G4 Digital Audio power supply [solved]

[DirtyHairy]

Hi everybody. Yesterday I have installed an IDE controller (SIL 0680) based together with two 500 GB harddrivs into my powermac to function as a raid1. However, the drives (also the third one connected to the integrated controller) lock up frequently and get reset. The problems decrease if I unplug one of the drives and vanish if I additionally remove my USB2 controller card. Therefore I suspect the power supply (180 watt) to be the problem. However, before I start going through the trouble (and costs) of replacing it, I want to ask if anybody has a better idea what might be the problem. The hardware consists of a Sonnet G4/1.33GhZ CPU upgrade, three bars of RAM (2x128MB + 1x256MB), three hard drives, a Radeon 8500, an USB2 controller and the IDE controller mentioned above. Can this hardware really exhaust the power supply?

[eccerr0r]

In the PC world, 180W is fairly low.... However it sounds like you have a lot of hardware hooked up and is clearly a suspect. The R8500 graphics accelerator I've noticed on my system to consume quite a bit of power, plus the fact you have many DIMMs and quite a few hard drives.

One thing you can try is to see what happen if you tell the drives to spin down and stay spun down, and see how often you see lockups.

You could also see how hot your exhaust is, and if it's noticeably warm or warmer with the extra hardware.

If you have a multimeter, you should check the voltages and see if any line is dropping below tolerances.

What CPU speed was originally in the machine? Was it passively cooled and is it still passively cooled?

Many PSUs don't even live up to their 180W and is simply a 'peak' number, which it will derate fairly quickly if run in that region.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[DirtyHairy]

Thanks for the reply. Actually the 180 Watts were a mistake by me, I've checked again, the supply only says 113 Watts for the non-board connectors and gives no total number, which makes it something around 200 Watts I guess. The original CPU was a passivly cooled G4 at 466 MhZ, while the sonnet CPU has a fan and a fairly big heatsink. Also, the original graphics were a passivly cooled Rage 128 pro, while the Radeon also has a fan. I was going to grab one of these induction amperemeters over christmas holydays and check the total power consumption of the thing; however, I have noticed that the fans (which are pretty noicy) audibly reduce their speed when I do

[eccerr0r]

Yes, your hunch is right. One of the key things for computer PSUs to do is to keep voltage constant. If your non-temperature controlled fans are audibly slowing down when power consumption goes up, the voltage must have drooped, and thus your PSU has failed to do its job, likely due to not being capable of doing so.

If you're going from a G4-466 to G4-1.33 that's almost a 3x increase in power consumption at worst case all other things being equal (ignoring leakage, and don't know anything about silicon implementation. Just a rule of thumb that power is directly proportional to frequency).

Sometimes I like off-the-shelf PC components just because of the fact, even if it is potentially noisy, ugly, ... -- at least they're readily available to swap in and out, and don't have to worry about "hacking" it to fit and still cheap to boot.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[DirtyHairy]

Well, the conversion kit has arrived. I have installed it, and now my mac is running happily on an ATX power supply; it's quieter now, too . The hickups on the disk running on the internal controller are gone now (so power DID play a role), but I stlll have problems with the two RAID disks attached to the SILO680.

If I attach them both as masters to the two channels, writing is OK, but heavy reads tends to cause DMA timeouts, especially if the system reads from both drives simultaneously; doing a RAID1 check is virtually impossible. Putting them as master/slave on one channel kinda works; this way timeouts only very rarely occur on heavy reads with additional activity on the bus, like doing an emerge (however, the failues are very rare, so everyday use of the RAID might be possible, while a check still ist likely to cause 1-2 timeouts with requiring DMA to be enabled again). Rummaging around the forum, I have several other posts of people having similar issues with the SILO680, so I have ordered another controller with a promise chipset. Hopefully, I can add solved to the topic when THIS arrives.

In the meanwhile: has anyone another clue what might be wrong with my setup (I am pretty sure the cables are fine, and updating the kernel didn't help a tad, I'm on gentoo-2.6.26-r2 now)?

[DirtyHairy]

The promise controller has arrived and works like a beauty, all problems are gone now