160 Gig Drive in a DMS? What?!? How can that be?!?

[judb]

(Update: This Link is now taking donations for the DMS >127 gig project and Kernel mods to support that.
Sam will be in charge of getting the drive(s) if we can raise the money.)

Okay,

I want to know how much interest there would be in having the ability to use any size drive in the DMS..  Right now we are stuck at the 127 gig max size.  

If there are enough interested parties.. I'd like to collect say 10 bucks a person (you are free to donate more if you wish) to get at least one, if not two of these 160 gig drives (the Seagate Momentus 160 2.5 inch drives.. available at newegg).  The idea is to get a drive for Sam and I to do testing on.  These drives are about 350 bucks plus shipping from newegg.  I can't find them for sale elsewhere.

Up to today, Sam and I have not asked anyone to help cover the costs of hacking the phatbox.. we've done it as a hobby / something to be done for fun.  However, I've personally tied up quite a bit of money in these boxes over the last year while we did this project and can't pick up a drive like this for quite a while.  

Any money that is collected for this purpose would be used to purchase no more than two drives (one for Sbinger and one for Judb) and any money over that amount would be forwarded to the Phathack Forum Support Fund.

Does this sound like a good idea to you guys?  If it does then we'll set up a donation method for that purpose.. be it via Pauls account with instructions for the money to go to the drive fund, or perhaps a second temp account for the express purpose of the drive upgrade.

As it stands now, we don't have an agreed on plan for collecting these funds, but please feel free to donate to Paul for his efforts to give us a working forum and keep it running smoothly.

To do this we'll have to build and distribute a new kernel image.. The intent is to see if we can squeeze any better performance out of it, or perhaps bring us up to a 2.6 kernel version.

Thanks folks!

[Terry_Kennedy]

I want to know how much interest there would be in having the ability to use any size drive in the DMS..  Right now we are stuck at the 127 gig max size.  

I'm very interested, though I won't be buying a drive until the capacity hits 200GB. That 160GB drive is just Seagate's way of putting their toe in the water - with traditional (non-perpendicular-recording) notebook drives at 120GB now, we should easily see perpendicular drives in the 200's. Back when I was testing this (in the early 80's) we had a 10MB 3.5" floppy, so 8x current capacities should be possible.

If you don't mind futzing with adapter cables, I could send you guys a couple 250GB 3.5" drives - as long as you're working on a bench and not in a car, the added size / power / mess shouldn't matter.

To do this we'll have to build and distribute a new kernel image.. The intent is to see if we can squeeze any better performance out of it, or perhaps bring us up to a 2.6 kernel version.

I'm normally "Mr Latest Software Version", but given that the PhatBox is an embedded system, I don't see the need for a newer kernel unless it makes it easier to pick up the latest drivers for larger disks.

What I'm interested in:

1) > 128GB support
2) Kernel built with logging to the serial port on the PhatBox
3) If the OS random number function is used by the player, fix it - I don't care what PhatNoise says, "Random" should not play the same song 3 times in a row...

[judb]

Great point on the random function.. I don't know if anyone has looked into how phatd does random.. we should fix that pronto.. i hate the current "random" it isn't very random at all.

[todd1010]

I'm interested, and will donate but I won't buy a 160gb. I'll wait for the 200gb drives to be coming soon. This will just us ahead of the game.

Let where to donate. Also, what  if there's not enough interest in this project and no one donates. Or there's a few and it no where near enough to cover the cost of the drive?

[judb]

Terry, I do have a phatbox set up on a test bench and don't mind messing with adaptor cables... however on the test bench (my desk) I don't have a good way to listen to the playback, so I can't really test any changes in responsiveness of the system.  I could go down to Radio Shack and get some cables to hook the RCA outs up to my PC's line in or swap the cables into my PC speakers but that is a pain in the buttox. heh.  

[judb]

Please Donate Here...

Sam is collecting the money via the SourceForge account for PhatOS...  We'll try and raise 700 dollars to pay for TWO drives.

Please, if you can, donate a few bucks.. if you can't.. check out the PhatOS sourceforge site and maybe contribute some changes.

we'll be looking for other folks to test the kernel changes out on smaller drives if we can get these 160 gig drives to make sure all the boxes work well with it.

I also expect to release an updated version of the phathack CD / floppy along with the new kernel.

Please respond to this thread after you make a donation and let us know what you contributed.. since we can't track it via the forums any other way.

[Terry_Kennedy]

Please respond to this thread after you make a donation and let us know what you contributed.. since we can't track it via the forums any other way.

I just sent $50.

[S80_UK]

I will be happy to help with kernel testing when the time comes.  I have a 20 gig original DMS and an 80 gig Western Digital drive in a second built from a bunch of parts.  Later I may want to go to 160 or 200 gig as well, but will wait for the prices to fall (when there is competition from other HDD suppliers).

As far as a rebuild of the kernel goes, one area of possible improvement may be in the caching.  I know almost nothing about Linux at this point, but I imagine that it may be possible to say how much cache is allocated, possibly on a partition by partition basis.  This may help with some of the glitching that I and others have sometimes seen.

[judb]

Thanks for the dontation Terry!

[A543]

I'm not trying to be a wet blanket on the fun here but just wanted to throw this in for opinions. I don't know anything about IDE controller chips, but from what I've seen, I'm still not sure that the aspect of the hardware supporting >120GB has been absolutely anwered.  I do know that PC motherboards can't be upgraded to work with 48bit LBA by just a BIOS or driver update. Wouldn't it be better to see if a desktop 160GB (~$60.00) drive can be made to at least access the full 160GB of data before spending $700.00 on laptop drives?

[SteveC]

I'm not trying to be a wet blanket on the fun here but just wanted to throw this in for opinions. I don't know anything about IDE controller chips, but from what I've seen, I'm still not sure that the aspect of the hardware supporting >120GB has been absolutely anwered.  I do know that PC motherboards can't be upgraded to work with 48bit LBA by just a BIOS or driver update. Wouldn't it be better to see if a desktop 160GB (~$60.00) drive can be made to at least access the full 160GB of data before spending $700.00 on laptop drives?

Actually, there were quite a few that could be LBA enabled with a BIOS update. The hardware led the software by several months, and while the drive controllers installed on the motherboard provided LBA compatability, the BIOS's took longer. However, this was during a fairly narrow transition period (about 6 months, IIRC), but it really does depend.

[judb]

Sam and I have no doubt that it will work, just with a kernel update and some tweaking.

The real issue I am concerned about has to do with performance and larger drives.  I've seen several people mention playback issues with their drives after the upgrade.. I think we could be running into some issues with caching or drive power management issues... Unfortunatly I don't have a drive that is causing me any issues right now, nor do I have a laptop drive that is over 120 gigs in size.

A regular drive does not always behave the same (thermally, or acoustic managment-ly, or performance wise) as a laptop drive, so while we *CAN* use a 400 gig IDE drive to do the testing with by creating some hacked together cables, it does not mean that we will reveal any issues that would show up in the car.

Of couse, there is no way to know for sure if that seagate drive would do it either.

So, if you think that its a waste of money, thats fine, don't contribute.  Which it seems a lot of folks feel that way considering the view count of the thread and the one donation repsonse in so far.

[SteveC]

I think the biggest potential problem is memory management and the database. If everyone uses the larger drives to increase the quality of their music (i.e. FLAC), then the impact would be minimal. But once you start shoving many tens of thousands of songs onto a DMS, I don't see how their database will be able to perform well, and caching it will certainly impact things.

A good experiment would be to create an artificial song database - maybe 30-50,000 small MP3's - on a normal DMS.

Just a thought

[judb]

Thats a good point..  

Something else I'd like to look into is taking my test unit on my desk and seeing if I can add more ram to the pads that are there for it and have the box recognize it.  That would be a much more serious mod where most users couldn't do it.

An alternate option would be finding out if we can compress the database in memory and decompress on the fly ... it might be managable to just load parts of the database unless you are in a random mode.  who knows.

[sbingner]

Well, I couldn't talk myself into waiting until we got enough donations, so I bought one and got it working...  working kernel is posted on the sourceforge page (http://sf.net/projects/phatos) and on downloads (http://downloads.phathack.com/sbingner/linux-lba48.bz2)

[judb]

We'll be putting together a new CD for making your DMS based off some new scripts and it will have the new kernel as an option perhaps.

The floppy won't be able to hold that though.. not enough space.

The new patcher that we'll be distributing may be of use for everyone to run again as it changes some additional things that the older patchers did not...

Don't expect the new release to be in the next few days.. but it will be coming soon.

[S80_UK]

Thats a good point..  

Something else I'd like to look into is taking my test unit on my desk and seeing if I can add more ram to the pads that are there for it and have the box recognize it.  That would be a much more serious mod where most users couldn't do it.

Of course, the 710 version PCB doesn't have any extra pads for additional chips.  Actually, on the 910 PCB, since there are a different number of pins on the unused pads, I would assume they are for an alternate RAM type rather than for extra, and it may not be possible to use both sets together.

On the 710 PCB, as per the VW boxes, the two RAM chips are 64Mbits each, organised as 4 Mbit by 16, giving 8 Mbytes each chip.  Technhically it should be possible to replace them with 128Mbit chips (8 Mbit by 16), since these are available in a pin compatible form.   This would increase the total RAM to 32 Mbytes.  However, this kind of mod would obviously only be for those with experience of hacking surface mount PCBs (not many of us I guess), and I have no way to know whether the SDRAM controller in the processor is set up to handle these, or how easy it would be to change the software to make that possible.  Personally I wouldn't want to try this mod without a better understanding of the likely chances of success, even if I was doing this on a spare box.

Any thoughts on this?

[Terry_Kennedy]

Of course, the 710 version PCB doesn't have any extra pads for additional chips.  Actually, on the 910 PCB, since there are a different number of pins on the unused pads, I would assume they are for an alternate RAM type rather than for extra, and it may not be possible to use both sets together.
Any thoughts on this?

The easiest thing to do is probably to stack an additional chip of the same type and stack it on top of the existing one, then lift the chip select / output enable / whatever the chip uses and wire it up to the next higher address line.

I can do the rework to add the parts, but don't have the time to figure out the memory map. This doesn't mean I'm volunteering to do it for everyone, but can play with a unit as a feasibility study.

[judb]

do you have one or do you need one?

[Terry_Kennedy]

do you have one or do you need one?

Me? I have some PhatBoxes, I'd need the chips.