µATX P-M Motherboard : DFI 855GME-MGF

The forum for non-component-related silent pc discussions.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

zouav
Posts: 17
Joined: Wed Nov 03, 2004 12:38 am

µATX P-M Motherboard : DFI 855GME-MGF

Post by zouav » Wed Nov 03, 2004 1:10 am

Hi all, first post on SPCR from France :)
Just read a review on this motherboard on X86-secret. Could be an other solution to get silence by using excellent ratio heat dissipation/performance Pentium M, just as with Aaopen motherboard mentionned in an other thread.
This motherboard is the sister of G5M300-N dedicated to the industrial market and confirms the great potential of using Pentium M in desktop system in order to achieve silence computing. The main difference with the Aaopen i855GMEm-LFS is that the DFI doesn't support standart P4 heatsink ( too risky with destructing the core is the reason mentionned) BUT do support Northbridge's heatsink.

Full Specifications :
i855GME(6300ESB)
microATX
VGA
6chSound(AC'97)
1000Base-TLAN(RTL8110S)
IEEE1394(VT6307)
AGPx1 (4x)
PCIx2
PCI-X
DDRDIMMx2

Lifecycle
Posts: 118
Joined: Thu Nov 13, 2003 1:29 am
Location: London, UK

Post by Lifecycle » Wed Nov 03, 2004 1:35 am

Welcome to SPCR!!

Here are some English translations of those pages, thanks to Google:

Review in English

Overclocking/underclocking in English

Ralf Hutter
SPCR Reviewer
Posts: 8636
Joined: Sat Nov 23, 2002 6:33 am
Location: Sunny SoCal

Post by Ralf Hutter » Wed Nov 03, 2004 5:08 am

zouav - WELCOME TO SPCR!!!

Great first post, keep up the good work!

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Nov 04, 2004 4:58 am

From Anandtech:
AOpen and DFI both have motherboards ready, and are both targeting the Japanese market first. DFI built their board for a particular customer and is planning an enthusiast level board based on the desktop 915 chipset with some overclocking features in the near future. We know that Shuttle has been working on a SFF based on the Pentium M for quite some time now but have yet to see anything from them.
Edit: Do you think the PM/CM CPU is more fragile than a AXP for instance? If I can use a heavy Zalman 7000 Cu/(Chew) on a AXP, why shouldn't it be possible with this one? I'd rather see a P4 retention bracket, it gives a lot more possibilities even though it got the wrong height. This gives the AOpen board more clearence around the CPU, that's always good.
Last edited by Mats on Thu Nov 04, 2004 5:19 am, edited 1 time in total.

ChrisH
Friend of SPCR
Posts: 65
Joined: Wed Oct 30, 2002 10:44 am
Location: Charlotte, NC USA - Go Panthers!

Post by ChrisH » Thu Nov 04, 2004 5:08 am

Here is a link to the official press release. It says the board will begin shipping in North America starting November 8, 2004.

Pakkapakka
Posts: 36
Joined: Mon Jan 05, 2004 5:26 am

Post by Pakkapakka » Thu Nov 04, 2004 5:29 am

A great review to read, thanks.

I'd really like to see a comparative review of this board and the AOpen i885GME.

Does the AOpen BIOS include similar multiplier/voltage options?
Does the AOpen Promise SATA chip support NCQ?
Do any of the two boards have other unique features?
DFI built their board for a particular customer and is planning an enthusiast level board based on the desktop 915 chipset with some overclocking features in the near future.
I assume that this 915-based board is a completely different product. I'm quite curious. Wouldn't it make more sense to wait for Alviso-GM in 05Q1?

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Nov 04, 2004 5:40 am

I still don't know the difference between socket 478 (I'M NOT TALKING ABOUT P4) and socket 479 for PM/CM.
Do they both fit the AOpen and the DFI board, both made for socket 479? If not, how can those french guys use a PM 755 which is listed as socket 478 only at Intel? The newer PM 765 are available with both sockets for instance.

Can someone explain this?

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Thu Nov 04, 2004 5:51 am

Grrreat news!

Now all we needs is a new chipset with low-voltage options, PCI-E support, SATA II, DDR400 (or higher), etc :)

Nah, just kidding, although I'm tempted to think what a better chipset/memory interface would do to the performance of that chip.

I hope this comes available soon and the prices for Dothan's in retail come down more.

We've waited for over a year for this and it's slowly becoming reality.

Pakkapakka
Posts: 36
Joined: Mon Jan 05, 2004 5:26 am

Post by Pakkapakka » Thu Nov 04, 2004 6:06 am

halcyon wrote: Now all we needs is a new chipset with low-voltage options, PCI-E support, SATA II, DDR400 (or higher), etc :)

Nah, just kidding, although I'm tempted to think what a better chipset/memory interface would do to the performance of that chip.
Some of that will be provided by the Alviso chipset platform early next year (unless the embedded versions lag the mobile platform):
-dual channel DDR2
-PCI-E
-more power saving technology
-higher FSB (533), although this doesn't help power consumption. (Seems Dothans running on 533 consume 5W or more extra power)

The current 915/ICH6 chipsets support the SATA2 NCQ extension, but not the 300 transfer speed. It is possible that the same will be true for the Alviso chipset, since these will use a modified ICH6 southbridge.
halcyon wrote: I hope this comes available soon and the prices for Dothan's in retail come down more.

We've waited for over a year for this and it's slowly becoming reality.
I share your sentiment :)

I could also wish for some additional items:
-PGA packaged LV Dothans. IIRC, Banias LV was available in PGA package.
-Heatsink/heatpipe designs and cases tailored for fanless operation.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Nov 04, 2004 6:57 am

halcyon wrote: I hope this comes available soon and the prices for Dothan's in retail come down more.
I've been watching the prices for the last year now at itbutikken.dk (our favoríte!!! :lol: :lol: ) and the most expensive PM have gone from 6600 SEK to 4000 SEK in just maybe one month or two, and that's long before next model have showed up. The cheapest one, CM 1.5 GHz, goes for 840 SEK now.

Wraith
Posts: 139
Joined: Sat Dec 27, 2003 1:57 pm
Location: Minnesota

Post by Wraith » Thu Nov 04, 2004 8:01 am

Link to the INQ article about the DFI board.

zouav
Posts: 17
Joined: Wed Nov 03, 2004 12:38 am

Post by zouav » Thu Nov 04, 2004 8:14 am

Mats : sure the Aaopen has got the bracket advantage but certainly not the DFI's great overclocking potential. The overclocking in silence is here a reality.
Who heard about a mixture of both coming from a third company ? :D Or more realistic, ask DFI two different versions (X86-secret did participate to the building process and could eventually ask about that feature ?)
Just dreaming... :roll:

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Thu Nov 04, 2004 11:21 am

@Mats: "Socket 479" is a moniker given to the PGA P-Ms by the enthusiast community, because they don't fit in the official Socket 478, and the BGA (ball grid array - meant to be soldered, not socketed) chips have 479 balls. However, Intel still calls them 478 because the PGA chips have one less pin than the BGA chips have balls. The only reason we call them 479 is because 478 was already taken, and saying "OMG, look, it's got Socket 478" makes everyone look like a bunch of douchebags, because they'd apparently be impressed by a standard P4 board ;-) Maybe Socket M would make more sense...

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Thu Nov 04, 2004 2:53 pm

Reading the machine translation of the X-86 review of the DFI board, which I’ve copied below, I’m not clear on whether the BIOS supports Speedstep fully. i.e. will it dynamically alter frequency and voltage.
It’s great to see desktop boards arriving, but it’ll be a shame if they don’t support Speedstep. Ideally, it would be good to be able to alter the parameters for each individual power state within the BIOS. Although I may be going over the top here :)

<<FID/VID Control: This makes it possible to control the SpeedStep registers directly in the BIOS. One thus can selectionner a multiplying coefficient ranging between 6 and the maximum value (15 for one 1.5 GHz, 16 for one 1.6 GHz... etc). Even thing for the VID which makes it possible to choose the supply voltage between 0.7V and the possible maximum. This menu is not available when Celeron M is used, this processor not supporting the EIST.>>

The corresponding BIOS screenshot only shows a single set of values to change, which doesn’t auger well.

Crow

zouav
Posts: 17
Joined: Wed Nov 03, 2004 12:38 am

Post by zouav » Thu Nov 04, 2004 3:33 pm

If I retranslate your quote it gives this :
"FID/VID Control: This makes it possible to control the SpeedStep registers directly in the BIOS. One thus can select a multiplying coefficient ranging between 6 and the maximum value (15 for one 1.5 GHz, 16 for one 1.6 GHz... etc). Same thing for the VID which makes it possible to choose the supply voltage between 0.7V and the possible maximum. This menu is not available when Celeron M is used because this processor doesn't support the EIST."
Anyway, I'm note sure it answers your question about "dynamically atler frequency and voltage". Nothing about that is mentionned in the BIOS&Software page.

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Thu Nov 04, 2004 3:37 pm

OK, I understand it as:

You can manipulate the CPU multiplier and voltage from the BIOS, but you can't really do it on a Celeron M.

Now, if Intel's SpeedStep drivers work with the BIOS, you're good. Also, even if not, DFI (or x86-secret - I've heard that they're partially responsible for the development of the board, especially the BIOS) could add it to the BIOS.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Nov 04, 2004 4:37 pm

zouav wrote:Mats : sure the Aaopen has got the bracket advantage but certainly not the DFI's great overclocking potential.
Why do you say that? I don't see ANY advantage with the DFI solution! (YET)
___________________
bhtooefr wrote:@Mats: "Socket 479" is a moniker given to the PGA P-Ms by the enthusiast community, because they don't fit in the official Socket 478, and the BGA (ball grid array - meant to be soldered, not socketed) chips have 479 balls. However, Intel still calls them 478 because the PGA chips have one less pin than the BGA chips have balls. The only reason we call them 479 is because 478 was already taken, and saying "OMG, look, it's got Socket 478" makes everyone look like a bunch of douchebags, because they'd apparently be impressed by a standard P4 board ;-) Maybe Socket M would make more sense...
I just don't believe you, because you couldn't answer my previous question how they fitted a 478 pin Pentium M 755 in the DFI board (that is, IF we can trust Intels technical specifications...).
____________________
smilingcrow wrote:Reading the machine translation of the X-86 review of the DFI board, which I’ve copied below, I’m not clear on whether the BIOS supports Speedstep fully. i.e. will it dynamically alter frequency and voltage.
Well at least the AOpen board got it, probably thanks to the 855 chipset. In the lower left corner in this picture you see a program that controls overclocking, speedstep, fans and more. It's a utility that's included with the mobo, you can of course download it as well. I don't know how good it is, but it looks good! :D I think that the DFI can be controlled in the same way, otherwise there should be no big reason for using the 865 chipset.

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Thu Nov 04, 2004 5:59 pm

ftp://download.intel.com/design/mobile/ ... 218904.pdf page 47 is the P-M pinout (this is the Dothan datasheet, but the Banias has the same pinout - also, pin B2 doesn't exist on the PGA version (the interesting one), making it 478 pins).

ftp://download.intel.com/design/Pentium ... 056102.pdf page 37 is the P4 pinout (Prescott datasheet, but S478 Williamette and Northwood have same pinout).

They are clearly different. Pins A1 and B1 are depopulated on the P4, pins A1 and B2 are depopulated on the P-M. They are NOT compatible. Here's what the corner of a P-M looks like (O is a pin):

Code: Select all

  O
O
The corner of a P4:

Code: Select all

  O
  O
Why the P-M socket is called Socket 479 - the corner of a socket (O is a hole, this time):

Code: Select all

  O
O O
Basically, there's nothing physically stopping a 478-pin P-M from plugging into Socket 479. There IS something physically stopping such a P-M from plugging into Socket 478. Unfortunately, there's nothing physically stopping a P4 from going into Socket 479, so fried processors are a possibility.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Nov 04, 2004 6:51 pm

Yeah, they're different, I know. And now I checked the CPUs they used once more and found something, look! The first time I saw the core speed (2555.8 MHz) and the specification (2.0 GHz, just like AXP 2500+ becomes 3200+ with raised FSB) BUT I missed the name which is 735 (=1.7 GHz). It must be a typo or maybe they did the same mistake since the text under the pic says 2.0 GHz. That's why I thought it was a 755.

I'm sorry for being such a bitch. The only reason I want to know for sure is because I don't want to end up with a useless CPU! So the question is when can you use a PM 2.0 GHz? It's available in some stores, but does it fit this board since it's 478 pin? I don't think Intel are being lazy with updating the specs, although it's not impossible... ..because they don't list the 735 with socket 479! I don't get it! AAARRRGGHHH!!!!

Code: Select all

SL7UZ  2.10 GHz 400 MHz 90 nm  B1 2 MB 479 pin H-PBGA 
SL7V3  2.10 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL7EM  2.00 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL7EN  1.80 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL6N9  1.70 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL6N5  1.70 GHz 400 MHz 130 nm B1 1 MB 478 pin PPGA FC-PGA2 
SL7EP  1.70 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL6FA  1.60 GHz 400 MHz 130 nm B1 1 MB 478 pin PPGA FC-PGA2 
SL6F7  1.60 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL7EG  1.60 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL6F6  1.50 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL7GL  1.50 GHz 400 MHz 90 nm  B1 2 MB 478 pin PPGA 
SL6F9  1.50 GHz 400 MHz 130 nm B1 1 MB 478 pin PPGA FC-PGA2 
SL6F8  1.40 GHz 400 MHz 130 nm B1 1 MB 478 pin PPGA FC-PGA2 
SL7F3  1.40 GHz 400 MHz 90 nm  B1 2 MB 479 pin H-PBGA 
SL6F5  1.40 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL6N4  1.30 GHz 400 MHz 130 nm B1 1 MB 478 pin PPGA FC-PGA2 
SL6N8  1.30 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL6NB  1.20 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL6NC  1.10 GHz 400 MHz 130 nm B1 1 MB 479 pin H-PBGA FC-BGA2 
SL7F4  1.10 GHz 400 MHz 90 nm   B1 2 MB 479 pin H-PBGA 
SL6NJ  900 MHz  400 MHz 130 nm  B1 1 MB 479 pin H-PBGA FC-BGA2 
Ok, this is what I get (CONFUSION) for even thinking about buying Intel instead of AMD....
_______________________

BTW, why didn't they overclock the P4 and A64 to the max? I'd like to see an overclocked Winchester! Very unfair!

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Thu Nov 04, 2004 7:07 pm

Basically, treat anything with "H-PBGA" as unusable, as you don't have the equipment to do it (most likely).

If it says 478 pin PPGA or 478 pin PPGA FC-PGA2, and it is a Pentium M, it plugs into Socket 479.

Proposal: How 'bout we all call Socket 479 by Socket M, to prevent confusion? All PPGA Pentium Ms would plug into Socket M.

DanceMan
Posts: 287
Joined: Sun Aug 11, 2002 3:26 pm
Location: Burnaby, BC, Canada

Post by DanceMan » Thu Nov 04, 2004 11:56 pm

bhtooefr wrote:Basically, treat anything with "H-PBGA" as unusable, as you don't have the equipment to do it (most likely).
Are you saying that "H-PBGA" is a solder only, intended for notebooks?

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Fri Nov 05, 2004 2:31 am

Not just notebooks, but yes, solder only.

It's most common use is thin-and-light notebooks, where there's no room for a socket. Shelton is also a 479-pin H-PBGA chip, BTW.

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Fri Nov 05, 2004 5:46 am

There’s a link at the URL below to a version of Clockgen for the DFI board:

http://forum.x86-secret.com/viewtopic.p ... c93173d12a

If Clockgen works with the board it bodes well for Speedstep working, provided you load the driver.
Speedswitch might be a good solution here, as it is more configurable than XP’s built in utility.

http://www.diefer.de/speedswitchxp/

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Fri Nov 05, 2004 7:04 am

Thanks a lot bhtooefr for your help and WELCOME TO SPCR!!!

Now my only question left is not that important:

Why do Intel make both 478 and 479 pin versions of PM/CM? Is there some kind of difference in functionality?

bhtooefr
Posts: 8
Joined: Thu Nov 04, 2004 11:16 am

Post by bhtooefr » Fri Nov 05, 2004 10:53 am

I don't know why the 479-pin versions have the other pin. It might be for ULV support or something.

zouav
Posts: 17
Joined: Wed Nov 03, 2004 12:38 am

Post by zouav » Sat Nov 06, 2004 9:26 am

final version should look like this :
Image
Shipping to retailers should begin on November 8th
official page on DFI is here : http://www.dfi.com.tw/Product/xx_produc ... MB&SITE=US

Schroinx
Posts: 218
Joined: Wed Apr 09, 2003 3:21 am
Location: Denmark

Post by Schroinx » Sun Nov 07, 2004 2:20 am

There are 479 pins, to prevent you from insearting a PM in a 478 board. Since P4 and PM is electrical compatible, PM does not need an extra pin. But since the pin out is somewhat different, Intel don't want n00bs to return fried PM's because they tried it in a 478.

Schlotkins
Posts: 278
Joined: Thu Mar 27, 2003 5:30 am

Post by Schlotkins » Sun Nov 07, 2004 4:47 am

What heatsink can you use on that think? A socket 370 one?

Ralf Hutter
SPCR Reviewer
Posts: 8636
Joined: Sat Nov 23, 2002 6:33 am
Location: Sunny SoCal

Post by Ralf Hutter » Sun Nov 07, 2004 7:02 am

Schlotkins wrote:What heatsink can you use on that think? A socket 370 one?
The board comes with some cheesy looking proprietary fan/heatsink/mounting bracket. There's a stamped "X" bracket that goes beneath the board and a cheap looking aluminum heatsink with what looks like a 60x15mm or 70x15mm cooling fan that attaches to the top of the heatsink. :shock:

Hopefully that hole pattern is the same size as the LGA775 heatsink pattern so some HSF swapping can be done. If not, well, it's on to plan "C".

One more potential issue is that, according to the pdf manual, there is no adjustment for Vcore or FSB/multiplier, only some thermal ramping sort of thing. :(

Schroinx
Posts: 218
Joined: Wed Apr 09, 2003 3:21 am
Location: Denmark

Post by Schroinx » Sun Nov 07, 2004 1:19 pm

The DFI board does not support standart HSF. See for youself: http://www.x86-secret.com/articles/cm/d ... i855-4.htm
In this case there is used a 40mm fan and a NB HS.

The Aopen PM board is compatible with a P4 HFS. http://www.aopen.nl/products/mb/i855GMEm-LFS.htm

/Schroinx

Post Reply