The Ultimate Motherboard: Initial design V15

The SB450 southbridge is pretty poor, but ATi released the SB460 recently which can hook into RD480/RD580 northbridges and is pin-compatible with the upcoming SB600. That southbridge is released around the launch time of Socket AM2, so we'll probably see boards with a full ATI complement of chips around then.

 

ATI's Radeon Xpress 3200 chipset can be read up on on ATI's website. I dont know much about the southbridges yet.

 

get rid of the legacy ports IMO.

 

The serial port is already a header with a little break-out cable, right? Do that for the parallel port, too. I have found that parallel ports are lowering in usage, really. You can get USB to parallel adapters, you know. If both the serial and parallel ports are made internal headers with cable-break-outs, then maybe you could include a backplate which they screw into if someone needs them externally.

In reference to the sound-card, having that kind of slot seems a bit... backwards. Proprietary slots is so... CNR/all those other random slots. Perhaps you could have a standard x1 slot, and then contract a sound card (x-fi type?) from a sound card maker to come with the motherboard. Then it's upgradeable and truly separate from the motherboard. There would be two models, an onboard sound version (which would just use that x1 as a riser) and an X-fi (example) version. Sorry if that didn't make sense.

The USB ports look like they might suffer from the same problem as front ports-too close. Are they all standard distance?

The Firewire 4-pin and 6-pin are much too close. If they are on the right scale, then only one will be used at a time. I find that a 6 to 4-pin cable is much more handy than a 4 to 4-pin. Perhaps keep the 6-pin and add another on the back-plate?

Just thoughts. This is still looking great!

 

The problem with ANYTHING on the backpanel is that there isnt enough space as things stand. The design as it is for the audio riser is just to free up backpanel space. Otherwise, it could probably be substituted in with conventional audio jacks and I could remove that audio riser altogether.

 

thats what i said and it is how it should be means parallel LCDs dont need a cable hooping back inside the case too

 

If you are talking about my suggestion for firewire, is there not space on top of the parallel? Or if it is removed, then where it is now?

The audio riser and slot is a great idea, and I don't think the audio should be re-incorporated onto the back panel. I think the proprietary slot should be made into a standard x1, and then the audio card should be placed into it. If it is an audio solution such as a custom X-fi, then that's what it is. If it is a standard AC'97 or Realtek or any other audio solution, then it is that as well. What the slot does (x1 or proprietary) could be changed in the BIOS. This would basically free up that slot to be used normally. Perhaps the user would like to use that slot for something else, but still wants sound. This would allow that. If this doesn't make sense or was pointless, sorry.

 

There is no way I am going to be able to get Creative to create a custom PCI Express 1x audio card. The design is what it is in order to allow the audio signals to use the convenient 1x connection configuration for a different purpose whilst preventing people from using it as a 1x slot. I will probably end up removing the parallel port and relocating the firewire header B connection.

 

Why not just remove the 4 pin firewire alltogether? If people really want a 4 pin you could include a 6 pin to 4 pin adapter in the box.

 

The backplane space is an issue, expansion cards are just as large of one. The best idea I have seen to similar issues would be a breakout box that uses some sort of a submini connector on the backplane.

As for legacy ports, the problem, as I stated before, is that they are there wether you like it or not on the super IO chip as is a floppy controller and other assorted legacy stuff. There are uses for those ports such asa disaster recovery. That said, the ports don't need to be on the backplane, there are other solutions out there. 1u server mainboards are a good place to look.

 

from what i can see from the first post that u've updated.

- Lose the Serial port
- re-align the firewire ports, don't think they'll work like that can't plug 2 in at the same time even with 6pin and 4pin variations.
- Where the serial was add another gig nic port and 2 more USBs 2 firewire.
- Bring back soundstorm as it ripped the pants off of any creative machine that would be better then having onboard xfi.
- lose the IDE ports and floppy port as u can use a USB floppy if u need to flash anything these days.
- move the LED, hdd led, reset headers to the right hand edge as that would make for better cable managment.
- Loose the bottom PCI slots they'll be redundent soon.

What i want to see.
onboard scsi, better easier mounting, flip the motherboard so that the cpu is at the bottom, space the PCI-e slots out, better cooling no need for heatpipes its the latest gimik a heatsink will do the job fine if u place the chips where you can get larger sinks on them u don't need a stupid pipe to route the heat somewhere else.
Add more SATAII ports on the board 8 atleast.

Should do the trick


Morgan.

 

You say that PCI is going to be reduntant soon, then go on to say you want onboard SCSI, that just doesn't make sense.

Also, PCI is no where near reduntant yet, there are still loads of PCI cards being manufactuered.

4 SATA/SATAII Ports is enough for ANYBODY.

Soundstorm has NOTHING on the X-Fi, and Roto is using an ATI chipset, so that just wont happen.

Don't lose IDE either, nothing wrong with it, especially for Optical drives.

I agree with the floppy header, I would rather have a USB floppy drive that I only need to see when I need to use it.

 

i still think the serial port and the parallel port should be on headers, not on the back plate...as everyone says it gives more space for USB and more space for the Fire Wire sockets

 

having 10USBs and 2GiG lans on the IO panel would do me fine ;D. Also it makes a whole lot more sense to have both the 8pin +12v and ATX cable together, and then have CPU power circuitry only on the right side of the board - like on the DFI nf4 boards.

and whats the heatsink on the end of the heatpipe up the top left of the board cooling? the back of the parallel port? heh lcearly the vCore mosfets are already sinked (plus whats the deal with a vcore circuit on each side of the CPU? or is one for vdimm?)

 

I dont know if its possible to break it up into 10 USB ports, but who knows, it may be possible

As for the power connectors, Yes, it would be nice - but being realistic, its probably extremely difficult - if not impossible - to implement and still maintain a good board layout.

The heatsink there is for the northbridge. The heatpipe cooler brings the heat from it to an area where theres space. As you can see, you cant fit any aftermarket heatsinks if a heatsink sits there on its own

The regulator circuits are on either side in order to provide 8 phase power regulation in two four stage steps to provide the cleanest power possible. It wasnt feasible to provide the regulation in one continuous chain.

 

If they had carried on developing the soundstorm it would be much better then the x-fi instead creative bought it out or something.

Anyhoo why do u need IDE when SATA does the same job much faster and alot neater with cables being a 1/4 of the size. Lets also look at 4 sata ports 2 optical 2 hard drives wow your now onto pci upgrades, 8 would be plenty there are limitations on how many drives u can have but having 8 ports gives you lots of head room.

SCSI doesn't require a PCI slot just the control chip and if its built in u320 then you are looking at biparsing the PCI controler as there wouldn't be enough bandwidth much like the Gig nics they can't run through the PCI due to the limitations.

I don't use my PCI so i wouldn't need it so hence my perfect board wouldn't have pci . PCI_X on the other hand now where talking ....



Morgan.

 

I think SCSI is best left out of this. Its too expencive, and imo the cons outweight the pro's.

You have to think about what other people would use, just because you use SCSI, it doesn't mean that everyone else has to have a couple of SCSI headers doing nothing.

IMO 4 SATAII ports is enough, obviously the more the merrier, but when you think that most cases can only take 6 drives without a floppy drive, it starts getting a bit pointless.

Ok, maybe you'd use an SATA Optical drive, but they're really not very popular yet.

IDE is still required as the Optical drives would use it, plus one day I guarentee that you will need to use the IDE channel.

 

If you want ed to integrate SCSI, I'd put a SAS (serial-attached -SCSI) controller on there. That way you could run SAS or SATA drives off the SAS controller if you wanted. SAS isn't cheap though and I don't really think it's needed.

As far as disc controllers I'd like 8 SATA II/3gbs, 2 IDE 133 channels, 1 Floppy as a minimum, that's enought to fill up just about any case (PC-V1000 for instance). Also if possible I'd want to see all seven expansion slots used, preferablly 4x PCI-E 16x slots with 3x PCI-X (or PCI) slots between each one (though you'd have to use a different chipset for PCI-X). That way you'd have room for cooling on the PCI-E slots if you didn't use PCI-X cards and you'd have PCI-X for your SCSI/SAS controller cards or enough room to potentially put 3 gigabyte I-Ram cards in the system (since they take two slots each).

 

Chances are, you won't be able to kill the legacy ports unless you want to write the BIOS and do it nearly blind. BIOS developers strongly request at least a serial port. the circuitry is already there, just look up the spec sheets for the super I/O chips (and you might be surprised to find ISA is still there too). But this stuff does not need to take up much board space. Start poking around on www.amp.com for some high density, submini connectors. The actual ports use a proprietary cable. And more than just BIOS befelopers will thankyou, look at the otheer hardware development out there that uses serial, PCI programmers, basic stamp interfaces, homebrew TI link cables, interfaces to aome automotice electronics etc. IT may surprise you how many gamers like to develop hardware and software too, but after spending a bundle on their rig, they dont want to spend money on old tech to do their development on.
You only need 9 pins for a serial port, using something like a flex ribbon cable that should not be much space at all.

As for a floppy drive, lookj around on the net for storys of flashes gone bad. There is a neat feature of some bios called recovery, if a flash has gone bad, chances are you still have the first 64K of you bios called boot block as that part of the flash chip is usally write protected. If the rest of the chip is FUBAR, this will still be OK. Some BIOS will automatically hunt for a legacy floppy, find an image of the BIOS with the proper filname, and use it as BIOS untill it can flash and fix your bad flash. For AMI BIOS, a standard feature is you can hit control-home even before you get video to force recovery. To have USB recovery severly complicates things. With a PS/2 keyboard and legacy floppy, all this can be done with very simple direct port IO, these things are ready to go as soon as the processor is running. With USB, you MUST initalise all your PCI bridges, program them (assigning them bus numbers, map those behing other bridges and other work), scan the buses for devices, and enumerate them. But the fun is not over yet, you then need to find the USB host controller(s), initilse them, set up the USB data area, and enumerate them. A USB floppy drive is just another USB mass storage device, so you go through the same process for a USB floppy drive or a flash drive. Once you have found a device with the proper class and subclass codes, you then need to determine the type of transfers to wor with the device, bulk transfers, control-bulk, control-bulk-interrupt, or any of the other transfer modes used by one of the many USB mass storage devices out there.
On top of all this, there are PCI cards that can totally mess up your USB floppy with its on board BIOS, and it is nearly impossible to tell this has happened. All you would know is that the floppy is not working for some mysterious reason. This can all be crammed into a recovery system, but the code size means you need to compress that code to make it fit, and that can have it's own risks.
The only way I would even consider going legacy free is on a high end server board witht eh reliability of a server board, the advanced recovery features of a server board, and the user fiddles with it as little as possible...like a server board.
This is a standard to connect a floppy using a small flex cable connector, these are typically found on 1U server boards.

Legacy free sounds nice, but the public is being sold a bogus bill of goods. In order to go legacy free, you would need a complete platform overhaul including leaving BIOS totally behind, and totally forgeting about any reverse compatability. We are almost to the point where we can leave all this stuff behind and emulate it all, but it's not quite ready yet. UEFI will help, vista will help, but if you want legacy free, then you have to be willing tot totally abandon everything you have ever know about PCs. Quite frankly, the only computer I have ever seen that is anything even close to legacy free is s G5 Mac. PCs are still lugging their luggage from 1985 around with them and the public is not willing to let it go.
Sorry for the rant, but I have been neck deep in this stuff for a bit. Legacy free is somthing of a joke from the manufacturing and system developer end of the industry. It usally causes us headaches simply to make the system look legacy free. Make no mistake, at this point it is mostly cosmetic, and you are still paying for the legacy ports, you just don't have the physical connectors.

 

man.. a motherboard manufacture would make so much cash in the enthusiast market with this board I think you better © it just in case