Fun with pin modding: Dell 9300 dothan 755@ 2.66, laptop wr

[lucidchaos]

After reading about the luck people were having with pin modded dothans in laptops, I tried my luck by replacing a dothan 750 with a 755, bsel[0] modded and vid modded. The results:

31 sec superpi 1M

http://valid.x86-secret.com/show_oc.php?id=19366 - my cpuid verification (complete with dell motherboard )
Since the voltage change is a pin mod, the voltage you see in CPUID (1.340) is not correct.

I had to play around with many small copper strands of wire to get it -=almost=- 100% prime stable @ 2.66. After trying out every voltage setting, I have determined 1.644v to be the most stable voltage. However, prime still fails after about 8 mins of 12k FFT testing. All the other tests are stable, and it runs every program I can throw at it like a champ. I ran prime on it overnight with the keyboard taken off the machine - it lowers the temp 3 degrees C.

Even if it were 100% stable I would still not use it regularly at full speed. The CPU temp peaks at 78 C and the fans kick on high when the CPU temp hits 70 (very loud on the 9300 on high!). So I scaled the multiplier back to 19x and am "settling" for 2.53 ghz. I can then drop the voltage down to 1.516. There the temp never breaches 65 C and the CPU fan stays on low. I've tested prime on this config twice overnight with no problems.


My socket w/ the jumpers (bsel0, vid0, vid1, and vid4)

I got the idea by reading this notebookforums thread: http://notebookforums.com/showthread...6&page=1&pp=15

Gotta give credit to ziddey for the great find for the pinmod and Spazilton for the vid mod. You can read about ziddey's 6000D in this thread: http://www.xtremesystems.org/forums/...ad.php?t=61875

The interesting thing about this vid mod is the speedstep still works, it's just crippled. At 6x multiplier, my chip is asking for .700v and getting 1.004. Not bad considering the chip ships stock running at .988v when stepped down.
BTW my idle temps actually dropped, since I removed the janky thermal pad it comes with and applied AS5 on both the CPU and GPU.

Oh and since this is my first post, I would like to add: thank you everyone for all the good reading! You have all been an inspiration to me!

Edit: my 3dmark '01 run: 25698!
http://service.futuremark.com/compare?2k1=8587678

[The Mofo]

Thats pretty kickass man

[metro.cl]

awsome 31.xxx sec 1m super pi on a notebook is insane

[charlie]

wtg... whatcha gonna do with the 750??

C

[lucidchaos]

Probably just sell it on ebay.

[mr_knowitall15]

jesus!!! did that say the cpu temp was 73 degrees???

[lucidchaos]

Yeah, the dothans can get pretty hot without problems. The thermal spec is 100 C. I had it up to 83 w/ the voltage @ 1.692 before it started locking up, and even then it took a good hour and a half of prime to do it.

I didn't feel too uncomfortable running a mobile chip at those temps. I've seen my Pentium 4 M laptop running at 90 C at stock clocks! I had no idea it was running at those temps until I noticed it throttling to half speed after a bit of gaming. My heat sink was clogged with dust. They have to design laptops with that in mind, I'd imagine.

[Pejsen]

That frickkin' insane in a good way

How about us who have latops with 855 chipset, and are ofcourse using 400fsb mobiles, can we mod the chip to force it @ 133fsb or is it only the newer chipset which support 133fsb from birth?

My 2.0 would sure love to run @ 2.66GHz :P

[TSFroggy]

That frickkin' insane in a good way

How about us who have latops with 855 chipset, and are ofcourse using 400fsb mobiles, can we mod the chip to force it @ 133fsb or is it only the newer chipset which support 133fsb from birth?

My 2.0 would sure love to run @ 2.66GHz :P

It's only for 915GM because it's the only chipset that can support running 533 FSB.

[Pejsen]

It's only for 915GM because it's the only chipset that can support running 533 FSB.

Thx

*me runs of to get a new laptop*

[lucidchaos]

Yeah, you could pick up a new one that runs native at 533 fsb sometime with like a 1.6 in it and swap the two processors. Although then your old machine would take a speed hit cause you'd be running a 533 chip at 400mhz. If your old laptop can use a clockgen you can work around that though...

Update and question for all: I was able to change my memory timings from 4-4-4-11 to 4-2-2-4, been running it for a while with no problems. Haven't noticed a performance boost in memory tests or anything, but cpu-z does indeed report it running at 4-2-2-4. Should I notice a boost in performance? Am I putting myself at risk of instability?

[BitpowerPM]

It's only for 915GM because it's the only chipset that can support running 533 FSB.

But what could happen if setting up BSEL0 on 855-based Notebooks?
Does it switch to FSB533 or not?

I think most of i855 Chips getting FSB533 easily.

cu, BitpowerPM

[[XC] moddolicous]

wow, nice job. Did u apply AS5 maybe to lower temps a little more??

[Vapor]

Is it just me or are the two images of the socket with the wiremods different? The one with the overlay seems to have the wires one row lower? Which is correct?

[dippyskoodlez]

I'd hope it was the top one, since its an actual pic, and its probably what he's running

[A Scream]

Is it just me or are the two images of the socket with the wiremods different? The one with the overlay seems to have the wires one row lower? Which is correct?

Yeah... I noticed that to

[lucidchaos]

But what could happen if setting up BSEL0 on 855-based Notebooks?
Does it switch to FSB533 or not?

I think most of i855 Chips getting FSB533 easily.

cu, BitpowerPM

No, 855 chipset mobo's don't even check bsel0, they just assume it's a 400fsb chip. In fact, if you put a pentium m 7x0 chip and stick it in your motherboard it will run at 400 rather than 533 (for example it will downclock a 760 @ 2.0 to 1.5 ghz)

wow, nice job. Did u apply AS5 maybe to lower temps a little more??

Yeah, I applied AS5 on both my CPU and GPU in an attempt to lower my temps. I even varied the thickness a little bit to see if I could get any colder. (to no avail.)

Is it just me or are the two images of the socket with the wiremods different? The one with the overlay seems to have the wires one row lower? Which is correct?

Whoops, glad I spent more time on the wiremodding than I did the photoshop

The picture of my socket is the accurate one. Thanks for pointing that out. I took the picture down (hopefully before people started following my diagram).

[mcbarnet007]

how did you change the timing of the ram?

[lucidchaos]

RightMark Memory Analyzer. Although the timing change registered with cpuz, I don't see any percievable performance difference when measured with sisoftware sandra or RMMA.

[dippyskoodlez]

RightMark Memory Analyzer. Although the timing change registered with cpuz, I don't see any percievable performance difference when measured with sisoftware sandra or RMMA.

tiomings most of the time have little to no impact on most benchmarks. critical programs like superpi will show a difference, however.

[bigjohns97]

25k on a notebook is awesome, makes my pc feel E witty bitty.

[hitmanx2]

I just want to know what if your cpu die in inspiron 9300, will deal replace one for you?

[lucidchaos]

lol I think they would take the chip off and wonder "where is the thermal pad? no wonder it overheated, it has no thermal pad!

I'm keeping the stock 750 around just in case my 755 goes boom.

I think i'll try another superpi run with the new timings soon, was able to score a 31.1xx after a couple tweaks last time, maybe I can hit 30...

[AKURA]

I can't wait to do this mod on my gf's dell 1.6Ghz lappy

[lucidchaos]

Ok I'm now in 30 second territory:


I found a clock gen that <sort of> works with the lappy. I can pump up the fsb from 133 to 134 :/

Now I need to figure out why that is the limitation. I tried setting my multiplier to 12x temporarily to see if I could raise the bus speed more but to no avail. I tried locking my PCI at 33. I tried doing a rain dance before the FSB change. No results. It still craps out within a couple of MHZ. Even in the other direction, dropping the bus speed from default. And here's the kicker: if I raise it just a little more than it should be, fail superpi 1m at the 16th iteration, and then set it back to normal, superpi fails IMMEDIATELY.

What is the problem here?