M11x clock generator ICS9LPRS387BKLF

Oh god over 3ghz and everything is stable, well... ethernet was a pos and went out at like 4% OC, but the wireless card still works so who cares.

UPDATE:
Not enough power! I don't mean cpu voltage. Ran some 3dmark and graphics card can't run at max clocks at anything over 2.7ghz. When it crashes and runs throttle clocks and voltage I can run over 3ghz. Never managed to hit max cpu clock at stock volts. CPU and mem are stable at over 3.2ghz then all PCI-E devices start to randomly fail. Ex. graphics card driver crashes, wifi card loses connection, and on reboot system can't find HDD. Also noticed that the system can never successfully restart while overclocked, so I'm thinking that when the ethernet goes crazy that it refuses to post.

The graphics card crashing is kinda funny since it supports pci-e 2.0, which from my understanding is 2x the frequency of pci-e 1.1. I wonder if power issues are solved if I can overvolt the southbridge for PCI-E device stability. If 20%+ OC gets stable I'm getting a PE4H.

 

That doesn't sound good. See if changing reg 0 to 6 instead of 2 helps. 2 switches the sata clock to PLL2 so it's not affected by the OC on PLL1. 6 also moves the SRC to PLL3. I would have thought it's divider would be disabled by reg 10 but try 6 just in case.

If the PLL isn't resetting after an unstable reboot, power needs to be removed for a reset.

I am still looking to do something with my RAM for my OC, unfortunately it's DDR2. DDR3 would have been so much better.

setFSB is still in production but no longer freeware. Still at ~$5 it's good value for those that need a simple way to software OC. Maybe rpg-XPS who provided the 387 datasheet should email Abo. Who knows, maybe he will give a free key in exchange. Seems everything is there except reg 21 and the M/N divider registers, will try to look into it later.

 

Ok... so for those of us who are not quite as technical as others... we are going to have to pin mod our alienware and use setfsb to get an extra 200mhz or more?

Thanks,
stevenX

 

Changing it to 6 didn't help. The ethernet is on the PCI-E bus so both sata and PCI-E would have to be locked to bring ethernet back I think.

BTW got around to testing max stock volts stable cpu clocks. So far it looks like 2936mhz is it. All components seem to be staying in line besides ethernet. Definitely having power issues with overclocked cpu and overclocked gfx card on both battery and AC so gfx card is at stock for now. Going to drop the multiplier after I test these clocks a bit more and see if I'm cpu or mem/pci-e bus limited. I'm hoping cpu since that's something I can work with pretty easily. Can't edit SPD table with my particular memory.

 

even though I have no idea what you guys are talking about, keep up the good work!

hope there will be an easier way for us to overclock the CPU in the near future

 

I'm not sure if you would need to pin mod unless you need some extra volts for the cpu. You might need to remove a resistor that sets TME to prevent software overclocking and hope that Abo will update setFSB with the 387.


Khenglish, I don't know what's going on with your ethernet. I'm not seeing anything like that and am able to still use the LAN for validation up to 360MHz bus clock. For now I have just been using the registers in clockgen to increase clocks, not that it should make any difference.



http://valid.canardpc.com/show_oc.php?id=1147905

 

Hopefully this brings good news for the M11x, I don't have my overclocking and benchmarking drive in so I can't really try anything at the moment.

 

Wow your cpu blows mine away. Mine can only manage 2938mhz at 1.263V at only 58C (low for a laptop). If I go for a suicide shot I don't think I can get much over 3.1ghz. I looked into getting an E0 revision but it's just too expensive to justify. Your fsb is also ridiculous. When I drop the multi my bus can get up to 327mhz. Any higher then the gfx card driver starts crashing constantly. Gonna look into what I can do with the southbridge since I'm thinking that the PCI-E controller is the issue and not the graphics card since the card supports 2.0 spec.

 

Well I undid my little gfx vmod and now I can run both the cpu and gfx overclocked at the same time, just the gfx is a bit slower than it was.

The cpu is only stable up to a 309mhz bus while 327mhz is stable when I drop the multiplier. Loosening memory timings and chilling the system with an ice pack didn't help at all for trying to get over 327. I grounded VID0 and VID1 to try to get the cpu stable at the 327 bus it seems that nothing changed. Intel spec finder and cpu-z list different voltages for my cpu. Grounding those 2 pins should have given me a voltage increase for the voltages that both reported so I'm a bit confused.

 

I just run that validation with a high bus clock to see if I would get the problems you had with the LAN. I'm still thinking what I am going to do with the RAM so haven't run for stability yet. Good to see your getting on top of it.

Seems like a busy weekend but will try and get back with some more info on the 9LPRS387 for the M11x users a little later on.

 

I was trying to follow the posts, but I'm kinda lost here. So did someone actually OC their SU7300 over 3.0GHz? How?

 

thats a different cpu, they are talking about getting up to 1.9 on th m11x.

 

No, there are a couple of guys here with a Sager I think, which has the same PLL as the M11x. TME is disabled on that Sager, and they just figured out how to overclock their CPU via the PLL. Which is good news.

So to conclude, M11x still needs to be pinmodded to disable TME, and after that it should be possible to overclock the SU7300 I guess, preferably via SetFSB if the author of that program decides to support this PLL. Could maybe reach 1.9 GHz, as pointed out before. There is no way to increase voltage yet though. But if anyone has an idea for how to do that, would be interesting.

 

Ok, thanks for clarification. If there's a pin mod though, forget it, I'm out. Already tried that on another machine and fubar'd it. 1.9GHz would be nice, but not enough IMHO for the hardware mod risk.

Now if they can boost the voltage, then definitely it might almost be worth it to pull it over 2GHz.

 

Overclocking the 9LPRS387 PLL

First a thank you to rpg-XPS for the datasheet.

This is based on testing without full documentation and may contain errors.
Use at your own risk.
Do not use it for other types of PLL.
Corrections are welcome.


setFSB is used in this example but you could just as easily use clockgen.

---

1.
Open setFSB
1. Select the Diagnosis tab.
2. Select the PLL diagnosis as the Clock Generator.
3. Press the "Get FSB" button and 13 bytes will be read in the "PLL Control Register"
4. Clicking on any of these bytes will load the byte into the Offset box for manipulation.

---

2.
Click on byte 09h, if the box shown is a 1 instead of 0 then overclocking is disabled. For my own PLL it was a 1 but by removing a 10k pull up resistor connected to pin 14 of the PLL and +3V, it became 0. No resistor was required to pull the signal to ground.

---

3.
Now we know there are more than 13 registers so where are they?. Well the byte count register 0Ch (12) is a read/write register. That means we can write in the number of bytes we want to read.
1. Click on byte 0Ch whose value is 0Dh (13) and change it to Hex 16h (which is 22 in decimal).
2. Click "Update"
3. Click "Apply".



We now have 22 registers showing (0-21).

---

4.
Next we need to set some bits in register 0. There are 8 bits from 7:0. We need to make sure the Sata clock runs on PLL2 which is a fixed clock by making sure bit 1 is set to 1. For good measure we could also move the main SRC to PLL3 by setting bit 2 to 1. This would leave just our CPU clock on PLL 1.
Click on byte 0, modify as necessary and click "Update". It doesn't matter if you click "Apply" at this time or later.

---

5.
Some bits in byte 0Ah (10) might need setting. By setting the bits 6:2 shown in the Bin box we will have enabled PLL3, enabled PLL2, disabled the SRC divider, disabled the PCI divider and enabled the CPU divider. In other words the only clocks we should be changing are that of the bus clock (CPU). Remember to use the "Update"

---

6.
Now our bits are set we can enable the M/N divider by setting bit 0 of byte 15h (21). Again use "Update" and now use "Apply"

---

7.
All that is left to do is start increasing our clock. The M/N divider for PLL1 resides in bytes 0Dh & 0Eh

---

8.
Here we have increased N by 10h (16) from 68h to 78h using byte 0Eh (14). Each step of N in this case is equivalent to a ~2.54MHz change in bus clock. The result is the bus clock has increased by 16 x 2.54 = 40.6MHz from 266MHz to 307MHz. With a 9x CPU multiplier that results in a CPU frequency change from 9 x 266 = 2394MHz to 9 x 307 = 2763MHz. You may find your steps sizes are different depending on the original clock, ie 266 or 200 etc. and the default values of M and N.
Use "Update" & "Apply" to make the changes.

---

Perhaps if this information were passed to Abo of setFSB with the datasheet he would be kind enough to included it in the PLL database, data permitting, thereby making the overclocking so much easier.

 

Dufus, +1 for the excellent write up. Do you know if this is compatible with the M11x? I don't have mine with me at the moment so I can't try it.

 

Excellent find Dufus!!! This hehehe higher m11x OC!!!

 

I checked the overclocking value on mine - it's set to a "1" - which indicates that the overclocking is disabled? (screenshot)

 

@Dufus -

Whoa, dude, I'm not worthy

 

@Dufus +rep.

 

keep up the good work guys!

 

Yes, so you will have to pin mod the PLL to enable overclocking.
Not too hard:
http://forum.notebookreview.com/ali...-generator-ics9lprs387bklf-7.html#post6030240

Pin 14 is your target.

 

It should be compatible with any netbook, notebook or PC that uses the 9LPRS387 clock generator. It does seem though that the M11x software overclock is disabled by TME so the chip would need to be modded. Shouldn't be too bad, perhaps just a case of removing the pull-up resistor. Once someone with a M11x does that then testing can be done to see what sort of overclock can be achieved without raising vcore. Of course if you wanted to raise vcore as well then Nando4 and Moral have already posted with some methods for generally doing this.


Looking at the picture duffyanneal provided it looks like pin 14 goes to a pad and on. A shot from that side might possibly provide more info.

 

Does anyone know if any shops do this type of work? What about in Texas? Thanks, StevenX

 

Yeah, I'd rather have a pro do it. Although that doesn't look as bad as my Acer netbook where you had lift one end of a resister the size of a spec of dust.

If those things are really that close, seems you could dig a small valley in the PCB and fill it with solder to make the connection.

 

Hmm so I did some more testing. I can't set byte 10(decimal form) bit 7 to 0, it just stays at 1. Otherwise my settings are what dufus wrote. My pci-e is still getting overclocked, and my graphics card gets unstable over 20% OC. Apparently PCI-E 2.0 is the same 100mhz frequency as 1.0 and just magically has more bandwidth, so the fact that my card can run 2.0 doesn't help. I'm wondering if I get an external card if I could turn off the internal card so it doesn't limit me, or if there is some way to make the card capable of higher pci-e speeds. I'm also concerned if the external card could be a dud in terms of max pci-e frequency.

Hmm if I could find out how to overvolt a graphics card's pci-e controller. Could be interesting. I do have a voltage regulator on my graphics card that I dunno what it's for (modding the other raised both mem and core voltages). Might pencil mod it slightly and see what happens.

Update: Pencil modded the voltage regulator and it increased core and mem voltages. PCI-E controller seems to be within the GPU core since PCI-E can now run higher

 

Hey Duffyanneal, you mind telling me how you fixed your right speaker? Mine is dead too. I'm pretty sure the soundcard is fine as earphones work perfectly fine. Anyways, I could really use some details/pointers on what to look for and how to open the pc without breaking anything. Thanks!

 

Hi everyone, sorry for the slight OFF topic, but does anyone know about the M15x (i7) PLLs?

 

Guys, what's the status? We know what to do, but there is no one who would like to use its m11x as a test subject?

 

http://dl.dropbox.com/u/5020311/m11x/8656868.png

So its a lie about ddr3 1066?

 

do we know what we need to do? I had actually forgotten about this thread.... very interested though

 

As far as I understood, we just have to cut one Lane leading from the PLL to do the TME unlock.

Afterwards we should be able to soft-OC the M11X via SetFsB.

 

Ok i tried with clockgen and i don't see a difference at all!! Be nice to see something and not have to update to the core i CLUVS.

 

We still don't have any guinea pigs for this experiment? I'd really love to squeeze that last ounce out of my M11x.

 

So do it, and let us know how it goes

 

If we can't change FSB using soft, can we change multiplier? I imagine to disable Speedstep with OC enabled. This leaves us 266MHz bus at fixed multiplier 6. Then use Crystalcpuid to change it to 6.5.
Any volunteers?

 

I tried an almost identical mod on an acer 1410. You know when you look at photos of the resistors and solder points that need modding, you think to yourself, well thats small, but I can do it..

Then you open up yours and look at them and are now thinking, damn those are fricken tiny! But I can still do it...

Then you get the solder out and have a dangerously hot iron in your hand. Now your thinking , god damn , whered it go?


Think of it this way, maybe a 0.3 chance it will work, 0.7 chance you destroy say 1000$ worth of something, and replace it with the same thing of 1000$. On averate thats 0.7 x -2000$ = -1400 . Or you could just sell your R1 ( say +700$) and buy an R2 (say - 1500$) = -800$

Not an exact math, but that's what I would be subconsciencly thinking in my head!


In a nutshell, I won't be a guinea pig, nor have I actually contributed anything to this thread, apart from excellent logic

 

@Me-How - The multiplier is 6.5. This is well known and in many threads. You can also look at the specs on Intel site. Both the su7300 and su4100 are both 6.5 x 266 (oc in bios) = 1729 +/- depending on bios version.

The fact is the FSB can not be OC'd any more than where it's at without major mods to the TME on chip. Nobody is willing to do that at this time for an extra 100-200mhz.

Best wishes,

StevenX

 

@SeteveX - Well after reading some threads here nobody was able to confirm for sure what is SU7300 actually doing when oc'd.
It's just that CrystalCPUid recons multiplier is 6 when oc'd (max 6.5 when not). I'd have tried it weeks ago if only this wasn't my only computer at present where I am. We all know that different software identifies CPU speed diffrently...

 

The su4100 and the su7300 have two P levels. Multiplier 6 and 6.5. The way the motherboard and bios is configured for us the P level, chosen by Alienware is 6.5. If you look at advertised speed for the R1, which is 1.3 mhz, it is simple to calculate 200fsb x 6.5 = 1300. When you choose the oc option in bios it changes the fsb to 266.x (x = different numbers depending on bios). So 6.5 x 266x = 1729

 

I'm familiar with all that. There seem to be also multiplier of 8 (hwinfo can see this one) used to reduce speed/power usage simply 8*100=800MHz.
I'm just saying, is it worth trying to disable speedstep and yank multiplier to 6.5 in crystalcpuid?

 

If you can get it to go to 8x266 then you would have HERO status here.. by all means. Give it a try. I have not seen the 8 multiplier myself. I say no effort is wasted effort if you learn something new.

Good luck,

StevenX

 

did anyone try throttlestop with the EIST trick?

 

I did... the series of cpu's in our notebooks do not support IDA. I tried it.. but hey, I make mistakes so please try it again. I would be the first to admit when I am wrong and if strides are made here then I say bravo... Let us know what you find out!

Good luck,

StevenX

 

Just a quick question, but why is there a variation of overclocks among M11x models? My one is one of the ones that go up to 1.6GHz, while others are able to go up to 1.73GHz. What's the deal with that?

 

@noob - The real issue is the software you might be using to report cpu speed. GPU-z apparently is not supported by the m11x and reports incorrect speeds. What does your notebook say if you click start, right click on computers, go to properties... Look at the system area. What does it report there?

Best wishes, StevenX

 

FYI: A new version of CPU-Z (1.55) was released yesterday. It's added support for the i5-540UM, but I don't see any of the specific UM processors or the QS57 chipset that the R2 uses listed yet. On the other hand this version is displaying more detailed information than 1.54 did. Hopefully the next revision will cover it all soon.

 

Don't worry, they all are actually at 1.73GHz (with the overclock enabled through the bios).

Check with CPU TrueSpeed.

It works differently from other software, and always measures the actual speed, instead of knowing the fsb and multiplier, which be easily misreported. The issue with CPU-Z was it was saying it was at 6.0 instead of 6.5, causing the 1.6GHz and 1.73GHz confusion.

 

Thanks heaps guys! +1 for you all! Seems like CPU-Z is a bit weird eh... doesn't matter too much. I hope SetFSB will be working some time in the future. That way we can all under/overclock our CPUs easily. I remember using this for my old Asus N10, where I underclocked my Atom CPU to save power, as well as overclocking to play CoD4.

 

@stevenx: I'm pretty sure it's actually running at the lowest multiplier due to the alienware's choice of BSEL mod.
Feel free to benchmark normally, then force power settings to only use lowest multiplier if you want to prove you're right though.