This thread is about a very very quirky Asus PC-DL.

I bought a supposedly brand new Asus PC-DL motherboard on Ebay a week ago and I'm having an unique issue I don't think anyone else have had, at least I can not find any information on the Internet fully matching my problem. I have read the huge 376 pages long thread about the Asus PC-DL at 2cpu.com.

The issue is that the motherboard will only POST when forced to 100 MHz FSB with the FSB selection jumpers. Trying to POST at the 133 MHz FSB (default / auto) the motherboard is supposed to run at with common Prestonia and Gallatin DP CPUs will result in the motherboard halting at either random "reserved" diagnostic codes or just not doing anything at all except for powering on fans and such depending and the CPU(s) used. No diagnostic beeps with any of the CPUs at 133 MHz FSB though but the beeping is working when completing POST at 100 MHZ FSB.

The CPUs I have tried are.

1 or 2 Prestona 2800 CPUs (21x133 MHz) 1.5V. These work fine when forced to 100 MHz FSB but the motherboard does nothing when trying to run at 133 MHz FSB.

1 or 2 Prestonia 3066 CPUs (23x133 MHz) 1.525V. These work fine when forced to 100 MHz FSB but the motherboard halts at random "reserved" diagnostic POST codes during POST at 133 MHz FSB.

1 or 2 Gallatin DP 1M 3200 CPUs (24x133 MHz) 1.525V. They behave like the Prestonia 3066.

All these CPUs work in other Socket 604 motherboards. I own an Asus NCCH-DL motherboard, an Intel SE7505VB2 motherboard and an OEM system with an E7505 chipset motherboard I will talk more about in a future forum post. To sum it up I have really checked that all these CPUs are fine.

Here are the CPUs pictured.

I have also tried many different memory kits spanning most common memory chips, both in single channel mode with a single module and in dual channel mode with 2 or 4 modules. This motherboard does not support registred memory as it uses the Intel 875P chipset but almost any PC2700 or PC3200 memory should work as the memory runs at 1:1 = PC2100 or 4:5 = PC2700 when POSTing at 133 MHz FSB. I have tried to relax the timings in the BIOS from default to 2.5-3-3-6, 2.5-3-3-7 or 2.5-3-3-8 to see if that makes a difference but it dosn't.

When POSTing at 100 MHz FSB the memory runs as PC2100 using the 3:4 divider. When using the modules SPD timings when POSTing at 100 MHz FSB with the memory at 133 MHz (the only memory divider available at 100 MHz FSB) the timings range from 2-2-2-5 to 2.5-3-3-6 deping on the modlues SPD tables and most modules can handle these timings even at 166 MHz with low memory voltage if PC2700 and not PC2100 happend to be the motherboards "safe defaults setting" when running at 133 MHz FSB and the modules SPD timings are as tightly programmed for PC2700 as for PC2100 (not very likely).

All this pictured memory is known to be good and work perfectly in the Asus PC-DL when the FSB is forced to 100 MHz with the FSB selection jumpers. I have tried even more memory but I clould only fit this much on the table... Some (not pictured) memory did not work even at 100 MHz FSB. The non working modules were mostly TwinMOS PC2700 (I think Mosel chips programmed for 2-2-2-5 needing more than 2.5V) and memory aimed at overclocking that probably only have CL3 in their programmed SPD tables but otherwise the memory compatibility with the 1009 BIOS seemed really good at 100 MHz FSB with the memory at 133 MHz (PC2100)

This is my test setup. Notice all the scratches on the northbride heat sink, very odd for a supposedly brand new motherboard... The board looks like a refurb, perhaps I'm not the first one trying to get this motherboard to work.

Things that could be wrong that I have thought of but already checked and ruled out.

CPU compatibility.

Already with the 1003 BIOS all tested CPUs (D1 and M0 steppings) were officially supported and I have updated the motherboard with the 1009 BIOS.

Memory compatibility.

It was decent with the 1003 BIOS and is really good with the 1009 BIOS, I think I have tested enough memory to rule this out. I have also tried relaxing the timings before moving the FSB selection jumpers to 133 MHz.

Shorts.

It's common for the Asus PC-DL to act strange if the backplates are shorting out traces on the back of the motherboard. I diddn't get any backplates and I use non conductive washers. I have also tried running with the CPU cooler just placed on top of the CPU without any mounting kit as Socket 604 is a PGA socket that dosn't need as much downward force as LGA sockets, the result was exactly the same. I have also tried running the motherboard without it beeing mounted on the motherboard tray.

The video card, the diagnostic POST code card, the HDD or the power supply.

I have tried 3 different video cards, two AGP and one PCI, I have tried with no HDD attached. I have tried 5 different PSUs among them a Corsair TX750, a Corsair AX750 and a Corsair AX850. I have of course tried POSTing at 133 MHz without the diagnostic POST code card just in case it somehow interferes with something during POST.

Things that could be wrong that I can not rule out with 100% certainty.

The memory voltage beeing too far out of specification during POST.

It's strange though that the motherboard behaves different with different CPUs if this should be the issue. I have also tested to run the memory at a much higher speed with the help of the SetFSB program for the Asus P4C800 (same clock generator) without issues. I think the memory voltage beeing the issue is unlikely but I can not rule it out as the Asus PC-DL is known for only providing 2.4V - 2-46V to the memory without hardware modification.

The CPU voltage too far out of specification during POST.

I think this is a more likely culprit than the memory voltage as the motherboard behaves different with CPUs that use 1.5V compared to the ones that use 1.525V. The CPU voltage is 0.08V lower than the specification at idle (BIOS HWmonitor and monitoring software agree) with the Prestonia 2800 (1.5V) CPUs and with these CPUs the motherboard shows no life when trying to post at 133 MHz FSB. The motherboard provides the Prestonia 3066 1.525V CPUs with 0.04V under the specification (this seems more normal) at idle and these CPUs at least makes the motherboard spit out random reserved diagnostic POST codes when trying to POST at 133 MHz. With the SetFSB software for the Asus P4C800 i875P Socket 478 motherboard (it uses the same clock generator as the Asus PC-DL) I can increase the FSB in Windows to the corret 133 MHz and even overclock the CPUs, this makes the too low CPU voltage hypothesis seem less likely.

Bad CPU sockets.

This is not uncommon with the Asus PC-DL and NCCH-DL as Asus had serious quality problems with the CPU sockets on their Xeon motherboard lineup. This was caused by the company making the sockets using some bad quality materials. This would seem like a good fit, perhaps the motherboard is misinterpreting the CPUs FSB pins and thinking the CPU wants 166 or 200 MHZ FSB. The motherboard can run at both these speeds but will never POST with a high multiplier CPU as the CPU always POST at the max multiplier regardles of the multiplier chosen in the BIOS as it's the BIOS that lowers the multiplier at POST. The thing making this unlikely is that I have tried forcing the motherboard to POST at 133 MHz using the non "auto" jumper setting for forcing the motherboard to 133 MHz FSB.

A bad apple i875 chipset somehow lacking the ability to use the strap for 133 MHz FSB.

As I can increase the FSB to 133+ MHz in Windows using the SetFSB softaware with the chipset running at the presumably tighter 100 MHz FSB strap the chipset shoud have the capability to run 133 MHz using the 133 MHz strap with ease but perhaps the 133 MHz strap somehow dosn't work.

What have I missed? If anyone could think of anything please let me know.

I bought this motherboard on Ebay and it was listed as new. These motherboards are not cheap, especially not when sold as new and I got what seems to be a defect refurbished motherboard not even capable of POSTing at the standard 133 MHz FSB the motherboard and CPUs are made for. I think it's not too much to ask for that a motherboard sold as new is capable of running at stock speed without resorting to using SetFSB in Windows.

As it is now the motherboard could probably be used with two Gallatin MP 4M 3.0 GHz as these CPUs uses 100 MHz FSB to get both decent performance and the ability to run full stock speed using for example Linux where using SetFSB is not possible. However my intention was to build the optimal year 2003 system using two Gallatin DP 1M 3.2 and a Geforce 5900 Ultra. To be able to use the motherboard as a strictly year 2003 gaming system I'm stuck with using SetFSB in Windows.

Sadly the motherboards FSB selection in the BIOS is only 100 - 132 MHz when POSTing at the 100 MHz setting. The BIOS does not set the AGP and PCI locks when POSTing at 100 MHz (known Asus PC-DL bug) so it's not possible to use the BIOS overclocking feature to get higher FSB than ~112 MHz without causing instability. Using the SetFSB software for the Asus P4C800 in Windows does work for enabling the PCI and AGP locks though.

I have spent many hours, even days trying to solve this, time I rather would have spent doing something else. I'm not sure I want to return the motherboard as it probably would end up at the scrappers but I have contacted the Ebay shop that sold the motherboard and I at least want some of the money I paid for the motherboard (160 euro + shipping) back.