pierocks wrote:... I suspect that some CPUs may be more picky about voltage, and a PSU that's in-spec, but barely, used with one of those CPUs, may result in noticeable flakiness. ...
That sounds very plausible. It might be one of the issues which the various mbd revisions, PROM updates and PSU changes sought to solve. It's just a pain that SGI never published any detailed info. Sometimes we know about particular hw support restrictions based on item PN, such as bigger RAM in Octane needing 1467 mbds, but in O2 it's all a bit of a muddle (especially since even the oldest mbd can have its PROM updated to 4.18), and I've found companies often have their own evolved ideas about hw dependencies, one of them being that there's something special/different/better about the serial port hw on the 1327 board, even though I've never seen anything to prove it, certainly nothing from SGI.
So yes, probably the optimal setup would be a moderately later mbd (1227, or the final 1038) with a lesser R5K, in a later edition chassis.
dexter1 wrote:I've modded both my O2's with EBM Papst and while the R12K270MHz stays are 34 degrees Celcius my R5KMHz runs happily with less that 28 degrees on the exhaust, while cutting the sound levels from the original Panaflo by 10dB.
The capacitors do wear out after prolonged heat and use and may cause certain setups to become unstable, as mentioned by Pierocks. ...
Fascinating, I'd not heard of EBM Papst before. Good point about the caps, maybe it's time I start thinking about refurbing PSUs for O2, after all they are now basically 20 years old.
uunix wrote:What the OT hasn't really explained is the complete details on the environment.
True, just that it'll be hot. If the skins aren't needed then there's much one could to do mod for better cooling, eg. cut a hole in the side of the chassis, fit a modern fan to blow air onto the whole mbd (or suck it out; not sure which would work best). Or if noise doesn't matter, just shove a Delta or lunatic Yate Loon inside the PSU.
uunix wrote:and they hate you forever and day until you pull the Dallas chip and then, they forget. I'm thinking HAL.
They're not as bad as Fuel.
Honestly though, I've seen plenty of issues with both, but of course they're different issues. For the Stoll users I've dealt with, the most common problems are dying serial ports, failed Dallas (naturally), bad XL board, bad PSU, dying disk, and even the occasional duff CPU (including a PC133) and overall dead mbd. Probably doesn't help that many of these systems become dust bunny hell, making heat issues worse and potentially providing conductive shorts across the boards (dust bunnies contain metallic particles; UK govt. report said 400M/year is wasted in the UK alone on repairs to equipment caused by dust buildup).
uunix wrote:I have an INDY up for 70 days in a hot office, I have never heard the fans come on.
Presumably a Sony. My guess is, its temp activation point is either just way too high, or it doesn't function correctly at all. I've come across a couple of companies that have fitted extra internal fans with a molex mod to power them. One was a 12cm just sitting loosely on the XL8.
miod wrote:Assuming the machine is not diskless, the indy/challenge will require a narrow SCSI disk, which are hard to get by nowadays and are limited in capacity, while the O2 will use SCA disks, allowing for more modern (larger capacity, lower power consumption) disks.
That's why I fit Indy/Indigo2 as standard with a particular model of Seagate 36GB/15K SCA + adapter, it has a good speed/access, but is very quiet indeed and very reliable (not immune, no 2nd-hand disk is, but I cover them for much longer than other disks), though it does run kinda hot. I would use a Fujitsu I've mentioned elsewhere, but like several makes the SCA Fujitsu does not support narrow SCSI (the 68pin works ok, but it's harder to find, costs more and the adapter is more expensive). In Indy though fitting two such disks does need an extra fan.
However, there are also a number of good 10K 18GB/36GB drives which create far less heat and are also quiet, but I have less experience with their reliability; one could easily use two of them without a fan. By contrast, I bought several hundred of the above Seagate 36GB 15K.
miod wrote:Of course, one can use an SCA disk behind an adapter in an indy or challenge, but the adapter needs to fit in the chassis.
It just needs to be the low-profile version, direct 50/80pin (I bought 200 to have a good supply). The combo 50/68/80pin adapter works ok, but its bulk prevents the fitting of a 2nd disk. Not an issue though if there's just going to be a single disk.
josehill wrote:FWIW, my experience with Challenge S systems has been similar: rock solid reliability. I had one that worked as a corporate intranet server for years and was offline only when an OS update required a reboot.
It's certainly very reliable. As I understand it from old PR, Netscape started its business with ChallengeS (ah the old days of WebFORCE). Hmm, I wonder if there was a different degree of engineering robustness put into the ChallengeS design, because of the expected target market. Indys would often be turned off each day (I used the auto power on/off feature for the student lab of 20 systems I ran), but a lot of ChallengeS units would be going 24/7, though I do know of Indys that were left to run non-stop for 20 years until a PSU or something finally died, the original tech admin person long since gone. One company told me the Indy was just, "the blue box in the corner", which ran all the time, keeping their factory floor going; nobody even knew what it was until the PSU failed (you shoulda seen the dust inside, record breaking bunnies, think I have a picture somewhere).
Would be very interesting to know SGI's onsite MTBF data collated from returns/repairs. I suppose Indy did have a bit more inside that could go wrong, but even then.
uunix wrote:Anyone know what temp the fans cut in? I could get a heater in there and test.
Interesting question, I've never thought to test, and I wonder if ChallengeS behaves the same. Maybe the cut-in temp is somehow determined by the mbd rather than the PSU itself? I don't know. Best thing to do with a Sony is force the fan to be on all the time, and perhaps fit a better fan.