine with moded bios.

Next step was installation of 1066FSB cpu (T9550 in my case), just for curiosity tried to power on, but system did not post (of course)

So I went ahead with chipset at 200mhz mod, however, my 14.1" wide screen T61 motherboard is slightly different from the one in how to, but I've traced correct resistor, here is my case:
Chipset at 200mhz (Very big picture)

After this mod, system posted, however T9550 was running at 2Ghz (200Mhz).

At this point, it would have been better to flash moded SPD table into ram, however, I went ahead skipping this step. So I've performed a pin mod to SLG8LP564V clock. I found line, that went from pin 7 and cut it carefully with a sharp scalpel. After this, my FSB was overclocked to 266Mhz, there for ram modules were running at 440Mhz (DDR880), and cpu at it's native 1066Mhz.

It was time to run some tests. Memtest86+ ran stable, prime95 ran stable as well, but once I've launched X3 Teran conflict rolling demo, my Intel graphic card snapped. So I had to flash my ram modules as described in this thread

After this, everything runs nicely. Only issue is, that Linux cpufreq-acpi module does not recognize 2.66Ghz cpu, and thinks it is running at 2.0Ghz (I suspect custom acpi tables would fix this), however performance wise everything works as it should be for 2.66 T9550 cpu.

Thank you once again el-sahef, your translated instructions helped me a lot

[RMSMajestic]

the modified BIOS is based on 2.27 middleton BIOS---- has full SATA 2, whitelist removed and practically everything you need (even SLIC 2.1)
and I've just discovered a better way---- pull up the CPU side of R1025 to C67 and pull down Northbridge side to ground, it will come with 1066Mhz straight
I'll upload pictures when my exams finished

[RMSMajestic]

@el-sahef
not sure if this is realistic, but I reckon you can even overclock your Q9000 by pull up the CPU side of R1026(BSEL[2]), pull down the CPU side of R1025 to form a H-L-L BSEL encoding combination and let the Q9000 running on 333Mhz external clock (OC to 2.5Ghz).

You might need to pull down the northbridge side of R1026 as well cos it's highly like that lenovo blocked the 333Mhz mode in BIOS.

P.S. I saw in your signature that you are using 8GB RAM, what's the manufacture?



Guys, shall we name our modified T6 the T63p

[Totoro-kun]

T63p sounds nice! Mine would be normal T63 tough

[el-sahef]

I saw in your signature that you are using 8GB RAM, what's the manufacture?

It is from CSX, however, in the SPD it says that Micron is the manufacturer. Don't know if this is true, but I don't care anyway because they work without errors at 440 MHz (but I modified the SPD so that the timings for the DDR2-800 profile are also used in the DDR2-667 profile, which is what the BIOS will use).

They are sold on ebay here but I got mine for 80€ from a private seller.

not sure if this is realistic, but I reckon you can even overclock your Q9000 ... running on 333Mhz external clock

It might work. RAM could be flashed to DDR2-533 with the timings from the DDR2-800 profile. This would result in DRAM = 1,33 * FSB and with the FSB set to 333 MHz, DRAM would be 442 MHz like it is now. But the chipset and the integrated graphics will be running faster. I might try that out sometime in the future but not know since cooling is already tough at stock speed. Maybe I will try it if I ever get a T500 cooler where I can plan the surface and do some testing which heatsink-fan-combination from T500 and T61 cooler parts is the best.

According to this post the QAVR and QAVS engineering samples have unlocked multiplier and working temp sensors, so they are an alternative to QX9300. Unlocked QAJF costs "only" 100€ but has no temp sensors.

[RMSMajestic]

@Totoro-kun
actually mine should be T63 as well
the performance subscript "p" should be given to the quad core variant
a QX9300 (if overclockable to 3.1Ghz) would crush all T400, T410 and T420s

@el-sahef
I got 4 crucial 4GB DIMMs from USA for 125 GBP about 2 month ago. hope that I can modify the SPD for these
BTW how comes that you are able to read SPD information on T61?

I think GM965 should be able to handle 1333Mhz FSB as their desktop variants. And also, the biggest improvement of nehalem/SNB processors are their memory controllers, so I think OC the FSB is a good idea to try, also given that the clock speed of Q9100 and Q9000 is quite low.

and do NOT get those ES CPUs from ebay.... a QAVR/QAVS only cost at most 50GBP on chinese ebay (taobao)
and there are a handful of QX9300s (not engineering samples) selling for less than 80 GBP

I've also ordered two W500 heatsinks from USA.....(different FRUs and mountings to see which one's better)
can't wait for my friends to bring back the QX9300s and Q9100
and can't wait for my exams to finish

[el-sahef]

not sure if this is realistic, but I reckon you can even overclock your Q9000 by..

I tried this today and it did not work. First I made a DDR2-profile for DDR2-533 with the timings from the DDR2-800-profile (did not work as intended because for tRCD the max value was 255,75 (69T) and it should have been 81T. Flashed this anyway and verified the timings and that the RAM clock was 354 MHz. I cut the trace between R556 and BSEL socket contact (so that the CPU BSEL2 pin still pulls the chipset BSEL2 pin low) and connected the resistor to VCC1R05B to pull FS_C high. Powered on but did not show an image on the screen.

I think that the resulting display clock for the integrated graphics is too high. With the RAM flashed to DDR2-533, render clock is the same as with DDR2-667 at FSB 1066 but display clock goes up further with FSB 1333.
See this table from the intel GM965 datasheet: http://abload.de/image.php?img=tabelle31dbulr.jpg (128 kB image)
and this table with the clocks resulting from pinmod: http://abload.de/image.php?img=tabelleoct0unm.jpg (55 kB image)

This can be verified if someone with a discrete Nvidia card is able to do the overclocking to FSB 1333 successfully.

At the german thinkpad forum, we also tested all possible coolers (T61/T500 Intel, T61 Nvidia, T500/W500 ATI) and the performance of T61 Intel cooler was 10°C worse that the others. T61 Nvidia and T500 ATI performed about the same (ca. 74°C at full load) but the T500 ATI cooler seems to be a bit quieter. To mount the silver metal piece again (that presses down the cooler surface for the chipset) when using a T61 Nvidia or T500 ATI cooler in a 15" frankenpad, some minor modification to the heatsink has to be made (filing away the metal block above the cooler surface for the discrete graphics chip).

[RMSMajestic]

I tried this today and it did not work. First I made a DDR2-profile for .......

I think probably you cut the wrong trace, and in my case previously I did not cut the trace, simply pulled up and pulled down the sides of R1025 and it worked (200 mhz=>266mhz). I can't find the circuit diagram now but I'm quite sure that R1024- R1026 corresponds to BSEL[2]- BSEL[0]. And all R1024,1025 and 1026 are not marked on the motherboard but I think R1024 is pretty close to R1025, which was what you've already modified.

pull up the CPU side of R1026(BSEL[2]), pull down the CPU side of R1025 to form a H-L-L BSEL encoding combination

Sorry that I made a mistake here, it should be "pull up the CPU side of R1024 (BSEL[2]) , pull down the NB side of R1024" and you should try this with the modifications to R1025 as well

PICTURE of what I really meant HERE
https://fbcdn-sphotos-d-a.akamaihd.net/ ... 7525_o.jpg
apologies for my mistakes and sorry for the confusions
I'll attatch a photo of actually soldering when my exams are done (cos I still haven't done it yet)
I've also used your photos for demonstration
https://fbcdn-sphotos-b-a.akamaihd.net/ ... 6329_n.jpg
https://fbcdn-sphotos-h-a.akamaihd.net/ ... 3697_o.jpg



http://download.intel.com/design/proces ... 318732.pdf
on P32, I was also assuming that the laptop CPUs have the same BSEL encoding for the desktop ones as in the datasheets for T9300/9500 etc
http://download.intel.com/design/mobile ... 891401.pdf on page p27
and for T9900 etc
http://www.intel.com/content/dam/www/pu ... asheet.pdf
on page 29
all BSEL encodings for Core 2 seems the same

Besides my friends have brought me the CPUs----- 2x QX9300 and 2x Q9100, one more QX9300 is coming next week
https://fbcdn-sphotos-f-a.akamaihd.net/ ... 0617_n.jpg
QX9300s are 85 GBP each and Q9100s are 40GBP each (I'll only sell to TPF members at this price). Just need to wait for my exam to finish and I'll try them Exam finishes on May.22nd and I won't be back until then

[bmwman91]

Amazing work guys! I have been itching to figure out some additional upgrades for my Frankenpad, and this might be one!

[ntp2000]

not sure if this is realistic, but I reckon you can even overclock your Q9000 by..

I tried this today and it did not work. First I made a DDR2-profile for DDR2-533 with the timings from the DDR2-800-profile (did not work as intended because for tRCD the max value was 255,75 (69T) and it should have been 81T. Flashed this anyway and verified the timings and that the RAM clock was 354 MHz. I cut the trace between R556 and BSEL socket contact (so that the CPU BSEL2 pin still pulls the chipset BSEL2 pin low) and connected the resistor to VCC1R05B to pull FS_C high. Powered on but did not show an image on the screen.

I think that the resulting display clock for the integrated graphics is too high. With the RAM flashed to DDR2-533, render clock is the same as with DDR2-667 at FSB 1066 but display clock goes up further with FSB 1333.

What is your mean "Powered on"? Is the light twinkling ok and just no display? If not, I doubt it is the power supply's problem.