BSEL Mod on a socket P explained with photos

Will do.

Any danger of this bringing the intermediate voltage steps over the 1.2v limit the CPU has by default?

 

No risk since the plage of voltages that can be accessed by the CPU after the mod is between 0.9125v and 1.1v.
...

OK, If I take your previous post as an example and assuming that you wont be using an undervolting software with the mod:
State 0: 06x 1.000v ----- will remain the same after mod
State 1: 06x 1.050v ----- will remain the same after mod
State 2: 07x 1.125v ----- will drop to 1.1v or below after mod
State 3: 08x 1.125v ----- will drop to 1.1v or below after mod
State 4: 09x 1.200v ----- will drop to 1.0v
State 5: 10x 1.000v ----- will remain the same after mod

But since the above values are from an undervolting software the only thing that I'm sure of is that x12 will use 1v, and x6 to x9 will use a vcore between 0.9125 and 1.1000v. This is why I recommanded that you should use an undervolting software, because with those you can control the vcore for all the multipliers from x6 all the way to x12.

 

Sorry for my english,i’m a french guy!

I have a problem with my T9800 on my Packard bell ipower GX m 102fr (P-7811FX in usa), impossible to undervolt less than the minimal gave by Intel (1.050V),when i use CPUgenie (RMclock is the same thing),i can select less but when i watch with coretemp or cpuZ, but the cpu is always at 1.050V,it's possible to underclock it with your pin mod?
Someone can help me?

Thanks

Nico

 

Your English is OK

I'm guessing that 1.05 v is when your CPU is running at 2.93Ghz. This is the minimum for your CPU to run stable so you won't be able to go any lower for multiplier x11 even with the mod.

 

Thanks for your quick answer!
My cpu is stable at 1.10V for multiplier X11 but i would like undervolt it when the laptop is on battery for multiplier X6.

 

That's great, thank-you. CPU-Z reports 1.2v when running under 100% load with the other software disabled...

Too late to start messing around with it tonight - it's nearly 11.30pm and I've a long day tomorrow. May be able to have a stab tomorrow night, failing that Thursday night.

Thanks again for all your help!

 

no worries

With RMclock you can undervolting for the battery too. Also you can use one power profile when you're using the power adapter and another one when you're using the battery. Below is a good tutorial to set up RMclock.
http://forum.notebookreview.com/showthread.php?t=235824

Follow the above tutorial to undervolt all your multipliers; from x6 to x11. Then, please post the voltages for all your CPU multipliers as set in RMclock. After that I'll tell you what to do next

 

My voltages are :
State 0: 06x 1.050v
State 1: 07x 1.050v
State 2: 08x 1.050v
State 3: 09x 1.050v
State 4: 10x 1.050v
State 5: 11x 1.100v

 

What? 1.05v is the minimum you can access in RMclock? You should be able to go lower than this (somewhere in the region of 0.8xxxv).

Uncheck "auto-adjust intermediate states VID". Then select one multiplier at the time (for example x7). lower the VID and test with orthos. Do that again and again until your CPU is unstable. When that happens Orthos will display a red error message. When this happens up the ViD by two values and stress again your CPU with orthos for at least 30 minutes. If you don't see any error message then move to the next multiplier.

 

This is why i want undervolt it, i can select less than 1,050v but when i check with coretemp or cpuz,vcore is always 1,050v.I think my cpu is lock at this tension.

 

What's the min and max VID for your CPU as listed in the "CPU info" tab of RMclock?

 

VID:
current:1.1625
min:1.2125
max:1.3375
But i think my minimal VID is 1.050V

 

As I thought the min Vid for the T9800 is 0.95v. Check this
http://forum.notebookreview.com/showthread.php?t=365714

I don't understand why yours is showing with a min of 1.2125v. Are you using the latest version of RMclock 2.35?

I won't be able to help you with the mod until we figure out why you can't access 0.95v in RMclock.

 

I have windows seven pro 64,i use the last version of RMclock,but the min vid for the 9800 is not 0.95V but 1.050V,in that thread : http://forum.notebookreview.com/showthread.php?t=365714 , the guy have a QS version, it's a beta version of the cpu, look this :
http://processorfinder.intel.com/details.aspx?sSpec=SLGES

 

Niko

I'm confused, really. If your CPU is the same as in here your max vid should be betwen 1.05v and 1.2125v not 1.3375v.

Assuming that your max vid is really 1.3375v I'm afraid no pin mod will work.
VID5 will drop your max vid to 0.9375 and that's to low for a CPU running @2.93Ghz
VID4 does not do much since it will block the access to your CPU to all the voltages between 0.9125v and 1.1000v

After reading this I think that RMclock is not displaying the right Vcores for your CPU. It's not 100% compatible with windows vista/7 64 bit.

You can try this and see it that fixes it.

 

I use cpugenie because rmclock have bugs with my seven 64.

Cpugenie works like a charm because when i put 1.100V, coretemp and cpuz give me the same vcore, so i see than the limits are 1.050V for the mini and 1.2250V for the maxi.
I need for my T9800 a maximum voltage of 1.100V because it's full stable @ 2.93Ghz.

 

1.2250v as a max VID is good since you can try the VID3 mod. Once this is done set CPUgenie as follow:
x11 --- 1.1125v (1.1000 will crash your laptop because after mode 1.1000 = 1.0000v)
x10 --- try 1.1000 (i.e. 1v after mod) if your laptop crashes up it to 1.1125v
x09 --- should work with 1.1000v (i.e. 1v after mod) or less
x08 --- should work with 1.1000v (i.e. 1v after mod) or less
x07 --- should work with 1.1000v (i.e. 1v after mod) or less
x06 --- should work with 1.1000v (i.e. 1v after mod) or less

NOTES:
CPUgenie/RMclock/CPUZ... will not show voltage - 0.1v after mod. I mean the volt mod will work but those software will not detect it and display the new vid accordingly.

Doing the mod your CPU will be losing access to the voltages between 1.0500v and 1.1000v but gaining access to the voltages between 0.9500 to 1.000v.

voltages between 1.2125 and 1.2250 will drop to 1.1125 and 1.1250 after mod
voltages between 1.1125 and 1.2000 will remain the same after mod
voltages between 1.0500 and 1.1000 will drop to 0.9500 and 1.0000 after mod

 

Righty, it's very late here and I've lost interest right now.

I know why my fans are going into noisy-mode when under full load. After much trial and error, followed by investigating and eventually hooking my meter up to Vcc sense pin, this CPU is faulty. I think vid4 is set to a constant low state, resulting in the CPU working at odd voltages - it'll swing from 0.80v to 1.47v, and when adjusting the voltage inside CPUGenie(I earlier referred to it as "CrystalCPU", mistakenly), selecting voltages between 1.1v and <1.2v will cause it to go right up to 1.35v and above.

I can't remember exactly which ones work correctly, but but it's repeatable. To be honest, it could've been vid3 that I narrowed it down to.

Anyway... this is a throwaway CPU. It works fine - in so much as it'll run under full-load for hours, though I'm not sure what voltage it's actually been operating under. My main query at the moment is can I snap off pins vod[0] through to vid[6], and manipulate them manually, leaving them NC for low and ->Vss for high? If so, will it only work at the one pre-set voltage, and not attempt to use alternative voltages in the speed-step profiles?

I'm really not fussed about killing the CPU, but if I can manipulate it to suit my needs then it'll save me parting with £50 for a replacement processor...

Motherboard and VRM should be fine. I tossed in my old Celeron and it's putting out a steady 1.256v on Vss under all loads.

Thanks.

 

That's too odd

What did you use for the monitoring?

without any mod that's odd

In theory it should work but in practice I can't say. Few people around the net reported removing a BSEL Pin when they attempted to do an FSB mod. I tried to remove a BSEL pin when I first started experimenting with BSEL and VID. It was with a Celeron CPU. By mistake I removed the wrong pin, and nothing happed; I mean I didn't kill my CPU. Then I removed a second pin (this time the right BSEL pin), and again nothing happend. I didn't kill my CPU but didn't succeed in pushing up the FSB from 133 to 166Mhz.

The reason I removed the pin was because of some misleading information laying around the net stating that a CPU can blow up if a VID is connected to VCC. That information is a lie since I and a lot after me (can't be sure but maybe even some before me) have safely connect a BSEL pin to both Vcc and Vss, and a VID pin to both Vcc and Vss.

A guy here in the NBR forum didn't remove the VID pins but has connected if I remember well 4 of them together to a Vss hole when he overclocked his CPU.

Bottom line I don't think you should remove the pins because you won't be able to go back. The adventage of the Pin mod is that's reversable.

I think you mean 1.2625
The celeron is a good condidate for a BSEL/FSB mod if you're interested.
You can also undervolt it. I did that for my present Acer laptop before upgrading its Celeron T1400 to a C2D T5250
To undervolt it you can try VID3 for 1.1625, or VID3+VID2 for 1.1125.

 

Multimeter hooked up to Vcc_sense.

I don't understand this CPU... anyway.

Do we know if the voltage presented to the CPU is changed on-the-fly by the CPU, using the VID pins?

 

The Vcc_sense pin in the CPU?

I'm looking at the datasheet for the T5800 and similar CPU, and I can't see the voltage of Vcc_sense / Vss_sense. The only useful info is the value of Vcca (PLL supply voltage). It's tipical value is 1.5v (min = 1.425v and max 1.575v).

This seems to confirm that the CPU recieves 1.5v (i.e. VID 6 5 4 3 2 1 0 = 0000000) and changes its inner voltage on-the-fly.

Maybe Vcc_sense is not the right pin to read the voltage. Do you see a considerable increase in temperature when the Vcore goes to 1.3v plus? At this voltages your temperature should easily reach 85c to 90c?!

 

I try to undervolt my T9800 by the VID3 method...It's weird but the temperatures are the same and at 2.93Ghz it's stable @ 1.125V but when i put 1.100V laptop crash...before mod it was stable @ 1.100V.

 

1.1125v and 1v are about the same so the difference in the temperature should no be more then 1 to 2c max.

It's normal that you laptop crashes when you select 1.1v. Because 1.1v after mod is equal to 1v... This proves that the mod is working. As I said before this what happens after the mod:
voltages between 1.2125 and 1.2250 will drop to 1.1125 and 1.1250
voltages between 1.1125 and 1.2000 will remain the same
voltages between 1.0500 and 1.1000 will drop to 0.9500 and 1.0000

use this to adjust the voltages in RMclock:
x11 --- 1.1125v (1.1000 will crash your laptop because after mode 1.1000 = 1.0000v)
x10 --- try 1.1000 (i.e. 1v after mod) if your laptop crashes use 1.1125v instead
x09 --- should work with 1.1000v (i.e. 1v after mod) or less
x08 --- should work with 1.1000v (i.e. 1v after mod) or less
x07 --- should work with 1.1000v (i.e. 1v after mod) or less
x06 --- should work with 1.1000v (i.e. 1v after mod) or less

 

Yep. Oddly, if I lowered the profile from maximum voltage and speed to one of the intermediates, the temperature would increase as the measured voltage went higher, despite the clock being lower(but still under 100% load).

From the datasheet:-

Sadly, I sold my 'scope last year. Whilst the measurements on my meter won't be 100% accurate, I'd say they're painting a pretty good idea of what is happening with the voltage...

Default startup voltage across all CPUs is 1.2v, from what I've seen in the table just above the quote I've posted just now. I guess it'll then read the VID pins and adjust the core voltage accordingly once the CPU is up and running?

This is all quite interesting... it's a pain in the , but I've picked up a bit.

 

Me too I've learned a lot like the porpuse of Vcc_sense and Vss_sence and that the startup/boot voltage for socket P C2D is 1.2v.

For the Vcc_sense / Vss_sense did you do the reading at the bottom of the CPU socket or did you connect two wires to the socket holes and to a voltmetre?

 

I used two small lengths of Kynar wire(0.25mm thick/30awg, insulated), approximately 30-40mm long, hooked on to my multimeters probes.

I do think those readings were pretty accurate, to be honest...

 

If you have any doubt you can always test with the Celeron . Aren't you intersted on BSEL Modding it?

The Celeron M 530 I modded from FSB 133 to 200Mhz (i.e. from 1.73Ghz to 2.6Ghz) did very well in Sandra sisoft benchmarks. It was faster then a Pentium Dual Core T2060. In one of the benchmarks it was even faster than a Core Duo T2300 Not bad for a Celeron

 

Nah, not at all. It's an Intel chipset, so it likely wouldn't work anyway. Plus it's the 550, with a native 2ghz on a 15x multiplier. There's a good chance it'd crap out at the higher clockspeed.

Benchmarks are great... but in real-world use I'd take 2x2ghz cores over 1x2.5ghz or even 1x3ghz... I'm a keen overclocker, I've taken everything to the stable limit for over a decade now. Typing to you now on a dual-core E5200(2.5ghz) running at 3.6ghz...

 

Check this out http://forum.notebookreview.com/showpost.php?p=5404511&postcount=151
You might have to overvolt a bit to get the M550 stable @ 3Ghz

I'm not really an expert in overclocking. A while back I tried overclocking an Athlon Xp 1800+ but that didn't work. Anything over 100Mhz over the stock speed made the CPU instable. Upping the voltage didn't help. About a year ago I overclocked an AMD Geode NX 1750+ for a friend. The stock speed was inceased from 1.4Ghz to a little over 1.8Ghz. I know could have tried a higher speed but I didn't want to risk it since it is his office desktop.

I pin moded one or two Celeron M 360 from 1.4 to 1.86Ghz. I pin moded a Pentium M 715 from 1.5 to 2 Ghz, but my best achievment is the Celeron M 530 @ 2.6Ghz

 

I've probably forgotten everything I've overclocked, over the years. It's a good sport, though.

There have been times in the past where I've bent a pin to send it low(NC), and it's worked without issue. I'll have a look over the datasheet tonight and see what I figure... chances are it'd still work fine.

 

Aha... P18 of the datasheet talks of Enhanced Intel Speedstep Technology...

There we go. If I remove the pins and hardwire the VID pins to Vss(ground/low-level signal) and Vcc(high-level signal), it should always remain in the power state set by the configuration of VID0:6. Excellent. I'll snip the pins off tonight and give it a go.

I tend to think leaving them open-circuit for low-level would still work, but it doesn't really seem worth bothering with. It's a pain to strip and rebuild with each test case...

Edit - attached images showing hard-wires for 0.90v, 0.95v, 1.00v & 1.10v. No promises, not tested, your own risk, etc.

 

Just rebuilt. Hard-wired for 1.00v after snipping the pins from the CPU(using a retractable/neverending pencil is great for this).

System is up and running, appears stable(though not conclusive yet). It's been running for 15 minutes under full-load, and oscillates between no fan at all, to silent fan mode. CPU temperature according to RMClock's monitoring section goes from - under full load - 52.5C(when the fan cuts off) to 62C(when the silent fan-mode starts).

Pretty good, overall. Am tempted to pull the CPU and go for 0.9v, see how it goes... but I run the risk of mucking it up. I've reached my goal, and shouldn't tinker further...

No core voltage measurements. I didn't want to disturb the CPU if possible, so opted to run it without a battery, with my hand on the AC adapter lead to yank it out quickly if anything seemed wrong...

Edit - decided to leave it at 1.00v, rather than try and get it lower. Currently have the machine running under load to check stability...

Picture shows the pins to remove if wanting to manually set the CPU voltage... Sorry about the quality, I had a hard time getting the pin legs to focus on my phone!

 

It's a wise decition.
I have to say that you did a very good job here

 

Been under 100% load now for some 14 hours, seems fine. Am tempted to go for 0.95 or even 0.9v... but I'll be sensible and leave it.

Cheers for all your advice.

 

does it work with acer aspire 3680, intel celeron [email protected]?

 

What's your chipset? is it the intel i940 or i943?

If it's the i940 the BSEL mod won't work.

If it's the i943 or i945 it will work. You can try the BSEL mod from 133Mhz to 166Mhz. This will allow your Celeron M410 to run at 1.83Ghz. The details of the mod are in the first page

 

unfortunately, mine is i940..

 

you're prety much out of luck. You can't even upgrade to a dual core CPU

You can still gain a lot of performance by upgrading to a Celeron M450 (2Ghz) or a Core Solo T1350

 

Since the BSEL mod on the CPU won't work, try a FSLx PLL pinmod.

 

Would that work on a chipset that doesn't support anything over 133Mhz FSB?

 

*Maybe*. Depends how the chipset would handle being overclocked from 133->166. RAM can be accomodated with slower SPDtable entries, but northbridge/GPUs can't. Only way to know is to try.

 

I guess you're right

 

i dont know whether its i940 or i943.. CPU-z says its intel i943/940GML. can i try it without harming my laptop?

 

sure you can

There is a thread on the forum about the Acer AS 3680 which had few tips on how to identify if the chipset is i940 or i943... let me find it for you

 

here we go http://forum.notebookreview.com/showthread.php?t=206137
Look for you motherboard serial number:

 

if i do this, and let say my chipset is i940, what will happen to my laptop? is it won't start?

then,is the BSEL mod(as u said) is the same with the FSB mod as in the first page?

 

Worst case the laptop won't start until you undo the mod. Otherwise nothing will happen; your laptop will start as if you didn't do the mod.

If you have the i940 there is no much point on doing the mod.

BSEL mod = FSB mod

 

o... you said you're using the wire from the IDE cable. do i have to remove the insulator of the wire?

or just remove the upper and lower part?

 

I'm not sure what do you mean by the above? mind explaining?

 

it is on cpu socket isn't it?