WIP: Karaktu's (and Zap's!) Unofficial Guide to mobile Celeron overclocking success!

[aidanjm]

Originally posted by: BigBadBiologist
While sanding down a penny may require a little bit of elbow grease, it probably is faster than going out and finding a railroad track, putting the penny there, waiting for a train, getting the penny back, coming home, and giving it a final quick sanding.

not as fun, tho

 

[aidanjm]

Originally posted by: Karaktu
Anyone tried a P4S800-MX 1.04 yet?

I'm about to buy a p4s800-mx for my 1.8 ghz mobile p4. I'll let you knwo how it goes (assuming I get the 1.04 revision).

 

[mercdrive]

Just got my stuffz form Joe, now I can PLAY WWHHHEEEEEEEEEEEeeeeeeee

and I got no Kidz this weekend WHOOOOOO HHOOOOOOOOOOO


Thanx For Everything Joe you are a great guy

anybody ever messed with this CPU???

Results

sSpec Number
SL6J2
Processor Frequency
1.60 GHz CPUID String
0F27h
Package Type
478 pin PPGA Core Voltage
1.300
Bus Speed
400 MHz Thermal Guideline
30.0W
Core Stepping
C1 Thermal Spec
100°C
L2 Cache Size
256 KB Manufacturing Technology
0.13 micron
L2 Cache Speed
1.60 GHz Bus/Core Ratio
16
Spec Update
http://developer.intel.com/design/mobile/celeron/documentation.htm
Product Order Codes


I have one and want to try it on the asus p4p800 mx



anyone who has experiance with this cpu LMK

also has anyone tried using a aopen ax4spe max MB w/ any mobile / M cpus?



Thanx Joe and all the great people on this site !!!

PS Joe u got PM and HEAT

 

[mercdrive]

~~~~~~~~~~~ u got PM and Heat

 

[Karaktu]

Added info on the Shuttle XPC Model SB65G2.

 

[TStep]

Great thread, tons of useful info. I've only been experimenting P4Ms to date, but I think I'll have to give a Celery a go. Looks interesting.

 

[quentrm250]

So I got a question.
If I put a cpu in a mobo that doesn't recognize and default to 1.2v I could break off vid4 and connect vid3 and vid2 to get 1.6v correct?
Then maybe do the fsb trick too. I would only do this on a chip that I knew worked at 533 in another board of course.
Thanks,
Quentin

 

[jbmx4life]

can someone better explain to me the table on the VR-Zone website about the vcore mod to the CPU???? The default vcore for my Celeron 2.0 is 1.525.... how much do ya think it will need to get it to 133FSB stable? Which pins do I need to connect? Thanks for the help.

 

[jodang01]

Awsome thread.

I have some questions: Anyone ever try the 1.3 Ghz pentium4 M with the 1mb L2 cache and B1 stepping (SL6N4)? Would it work like the 1.5 P4 M's and be able to hit the 2.4ghz on 200 fsb? If so, shouldn't that perform better that the 1.5's with only 512kb L2 cache at the same overclock? Would there be a dramatic difference in heat production between the two at 200 mhz fsb (provided the SL6N4's can reach that speed)?

I currently have a 2.0 Ghz celeron o'ced @ 3.0 Ghz (desktop CPU, 128kb L2 Cache). Would a p4 M at 2.4 Ghz 512kb L2 cache be better than my current cpu? How likey is a 1.5 and 1.6 Ghz celeron mobile to reach the 200 fsb (800 quad pumped)? Do they last long at that setting?

I know that's a lot of questions, but I'm seriously looking into this mod instead of buying the 2.4 Ghz Celeron D.

Thanks

Joe

 

[imported_Kensai]

Great guide there.

 

[Karaktu]

I have no experience with P4-Ms, so I can't answer any of your questions about them (that's why this is a Mobile Celeron Guide); however, ANYTHING would be better than that desktop Celeron.

Most 1.5 and 1.6GHz CPUs can be expected to run at 200FSB around 1.65v, and so far have lasted about a year-and-a-half. The real issue in life expectancy is heat, so the cooler you can keep it the better and longer it will serve you.

Celeron "D" CPUs will often hit higher speeds (up to 3.6GHz) than these Northwood mobile Celerons, but the trade off is heat. You can run a mobile Celeron at 3.2/3.4Ghz and up to 1.7v with a good heat sink and silent fan.

For the $$$ investment, I would have to (biased, yes) recommend mobile Celerons for now and a jump on Athlon 64s when the 64-bit version of Windows comes out (if you feel the need to upgrade). I'm singularly unimpressed with Intel's Prescott-core CPU, and my experience has been that an Athlon 64 is not only faster, but much cooler.

 

[TStep]

To: jodang01

The overclocked Celeron 128k even @3.0 just doesn't perform well in most tasks comparison to the other options available. OC'ed XP or P4's or mobile Celerons(256k) are much better choices. Read here at Sharkey's for a comparison. I remember XBit also having a good comparison as well. So yes, a P4M @2.4 will outclass the celery128k @3.0.

I don't have the mobile celery experience to answer as far as the 200fsb being achieveable, but consider this: a B0 stepping is probably going to make 2.4G and a C1 stepping should get close to 3.0G in a broad sense most of the time. All of them are Northwoods, so the broad generality should hold true in most cases. Celeries are multi locked so just do the division to determine the realistic fsb. I have an 1.5 SL6M5 C1 on it's way, so I'll post back when I get some testing done to see if it will make the 3.0Ghz mark.

P4Ms are great chips with the right motherboard. Since they all default to 12x100 this causes several problems. A very high fsb is needed to get great results. I had a P4M B0 stepping on an AI7 and that all worked out well as the AI7 was high fsb capable. I think I maxed it out around 2.7-2.8 (fsb @ 225-230ish). Pop a P4M C1 in there and you are looking at all the complications associated with the 2.4C. Fsb will range from 250 on up to get the most out of the cpu. Obviously you are looking at needing a high $ quality motherboard (and ram if you are trying to go 1:1, which most lower end boards limit you to on the 100fsb detect) when you are looking at building a budget system, kind of an oxymoron.

Also, since the mobos detect 12x100, most of the lower cost mobos limit you ram divisors to 1:1 or 4:3. At 250, your board is telling your ram to run at 250(500) or 333(666) which is very unrealistic. This will cause you to start pin modding so the cpu is detected as a 133 or 200 fsb cpu. It's fun, don't get me wrong, but it takes alot of tweaking.

 

[rogue1979]

Originally posted by: jodang01
Awsome thread.

I have some questions: Anyone ever try the 1.3 Ghz pentium4 M with the 1mb L2 cache and B1 stepping (SL6N4)? Would it work like the 1.5 P4 M's and be able to hit the 2.4ghz on 200 fsb? If so, shouldn't that perform better that the 1.5's with only 512kb L2 cache at the same overclock? Would there be a dramatic difference in heat production between the two at 200 mhz fsb (provided the SL6N4's can reach that speed)?

I currently have a 2.0 Ghz celeron o'ced @ 3.0 Ghz (desktop CPU, 128kb L2 Cache). Would a p4 M at 2.4 Ghz 512kb L2 cache be better than my current cpu? How likey is a 1.5 and 1.6 Ghz celeron mobile to reach the 200 fsb (800 quad pumped)? Do they last long at that setting?

I know that's a lot of questions, but I'm seriously looking into this mod instead of buying the 2.4 Ghz Celeron D.

Thanks

Joe

Welcome to Anandtech Forums

Yes, a P4-M at 2.4GHz or a mobile Celeron at 3.0GHz would be much faster than a 128k Celeron at 3.0GHz.

 

[jodang01]

Thanks guys.

I looked at the sharkys review, as well as those from toms hardware and Xbit. All the tests pretty much confirmed what we all know about desktop celerons: they stink at games and 3d graphics (basically any L2 cache intensive programs).

I don't actually play any games or use 3d intensive programs. I suppose I should have stated that my primary use for this rig will be video capture, dvd rendering, mp3 ripping and mpeg encoding. In most benchmarks of this type I've seen so far, actual cpu speed and bandwith tend to be of more importance than cache size. I just don't want to spend $$ on a CPU if I stand little to gain on these tasks. With that in mind, do you guys still think that a p4 mobile with 512kb L2 @ 2.4 ghz will run circles around this cpu? The cely has less bandwith, but 600 mhz more procesor speed. I thought it might be a toss up. I'm sure I would see noticeable gains in other areas because of that 512 kb L2 cache though.

The mobile celeron would fit the bill in both cases: better L2 and the possibility of a high fsb and processor speed. Only problem for me is that my mobo (ECS pm800 m2) does not have agp/pci lock, and I can only o'c to 1mhz below the next bus speed (and 166 is not a supported bus, just the standard 100, 133, and 200). So, for a mobile celeron, it'll have to be an all or none deal: either it runs on the 800 mhz bus and hits 3.0-3.2 ghz, or it doesn't, and I'm stuck with a dead celeron. Pretty risky venture for me on the 1.5's and 1.6's. If a buy a 2.0 ghz celeron, I wonder if I could hit the 3.0 ghz like this desktop model did? That should be a slight improvement at least.

thoughts or comments?

Here's my rig currently:

Thermaltake Xaser III case, 350 watt ps
ECS mobo: PM800-M2
Celeron 2.0 ghz o'ced @ 3.0 ghz
1 Gb patriot signature DDR 400
60 Gb Maxtor ATA 133 7200 rpm
Winfast TV2000 XP expert
Aopen 16x DVD-RW (dual layer)
Lite-on 52x CD-RW
Windows XP Pro Corporate w/SP2

EDIT: I just ran sandra lite 2005; heres the specs on the 3.0ghz cely:

CPU arithmetic:
dhrystone: 7873
Whestone: FPU: 2155 MFLOPS iSSE2: 3980
(both comparable to a P4-C 3.0 Ghz 512 kb L2)

CPU Multi Media:
Integer x8 iSSE2: 18288 (once again comparable to P4-C 3.0 Ghz 512 kb L2)
Floating Point x4 iSSE2: 22866 (comparable to P4-E 550 3.4 Ghz 1 Mb L2)

Performance Rating (PR): 3300

I also found Sandra benchmarks for a pentium 4 mobile o'ced @ 2.4 Ghz:
http://www.vr-zone.com/guides/Intel/P4M/ (article came out Dec. 2002, so they were obviously using an older version of Sandra. Whether that makes a diff, I don't know.)

CPU arithmetic:
dhrystone: 4825
Whestone: FPU: 1316 MFLOPS iSSE2: 3054

CPU Multi Media:
Integer x8 iSSE2: 10008
Floating Point x4 iSSE2: 12166

At least in Sandra, my CPU beats the mobile P4. I didn't do a membench b/c my ram is still being shipped (just 256 mb pc2700 right now).

As far as these results actually equating to anything in the real world, who knows.

I would like to see someone use Sandra on a mobile celeron for comparison.

 

[rogue1979]

For what you are gonna use the rig for, might as well stick with the cpu you already have.

 

[TechnoButt]

lol, just to pipe in:

I picked up a P4-M 1.6 from bigbadbiologist in time to build a PC for Valentine's day for the girlfriend.

I put the processor in an ASUS P4P800-E Deluxe and opted to use the copper shim he provided me (looking strangely like a penny sized piece of copper) with some AS3 on both sides and a Zalman 7000AlCu.

It's run for over a month like a top. Defaulted to 1.6V or so and 12x (1200mhz). I have solid ram in the box, so I cranked it up to 200mhz immediately and it's been running flawlessly in her box as a 2.4C minus HT for a month (never gets powered off).

I'll run some tests and let you guys know the temps (idle and load)...

*edit*
Generic Cheap case with 80mm in and 80mm exhaust, with Fortron Source 300W PSU (single 120mm fan exhaust for psu), one WD400BB, one NEC ND-3500AG, ATI 8500DV AIW.

1.60V per MBM5
CPU/MOBO 41C/31C idle

Ran for ~10 mins SuperPi locking in CPU usage at 100%
CPU/MOBO 54C/35C under load (room temp, roughly 70F)

I haven't tried to push this an inch further, because I'm primarily interested in nobrainer stability for the GF, but I may play some tomorrow and see how far it will go (since I have a very overclockable mobo and Corsair twinx1024-4400 memory).

 

[Skipholiday]

Bookmarked

 

[jodang01]

Your temps look pretty good! Currently my celeron will idle at 28 C and mobo 30 C when fully warmed up, with the side off the case. Close the case up, and the temps jump to 33 C and 31 C respectively. An hour+ at full load with the side off, the CPU temp stays at 39 C, and with the side on I have seen it go from idle of 32 C to 40 C just from the time it takes to convert a cd to mp3 (maybe 5-10 minutes?). The mobo is practically indifferent though, rasing maybe 1 or 2 C when under load.

What I did find interesting though was the heat of the ram. The single pc2700 stick rose from 32 degrees to 54 degrees Celsius in just a matter of 5-10 minutes while using dvd shrink. The module does not have a heatspreader or any other cooling device. I got out a 6" fan and aimed it into the case, right at the memory. The ram and mobo dropped down to room temp. The CPU dropped from 39 to 33 C. Everything was still under full load.

The cpu and mobo temps were collected using mbprobe (usually reads 1 degree cooler than the bios will report); the ram was monitered via the Xaser's temp probe slid between the two ram slots on the mobo.

 

[Zap]

I wanted to play around with some more mobile Celerons, so picked some up off Eg@y. Would have gotten more from Karaktu, but wanted a higher multiplier (to use in older boards). Picked up two 2GHz and one 2.4GHz. The 2.4GHz is a risk with such a high multiplier, but I got a good deal for getting all three. I'll report back with my findings.

 

[Zap]

I tested all three chips for functionality, basically see if they would POST at 133MHz FSB on my MSI 645E Max board. All three were :thumbsup:

Today, got in an Asus P4S800-MX based on Karaktu's recommendation. This board really does seem to like the mobile Celerons. Supposedly has an AGP lock... how can we tell? I'm accustomed to 845PE and 865PE chipset boards where they'll tell you that the AGP is locked. Anyways... decided to test the 2.4GHz one some more since it has the highest multiplier. Installed it in the board with 2x512MB Kingmax and a Geforce 6800LE, 120GB Seagate and an LG 48X CDROM. Installed Windows and ran a bunch of 3DMark 2003 passes (was testing enabling more pipes and overclocking the video card). I know this isn't the best for testing CPU overclock, but it's what I had on hand and served my "other" purposes.

So far the only glitches have been video, when I enabled the bad set of pipes or overclocked the video too much. System has been stable even with glitching video, no lockups or reboots.

BTW, the motherboard choice looks like a winner. :thumbsup: to Karaktu.

 

[TStep]

After toying with P4Ms for years, I finally picked up a Mobile Celery 256k, SL6M5 C1. Following are some comments and a few comaprison results with a 2.4C. System as follows:

-Aopen AX4SG-UN mobo 865G dual channel
-Mushkin Level II 2x256
-AIW 9700 stock
-Intel hsf w/ raised copper core (works perfectly for lidless mobiles on several different boards -- Aopen AX4SG-UN, AsRock P4i65GV, and Asus P4S800).

First boot - 15x100 @ 0.875v. Was suprised it booted at such a low voltage, but it did. This particular mobo didn't take well to overclocking at the native 100 fsb speed. Seemed to max out at fsb=135 or so. Also the memory divider was locked at 3:4. so this was not ideal. Proceeded to pin mod the cpu to native fsb=133.

On second boot, the cpu would not go. After a little thought, it dawned on me that the cpu was trying to boot at 2.0 @ 0.875v. Proceeded to hack off the AE1 pin and it booted nicely to 2.0ghz @ 1.225v. Checked the memory dividers and I now had 1:1 and 4:5 options. Note to self: After trying the nail polish isolation method, I have found that I do not have the patience or fortitude to make that work.

A few quick comparisons to a Celery @2.4 and a stock P4 2.4C:

P4 2.4C (12x200, HT on) 1.51V, memory 1:1 (200) 2-2-2-6
---3DMark01 10x7x32....................14951
---3DMark01 6x4x32......................16753
---Super Pi, 1M..............................0m 55s
---Super Pi, 4M..............................4m 38s
---Prime95, 10 min, Toture Test 2....42C

Mobile Celery 1.5@2.4 (15x160) 1.29V, memory 4:5 (200) 2-2-2-6
---3DMark01 10x7x32....................14365
---3DMark01 6x4x32......................15594
---Super Pi, 1M..............................1m 12s
---Super Pi, 4M..............................5m 48s
---Prime95, 10 min, Toture Test 2....37C

Pretty decent performance for a low budget, quiet setup. If I decide to keep this for my office setup using the Intel Extreme Graphics, I'm going to experiment with a fanless SP94 I have lying around. I'll let you know when I blow the cpu

 

[ll0zz]

Very very intersesting !

I'm building a very silent and very low power comsuming machine from a Jetway miniQ450.
I'm going for a P4M to get lower multiplicator, I hope I can get better performances for the same consumption.
Do you think I'm going the right way ?
Should I go for a C1 or a B0 core for lowest vCore ? (If I can choose since here, in France, it's impossible to find)
BigBadBiologist spoke about 9x chips, which are they ?

I know this thread is theoritically supposed to be about celerons, but it seemed to me it isn't

 

[TStep]

On the miniQ450: Is it a SIS 651 chipset? Does the motherboard allow you to manipulated the fsb?

I am not positive, but isn't the SIS 651 a derivative of 648 chipset (not the 648FX) with onboard graphics? If so, I don't believe the 648 based solutions were capable of attaining a very high fsb, maybe 145-150 IIRC. If that is true, I don't believe the P4M would be the best solution.

A P4M will default to a 12x multiplier, so assuming your motherboard has fsb adjustments, you may be limited to 12x(145-150) for about 1700-1800 range performance. Most of the BO stepping P4Ms I've come across probably average around a 2.4Ghz limit, C1/D1 being higher. Ie: I don't think you will max out your cpu if that is your intention. D0 is a prescott stepping is it not? I am not positive that a 651 will accept that, but someone else may know.

A celeron mobile will retain its native multiplier when used in a desktop board. So assuming you have a C1 stepping capable of 2.4ghz nearly guaranteed and the possibility of 3.0ghz, knowing you probably have an fsb limit of 145ish, I would look for:

2400 / 133 = mulitplier of 18x
3000 / 150 = mulitplier of 20x

I would look for a C1 stepping Celeron mobile of around 1800-2000.

 

[ll0zz]

Originally posted by: TStep
On the miniQ450: Is it a SIS 651 chipset? Does the motherboard allow you to manipulated the fsb?

Yes and yes, from 100 to 200 by 1Mhz

I am not positive, but isn't the SIS 651 a derivative of 648 chipset (not the 648FX) with onboard graphics? If so, I don't believe the 648 based solutions were capable of attaining a very high fsb, maybe 145-150 IIRC. If that is true, I don't believe the P4M would be the best solution.

And if I disable the onboard graphic chip, could I go further ?

In fact I hope I can reduce voltage as much as possible, ans I know I can't hope keeping high frequences.
I thought about 100-150fsb, according to the best perf/wattage I can achieve with the chip I will find...
And a 9x would be even better, to have more perfs for the same wattage...
I omitted to say don't want an overkill machine, just have some servers running, internet station, and HiFi...

A P4M will default to a 12x multiplier, so assuming your motherboard has fsb adjustments, you may be limited to 12x(145-150) for about 1700-1800 range performance. Most of the BO stepping P4Ms I've come across probably average around a 2.4Ghz limit, C1/D1 being higher. Ie: I don't think you will max out your cpu if that is your intention. D0 is a prescott stepping is it not? I am not positive that a 651 will accept that, but someone else may know.

Sorry it's a typo, I meant B0. Corrected.

Thank you for your advice.

 

[TStep]

If cool @ P4 1.6 performance is your prime objective, check this link out.