|
Good news: owners of the 694D-Pro A can have a hardware-manager
display like this:
- it all started with & came from a Korean correspondent who's been working on - er - encouraging the onboard PDC 20265 ATA100 controller to be recognised & to work as a ATA100 RAID controller. Since
MSI list "AR" & "AIR" versions of this motherboard - with onboard RAID from the same PDC 20265 IDE controller fitted to model "A's" - it is a very simple modification indeed - much simpler than the mod for the Asus A7V published at tweakhardware
- the A7V [& K7V] have the same onboard 20265 but no ready-made RAID BIOS - which [due to the AR/AIR] the 694D has.
What you'll need:
Hardware: 1 x 100-250 Ohm/0.25-0.5Watt resistor [150 Ohm is said to be ideal]; 2 x ATA66/100 80-way flat cables -
provided in the box with our board; then 2 x matched ATA100 HD's. See our article on linking two cheap ATX PSU's if your existing PSU is heavily-loaded or if you intend at some point to run a 4-disk array - the startup current for
these will be a severe drain on the 12v rail of any normal PSU; especially when the +3.3/5v rail is heavily loaded with the demands of a dual-CPU motherboard [over 65W for the CPU's alone on our board - that's more than an AMD CPU 'clocked well over 1GHz].
Software:
An "AR-AIR" BIOS - 3.00 [6321vms.300] or later; RAID
driver-set version 1.30 Build 42 [or later] & a text-editor like notepad. You'll also want the Windows utilities for the RAID-controller.
|
Mods:
Hardware: All you need to do is use the resistor to bridge between pin 36 & a ground-pin [40 is conveniently nearby] of the
primary IDE channel [IDE#3] controlled by the PDC 20265 - sticking it into the right holes in the "slave" connector of the cable will at least allow you to see whether it all works - We show you a more secure but easily reversible way in Tweaks3
- obviously if you use 4 HD's or want a truly permanent solution you'll have to reach for the Single Malt & a soldering-iron.
|
|
Software: Expand the driver-set to a floppy; select a) TXTSETUP.OEM in the root, & b) *.INF in the OS-folder[s] you wish to use it
in. Change file-attributes to Archive; then replace all strings: "4D30" with "0D30" [all numeric bar "D"] - there are three or more instances in each *.INF & at least one in TXTSETUP.OEM. Hellish complex, eh? Save this, then change the file-attributes back to:
Read-Only & Archive.
Install Process: Save all data etc etc - it's much easier to do this with two new matched ATA100 HD's, & to keep your existing one
as is: if you're going to do this, remember to remove from "Device Manager" the "Win 2000 Promise Ultra100 (tm) IDE Controller (PDC20265)" you should have listed there, together with its drivers; then transfer your existing boot-HD to IDE#1 [primary VIA ATA66 channel]. If you
want to boot off this, set primary boot-HD order to HD#0, if you want to boot off the RAID-controller, set to SCSI.
Step 1: Update BIOS to 3.0 - it contains a U100 bios (for 20265 & 20267) & FT100 bios (for 20265) Do
not under any circumstances use the "Awardflash 783M" & update.bat file supplied with the RAID BIOS - this will not work on model "A" motherboards. Use Awardflash 770 & make an "UP300.BAT" file containing the single line:
AWDFL770 W6321VMS.300 /SN /PY
Copy: W6321VMS.300, UP300.BAT, & AWDF:LSH770.EXE to a minimum-file boot-floppy: reboot to this & at the A:> prompt type:
UP300
Wait till the process is complete & don't interrupt your computer.
Step 2: Power-off and insert resistor between pins 36 > 40 [or 36 > 2 /19/22/24/26/30 - they're all grounds] of the primary U100
connector cable - #3. Connect & jumper one HD as master IDE#3, t'other as master IDE#4
Step 3: Power-on and then set up a RAID-array: in your new FT100 "Lite" BIOS menu you have a choice of striped or mirrored
arrays; but, alas, no choice of stripe-size [64Kb fixed] nor full 4-disk support. Reboot to a full W9x system-floppy; FDISK & prepare at least one partition on your new array.
Step 4:
[new install, no HD on IDE#1] Set boot-order: Floppy 1st; SCSI 2nd: Reboot into the 1st of the 4-floppy Windows 2000 install-set OR:
Set boot-order CDROM 1st; SCSI 2nd & reboot into the Windows 2000 CDROM - I prefer to boot into the floppy-set. Install Windows 2000 , remembering to press F6 in the few seconds allowed to install your hacked FasTrak100 driver-set from the driver-set floppy you've prepared.
[Have 3rd HD on IDE#1] Set boot-order to HD#0 1st, boot into your existing Windows 2000 install & install drivers for your new
RAID-controller - then format & partition array as you please from W2K's disk-manager.
HD performance is far, far more important to overall system-speed than any other factor bar lots of RAM - as anyone with a whimpering wallet & a 10k LVD SCSI-HD knows - & this quite simple hack allows
users of this motherboard to enjoy the speed & low CPU utilization of hardware striped RAID, or the security of mirrored RAID; but not, alas, both.
To unlock the full functionality of the Promise 20265 controller you will need to go onto the "Stage 2" modification - this requires using an experimental 694D BIOS with the full PCI controller-card
RAID-BIOS inserted into it, & using a second resistor between pin 37 > ground. This offers full 4-disk support [ie RAID 0+1] & stripe-size adjustment - vital for maximum performance. It is a slightly more complex modification & requires you make & use an experimental BIOS - all
wholly at your own risk.
Obviously, the above modifications are likely to be discouraged by MSI or Promise, will void your warranty etc etc & should not be carried out by anyone unsure of their ability to do any part of this or
unwilling to accept all responsibility for their actions & all consequences therefrom. We publish this public-domain information without any endorsement or encouragement of its private or public application.
We also strongly advise using a hardware BIOS-backup device - & will be testing one soon.
It is fair comment to say the whole thing took [no joking] 15 minutes including downloads - & works, according to all the benchmarks we've taken so far, EXACTLY as well as a PCI RAID-controller using the
same settings. We advise - as usual with this motherboard - dropping FSB-speed/memory-speed to defaults when installing an OS - we had trouble at overclocked settings.
Voltage 2 : schizophrenia & the 694D:
The image below is of our system-board ready to go back into the fray after some idiot [guess] snapped a lug off a socket370.
|
It was easy enough to bend up & epoxy in a new metal lug to mobo & socket - not so easy to pare off the stump of the old one with a seriously sharp paring-chisel 0.5mm from the mobo & its delicate traces [yes,
I used a steel-shim - in fact a feeler-gauge - as a shield; but still . . . . a major bowel-loosener]. Anyone with first-hand experience of a better technique [a dremel with a router-base?] please mail in with details.
Seen in the very small - 7mm/0.25" - gap between socket370 & ATX socket are the twin ultra-thin wires leading to a bead-thermistor bonded to the base of CPU#1 - they lead to the top RH readout below- CPU#0 leads
to the top LH readout.
|
|
|
As you see, there's a fair bit of spaghetti here - allowing real-time monitoring of CPU-base, Intake-air & ambient temperatures. The relative accuracy of this setup is quite reasonable - see our "Taking
Your [computer's] Temperature" piece for details of how - & more importantly why - we do this.
The spaghetti in the right foreground is related to our stepless remote case-fan controlling system - details in an upcoming piece.
The strange little albino loch-ness monster between the GOrbs is a crane to hold the Intake-air temperature-sensor equidistant to both fansinks.
|
- The point of these pictures is to illustrate that we can monitor the temperatures of both CPU's at all times: while installing Windows 2000 it was noticed that CPU#2 - which goes in the secondary socket370
& which has a fluctuating vcore - is the "bootstrap" CPU.
When later tweaking the voltages [BIOS v 1.50, 1.62, 3.02, 3.00 - no obvious difference] it was noticed that a +0.05v adjustment to CPU#2 bumped the [default] 2.0vcore of CPU#1 to 2.29v. Time & again
adjustments to one affected the other in unpredictable ways. Whatever was tried; the vcore of CPU#2 cannot be adjusted accurately or held steady - it wanders randomly up & down over a range of some 0.05v.
If you have one CPU noticably better-behaved while overclocking than t'other [needing lower vcore/easier to 'clock] be sure to install this in socket370 #2
We are using a high-quality 300W "Athlon" PSU with the +3.3/+5v-rail providing over 120W - so there's little chance of this often-reported instability being due to anything but inadequate regulation or
control by the motherboard & its BIOS. This instability is enough to explain why this motherboard is a relatively poor bet for overclocking, despite the extraordinary headroom of Intel's remarkable cB0-stepping FC-PGA's.
|
