Now that I'm actually putting everything in the case, I'm getting into blivet issues. A blivet is a five-pound bag with ten pounds of stuff in it. Stuff being the polite term.

Put another way, in a build with so many components there is the possibility of not being able to hook up everything we ordered. Here are the mild trade offs I've had to make and the reasons why.

1. With the second 9800 GX2 in the third 16x PCIe slot, it's difficult to reach the front panel audio header with the cable for the front of the 3D Mercury case. Rather than stretch it to its max and run the possibility of it unseating other wires or connections in the future, I've opted to leave it disconnected. The on board HD audio has been disabled in favor of the Emu 1616m anyway, which has a very nice breakout box. If you need to attach a headset, simply enable the on board audio and use the back panel hook up. Or, get 1/8-inch to 1/4-inch phone jack adapters and use the 1616m breakout box's super high-quality I/O.

2. While I'd hoped to mount a USB floppy/card reader internally, there are only two USB headers, both of which go to the 3D Mercury's front panel, so I'm going with external units. If you plan to leave the chassis under your desk. this is the preferred option anyway.

3. Slot protectors? Forget it. You can't see the free slots anyway as they're buried beneath the 9800 GX2s. Considering the excellent performance of the cards, that's really no tradeoff at all.

4. The 3D Mercury chassis has a nice flip open/lock down pressure retainer for the cards. Unfortunately, the big box surrounding each 9800 GX2 runs right up to the back edge and interferes with it closing. I've opted for old-fashioned screws instead of "making it fit".

5. The power connectors for the second 9800 GX2 interfere with the lower liquid distribution unit. Hence, since I'm only using the liquid cooling on the CPU (there's no LC solution for the 9800 GX2 as of yet, the memory runs pretty cool, and nVidia ranted about their thermal solution) I've ditched both distribution units and run a single feed to the CPU. I've left the feed tubes the original lengths so that if the winner wants to reinstall the distro blocks they need only reroute them. This means a little less coolant in the system but a slightly shorter run to the fill tank and radiator.

There are other things that really aren't tradeoffs as such. For instance. As we're using an external USB floppy drive, the onboard floppy controller is disabled. Same deal with the onboard ATA controller. There aren't any plain ATA drives in the system so why waste the interrupts?

Just so the winner knows, the motherboard treats bootable USB thumb drives as hard drives, which is where you'll find them under the boot menu (invoked by pressing ESC at boot time).

SP1 seems to have stabilized Vista Ultimate 64 as regards the two SLI cards. I haven't had it blue screen at startup since I installed it. By the way, Vista Ultimate 32 is in the package in case you want to run that.

I"m still having problems with the BIOS and the Adaptec 3405. Believe it or not, whenever I press DEL to enter the BIOS, the 3405's own BIOS installs fine. Otherwise, it only installs about one out of five times. I think this may have started with the P03 790i BIOS flash so I'm hoping P04, which nVidia said would be out last Friday, fixes it. Otherwise, I may put these drives in a NAS box as a secondary backup unit and slam a couple of 750GB SATA drives inside in their stead. This might be a wise thing from a thermal standpoint as well.

Part of the problem may be the very large size of the 3405 BIOS--the flash update files total 2MB. I'm guessing it's a matter of free contiguous boot rom space, an issue you'll remember the Intel D5400XS suffering as well.

More Monday. And maybe pictures if that situation has been rectified.

Cheers, Jon

Since there's really nothing I could do to expedite the development of the extremely new hardware components involved in the build, I switched to the software front. As you may have read in my previous posts, the ATI kernel DLL was causing problems with the XP install, so I had to retreat from the 3870 X2s. With the two GPU boards removed, only software problems remained. Or at least so I thought.

Having installed Vista Ultimate 64 on one 80GB partition, I had 40GB left on the SSDs (striped in RAID 0) for XP SP2. Normally a not a big deal. However, when you're planning to employ a boot manager that's going to hide one partition to use the other, you need to make sure you install XP on what it thinks is, and will be seen as the C: drive. As it stood, Vista was on C:. No problem, simply hide the partition, right? Well that should have been easy.

Unfortunately, my old standby Acronis Disk Manager, a Linux-based partition utility, apparently lacked the drivers for the Intel RAID matrix and/or Marvel eSata controller and wouldn't boot. I then tried Microsoft's DISKPART utility which comes with the PE2-based Active @ Boot disc, however, it only allowed me to hide/unhide partitions en masse with an MBR-based hard drive. I.e., hide em' all or show em' all.

I was just about to give up and delete the Vista 64 partition and reinstall that OS later when Paragon's Partition Manager 9.0 came to the rescue. Using PE2 (windows PrE-install version 2) it picked up on both RAID controllers without my so much as having to load a driver. It's my new favorite and a real gem. The Adaptec 3405 isn't attached since I can't use it and the Marvell at the same time until Intel fixes the D5400XS BIOS.

This would seem to have solved my problem, however, XP's install came up with a CRC error which probably means that some of the files copied from the CD weren't recognized. This most likely means a RAID driver problem. I finally buckled, removed Vista 64 and devoted the entire disk to the XP installation, but that didn't help one iota. After switching the D5400XS from RAID to IDE mode, the XP install crashed immediately after the first reboot. I hate to say it, but I'm running out of time and patience. According to a reader, this whole deal is a known issue with this chipset so I'll investigate over the weekend. Rest assured, I'm NOT going to send out a system that won't let you install Windows XP!

Plan B

This might mean implementing a plan B based on an nVidia SLi system or Intel X38 Crossfire system. I would lean towards an Asus P5N-T Deluxe and two liquid-cooled EVGA Black Pearl 8800 Ultras (as suggested in the forum by number3124. thanks!) which would also make room for the return of the Emu 1616m sound card. Actually, simply using the Black Pearls on the D5400XS might make room for the EMU. I have an email in to EVGA. Yeah! Feel free to chime in on this folks--one of you will be winning it.

Mar 15th Addendum: No matter what card I use, I'll be looking for low-profile cooling blocks like you'll find at Koolance

We finally opted for the faster Sony BWU200S to replace the Plextor PX-900's. Partition Manager 9 is sitting on top of one of the BWU200's in the picture below.

I appreciate all the thoughts and suggestions folks. Keep em' coming.

Till Monday...

Jon

One of the things you get used to when you assemble and configure PCs is mystery. Turn the router and the modem off for ten minutes and a network problem that was vexing you all day suddenly disappears.Even stranger fixes are commonplace. My all time favorite was the tech report that instructed you to drop your Atari ST from a height of I believe 6 inches to reseat possibly loose memory chips.

In the end, I can't really tell you why the SSDs starting working with the D5400XS. I did play with the Drive Configuration in the BIOS, changing from Native to Legacy, AHCI to RAID to IDE. But I'd done that before to no good effect. Perhaps it was setting the IDE Pre-Delay to 5 seconds that turned the trick. Now running in Native/RAID mode with the SSDs striped in RAID 0, everything is working perfectly and Vista is now installed.
But even installing Vista wasn't easy. It kept yelping about no viable volume available, even after I'd partitioned and formatted. Once again, two reboots later the problem mysteriously resolved itself.

In the latest picture of the motherboard that you'll see below, you might notice that I've turned one CPU fan around. This is to blow air across the four Kingston 800MHz FB Dimms which otherwise run pretty hot. I'm going to have to really mind my thermal P's and Q's on this build. I may even add more fans for the 15K drives to make sure they play nice. I used to run Cheetah's back in the day, but I'd forgotten just how hot they can get. And... It takes a very long time to format a 1.1TB drive which is what you basically get with the four Seagates combined in striped RAID 0.

So how is it? I've never seen Windows operate like this. Normally, I can barely tell an entry-level machine from a gamer but this one is NOTICEABLY faster, even with the placeholder nVidia 8600GT and a standard VGA driver dealing pixels. With the nVidia driver installed, even better. Overall, the effect is WOW! Whoever wins this will have no complaints about the performance. At least after it's booted--it's still a bit slow in that regard having to check two CPUs and initialize the RAID card.

Yours less stressfully, Jon