Thinkpads Forum

Just a guess, but if the connection from GTLREF to GTLREF_2 is not correctly done, it could lead to CPU 2 and 3 not working (only 0 and 1). On DOS level, the additional CPUs are not used (only the first one) so everything works there, but when the higher level OS try to utilize the additional cores, they freeze. Other possibility would be the RAM. Even if you use DDR2-800, the timings might still be too low because the ones from DDR2-667-profile are used. If your RAM modules do not have their SPD locked and have 16 chips, you could try flashing one of the SPD profiles in this post. The are edited so that they use the higher timings from DDR2-800-profile also in the DDR2-667-profile. If you have a module with 4 GB, you could even use the 533-profile. That would reduce the RAM clock to 354 MHz (now probably 443 MHz). Unfortunately, I did not make one for 2 GB modules yet.

RMSMajestic, off-topic but the comment in your signature about your T61p needs a "like" button next to it

el-sahef wrote:Just a guess, but if the connection from GTLREF to GTLREF_2 is not correctly done, it could lead to CPU 2 and 3 not working (only 0 and 1). On DOS level, the additional CPUs are not used (only the first one) so everything works there, but when the higher level OS try to utilize the additional cores, they freeze. Other possibility would be the RAM. Even if you use DDR2-800, the timings might still be too low because the ones from DDR2-667-profile are used. If your RAM modules do not have their SPD locked and have 16 chips, you could try flashing one of the SPD profiles in this post. The are edited so that they use the higher timings from DDR2-800-profile also in the DDR2-667-profile. If you have a module with 4 GB, you could even use the 533-profile. That would reduce the RAM clock to 354 MHz (now probably 443 MHz). Unfortunately, I did not make one for 2 GB modules yet.

Thank you for the reply. I have tried resoldering the GTLREF to GTLREF_2 Again, and the same issue appears. Either i am incredible bad at soldering or i just have plain bad luck - unless its the RAM.

I have looked at the flashing Tool, however i am a Little unsure about how i am supposed to do it. I have not tried anything like that before.
When i enter the flashing Tool, i can load the SPD file (i have 2x2 GB block, 800MHz with 16 Blocks each). I can see there are three "profiles" that i can flash to, but i am unsure which one i am supposed to flash?

My german is rusty so i don't quite understand the guide.

Best regards

Andreas

On T61 and X61, access to the SPD-ROMs is blocked somehow. Instead of these, some other ROM contents appears in the SPD tool when you try to read out the SPD-ROMs on your memory modules and there is also no write access to them. To get around this, you have to download RWEverything portable (32-bit-version or 64-bit-version according to your operating system) and then proceed as shown in the picture:
http://abload.de/image.php?img=x61_spd_ ... sz4asv.jpg

Summary:
1) click on left icon in the toolbar
2) choose "intel corporation isa bridge" in the dropdown-list
3) in the new window at row 40 and column 09, right-click on the "11" and left-click "open IO Space"
4) in the new window at row 30 and column 09, double-click the "03"
5) in the new window, double-click the "0" below the "2" (should become a "1")
6) click "done"
7) minimize RWEverything

Now you can access the SPD-ROMs. This was found out by tcone in the german thinkpad-forum and I stole the picture from his post here .

Now you start SPD tool and click on file --> read. In case of two memory modules installed, you should see the possibility "module 0: installed" and "module 1: installed" if everything went well. Now open the file "Samsung 2GB 16C edited DDR2-667 CL6.spd" (in case of 16 Chip 2 GB module) and choose file --> write --> "module 0". After the write process finished, you should read the SPD from the module you chose to write to, so you can check that the contents of the .spd file were really written. Some memory vendors lock the SPD-ROM so that their content cannot be changed.

If the contents were successfully written, you can power off and then power on the machine again and check with CPU-Z in the memory-tab if the RAM timings are now 6-6-6-18. I did not make a profile for 2-GB-modules with 16 chips and DDR2-533 yet. That would be the next thing to try out.

PS: If you have a 64-bit windows on your machine, you have to boot with driver signature enforcement off. To do that, press F8 just before the windows bootscreen appears and choose "disable enforcement of driver signature"

Thank you for the guide - i got the RAM flashed, and they are currently at the rigth timings.

However, it didn't change the situation. I am still stuck, and cannot get into any higher level OS. Not even with core multi processing disabled.

The GTLREF to GTLREF_2 should be soldered correctly (otherwise it wouldn't be able to boot AFAIK).

Before i could boot Windows XP, but now it just gives me a BSOD with 0x0000007B - probably meaning a hardware fault.

EDIT; Thank you RBS, i forgot that when i was working with it With a P8400 i managed to install, and boot, Windows 7. But inserting the QC will bring me back to only Windows XP working - because Windows XP doesn't utily all four cores.

The conclusion must be a bad GTLREF to GTLREF_2 connection, that i -somehow- screwed up. I have a spare Intel X3100 mobo i will try to get the quadcore working on, and then i will install an Intel P8700 on this board. I won't risk loosing my '09 nvidia GPU once again A P8700 is also better than the X9000 i had before

BSOD 007b usually indicates that you need to change the SATA in BIOS from AHCI to Compatibility or v.v.

The conclusion must be a bad GTLREF to GTLREF_2 connection, that i -somehow- screwed up.

Most likely this is the reason that the quad does not work. But there might also be a problem with the pin insulation (especially if the material the insulation tubes are made of is too soft so that the socket contacts cut through it). You should check that too.

el-sahef wrote: Most likely this is the reason that the quad does not work. But there might also be a problem with the pin insulation (especially if the material the insulation tubes are made of is too soft so that the socket contacts cut through it). You should check that too.

I have insulated the pins by removing them - so they should be ok

How long did you put the cable into the GTLREF to GTLREF_2 holes? I suspect i didnt make the cable lon enough into the hole

I have insulated the pins by removing them - so they should be ok

Definitely, as long as you removed the right ones .

I suspect i didnt make the cable lond enough into the hole

I did not put the cable into the holes at all. I just soldered them to the vias (cable as short as possible). If you put them into the holes, the conductor might bent when it touches the underside of the socket (if the blank end of the cable is too long) and short two neighboring contacts.

el-sahef wrote: Definitely, as long as you removed the right ones .

I did not put the cable into the holes at all. I just soldered them to the vias (cable as short as possible). If you put them into the holes, the conductor might bent when it touches the underside of the socket (if the blank end of the cable is too long) and short two neighboring contacts.

It certinatly is the rigth ones i removed - at least i can be pretty sure about that! I have quadroble checked on that already

So, i tried doing the mod on another motherboard (15.4" X3100 penryn board)

I installed the updated BIOS, and checked everything worked.

I soldered the first connection, and checked everything still worked (With an Intel P8400)

I solderen the GTLREF to GTLREF_2 connection. I soldered the cable onto the vias, nothing in the holes. And still the same - i cannot boot Windows 7 or the like...

BIOS still recognizes the C2Q. Windows XP have two processor units in device manager, and 4 in taskmanager. In devicemanager it recognizes the processors as Q9100 @2.26GHz. When disabling core multi processing in BIOS, it shows 3 cores in taskmanager

So somehow, i dont think its the GTLREF to GTLREF_2.

I then thougth it migth be the RAM... But they are flashed. I pulled the stick out, and inserted a 1 GB PC5300 stick. It gives the same result
I then tried flashing a new SPD to a 667 MHz 4 GB stick - same result Again.

Prime95 in Windows XP gives a workload on all four cores, and its stable.

EDIT; Some further investigations; Windows Vista 32 Bit installs just fine. 64 bit can install, but get stuck at the booting screen with the green bars (and i cant boot into safemode because i havent finished the installation)

BIOS still recognizes the C2Q. Windows XP have two processor units in device manager, and 4 in taskmanager. In devicemanager it recognizes the processors as Q9100 @2.26GHz. When Disabling core multi processing in BIOS, it shows 3 cores in taskmanager

this is something others also observed. Seems to be normal.

I then tried flashing a new SPD to a 667 MHz 4 GB stick - same result Again.

Did you choose the DDR2-533-profile (link)?

Since prime95 is stable in Windows XP, I do not think that the problem is on the CPU-side. The last idea I have is the integrated graphics card. The render clock of the intel graphics is linked to the RAM speed. Using RAM flashed to DDR2 533 will lower the render clock. But since the render clock is already raised because of the higher FSB, it will be 33 MHz higher than default even with DDR2-533. With DDR2-667 it is 165 MHz higher than default. This table shows the clocks: http://abload.de/img/tabelleoct0unm.jpg (55,13 KB image link).

I had problems with graphics stability when the RAM was flashed to DDR2-667 and aero turned on in windows 7. This was the reason why I made a profile for DDR2-533.

el-sahef wrote: this is something others also observed. Seems to be normal.

That is a good sign

el-sahef wrote: Did you choose the DDR2-533-profile (link)?

Since prime95 is stable in Windows XP, I do not think that the problem is on the CPU-side. The last idea I have is the integrated graphics card. The render clock of the intel graphics is linked to the RAM speed. Using RAM flashed to DDR2 533 will lower the render clock. But since the render clock is already raised because of the higher FSB, it will be 33 MHz higher than default even with DDR2-533. With DDR2-667 it is 165 MHz higher than default. This table shows the clocks: http://abload.de/img/tabelleoct0unm.jpg (55,13 KB image link).

I had problems with graphics stability when the RAM was flashed to DDR2-667 and aero turned on in windows 7. This was the reason why I made a profile for DDR2-533.

I used the DDR2 667 profile, but i tried with the 533 profile just now. Nothing changed. I migth begin to think it is the processor - because at this stage i don't think it can be anything else

It would also make sense, since the other OS Works fine with a normal dual core

KJC@51nb suggested that northbridge chip is the actual bottleneck for higher Memory speed. He suggested a way of overvolting northbridge chip on T61 X61, picture below is the area near PCMCIA. Adding voltage from 1.05V to 1.15V while the voltage during idle remain the same

https://scontent-a.xx.fbcdn.net/hphotos ... FF200&dl=1

During the recent X61 P8800 and T9900 project, a few X61 managed to run dual channel DDR2 888@5-5-5-15 after northbridge overvolt. so hopefully this could help with the FSB 1333 project and achieve higher memory speed.

Enjoy the tweaks

Correct: The picture is for northbridge overvolt on X61, NOT T61

Maybe I will try it, but it depends if I find the time. I still did not try remove that power limit although I have it on my list.

Has anyone had success with RAM flashed @ DDR2-533 on a mainboard with nvidia? It seems that only Intel boards do accept DDR2-533, nvidia boards seem to require DDR2-667.

el-sahef wrote: Has anyone had success with RAM flashed @ DDR2-533 on a mainboard with nvidia? It seems that only Intel boards do accept DDR2-533, nvidia boards seem to require DDR2-667.

About two month ago, a friend of mine took my intel board away. So I had to use a nVIDIA board.
I did not changed the RAMs ( 2GB+4GB DDR2 711@6-6-6-18) and put those on my Nvidia board (NVS140m merom board), they runs perfectly fine

Can you please upload the SPD dumps of those two memory sticks that run with DDR2 711@6-6-6-18 on an nvidia board to zippyshare or some other OCH?

If I give FSB 1333 another try, I will have to use an nvidia board because I suspect that with an X3100 board, the resulting clock for the IGP would be too high.

el-sahef wrote:Can you please upload the SPD dumps of those two memory sticks that run with DDR2 711@6-6-6-18 on an nvidia board to zippyshare or some other OCH?

If I give FSB 1333 another try, I will have to use an nvidia board because I suspect that with an X3100 board, the resulting clock for the IGP would be too high.

Here u r:
http://www50.zippyshare.com/v/93864620/file.html
I did not change the spd on my 4GB stick (Integral RAM with Crucial modules). I just put it with my 2G Elpida RAM and they both work. The latency were set quite high cos I want a fully stable machine.
Regrading the Elpida 4G stick, I don't think that I've made any mistake during spd modification, I guess they just have spd partiallly or fully locked.

May I ask if the modded BIOS enables support for later 800Mz FSB Penryns like the Pentium Dual Core T4500 and Core 2 Duo T6670?

If yes, is it advisable to flash the modded BIOS on a stock T61 just for the CPU support?

axur-delmeria wrote:May I ask if the modded BIOS enables support for later 800Mz FSB Penryns like the Pentium Dual Core T4500 and Core 2 Duo T6670?

If yes, is it advisable to flash the modded BIOS on a stock T61 just for the CPU support?

Yes for T6670,T5670. "probably" for T4500
Not really for the second question. Although there has been no problem reported about the BIOS so far, still some people reported that they can't change back to the normal BIOSs.
Given that T8100 is as cheap as a 40GB SATA HDD now, I don't recommend flashing the modded BIOS just for T6670 or T4500

I flashed it on a stock T61 mainboard to enable support for a T6600 and had no problems so far. I do not see a reason for reverting to the original BIOS, but if you want to do this, then maybe you should not flash the modded BIOS.

@ RMSMajestic:

I tried to get the T61 to boot with 333 MHz FSB again (on a nvidia board). First, I tried your DDR2 533 profile for a 2 GB stick that had worked on your nvidia board. It was not accepted on mine, board beeped only. I found out that one of my DDR2 533 512MB sticks was accepted by the board, so I edited the SPD for this stick to fit 2GB and flashed it. The 2GB testing stick was still accepted. After that I changed the timings to 6-6-6-18 and it was still accepted. I have no idea why my nvidia board is so picky with DDR2 533 RAM and what is so special about this specific SPD of that 512 MB stick that it got accepted but anyway, I finally had a DDR2 533 profile that worked on a nvidia mainboard.

I also verified with a DDR2 667 profile that the stick was able to boot at DDR2 440. Next step was the CPU, a P7450. I verified on a pinmodded T500 mainboard that it was able to boot with 333 MHz FSB.

Since the CPU and the RAM were known to work with 333 MHz FSB now, it had to be the mainboard if it did not work again this time - which was unfortunately the case. CPU and RAM ran perfectly with 266 MHz FSB on that board. Then I wanted to try the northbridge overvolt mod but I found out that it overvolts only the GPU part of the northbridges that have an IGP, so it did not help me with my nvidia board, which has PM965 instead of GM965. VCCGFXCORE is connected to the nvidia chip on these boards.

I checked the intel datasheet of the northbridge and the voltage that has to be raised if someone wants to overvolt the part of the northbridge that talks to the CPU should be VCC1R05B. But this voltage is not only connected to the northbridge. The CPU and the ICH are also connected to it. The RAM controller of the northbridge should be ok, because the clock will be 440 MHz with the DDR2 533 profile and it is confirmed with the DDR2 667 profile @ 266 MHz FSB that the RAM controller works with that clockrate and the default voltage.

I soldered one end of a 100k potentiometer to R711 (FB1 of MAX1540 voltage regulator) and the other end to ground, powered on the board and went into the BIOS. Then I lowered the resistance slowly until I measured 1,15 V (instead of 1,05 V) at C808 above the expresscard slot (numbers refer to hisui 5W schematic).

It did not help with 333 MHz FSB. I could try to increase the voltage further, but I think this would put the CPU and mainboard under high risk.

large Image: http://abload.de/img/cimg0278fzcpy.jpg (1,5MB)

If I would have one more CPU with 1066 MHz FSB to test, then I would happily sacrifice the P7450 and the mainboard, because it had the nvidia bug already and was revived with the hot air gun, so it might possibly fail again anyway sooner or later.

Do you have some more pictures of "hybrid cooling" i want to try it on my T42P with M770 i hope

It seems a lot of the discussion is based on putting in a Quad Core CPU. If I just want to use a T9800 or T9550, is it as simple as installing the modified BIOS?

For T9550 or T9800, you have to do steps 1 and 2 of the How-To in this post. Use a BIOS from this post without the quadcore APIC.

I would like to do the mod, so I can put a T9550 into my frankenpad, but the cpu will arrive in 4 weeks. As Ive just ripped apart my T61 it would be convenient to do the mod now and even flash the new bios. However, after modifying the mobo and flashing the bios will my current T7100 work? or do I have to wait with the modification until my T9550 arrives? Thanks.
I am just planning to do step 1 and 2 cause it is sufficient for my new cpu (please see previous post).

However, after modifying the mobo and flashing the bios will my current T7100 work?

Yes, as long as you choose a BIOS from this post (without the quadcore APIC in your case) and not the initial BIOS version by xiaofei290.

GOOD NEWS to those of you who only want dual-core upgrade
amaninsh@51nb has modded a new BIOS based on 2.29 middleton BIOS with newer CPU microcode, supporting all T9 T7 P8 X9 T6 P7 CPUs. The newer microcode brings slightly better temperature controls.
I'll upload the BIOS tomorrow and post it here
http://www4.zippyshare.com/v/y8NHNybU/file.html

Hi there

Could you please tell me which side of the mobo is pictured here
http://abload.de/img/266-fsbe2cow.jpg

I have T61p with NVIDIA G84GLM Quadro FX 570M and 15'4 screen but I haven't seen similar place. I guess this is the side of the mobo that I would see when I remove it from the chassis, but I haven't been brave enough yet to disassemble everything. Could somebody confirm?

Anyway, thanks for your impressive work!


After a bit of research I suppose I found this area of my mobo here:
http://s14.postimg.org/gr3aesv8x/IMG_0493.jpg


but still couldn't find the yellow trace, I guess i have different mobo layout. Could anybody confirm and/or help?


MOD edit: your pictures are WAY to big! Tags removed. Please read the Forum Rules, especially Section 5. Thank you.

RMSMajestic wrote:I'll upload the BIOS tomorrow and post it here
http://www4.zippyshare.com/v/y8NHNybU/file.html

Could you please also post a valid md5sum hash of this ROM file?
I feel a bit uneasy flashing a BIOS with unknown md5sum value...

What about ASPM Power Management in this BIOS?