DFI 855GME-MGF Motherboard Review - Layout and overclocking
(Page 6 of 12 )
As for layout, I mentioned above that there simply isn't much room to work with on the PCB. They did manage a pretty decent layout however, without getting too many wires in the way. One factor that helps is the lack of a P4 12V four pin connector hanging over by the socket. With that out of the way, and with the PATA and SATA connectors intelligently placed at the edges of the board, it requires only a minimum of fuss to keep good air flow over the board. The ATX power connector is also off to the top right hand side of the PCB.
One other question needs to be answered before we move onto the benchmarks. Overclocking is obviously an important factor for many people, especially on a relatively new platform that not many people have had a chance to OC yet. Laptops aren't exactly known as vehicles for all out performance. Put that processor in an environment where much more cooling is available, not to mention the addition of more power from the wall socket, and we are set to look good, right?
Well, the "upside" as Hubie Brown would put it, is there. The problem arises that there is very little bios support at this stage in the game from DFI. The clock options are there, and you have a full range of FSB frequencies, as well as every multiplier downward from stock to 6. You can thank SpeedStep for that one. DFI also opens up the memory options, but they are essentially useless without voltage adjustments. I don't know of many RAM sticks that run 240MHz at 2.5V -- or 220MHz for that matter. So you are stuck using 3:4 instead of the 3:5 option that would be very helpful when you start increasing the front side bus clock.
On my board I found that it was stable up until around 143 MHz; after that it wouldn't pass Prime95 for extended periods of time. I'm going to guess that two things are holding it back: chipset voltage and cooling. After all, this is coming from a laptop that is meant to run the minimum amount of voltage with full stability at stock speed. Since the north bridge also features the memory controller (unlike A64 solutions) there is a fair amount of circuitry there demanding power and cooling, not to mention the graphics core should you choose to use the onboard video.
It does get fairly toasty under load, I'm not going to deny that. DFI also knows how to design a PCB, so I don't think that traces are not capable of handling more than 143MHz, not when they have been driving their other boards well past double that. There was originally a PCI/AGP clock issue when I first got the board, but a bios update seems to have cleared that up, at least using the 36/72 lock. I still experience redraw errors periodically beyond 143MHz, letting me know something isn't right on the AGP bus at least. A dose of voltage might snuff this to at least get beyond 150MHz FSB, hopefully closer to 160.
As for the CPU itself, playing with multipliers was a great way to help alleviate that bandwidth problem manually. I ended up settling for 141x16, as opposed to the stock 100x17. This amounts to a clock of 2.26GHz, from 1.7GHz. That, may I also remind you, was with stock voltage to everything on the board. This is why I see so much potential in the platform. A 33 percent overclock at stock voltage is nothing to turn your nose up at, especially with the meagre cooling available.
Once I get the water hooked up and functioning, the soldering iron will make an appearance to correct some of the oversights from the factory. After looking at the VID pin layout care of www.intel.com, the same trick using a piece of thin copper wire to tie pins to ground is available to easily modify what the motherboard sees as the stock voltage of the CPU. The kind folks at www.xtremesystems.com and www.x-86secrets.com have already found methods to raise the chipset and vdimm on this board fairly easily. I'll highlight these mods when I do them in a few weeks.
Next: Specifications and benchmarks: Adobe After Effects, STARS CFD Solver >>
More Motherboards Articles
More By DMOS
