Need help for PM/GM965 FSB downclock

Hi everyone,

unfortunately the original thread has been locked, so I'll try to get the more relevant information here.

TL;DR: You can use 1066Mhz Core2Duo (and quads) on motherboards with the PM/GM965 chipset (and others) by "downclocking the FSB to 800Mhz from 1066Mhz" if the MB bios has the necessary microcodes for the new CPU.

Original thread where all the pictures are sadly gone:
http://forum.notebookreview.com/thr...-and-gl40-useful-info-for-pll-modders.605383/

Actual working picture of the mod:
http://forum.notebookreview.com/thr...-for-pll-modders.605383/page-22#post-10278099

So it's just a matter of making a bridge between two socket holes... but there are things which are not 100% clear to me.

In this post:
http://forum.notebookreview.com/thr...fo-for-pll-modders.605383/page-6#post-8093213

user @RickiBerlin stated that by isolating BSEL pins you're going to make them stuck high.
This is something which, for me, is miles ahead better than fiddling with the socket because Core2Duo P9500 and such are ~10€ online so I'd rather just isolate or cut the pin on the CPU instead of risking damaging the socket on the motherboard.

Can anyone confirm this?
In order to downclock the FSB from 1066Mhz to 800Mhz do I just have to cut/isolate the BSEL 1 pin on the CPU?

Onto another matter:

User @remdale stated in this post:
http://forum.notebookreview.com/thr...-for-pll-modders.605383/page-29#post-10902732

That the whole thread was basically useless because

This is really confusing.

What does this mod involve? Cutting/isolating a pin? Tying it to ground or high?
Which is the pin which needs to be "modded"? On the CPU? The motherboard? The northbridge? Which one is it?

The solution by Remdale is really interesting because it looks like it will make the CPU work at its full 1066Mhz FSB without downclocking it, which would save the hassle of using SetFSB to get the FSB back up to its original frequency (or close to it).

I'd be immensely grateful if someone could come here and shed some light on the topic.

Thanks!

 

I need to see schematics of your motherboard. Because BSEL1 pin can be forced by being attached to power. What you need to do is to break the connection between the north bridge and the CPU on BSEL1 pin. But you have to make sure the same pin is not forced by power on the PLL too.

You can still use it after completing the mod to figure out the highest stable FSB frequency.

 

Thanks for chiming in.

There are no diagrams that I know online of this notebook, it's a Packard Bell BG45-P-027.

But I can definitely follow the trace and make the mod.

So what I need to do is to cut the trace that goes from the northbridge to the BSEL1 pin on the cpu socket and then tie the BSEL1 pin high, right?

EDIT: I've re-read your post carefully.
So the first part is the classic BSEL mod, where you either jump the BSEL1 pin to Vcc (or, in this case, I can just break it and it should be stuck high).

The second part... I've kind of lost it.
If I just break the pin on the CPU there's going to be no connection between it and the rest of the board, so we're set on that part.

Regarding the PLL, I guess I have to tie it to ground then? Because now, due to the broken CPU pin, it is floating, hence I should just make a bridge in a convenient spot to ground.

Correct?

 

Your platform is ASUS T32E. This is where I found it. http://vlab.su/viewtopic.php?t=3620&f=172
Look for T32E (which is printed on your board)
And this is the only source that has its schematics https://remont-aud.net/load/noutbuki/platforma_asus/asus_t32e_rev_2_0/313-1-0-52957?l4bViI
Grab it from my mediafire account http://www.mediafire.com/file/f0v6tut07j575hx/ASUS_T32E_%28Packard_Bell_BG45%29.pdf/file
Is it the same board?

Your video core is built into the chipset. Never dealt with integrated video cores, so I'm not 100% sure if it works out. But I've heard it might not because when overcloking chipset, video core will be influenced too and who knows if it can work at much higher frequencies. But we can give it a try.


You need to remove R2957 and move R2922 so that 1 side gets disconnected from the CPU line and the other side is still connected to the MCH line. Then just attach VCCP to the unconnected side of R2922. So the MCH will still be thinking it's operating at 800. It should suffice.

 

Wow, thank you so much!!

I need to check if the boards match. Pretty strange as Packard Bell was part of the Acer brand, but whatever.

I'll be away from home for a few days, I'll report back as soon as I can.

EDIT: btw, if I just break the BSEL1 pin on the CPU, it should just work by bridging the 2957 resistor

 

I'm curious if it works out. So waiting patiently. But don't install a quad.

 

My actual goal is to take advantage of the SSE4.1 instructions and lower TDP so I'll probably put in either a P8800 or a P9500.

 

I would recommend you to get X9100 for better performance. But you won't be able to overclock it much because of weak VCORE power source for the CPU.

 

This notebook ain't gonna win any speed contest, it's my windows backup machine if/when things will go down burning (I mostly use Macs but sometimes I need a windows PC).

The X9100 is 30€ on the ebay, the P8800 is 6€.
Now it has a T5750 which runs hot and it's not that speedy, with a P8800 it's going to run Win10 nice and quick with a SSD.
And, as a nice bonus, the battery should last longer.

Not going to invest any more than that on this notebook.

 

I wouldn't either

 

I've found kind of a service manual, it might prove useful for you to figure out if that schematic matches my board: http://tim.id.au/laptops/packardbell/easynote bg.pdf

 

Sorry, it's not helpful at all. Just look at the schematics and try to locate some specific resistors or chips by their inscriptions on the board. If they match or not.

 

Yeah I guessed as much, couldn't find anything else unfortunately.

When I get back home I'll take a look, although, even if the boards do not match, the concept is pretty straightforward: float the BSEL1 pin on the CPU (which will set it high) and set the corresponding MCH high.

It's just going to be hell to trace through the board as it's probably multilayered, I'll have to spend lots of time checking for continuity between points.

 

BSEL pins on the CPU are all output, not input. You just need to cut it off from the MCH,
R2957 resistor is also forcing the PLL to work at 200. So take it off from the board to let the CPU decide the FSB frequency.

You don't have to as long as you have the schematics

 

Ahhhhh that's one very important point I was missing. I was under the impression that the BSEL pins were inputs.
That makes perfect sense now, I'm finally getting the whole picture.

On another note, I'm still waiting for the new CPU to arrive, here's hope it didn't get lost.

 

It shouldn't get lost, let's wait

 

I mean, worst case I just buy another one, those old CPUs with the exception of the very high end ones are a dime a dozen those days.

 

I bring bad news:

First, the board is different from the one in the schematics and all the various compontents have no markings near them making it impossible to really tell which goes where.

Second, it won't work with the P8800 CPU, no matter what.

I've tried first with the CPU as it was, then breaking the BSEL1 pin, no change.

I've even updated the microcodes in the BIOS to make sure it had the correct ones for the new CPU.

No change.

Oh well, I tried I guess

 

Wait, I finally got it to at least boot to the bios splash screen.

I had to put the jumper wire as described in the other thread.

The CPU speed reported is, correctly, slower than intended, the problem is that it locks up right after the first bios screen, I can't even get into the bios!

After the RAM check it just stops dead, it doesn't even reach the IDE/SATA detection.

What could it be?

EDIT: and one more interesting thing: it looks like the RAM is still working at 166Mhz instead of 200Mhz (667Mhz instead of 800Mhz), even with a 800Mhz stick.

Mumble...

 

So you are saying there are no names printed on the board for each component? How did you break the BSEL1 connection?

 

Unfortunately there is absolutely nothing written next to each component, which is disheartening.

Of course, the PCB is multilayered, which makes trying to trace the pins too much work for me. It is doable, but it would require an immense amount of effort I'm not willing to put in.

As for the BSEL1, I just cut the pin on the CPU, but it made no difference.
I have two P8800 here, one with pin broken, they show no differences: they both need the jumper from BSEL1 to VCC, otherwise BSEL1 is set low.

I was reading again through the original thread and the OP said that there was a notebook, a Toshiba Equium L40, which did the exact same thing as mine: it hung on the BIOS splash screen, which, he said, it's a symptom that the BIOS does not support newer C2D CPUs.

This might be the case here and not even putting inside the BIOS the necessary microcodes makes a difference.

 

Because the pin is forced by VCC on the board via the R2957 resistor. You have to find and remove it. Removing the BSEL1 pin from the CPU is the least thing you would want to do. But it should not effect anything if you remove VCC from BSEL1.

This sucks. Can you at least send me high resolution photos of both sides of the board?

P8800 should be outputting logical 0 there. Don't attach VCC.
You will probably need to update BIOS microcodes, but I don't think the problem is there. Try to get the cheapest 1066 CPU and use it instead.

Can you send me a photo where you cut it? I need to see a full photo of the bottom side of the CPU

 

The P8800 *is* outputting low on BSEL1, I've done a test by setting BSEL1 both high and low through the socket.

When set high, the machine will boot and get stuck at the bios splash screen.
The CPU is properly recognized as P8800 and with the current speed set at 2.0Ghz (200Mhz x10 instead of 266Mhz x10 – the P8800 has a x10 multiplier), but everything hangs after the RAM check.

When set low (same as having the CPU with or without the BSEL1 pin without any other mods) the notebook will power up without displaying anything and then shut itself off after 4-5 seconds.

The pin has been removed cleanly by using one of the best ways I know: use an exacto knife to slightly bend the pin from its base back and forward until it breaks off. It's cleanly broken at the very base.
Anyway, that's really a moot point as the CPU with the BSEL1 pin intact behaves the same.

The BIOS microcodes has already been updated and made no difference.

This is not true, the board sets it to low.
As I said, I've tested this.
If you were right, I would have just needed to snip the BSEL1 pin.

As for the pictures: I've put the whole notebook together yesterday and wasn't planning on tearing it down again. I can definitely give you a picture of the CPU with the broken pin if you wish although I have no macro lenses on my mirrorless camera, I'll see what I can do to give you a picture with enough definition.

About buying another 1066Mhz processor: I think I've invested enough time in this project and don't want to invest more.
Maybe I'll give it one last go this weekend to give you some high quality pictures of the board.

I do appreciate all your help so far, but I think that this notebook is a no-go for the 1066Mhz processors.
Maybe the 800Mhz one will work but at that point I'd rather just keep it as it is now.

 

Yes, it's outputting low there.
I'm sorry, I didn't notice one thing. Actually, those resistors are absent there because there's a @ symbol present next to each of them. It means "not connected". I forgot about this. The other schematics say NC instead, so you can easily notice when the element is there or not. Yours has just pads for the resistors so that you could force FSB yourself.
I got the point now. When you insert a 266 CPU, the chipset starts working at this frequency too. But the problem is that it can't recognize this FSB setup which is 000 and so according to the datasheet, the chipset shuts the PLL down and this is probably why your machine freezes up.

I'm just wondering how it can boot at all. Mine couldn't.

What you need to do now is to find R2922 and move it closer to the chipset to break the connection of the BSEL1 line for the chipset and attach VCCP to the floating end of the resistor to fool the chipset.
It should work out.
But it might also be the case that the video core cannot hold the increased frequency.

 

Just to make sure we're understanding each other...

This is where I'm currently standing:



As you can see I'm already feeding the BSEL1 pin VCC and that's why I'm able to at least reach the BIOS boot screen, where it recognizes my CPU and says it's running at 2.00Ghz (instead of 2.66Ghz due to the FSB forced to 200Mhz via BSEL1).

Just for fun, I've quickly tried to tie BSEL0 high as well to make the FSB run at 166Mhz and nothing changed (apart from the reported CPU speed of course): it still hang on the bios screen.

This means clearly that the microcode is not recognized and I have a feeling I know what's wrong with it. I might try later doing some hex editing job on it and will report back.

 

Is it working well at 200?
If your goal is 266, you have to do as I'm saying if the CPU is working fine at 200.

 

Nothing, even by manually doing some edits in the BIOS to make it recognize the CPU it will fail to POST.

It looks the BIOS just won't work with the newer 1066Mhz CPU, it doesn't support them.

Oh well, I think it's time to call it a day, took a hit for the team so everyone now knows that a Packard Bell BG45 notebook won't boot with 45nm CPUs

 

Santa Rosa can work fine with 1066 CPUs. I have a solid proof. You have to dig deeper and try different ways.

 

Try to install your 800 CPU back, remove R2921 or BSEL1 pin on the CPU and do what I said to set up 266 MHz for the PLL to see if your system can work at 266. You might have to increase CPU voltage though

 

My other CPU (the one that came with the notebook) is a T5750 which is 667Mhz, not 800.

 

My PM965 can work with 1066 CPUs as far as you remember. The question is if GM965 can either? The first thing I want to make sure about if your P8800 can work at 800 there. Actually, it should. But if you are having problems with it, there must be a reason for this. Not sure if it's about BIOS at all. If you are familiar with CPU manufacturing technology, you should know that basically CPUs with different FSB use same internals, they are just programmed differently to report different frequencies and work at those frequencies. So your problem needs to be investigated.

 

I'm 99,9999999999% this is a software limitation, otherwise the notebook would not boot at all.

The fact that it reaches the first part of the BIOS and shows correct processor name and frequency (albeit 2Ghz instead of 2.66Ghz) and even manages to make a RAM test makes me think that the whole issue resides in the BIOS.

And, as a matter of fact, I think I've even found where the limitation lies: this BIOS can accept microcodes only up to 4KB and the 1067A is 8KB.

I'm 99% sure that if I had put a P9500 (which has a CPUID of 10676) in it would have worked without issues, probably without even having to update microcodes.
But honestly I'm not in the mood to waste more money on this laptop, so, as I stated earlier, this is it for me.

If I'll ever come across a P9500 or any other 10676 CPU I might give it a go.

 

After reading this thread: http://forum.notebookreview.com/threads/upgrading-cpu-on-obsolete-notebook.606225/

I'm 100% sure the problem is in the BIOS and that it doesn't accept 8KB microcodes.

He has the exact same notebook I have and he succeeded in using a P7350 which is the same as the P8800 but with one very important difference: the P7350 belongs to the 10676 CPUID Family and uses the 10676 microcode which is 4KB (and it's already inside the BIOS, in fact he didn't have to update the microcodes at all).
The P8800, on the other hand belongs to the newer 1067A family and needs the 1067A microcode which is 8KB and that's why it's not supported.

Now I'm really really tempted to get a P9500 (C0 stepping)...

 

What CPU IDs does your BIOS have? Name all of them. I would also appreciate if you could share your BIOS dump.
I'm thinking on replacing two useless microcodes with 1067A if your BIOS doesn't have it. Give it a try and let me know if it helps

 

Also, I forgot to mention that you will have to increase RAM voltage to avoid errors as it will have to run on higher frequency after doing FSB overclock.

 

The only ones which are pertinent in this context are the 10676 and 1067A.

My BIOS has the 10676 updated to January 2008.

And I've tried literally everything BIOS side including adding and modifying by hand (hex editing) the microcodes in order to make the CPU recognized and nothing worked.

I've even injected the 1067A microcodes into the 10676 hoping to fool the BIOS but it was all for nothing.

No need to increase RAM voltage as the system will decide on its own the FSB:RAM ratio, meaning that you could have a 800Mhz CPU chugging happily along with 533Mhz RAM.

Of course, if you want to bump the CPU FSB at 1066Mhz and keep a 1:1 ratio with the RAM making them work at 1066Mhz you'll need some VERY good RAMs and definitely need to raise the voltage on the DIMMs.

I'll share the BIOS later.

 

Here's my BIOS: https://mega.nz/file/RIBDDYyI#EK1L3icHXa_Iik_AliE2tacG2Rt3jHJbv1TPxzL8Qac

You can find the most updated microcodes here: https://github.com/platomav/CPUMicrocodes/tree/master/Intel

Older versions of microcodes (because you never know) here: https://www.bios-mods.com/resources/index.php?dir=Intel+Micro+Codes/

Otherwise you can just look for BIOSes updates of notebooks and extract the microcodes from there.

Here's the list of the microcodes included in my BIOS:



EDIT: also, one kind request.

While I do appreciate your help, I'd appreciate if before posting you'd go and re-read all my previous posts to avoid going over the same things several times.
I've mentioned about modding the BIOS in earlier posts.

 

Your RAM is running at 667Mhz with the correspondent timings. When you do FSB overclock, you have to fool the chipset to make it think it's running at 200. When doing so, the divider will stay the same and so the RAM frequency will increase as a result as the FSB does. The possible dividers for Crestine chipset (PM/GM965) are the following

The chipset doesn't support 800 Mhz for RAM. The cap is 667.
If you are using a 800 CPU and RAM is running at 667, your divider is 5:3. If you do 266 FSB overclock, the RAM frequency will become 886 MHz (266*5/3).

I couldn't find anything useful by the link, so I'm trying to get the full the picture piece by piece and trying to find a way out of the problem.
The microcode list of your actual BIOS is what I needed to get. Thanks.
I have an idea that you can replace 2 useless microcodes with the 8KB one and then see if it works. Can you do so?

 

Already done that: it's a no go, sadly.

For the RAM: I can just change the SPD to make them work at lower speeds to have a 4:3 ratio, then when overclocking I'll have 266 * (4 / 3) which is ~709Mhz (I'm pretty sure my 667Mhz sticks can reach that speed).

If possible I wanted to avoid buying 800Mhz sticks and I value FSB and CPU speed way more than RAM speeds.

 

This is sad. Does the BIOS boot up at all after doing that? If it's a dead end, then here you go
https://www.ebay.co.uk/itm/Intel-Co...z-6M-1066-MHz-Notebook-processor/302790252672
C0 stepping.

This is exactly what I have done. But the problem is that as the chipset thinks it's working with 533 RAM, it will set up correspondent preprogrammed timings for that frequency. You can probably set them up in SPD, but it didn't work in my case because it seems like the timings were programmed in BIOS.
So the RAM will be working with low latency at higher speed which is considered to be overclocking as well. It will result in RAM needing more power. So if you want to avoid RAM errors (like it was in my case), you will have to install a couple of resistors to get around 2v output. Luckily, U8300 (MAX8632) chip which is the 1.8v power controller doesn't have any resistors for setting the voltage. So you don't have to search for them on the board. But you would have to break the connection between AVDD and FB pins because their connection sets 1.8v output. Then you have to calculate resistances for both resistors and connect them accordingly as shown in the MAX8632 datasheet.

As for the RAM divider, it cannot be set up freely (unlike some server boards). You can only use 4 available options.

 

The nice thing is that notmatter how much I've butchered the BIOS, it has always booted (albeit then getting stuck with the P8800).
Although I have to admit that I've always been very careful about what I touched and how.

Anyway... the SPDs are always read from the RAMs (at least in my experience), you just have to tailor them and flash them.
Start from JEDEC standard and work from there. I generally take the more relaxed timings of higher frequencies and slap them on the slower ones, so I have room to get the frequency back up without issues.

Google around for SPDtool and have fun (and be careful because you might brick a RAM stick if you go too ham).

 

I've got all the needed tools. I've also used Thaiphoon for changing and flashing the timings, but the CPU-Z always showed the same standard timings for the particular frequency the RAM was currently working at. This is why I came to the idea that they are programmed in the BIOS.
So are you going to grab that P9500 from Ebay?

 

I've already wasted money on 2 P8800 (I bought two in case I mangled one CPU when breaking the pin), I'm not really looking forward getting another one right now.

Maybe I'll see if I have some of my friends who work as techs a Penryn-3M CPU to try and see if it works (I'm 99% sure it will).

For now it will stay with the T5750.
Hell, Windows 10 loads up from cold boot to desktop in less than 10 seconds and everything opens up in less than 3 seconds. That's more than enough for a laptop which will barely see any use in the future.

About RAM and timings: be careful that if you use two RAM sticks, the one with the lowest frequency will always be picked, regardless of timings.
When testing always go for one RAM stick at time.

If you end up bricking a RAM, there's a risky way to try and flash it again, which is start with one good module, load up windows, hibernate, put the bad RAM in and get out of hibernation.
SPDtool should be able to detect the new stick.
This is risky and could potentially harm your motherboard, so YMMV.

 

Yeah, I remember it well. Actually, I have a USB flash programmer, so I'm pretty safe with doing such things.
Let me know if you need help

 

Quick update: with a stroke of luck I've found out that one of my parent's current notebook was running a P8600 with a stepping of SLB3S which has CPUID 10676.

To put it short: I've put the P8800 in my parent's notebook upgrading their P8600 and the P8600 in my Packard Bell and it worked first try.

Now, I want to see if SSE4.1 instructions are enabled and all that stuff, but I can definitely say that this experiment turned out a success.

 

Nice! What FSB frequency is it running on?

 

800Mhz of course.

I'll try later boosting the FSB to get back to full speed.

 

Ok, good luck with that. Let me know later

 

I've managed to do quite a substantial overclock, but stopped at 800Mhz for RAM, reached 2.1Ghz on the CPU.

I need to SPDtool the RAMs and bring them down to 533Mhz to obtain 3:4 ratio so I can OC more.

I'm asking a friend who dabbles in assembly language to craft me a small program to test SSE4.1 instructions since I can't find anything online that it will.
All the software I've found just checks the CPU and tells me what instructions it supports without actually testing them.

As for the RAMs, I'll probably use the more relaxed 667Mhz timing and set the speed to 533Mhz.

EDIT: yes, SSE4.1 instructions work without even needing a microcode update. That's huge.

 

After you bring the CPU back to the stock 1066 FSB, try to run memtest and see if any errors show up without raising the RAM voltage