Pentium M and WC

DMOS · 02-26-2005, 05:39 PM

I figured you guys might be interested in a little project of mine that I've now completed. I sucessfully managed to water cool my desktop Dothan setup, a little something different than the typical A64/P4 stuff you guys get to play with most of the time.

The DFI 855GME has a socket layout with dimensions similar to an AMD north bridge cooler, so as you can probably guess a Swiffy MCW20 fits in just fine. One one of the posts the nut that hold in a bolt is tight, but the mininut they supply just barely fits. Other than that, there's sufficient clearance. Going to anything much bigger would hit caps, and you can only have it oriented one way. The AOpen board would be a lot simpler, since it uses a typical P4 retention frame.

With the overkill provided by an MCP650 and a dual 120mm rad (leftovers from my previous WC setup) it has no issues bringing the CPU to 2.6GHz from it's initial 1.7, after volt mod for vcore. I'm going to fudge with it some more, and see what else I can get out of it.

pHaestus · 02-26-2005, 06:56 PM

That's really slick. Seeing the wb only supported by 2 holes out of 4 makes me nervous (esp with the hose bends) but if it works then it works. My thinkpad's pretty speedy at 1.7GHz; a 2.6GHz Dothan chip must be damn quick. How much cache do the desktop chips have?

redleader · 02-26-2005, 07:09 PM

There are no desktop ones, just moble chips people put in special boards, so it'll have 2MB cache.

snowwie · 02-26-2005, 07:13 PM

yeah, mounting seems a bit iffy, kinda like A64 boards that have the socket sideways (not per ref design), especially in a tower

DMOS · 02-26-2005, 09:54 PM

Yeah, I'm not terribly impressed with the mounting. I fear going to work monday and coming home to it sitting on the bottom of my case. The P-M has some fancy life saving circuitry though, so even if it falls right off it shouldn't die (I hope). And it's not moving anywhere, that case is a pain.

There are still some things its not good at. Like anything that requires huge memory bandwidth. That'll be fixed when I get an Alviso based board.

snowwie · 02-26-2005, 10:47 PM

still very cool to have a 2.6GHz dothan, very cool

...is there a dual channel chipset for s479? or is this something coming out soon?

DMOS · 02-26-2005, 11:12 PM

Yes, that's what Alviso is. DDR2 and dual channel. From the looks of it you can also use DDR1, but only in single channel (kinda defeating the purpose). Also brings PCIe (so I don't feel bad about buying a new gfx card, which I only do every three-four years).

2.6 is just a startng point. Once I get a new floppy drive, I'll update the bios to the beta one, and I should have a bit more headroom to work with. Apparently that clears up some of the current issues the board has.

redleader · 02-26-2005, 11:24 PM

Alviso only has 533 FSB, which matches perfectly with single channel DDR2 533. If it supports dual channel, its probably just for an integrated graphics card I'd think.

MaxxxRacer · 02-27-2005, 09:11 AM

looking good man.. what kind of rad to you ahve on it?? I suppose you can run it passive, can you not?

DMOS · 02-27-2005, 11:46 AM

First comment, on the bus situation. With the 533 bus, yes, it's equivalent to the "theoretical" max bandwidth of a single channel DDRII 533 module. Theoretical being the key word. Dual channel should allow it to max out that bus in real terms. Perfect solution? No dual channel and a 1066MHz bus would be, but you won't find that in a mobile platform (too expensive in terms of power) any time soon. In my testing this system is horribly bandwidth limited with single channel, bumping up the bus speed and going to 4:3 or 5:3 mem:fsb makes a huge difference. DC memory would be the next logical step.

The rad is a 2x120mm tranny oil cooler. I'm sure it could run passively, hell, I used the stock rinky dink heatsink passively at 1.7GHz. At the speeds and voltages I'm at now, and moving forwards to I don't know if I still could under heavy load.

MaxxxRacer · 02-27-2005, 04:30 PM

interesting...

leejsmith · 02-28-2005, 04:00 AM

What fsb are you running at the moment and can you change the multiplier with the board ?

Quote:

Originally Posted by DMOS

First comment, on the bus situation. With the 533 bus, yes, it's equivalent to the "theoretical" max bandwidth of a single channel DDRII 533 module. Theoretical being the key word. Dual channel should allow it to max out that bus in real terms. Perfect solution? No dual channel and a 1066MHz bus would be, but you won't find that in a mobile platform (too expensive in terms of power) any time soon. In my testing this system is horribly bandwidth limited with single channel, bumping up the bus speed and going to 4:3 or 5:3 mem:fsb makes a huge difference. DC memory would be the next logical step.

The rad is a 2x120mm tranny oil cooler. I'm sure it could run passively, hell, I used the stock rinky dink heatsink passively at 1.7GHz. At the speeds and voltages I'm at now, and moving forwards to I don't know if I still could under heavy load.

DMOS · 02-28-2005, 10:09 AM

I'm running at 150x17, and you can change the multi down, similar to A64s. When I get a new floppy drive today or tomorrow, I'll load up a new BIOS that supposedly smooths things out (above 150ish I get AGP and other errors, I think the PCI lock is fudged up). If that doesn't work, I'll break out the soldering iron and do the vAGP/chipset mod.

JamesAvery22 · 02-28-2005, 12:47 PM

Nice hardware DMOS, Ive been wanting to replace my P4C setup with a P-M.

MaxxxRacer beat me to it, that thing could easily run passive. SilentPCreview had that thing cooled passively by that large circular zalman heatsink (cnps7000?) without the fan that is integrated into it.
Have you tried taking the radiator and just sitting it outside of that giant cube case of yours? =) Maybe make a chimney. Try a ghetto temp one made out of cardboard. That chip really doesnt need much cooling.

Ruiner · 03-02-2005, 11:31 AM

Hmmm...I just heard that Dothan Celerons exist (1MB of L2) and only cost 90 bucks. That makes this platform fairly palatable to budget OCers...at least those willing to Vmod their cores.
Damn pu$$y bioses. What was DFI thinking?

JamesAvery22 · 03-02-2005, 12:39 PM

Quote:

Originally Posted by Ruiner

Hmmm...I just heard that Dothan Celerons exist (1MB of L2) and only cost 90 bucks. That makes this platform fairly palatable to budget OCers...at least those willing to Vmod their cores.
Damn pu$$y bioses. What was DFI thinking?

lol that no one was going to buy the board as the market for such a product is beyond tiny. So they put as little money/effort into it as possible maybe?

Ruiner · 03-02-2005, 01:02 PM

I'll buy that. Still, DFI makes enthusiast boards and it is just a bios.
Full pentium M's compete fairly well with FX55 chips when OCed. I wonder how the cache-strated celly M will do.

DMOS · 03-02-2005, 01:18 PM

I kinda wish they made this BIOS a bit more Ocer friendly. However, after talking to DFI, that's not the market they were really going for. It's for servers and HTPCs, neither of which have anything to do with overclocking. The fact that you can make it a solid gaming system is an added bonus. The next rev with Alviso will likely include more desktop oriented things like mounts for P4 heatsinks, and hopefully vdimm and vcore adjustments.

That being said, this fixes quite a few problems:

I'm about to do a few more, to get 1.7V. Right now 1.6V is limiting me to 2.6GHz. I have no idea what the long term effects of going to this much voltage are going to be. At least you can dial it down in the BIOS for when not benching.

Ruiner · 03-02-2005, 01:50 PM

That looks like pretty thick wire for what I'm used to seeing, at least back when I pinmodded s370's. Do you have anything thinner?
What is default Vcore? How are your temps at 2.6/1.6v?

JamesAvery22 · 03-02-2005, 02:59 PM

Quote:

Originally Posted by Ruiner

That looks like pretty thick wire for what I'm used to seeing, at least back when I pinmodded s370's. Do you have anything thinner?
What is default Vcore? How are your temps at 2.6/1.6v?

Does look really thick.

DMOS, why arent you dropping a U wire into the socket rather than wrapping it around the pins? Lot safer and easier that way.

DMOS · 03-02-2005, 03:02 PM

The reason it looks so thick is because of the rediculous magnification required to even see it and the pins. The differnce in density and size compared to s370 is quite a bit. This is the same wire I've used for s478, and it tends to work well. It's strands removed from 16gauge speaker wire.

I modded it to 1.66V now. I guess one of the loops I just put on didn't take, because I should have 1.69. Not that it's making any difference, the extra .5V didn't do shit. At idle it's around 24C. Load from P95 is around 36C.

Highest prime stable with memory at 4:3 is 158x16, the tradeoff in clock speed is worth the increase in RAM.

DMOS · 03-02-2005, 03:03 PM

Quote:

Originally Posted by JamesAvery22

Does look really thick.

DMOS, why arent you dropping a U wire into the socket rather than wrapping it around the pins? Lot safer and easier that way.

Tried that. Socket is too small. Even with a hair sized wire, it just didn't want to play nice. I should have taken a picture of the cpu beside an A64. It's really tiny, which makes things difficult to say the least.

02-26-2005, 05:39 PM	#1
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	Pentium M and WC I figured you guys might be interested in a little project of mine that I've now completed. I sucessfully managed to water cool my desktop Dothan setup, a little something different than the typical A64/P4 stuff you guys get to play with most of the time. The DFI 855GME has a socket layout with dimensions similar to an AMD north bridge cooler, so as you can probably guess a Swiffy MCW20 fits in just fine. One one of the posts the nut that hold in a bolt is tight, but the mininut they supply just barely fits. Other than that, there's sufficient clearance. Going to anything much bigger would hit caps, and you can only have it oriented one way. The AOpen board would be a lot simpler, since it uses a typical P4 retention frame. With the overkill provided by an MCP650 and a dual 120mm rad (leftovers from my previous WC setup) it has no issues bringing the CPU to 2.6GHz from it's initial 1.7, after volt mod for vcore. I'm going to fudge with it some more, and see what else I can get out of it. __________________ http://www.devhardware.com

02-26-2005, 09:54 PM	#5
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	Yeah, I'm not terribly impressed with the mounting. I fear going to work monday and coming home to it sitting on the bottom of my case. The P-M has some fancy life saving circuitry though, so even if it falls right off it shouldn't die (I hope). And it's not moving anywhere, that case is a pain. There are still some things its not good at. Like anything that requires huge memory bandwidth. That'll be fixed when I get an Alviso based board. __________________ http://www.devhardware.com

02-26-2005, 11:12 PM	#7
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	Yes, that's what Alviso is. DDR2 and dual channel. From the looks of it you can also use DDR1, but only in single channel (kinda defeating the purpose). Also brings PCIe (so I don't feel bad about buying a new gfx card, which I only do every three-four years). 2.6 is just a startng point. Once I get a new floppy drive, I'll update the bios to the beta one, and I should have a bit more headroom to work with. Apparently that clears up some of the current issues the board has. __________________ http://www.devhardware.com

02-27-2005, 09:11 AM	#9
MaxxxRacer Cooling Savant Join Date: Sep 2004 Location: Los Angeles, Ca Posts: 310	looking good man.. what kind of rad to you ahve on it?? I suppose you can run it passive, can you not? __________________ Excuse me, I believe you have my stapler.

02-27-2005, 11:46 AM	#10
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	First comment, on the bus situation. With the 533 bus, yes, it's equivalent to the "theoretical" max bandwidth of a single channel DDRII 533 module. Theoretical being the key word. Dual channel should allow it to max out that bus in real terms. Perfect solution? No dual channel and a 1066MHz bus would be, but you won't find that in a mobile platform (too expensive in terms of power) any time soon. In my testing this system is horribly bandwidth limited with single channel, bumping up the bus speed and going to 4:3 or 5:3 mem:fsb makes a huge difference. DC memory would be the next logical step. The rad is a 2x120mm tranny oil cooler. I'm sure it could run passively, hell, I used the stock rinky dink heatsink passively at 1.7GHz. At the speeds and voltages I'm at now, and moving forwards to I don't know if I still could under heavy load. __________________ http://www.devhardware.com

02-26-2005, 06:56 PM	#2
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	That's really slick. Seeing the wb only supported by 2 holes out of 4 makes me nervous (esp with the hose bends) but if it works then it works. My thinkpad's pretty speedy at 1.7GHz; a 2.6GHz Dothan chip must be damn quick. How much cache do the desktop chips have?

02-26-2005, 07:09 PM	#3
redleader Thermophile Join Date: Jun 2001 Location: The deserts of Tucson, Az Posts: 1,264	There are no desktop ones, just moble chips people put in special boards, so it'll have 2MB cache.

02-26-2005, 07:13 PM	#4
snowwie Cooling Savant Join Date: Mar 2003 Location: New Jersey Posts: 154	yeah, mounting seems a bit iffy, kinda like A64 boards that have the socket sideways (not per ref design), especially in a tower

02-26-2005, 10:47 PM	#6
snowwie Cooling Savant Join Date: Mar 2003 Location: New Jersey Posts: 154	still very cool to have a 2.6GHz dothan, very cool ...is there a dual channel chipset for s479? or is this something coming out soon?

02-26-2005, 11:24 PM	#8
redleader Thermophile Join Date: Jun 2001 Location: The deserts of Tucson, Az Posts: 1,264	Alviso only has 533 FSB, which matches perfectly with single channel DDR2 533. If it supports dual channel, its probably just for an integrated graphics card I'd think.

02-27-2005, 04:30 PM	#11
MaxxxRacer Cooling Savant Join Date: Sep 2004 Location: Los Angeles, Ca Posts: 310	interesting... __________________ Excuse me, I believe you have my stapler.

02-28-2005, 10:09 AM	#13
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	I'm running at 150x17, and you can change the multi down, similar to A64s. When I get a new floppy drive today or tomorrow, I'll load up a new BIOS that supposedly smooths things out (above 150ish I get AGP and other errors, I think the PCI lock is fudged up). If that doesn't work, I'll break out the soldering iron and do the vAGP/chipset mod. __________________ http://www.devhardware.com

02-28-2005, 12:47 PM	#14
JamesAvery22 Cooling Savant Join Date: May 2004 Location: Northern VA Posts: 383	Nice hardware DMOS, Ive been wanting to replace my P4C setup with a P-M. MaxxxRacer beat me to it, that thing could easily run passive. SilentPCreview had that thing cooled passively by that large circular zalman heatsink (cnps7000?) without the fan that is integrated into it. Have you tried taking the radiator and just sitting it outside of that giant cube case of yours? =) Maybe make a chimney. Try a ghetto temp one made out of cardboard. That chip really doesnt need much cooling.

03-02-2005, 11:31 AM	#15
Ruiner Cooling Savant Join Date: May 2003 Location: New Smyrna, FL Posts: 258	Hmmm...I just heard that Dothan Celerons exist (1MB of L2) and only cost 90 bucks. That makes this platform fairly palatable to budget OCers...at least those willing to Vmod their cores. Damn pu$$y bioses. What was DFI thinking?

03-02-2005, 01:02 PM	#17
Ruiner Cooling Savant Join Date: May 2003 Location: New Smyrna, FL Posts: 258	I'll buy that. Still, DFI makes enthusiast boards and it is just a bios. Full pentium M's compete fairly well with FX55 chips when OCed. I wonder how the cache-strated celly M will do.

03-02-2005, 01:18 PM	#18
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	I kinda wish they made this BIOS a bit more Ocer friendly. However, after talking to DFI, that's not the market they were really going for. It's for servers and HTPCs, neither of which have anything to do with overclocking. The fact that you can make it a solid gaming system is an added bonus. The next rev with Alviso will likely include more desktop oriented things like mounts for P4 heatsinks, and hopefully vdimm and vcore adjustments. That being said, this fixes quite a few problems: I'm about to do a few more, to get 1.7V. Right now 1.6V is limiting me to 2.6GHz. I have no idea what the long term effects of going to this much voltage are going to be. At least you can dial it down in the BIOS for when not benching. __________________ http://www.devhardware.com

03-02-2005, 01:50 PM	#19
Ruiner Cooling Savant Join Date: May 2003 Location: New Smyrna, FL Posts: 258	That looks like pretty thick wire for what I'm used to seeing, at least back when I pinmodded s370's. Do you have anything thinner? What is default Vcore? How are your temps at 2.6/1.6v?

03-02-2005, 03:02 PM	#21
DMOS Cooling Neophyte Join Date: Apr 2003 Location: Kingston, ON, Canada Posts: 31	The reason it looks so thick is because of the rediculous magnification required to even see it and the pins. The differnce in density and size compared to s370 is quite a bit. This is the same wire I've used for s478, and it tends to work well. It's strands removed from 16gauge speaker wire. I modded it to 1.66V now. I guess one of the loops I just put on didn't take, because I should have 1.69. Not that it's making any difference, the extra .5V didn't do shit. At idle it's around 24C. Load from P95 is around 36C. Highest prime stable with memory at 4:3 is 158x16, the tradeoff in clock speed is worth the increase in RAM. __________________ http://www.devhardware.com

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)