M11x clock generator ICS9LPRS387BKLF

Is it a BIOS option you change to overclock?
Also what BIOS do you have (any chance it's AMI)?

 

If FSLC is gated by the bios AND if you did figure out how to send a 1/0/0 to fslc/fslb/fsla to set 333Mhz mode, then you'd need sufficient voltage to be able to run at that overclock. BGA CPUS can't be voltmodded easily, though I did see the Acer Timeline folks modify the voltage regulator instead. In which case, *if* rmclock has higher voltage settings AND you can disable speedstep in the bios, it might be able to boot your OS up until the point that rmclock can take over with higher voltage and enabling speedstep.

 

You can disable speed step in the BIOS. And to overclock it is a bios option simply enable/disable

 

Yes, OverClock mode is either "Enabled" or "Disabled" and it's done through the BIOS. It's a Phoenix BIOS.

Ah yes, of course. Forgot about the VCore. I doubt it would make it to the OS @ 2.16GHz on stock voltage. It does seem tough, unfortunately the information you linked to above is a little out of my depth. With IDT not releasing the datasheet to us, it's really limiting our options.

 

Can you enable any higher voltages in rmclock than the default ones?

 

BIOS

 

Just curious, say if it wasn't pin 23, could we do irreversible damage?

 

Whoops.. other problem. Your GM45 chipset supports only up to 166/200/266Mhz FSB as determined by BSEL/FSLx pins (see series-4 chipset datasheet). Means no point in BIOS hacking as they would not be gating the FSLC pin. Seeing 333Mhz FSB would render the MCH disabled.

[ redundant - deleted ]

A different way of achieving portable gaming?

Before voiding your warranty, if need portability with a better CPU, consider a 12" SL9xxx HP 2530P or P8xxx Lenovo X200 + US$250 HD5750 DIY ViDock. Yes, you need an external LCD too. Either of which have had good recent deals and are smaller and lighter than a M11x and overall will give same or better performance than the G335M. The X200 may be able to do a x2 1.0 link in which case you'd get some serious framerates.

 

Just realized my post was stupid; edited

 

After sending about seven e-mails to various departments of IDT requesting the datasheet for this PLL, and being told (9 days ago) that it was a "custom device" and that "it can't be shared", I get this interesting message...

Straight and to the point! Now why couldn't Kyle be like this? ​

 

Nice work! You should forward it to Kyle and be like Booyakasha!

 

Software and hardware overclocking methods below based on the M11x's ics9lprs387 PLL datasheet.

Software overclocking M11x's 9lprs387 PLL

1. Disconnect TME pin 14 from the rest of the circuit by either lifting a leg of a resistor leading from the TME track OR cutting the track. Consider that you'll need to attach a wire at the disconnect point so a resistor leg is more convenient. 2. Re-route TME pin 14 to GND via a 10k resistor to allow software programming. There have been reports that by disconnecting the TME pin is sufficient, so this may not be strictly necessary. 3. The datasheet doesn't mention what data the PLL needs for programming. Configuration Register 12 is the last one it describes. Usually 13 and 14 are used for software programming. Suggest manually enter the values in the table in this post to see if can underclock the PLL from 200Mhz->183Mhz. If so, just need a values for > 266Mhz from another ICS9x PLL owner's system.

Pinmodding the PLL 266/1066->333/1333Mhz FSB (6.5x333=2.16Ghz)

1. Enable 1.73Ghz bios overclock. That sets FSLC/FSLB/FSLA to 0/0/0 (266Mhz) 2. Thaiphoon Burner the 1066Mhz SPDtable RAM entry to have 1333Mhz timings 3. Disable speedstep in bios. This may be needed to hold the CPU at lowest multiplier until Windows loads and rmclock can then enable higher voltages and multipliers. 4. Disconnect FSLC pin 7 from the rest of the circuit 5. Route FSLC to a 3.3V point via a 10k resistor to set FSLC/FSLB/FSLA to 1/0/0 (333Mhz). 6. Report results.

 

Wow, amazing stuff!

 

Looks like we're getting somewhere.

 

Now who really has the steady hand to do something like this. Those legs are so microscopic.

 

Hmm, I used to work for a guy that designed and built boards, might need to track him down, especially if somebody confirms this will work.

 

With my luck I'd brick the mobo and I don't think my warranty covers a botched pin mod. *sigh* If only there were a software option...

 

Maybe if you can figure out test mode.

 

.... realistically, if this involves soldering, it's not an option.

 

Has someone sent this to the CPUFSB guy? He said all he needed was a datasheet..

 

Yes, awaiting his response...

Like DR650SE, I'm not prepared to risk bricking a brand new system within warranty for a few extra MHz. As htwingnut mentioned, the legs are tiny, and I don't have the experience nor equipment to carry out the job confidently.

Moral hazard mentioned before that Abo (SetFSB creator) found a way of overcoming a TME-locked PLL through software only. So I'm eagerly awaiting to hear from Wolfram, and I'll update the thread with his response when I get it.

 

abo's software tme-workaround for the ics9lprs355 PLL is here.

 

All I know is that I tried a pin mod in a netbook and sneezed while soldering and the super tiny resistor disappeared. LOL. Luckily the machine still worked, but it's an impossible task unless you're really experienced and have the right tools.

 

You'd want to check if the pins of interest (FLSC pin7 and TME pin 14) have resistors leading off their tracks. It is much easier to attach a wire to a SMT resistor edge than a tiny PLL pin. Maybe someone can volunteer a detailed photo of those two pins?

 

I may not know anyone or can contribute in this area but I'm cheering you on RPG.

 

Someone linked to this in another thread:

http://tech.163.com/digi/10/0309/06/61AIAUF00016192R_3.html

It shows the clock generator, but maybe it's not zoomed in enough.

 

curious, if we did get it overclocked, are we pretty sure the CPU in our m11x's will work at that speed? And will it run at the stock voltage, or the neutered OC .9 settings we have now?

I'm not ruling out giving it a shot if we thought it might truly work...

 

Good point. If we unlock it, then it's of little use if you can't change voltage, unless you get lucky and get a decent overclock at stock voltage. Too bad these suckers are soldered to the motherboard. It would be nice to get a SU9600 in there and see if it would OC to 2.1GHz (8x266MHz)

 

That's my fear, finally getting it to overlock, but then it being locked at a low voltage that won't run! That would suck! Dissapointed it's not as easy as simply moving a slider in setFSB like on my other machine, that'd be awesome! I'd be more than happy to pay for a version of the software that would support our machines that gave us a bit more cpu speed.... so far VERY impressed, but more is always better

 

I agree. I'm not about to jump through hoops to make it work, but a fairly simple software interface for a nominal fee would be fine by me.

 

Received a response from CPUCool, and it doesn't look promising I'm afraid:

This seems to be similar to what nando4 mentioned earlier here:

Does anyone know (nando4 maybe) what exactly I need to ask him to do? Will I suggest to him manually entering the values in this post to see if we can downclock? I understand we'll still have to disable TME mode though.

I'll open up my laptop when I get a few minutes and take a detailed photo. I don't think the picture Mackan linked is zoomed in enough unfortunately.

 

Any more news on this?

Edit: More specifically,

1) Has the mentioned software tme-workaround (switching PLL1 to PLL3, etc as mentioned in linked thread) been tested yet, and if so, any results?

2) If tme was bypassed, did anyone try the mentioned downclock, to figure out values for register 13, 14?

 

Wow, actually found people that have the same PLL that I have. My laptop is a Sager np2096/Compal JHL90. The bus in this system is 266 and the CPU is 2.53ghz. When looking at the data from setfsb, a lot of my values are different. For example, byte 09 in my system is 25 instead of E5, which from my understanding means that my 9LPRS387BKLF is not TME locked. I tried different FSB values under the 355 and 365 PLLs, all of which resulted in my system instantly freezing.

When looking at the datasheet it looks like pin 6 is at 3.3V, and pin 7 for the 333fsb takes 3.3V to switch to 1, so I'm thinking that I may be able to solder pin 6 to pin 7 to at least try the higher fsb if software fails. I did this before on an inspiron 8200's PLL and the thing successfully went from 100 to 133mhz (so 533 effective fsb). The thing is the i8200 had a chipset that officially supported the higher system bus while the chipsets that we have do not. My CPU also is not CULV, so I'm wondering if that did miraculously work that I would be locked at a 6x multiplier.

Here's my info:

http://img11.imageshack.us/img11/1491/np20969lprs387bklf.jpg

I tried the downclock listed in the other forum, but when I hit apply the values would just revert to what they were originally.

 

Yes - TME is disabled on your system. The ICS9LPRS387BKLF datasheet is missing info on how to program the PLL. So you could try to program it like how a ics9plrs355 is programmed - manually change byte 14(OE) and 13 (OD) in the Diagnosis window. Suggest try small increments of byte14 to see if it alters the FSB. Some example values here.

You could also attempt a 266->333Mhz PLL pinmod. If the chipset sees BSEL of 333Mhz, it will refuse to run (PM45 only supports 200 and 266Mhz). If the CPU sees a BSEL of 333Mhz, it will multiplier lock you. So the solution is to separate BSEL-cpu and FSLx-pll signals. See FSLx pinmod here, comparing it to FSLx+BSEL.

Your experimentation could path the way for m11x owners to tme-unlock their PLLs if the PLL proves to be software programmable.

 

I tried a lot of values for 0D and 0E. These included what would set the 9LPRS355 to 133mhz and 200mhz. I also tried things like changing my 0E to values like 0x69 and 0x78, 0D to 0x08 (since 0x88 to 0x48 to 0x08 seemed like a possible pattern for 0D). I tried many more other combinations too, and every single time the fsb would not budge and the 0D and 0E values would reset. The only thing that would happen at all was that whenever I hit "apply", my cpu would temporarily jump to a 9.5x multiplier, then drop back down to 6x, which I'm assuming was just because it received some work to do from setfsb. I have little experience with setfsb, and the lack of change in the values made me wonder if I was doing things wrong. The way I attempted to change the values was by selecting one value, change it, click update, select the other, change it, click update, then hit apply. Was that the proper way to try to change the fsb?

When looking at the screenshot in the 9LPRS355 thread, I noticed that 0D was at 0xCF and 0E was at 0x44, even though the bus was running at 133mhz, which should correspond to 0D being 0x88 and 0E being 0xBF from my understanding. Am I somehow reading this wrong? Why did you have different 0D and 0E values from what the table in your post said you should have?

 

Hm, so what is the consensus here? The PLL seems not to be software programmable (outside, and inside BIOS?) no matter of TME status?

Actually, it would be interesting to clearly list what could be done in BIOS, assuming someone is good enough to mod it. Anyone with knowledge in the matter?

And what about other "tricks" like: DSDT, NVRAM monitoring, etc?

 

Asus ULx0VT owners can overclock their SU7300 to 1.9Ghz here by running turbo33 mode, then using setfsb. This means then that the only likely way to overclock the m11x beyond 1.73Ghz is to figure out how to program the PLL. It is unlikely the CPU could handle a 266->333Mhz PLL FSLx overclock (ie: 1.73Ghz->2.16Ghz).

 

Nice, I wonder if this will work on the M11x, I'm at work so I don't have access to the tools, but being a differant clock generator, I wonder if it's possible.

 

Khenglish's notebook has the same PLL as the M11x, and the TME was disabled. Yet, when he tried, the PLL didn't answer to anything he gave it. So far. I am not sure if that means the PLL is not software programmable, but it doesn't look good I guess.

 

I might try soldering together the 6 & 7 pins on the PLL soon and see what happens. It probably won't work but i could try it. I forgot where the PLL is though so I have some exploring to do.

 

keep us updated~ thanks all

 

Nah, my guess is Khenglish probably didn't enable the M/N divider by setting bit 1 of register 21.

EDIT: BIG WARNING, just tried this with my Acer and although it was successful increasing the CPU clock it may also affect some other clocks. Ended up with hard disk not responding, pulling the power and a corrupt and bricked BIOS

Insyde BIOS recovery to the rescue, phew... Have to do some further investigation.


EDIT2: Okay, problem fixed. Although I had configured reg 10 I missed reg 0.

Seems to work well now. Not sure if pulling the power or bad clocking caused the BIOS to brick.

Still have a lot to do to get the OC up to speed, might even have to write some SW for the PLL but be happy that the 387 OC's okay.

 

It works! This thing has a bios backup in read only memory so I don't have to worry about the bios dying and bricking the laptop. What values have you found correspond to specific frequencies? Only one I got working so far is 0E at 0x69 for 1% OC. Judging by the crashing it looks like this also increases the PCI-E and SATA busses which is great for me since I want to get an external card and need all the PCI-E bandwidth I can get.

EDIT: hmm even 1% OC is making HD unstable. I highly doubt that 1% is too much for it so I'm wondering what's going on.

 

[Unrelated to the M11x as likely can't be clocked to 333Mhz FSB]

Found this excellent post that sets the PLL's FSLx pins to successfully run at 333Mhz on a PM45 chipset. Required increasing the CPU voltage AND changing RAM timings. Your T9400-2.53Gz would be running at 3.16Gz if successful.

 

Can you post a screen shot dump of your PLL registers using the PLL Diagnosis (PLL Diagnosis selected) in setFSB or simple use clockgen PLL setup "Read".

 

Already did on the page before this.

 

Ha nice find. Do you know why he used the 6.8k resistor? I was thinking of doing something similar but without the resistor. I just ordered some conductive glue to try it out. Glue seems safer than solder since it takes forever to dry so I can just wipe it off it I don't get it on right or it doesn't work, and I don't have to worry about overheating and frying the insides of the PLL. Not as strong of a bond but it would just be connecting 2 adjacent pins.

 

Oops...

Try the following,

Reg 0 (00) 01 to 02

Reg 10 (0A) FD to E5

Reg 21 (15) 00 to 01

Use reg 14 (68) increase to change frequency.

Got to go out, will try to explain later. GL

 

Works beautifully! TY!

 

Great!!

FYI: setfsb is being developed again. Can one of you guys email the details of what is being done so abo can add this PLL to his software? Means then m11x users who tme-unlock their PLLs with a pinmod could see a 1.9Ghz stable overclock like Asus UL30VT/UL80VT users are getting.