X40 overclock ?

[Droider]

Hi everyone,

I have been trying to overclock my X40 with setfsb without success. Is there a way to squeeze a little bit of more processing power ? Mine has 1.2ghz banias board.

[Screamer]

I assume that your X40 has an ICS950813YGT phase-locked loop chip? That might be the reason why SetFSB was not able to overclock it.

Yes, by soldering pin 55 and pin 54 together on the phase-locked loop chip. This would give you an effective front-side bus speed of 133MHz (133MHz x12 multiplier = 1596 MHz), but it also has an unintended side effect of overclocking the memory controller hub.

[Droider]

Hello screamer, is there any way to overclock it a little without soldering ?

[Screamer]

As far as I know, no.

There is no overclocking utility that I am aware of which supports the ICS950813YGT, so there is no way to overclock it via software. I'm afraid that you would have to resort to soldering those pins, if you seriously need the increased performance.

The other option would be getting an X40 motherboard with a CY28346-2ZCT, but it might be somewhat difficult to find.

[Droider]

I bought an X41 with a broken screen and housing for dirt cheap today. Just finished swapping the mobo into the X40. So far, everything is working as it should..Pentium M 1.5ghz dothan, 512mb ob ram, 8 cell battery (holds charge), mint keyboard. Upgraded the ram to 2.5gb right away. I am installing windows xp now just to update its bios. It is still on the initial bios release

Lucky me kinda overclocked the X40, one way or another

ps : i was so angry and careless after wasting 2 hours to take out 2 worn out screws, i managed to break the x41 fan connector pin while pulling it out. This video helped me a lot to salvage the 3pin connector from x40 fan and put it back to x41 fan properly.

https://www.youtube.com/watch?v=S9xeg_3eREc

[Scrolls]

As far as I know, no.

There is no overclocking utility that I am aware of which supports the ICS950813YGT, so there is no way to overclock it via software. I'm afraid that you would have to resort to soldering those pins, if you seriously need the increased performance.

The other option would be getting an X40 motherboard with a CY28346-2ZCT, but it might be somewhat difficult to find.

Sorry to necro this thread, but I bought an X40 recently. any chance you have/could make a diagram showing where that chip is on the board? An overclocked thin and light from 2004 would be hilarious.

Cheers!

[Screamer]

Unfortunately, I've never owned an X40 in my life. That means I have to resort to guessing the phase-locked loop's position, and additionally I could be wrong as well.

You'll have to figure it out by yourself, or hopefully somebody else could help you out.

[kfzhu1229]

Well I just picked up a Dell Latitude D400 with the same low specs and the same problem (1.3Ghz) with the same clock generator, for the cost of e-waste.
I think I will try the same clock generator short trick to see if it can be overclocked as well.
Overclocking the memory controller hub should be fine by that much overclock as long as my RAM can handle it right? Fortunately for the D400, there are no soldered RAM and have 2 full DIMM slots.

[Screamer]

That would depend on your Latitude D400's memory controller.

If it's an 82855GME, you'll be forced to use enthusiast-class DDR400 RAM modules to maximize stability. If it's not an 82855GME and it's just a plain 82855GM, you'll need the average/regular DDR400 RAM modules.

However, if you are familiar with SPDTool, you can skip the above and flash the SPD of the RAM modules that you have to DDR200 specifications. By the way, this is assuming that you have RAM modules with an unlocked SPD chip.

[kfzhu1229]

That would depend on your Latitude D400's memory controller.

So it turns out it's the 855GM so I guess that's good?
But I will need SPDTool for sure, because as memtest86 tells me none of the RAM sticks I have are stable on this thing. I don't have DDR400 modules - only DDR333 and 266 modules.
So what you are saying is I should flash my DDR333 modules to run at the speed of DDR200 and then the overclock will bring it back to 266? Also am I supposed to do the flashing on that laptop itself or I can be safe and do it on another laptop that won't crash in a few minutes?

And well you can see here, I have basically torn apart the entire laptop for nothing, because by the schematics, the pull up resistors for the pin 54 and pin 55 are conveniently located under easily accessible RAM door. Soldering them together overclocks the CPU, and is easily reversible if things do go wrong.
https://imgur.com/XikRWbB

[Screamer]

So what you are saying is I should flash my DDR333 modules to run at the speed of DDR200 and then the overclock will bring it back to 266?

According to the memory dividers available for the 82855GM, yes.

Also am I supposed to do the flashing on that laptop itself or I can be safe and do it on another laptop that won't crash in a few minutes?

It's your choice, the main thing that has to be done is the SPD flash.

Either way works as long as it gets the job done.

[kfzhu1229]

It's your choice, the main thing that has to be done is the SPD flash.

Well I end up having lots of problems with spdflash recognising the chipsets. For some reason none of the machines that I have using DDR and also running Windows 7 work. I end up digging out a oven-baked T42 that has Windows XP and only that thing works.
And that D400 itself after the mod is just too unstable to even boot to any GUI so can't do it there.
So I took out an eBay 1GB DDR RAM stick that already had its SPD modified by those ppl to fake it to a Crucial branded RAM, and modded that one to run at 100Mhz (200Mhz in DDR). I left everything as is except for the speed, which I lowered to 100Mhz. Sure enough when I put it in the D400 again, it finally booted and when I open CPU-Z, it's running at 333Mhz.
Now when I run memtest86 again, it passes!
Well thanks for your help!
Also for this thing, why do I have to modify the RAM speed for both DIMMs to get it to work stable? I thought the motherboard will run both RAM sticks at the slower rate if one of them is 100Mhz? Yet even if I put in another identical stick it starts acting up all over again.
Also on a side note, for a machine like X40 with one DIMM soldered in place, doesn't that mean you can't just get around it with 400Mhz RAM sticks anyway?
https://imgur.com/a/7HeMLFZ

[kfzhu1229]

Okay well there are apparently some hurtles that I have to deal with.
1. Are you sure that we are supposed to underclock the RAM speed to like 200Mhz? The issue with that is even when I only touch the frequency in SPDTool to 200Mhz, it will no longer work in my ThinkPad T42 (No POST, beeps indicating a RAM error), but still works fine on my other Dell laptops, which is an issue because that means I have a real hard time adjusting afterwards. Changing the latencies as well will no longer work with those too, and throw a POST error message saying that 200Mhz RAM is installed and not supported by this (Pentium M) platform.
2. It seems like this program really doesn't like Dell computers and all but one of my DDR1 based Dells will throw me SMBus error, because somehow the SMBus system device simply doesn't show up on the device manager! This might have to do with the way Dell's old BIOS (which kept the same GUI from all the way back in 1994, which is only changed on the DDR2 based machines) works because the one that worked is a crap one with a generic Phoenix BIOS, and that thing has a crap chipset that won't POST if more than 512MB of RAM is inserted. So in this way, now I am stuck with a 1GB stick that will only work on my Pentium 4-M based systems.
3. Unfortunately, it seems like the CPU isn't stable either. Unlike the X40, this thing features a standard mobile Banias chip (but in a BGA package like the X40) and not a LV one, so maybe the chip just isn't that well binned? I can verify this because if I start the computer with warmth from a previous startup, it will freeze in minutes. I then use notebook hardware control to lower the multiplier to 12 and kept the same voltage and that solves the issue - providing it can boot that far.
I wanted to use Throttlestop but even without the overclock throttlestop immediately throws a BSOD on this thing. And oddly enough I can't overvolt with notebook hardware control.
So is there any other software that can be triggered at startup and can overvolt on these things?

[Screamer]

Also for this thing, why do I have to modify the RAM speed for both DIMMs to get it to work stable?

I don't know why. It could be the overclock that's doing something, or the RAM modules of questionable quality that you have.

Remember, this is a Latitude that we're messing with. This is some hardware that I have not touched before, so I don't have any experience or knowledge to share.

I thought the motherboard will run both RAM sticks at the slower rate if one of them is 100Mhz? Yet even if I put in another identical stick it starts acting up all over again.

Again, I'm unfamiliar with the hardware that you have.

If I were you, I'd check the JEDEC profiles stored in the SPD. Any 'extra' profile that has a higher DDR speed would likely be your problem.

Also on a side note, for a machine like X40 with one DIMM soldered in place, doesn't that mean you can't just get around it with 400Mhz RAM sticks anyway?

Normally, you can't.

However, if one has a steady hand and a soldering iron, desoldering the RAM chips would be a decent suggestion. It just wouldn't be feasible, though. The most feasible way would be to change the DDR speed, but I'm not very sure if it works for soldered RAM. All I know is that there's an SPD chip for sure, but whether if it's locked or unlocked, that's the part that I don't know.

Are you sure that we are supposed to underclock the RAM speed to like 200Mhz?

No. This is not a must, but rather my preference for stability. I understand that not many value stability for reasons that I can't comprehend.

The issue with that is even when I only touch the frequency in SPDTool to 200Mhz, it will no longer work in my ThinkPad T42 (No POST, beeps indicating a RAM error), but still works fine on my other Dell laptops, which is an issue because that means I have a real hard time adjusting afterwards. Changing the latencies as well will no longer work with those too, and throw a POST error message saying that 200Mhz RAM is installed and not supported by this (Pentium M) platform.

That's unfortunate, and not quite expected.

The 82855PM supports DDR200 RAM modules in most Tecra notebooks that I've used, even the datasheet for it specifies support for DDR200 RAM modules. It seems that the T42 lacks an appropriate memory divider to use DDR200 modules, which is something that I don't understand either. Very stupid of IBM.

If it's such a hassle, you could use the DDR266 profile, but be warned that it will hit a speed of DDR354. This is why I suggested DDR400 RAM modules in the first place, or underclocking them to DDR200 specifications to maintain stability.

So in this way, now I am stuck with a 1GB stick that will only work on my Pentium 4-M based systems.

May I suggest getting better back-up machines of a wider variety? It would certainly prevent such cases like this from popping up.

Unfortunately, it seems like the CPU isn't stable either.

You can't expect a Banias to overclock that high without requiring an increase in voltage. By any chance, are you not aware of Banias' horrible overclocking characteristics?

Unlike the X40, this thing features a standard mobile Banias chip (but in a BGA package like the X40) and not a LV one, so maybe the chip just isn't that well binned?

Low-voltage or not, that doesn't determine if it's binned better or worse. I've seen low-voltage Xeon processors overclock worse on average than their standard voltage counterparts.

So is there any other software that can be triggered at startup and can overvolt on these things?

IBM_ECW would be my suggestion as it has the ability to control the Pentium M's multiplier, but it does not have the ability to overvolt it.

If you wish to overvolt the processor, you need to perform a hardware modification on the solder pads of the processor. If you're lucky, the solder pads should be located directly underneath the processor.

[kfzhu1229]

If I were you, I'd check the JEDEC profiles stored in the SPD. Any 'extra' profile that has a higher DDR speed would likely be your problem.

Well again, just changing the speed is enough to not POST on my ThinkPads, but changing the JEDEC results in my Pentium M based Dells (including this D400) not POSTing either with the error message on my Dells saying the RAM is 200Mhz and not compatible, and press F1 to shutdown.
If it doesn't work on my T42 there is no reason why it should POST on X40 either.

However, if one has a steady hand and a soldering iron, desoldering the RAM chips would be a decent suggestion. It just wouldn't be feasible, though. The most feasible way would be to change the DDR speed, but I'm not very sure if it works for soldered RAM. All I know is that there's an SPD chip for sure, but whether if it's locked or unlocked, that's the part that I don't know.

Desoldering those chips cannot be done with a soldering iron, you have to use hot air. And you have to seek a replacement that actually works with the SPD you have.
While I have replaced a ADP3420 chip on a MMC-2 CPU module before (it's exactly like the ADP3421 in the blink of death T2x devices) with my heat gun before, it was quite sketchy because my hands are shaky and my heat gun is a little too hot so I have to be careful to not burn the chip.

The 82855PM supports DDR200 RAM modules in most Tecra notebooks that I've used, even the datasheet for it specifies support for DDR200 RAM modules. It seems that the T42 lacks an appropriate memory divider to use DDR200 modules, which is something that I don't understand either. Very stupid of IBM.

If it's such a hassle, you could use the DDR266 profile, but be warned that it will hit a speed of DDR354. This is why I suggested DDR400 RAM modules in the first place, or underclocking them to DDR200 specifications to maintain stability.

I have a feeling it's also IBM omitting 200Mhz RAM support completely. I guess it's nice to know if someone actually comes with a X40, that way they know to NOT downclock to 200Mhz. It would be a nightmare to deal with I am sure if someone did that to the onboard RAM and the machine no longer POSTs.

May I suggest getting better back-up machines of a wider variety? It would certainly prevent such cases like this from popping up.

My backup machines are not bad machines, but they are not compatible with this SPDtool. I have 2 Dell Latitude C840's, very solid machines competing against the likes of an A31p. They will work with this RAM but I can't get SPDTool to work because somehow the SMBus driver is propietrary or something. And there is the low end Inspiron 2650 which does work with SPDTool but does not POST when any more than 512MB of RAM is installed. So I used that Inspiron to flash back my 512MB module but am unable to do so with my 1GB stick.
I have an A30p so I never bothered with A31p or T30

Well then I assume overvolting isn't an option for these CPU's short of the multiplier unlocked ones. I do so on my X9000 and took that for granted.
The other problem I am having is that Speedstep no longer works after this mod, no matter what I do. It will blast at multiplier 13 which is not stable as I said.
Using nhc and setting it to only run at multiplier 12 doesn't bring the multiplier down to 12 unless I fire up the stress test. Weird eh?
I doubt IBM_ECW will work on a Dell. I want to see if I can find another program that works like throttlestop if it doesn't give me BSOD on launching the program.
But anyway, thanks so much for your help. I think this thread might be helpful if someone comes with a X40 or X41 and wants to overclock this same way.

[kfzhu1229]

Okay so I think I have made some great progress today in getting the overclock work well.
So this is what happened:
First of all, it turns out that I was having both CPU and RAM instabilities. The way to differentiate between them is that RAM instabilities usually do not result in sudden freeze where your mouse cannot move until you restart the computer.
So, I took the advice of possibly altering the VID pins to change the voltage. Fortunately there is schematics for this Latitude D400 that points all the VID pins to the CPU power supply IC right beside it, SC1476ITS (since obviously this CPU is soldered so it's near impossible to find where, if at all, the solder ball actually connect to the back of the PCB), and that chip has all the VID pins as input and outputs the correct CPU voltage.
So what I ended up doing was I attempted fooling that chip by changing VID2 from 1 to 0, so that my corresponding VID of 1.388V gets shifted up to 1.425V. Still short of the 1.48V of the 1.5+Ghz variants of the CPU at max speed but well this is the easiest VID change I can make.
My first attempt was to connect the VID2 leg of that chip to ground. It turns out it was near impossible to solder, so what I ended up doing was I took out my conductive pen and drew a trace to nearby ground by a capacitor (which still turned out harder than I thought because it had to be so narrow). That worked for about 30 minutes, and then suddenly the machine clonked out and never turns back on again! That was scary as hell and it gave me the dreaded symptom of turning on for a few seconds, nothing happens, and then shuts back off, which on a Dell computer indicates something severely faulty on the motherboard that the SuperIO shuts down the computer!
So out of desperate measures, I scrubbed that entire area with alcohol on toothbrush and it came back thankfully!
It turns out what ended up happening was my way of drawing a trace like that results in too high of a resistance that the voltage control chip gets confused. The pen says 1Ohm/cm and I have 1cm of drawing, but I think it is also because my final result is about 1/8 the thickness of the regular pen drawing on paper so the resistance was too high.
So attempt 2 was to simply dab on the leg of the chip to connect to a surrounding leg to hopefully bring the leg to 0. The chip has pin 9-14 as VID5-0 respectively, and 1.388V has config as 0 1 0 1 0 0 from VID5 to VID 0 respectively, and 1.425V has config 0 1 0 0 0 0, so I ended up connecting pin 12 to pin 13 with a drop of that ink, while cleaning up spillovers carefully (shorting VID4 to 0 will send crazy amount of voltage to the CPU and fry it so I had to be super careful) and that did the trick! You can see pic below as what I did to that chip.
Now with the CPU overvolted, there are no signs whatsoever telling you that it actually is successful, but I know for darn sure it is, because I no longer get any frozen mouse cursor freezes, and I have to stress the computer with games or web browsing (filling up the RAM) to get a bluescreen.
So yeah the RAM was still a problem.
It turns out just setting the RAM speed to 200Mhz without touching anything else for this Latitude D400 just does nothing. The system will just force it to run at 266Mhz. Touching anything else at the same time results in that 200Mhz error message and tells you to shut down the computer.
And unfortunately even after I got my entire collection of DDR RAM on my hands I do NOT have 400Mhz sticks, so my only bet was to try the 333Mhz sticks one by one to see which one is stable when downclocked to 266Mhz and then overclocked to 266*533/400=354Mhz slightly over 333Mhz.
Lucky me I literally tested to the very last DDR RAM stick that has an unlocked SPD and that worked! It was a Kingston 512MB PC-2700S RAM stick that came out of a Toshiba Satellite A70 that I just sold a few weeks ago and that passes memteset86 and windows memory diagnostic. I quickly put metal vest on that thing to make sure it does not overheat with this overclock.
Then as what to put in DIMM B, it turns out a 256MB Infineon PC-2700S stick that is SPD locked actually works and passes memtest86 when used with this SPD modded Kingston stick that brings the RAM clock to 354Mhz.
So at last, I finally have a working subnotebook overclocked to a much more sensible 1.7Ghz clock speed.
I hope my experience with this might help ppl with their X40's to perform this overclock.
Also, if anyone has actual PC-3200S sticks lying around that would be great as I cannot find one for anywhere close to a reasonable price that is not a fake overclocked one from China.
Here is what I did to the chip (Keep a keen eye on pin 12 and 13 and you can see just a very faint bridge between those 2 legs):
https://1drv.ms/u/s!At7-rt1KJ7WviPxM0ug ... w?e=tG6eg1

An useful link potentially here for the list of VID settings with their corresponding voltages:
https://www.overclockers.com/forums/sho ... tmod-Guide

[kfzhu1229]

That would depend on your Latitude D400's memory controller.

If it's an 82855GME, you'll be forced to use enthusiast-class DDR400 RAM modules to maximize stability. If it's not an 82855GME and it's just a plain 82855GM, you'll need the average/regular DDR400 RAM modules.

However, if you are familiar with SPDTool, you can skip the above and flash the SPD of the RAM modules that you have to DDR200 specifications. By the way, this is assuming that you have RAM modules with an unlocked SPD chip.

Well after more experience with a bunch of other models doing this overclock, I think the thing with 400Mhz PC-3200S RAM sticks is that if you have one of those sticks without a 266Mhz (133Mhz) SPD profile, then the computer will automatically use the 333Mhz (166Mhz) SPD profile but run it at 266Mhz instead, and then most likely you end up with a stable system as a result of the loose timings.
Otherwise, there is still a good chance the 266Mhz timings are way too fast to be run at like 355Mhz, as I learnt the hard way with my Hynix PC-3200S 512MB RAM sticks, and I still have to modify its SPD timings, after transplanting an unlocked SPD chip inside.
I've also run into Micron PC-3200S RAM sticks that does not like the FSB overclock no matter what I modify in the SPD. Turns out even when I got another stick modded and working already, inserting this stick inside with stock SPD will still result in no POST or memory error on POST, so these sticks are lemons.