Castanet

I can't even get Linux to handle my NVIDIA GeForce GTS 360M out of the box let alone what the OP is pushing. I'm sure after mastering a crash course in Unix I'd get it working. I just meant to boot Linux live which most likely won't have the drivers included (without downloading) to max the OP's system as a test. But if problems still arise then it might be hardware and not Windows (e.g. OS corrupted, bad update, driver, virus etc.)

Loed wrote:Then just do a staged boot up process(install each extraneous piece at a time, boot, test, repeat).

Good suggestion! Unless the problem is random. If it is then do a stress test and push the system to max and monitor it. A can of Super Cold 134 is helpful! If the OP is running a true online UPS it might be overloaded resulting in a poor output sine wave which could cause abnormal output voltages from the power supply. Just a thought!

DANSPEED wrote:I can't even get Linux to handle my NVIDIA GeForce GTS 360M out of the box let alone what the OP is pushing.

So you’re saying that I’ve got some decently exotic hardware? I do take pride in getting my Frankensteins into operational status under hardware configurations that most other systems would simply go tits-up under.

DANSPEED wrote:If the OP is running a true online UPS it might be overloaded resulting in a poor output sine wave which could cause abnormal output voltages from the power supply. Just a thought!

I am running multiple UPSs (about one for each major machine), only because the circuit my machines are on tends to get overloaded *very* easily. I have to reset the breaker at least once a week.

I have two 1U servers that are each running two six-core Opteron 2400s (2435s, I think), 16Gb of DDR2 ECC REG RAM and four 250Gb drives in a RAID-5 array, one tower server running a Socket 768 dual-core Opteron (not sure of speed) with 4Gb DDR RAM, A “heavy lifter” workstation with two dual-core Opteron 285s (and another RAID-5 array) with 16Gb DDR ECC REG RAM and assorted laptops and smaller specific-purpose workstations. Oh, and a networked HP 4050TN that tends to draw a metric assload of juice when it fires up to do a print job (as opposed to when it’s in idle mode).

And don’t even get me started on my dual NAS units that are out in the living room, under the TV… for people who enjoy seeing hardware pushed to well beyond anything it was ever designed to do, my NAS units are a special treat.

Hmmm… pulled my “daily driver” (the one with the strange errors in the OP) off of my upper desk and stripped certain parts to turn my “heavy lifter” into more of a daily machine. I now have the old “daily driver” on my dissection table, plugged straight into a power strip. There is no NAS, no RAID, no primary drive. The ATI Radeon HD 3650 has been replaced with an ATI Radeon x850. The HP Gigabit PCI-X card has also been yanked, although the LSI 9550SX-12 remains, along with the basic PCI USB/Firewire card (not sure if it’s a VIA or Ali chip).

All of a sudden, I can get past the BIOS with more than 4Gb in the machine again. Although I seem to be stuck on 8Gb now (apparently two sticks are bad - F1 offset error with the BIOS). This is strange -- the 4Gb BIOS hang limit is a recent thing, but it’s sat for extended periods without power since. And yet, all of a sudden it’s getting past the BIOS with more than 4Gb again. Things might be looking up, as I was hardly ready to retire this beastie.

Will try hooking up my old drive tomorrow and checking the IDE channel stability. Will also check out the Intel processor software mentioned on the first page.

…and now my “heavy lifter” is acting like a psychotic schizophrenic diva, but just with the USB ports (so far). Even those straight off of the motherboard, as well as those off of the PCI USB/Firewire card.

*le sigh*

Looks like I’m disconnecting it from the UPS tomorrow morning…

rekabis wrote:…and now my “heavy lifter” is acting like a psychotic schizophrenic diva, but just with the USB ports (so far). Even those straight off of the motherboard, as well as those off of the PCI USB/Firewire card.

*le sigh*

Looks like I’m disconnecting it from the UPS tomorrow morning…

You still have the RAID card and PCI/USB combo in the mix -- maybe try removing those too. Your mobo has enough USB ports and for a test you can boot off the mobo somehow.

I dunno, being an intermittent problem like your describing maybe your mobo has a faulty RAM socket or CPU socket, or a iffy power connector plug or poor power cable, or cold solder, or a faullty PCI/AGP socket or possibly a crack on the mobo. You could even have a flip chip issue like with the Geforce 8600M where under certain conditions, mainly heat, parts of the die would separate from the substrate material. Hey on the plus side, you haven't mentioned a BSOD! Put it back together and poke around (cold spray, hair dryer, plastic stick) while doing a stress test and see if you can get it to screw up. I don't have much more to add.

rekabis wrote:Anyone happen to have a bead on a local SuperMicro X5DAE (or the SCSI variant) in the Okanagan? I know it’s a long shot, but the only people on eBay selling that model for under $USD300 (as in, $USD40-80 before shipping) don’t ship outside of the continental US. And they’re adamant about it.

You could always get it shipped to one of the border service companies across the border from Osoyoos, and go down to pick it up and bring it across the border yourself.

underscore wrote:You could always get it shipped to one of the border service companies across the border from Osoyoos, and go down to pick it up and bring it across the border yourself.

All the sources that are selling them for decent prices (i.e., NOT $300-500) require a confirmed address. Ergo, a border service won’t work.

DANSPEED wrote:I dunno, being an intermittent problem like your describing maybe your mobo has a faulty RAM socket or CPU socket, or a iffy power connector plug or poor power cable, or cold solder, or a faullty PCI/AGP socket or possibly a crack on the mobo. You could even have a flip chip issue like with the Geforce 8600M where under certain conditions, mainly heat, parts of the die would separate from the substrate material. Hey on the plus side, you haven't mentioned a BSOD! Put it back together and poke around (cold spray, hair dryer, plastic stick) while doing a stress test and see if you can get it to screw up. I don't have much more to add.

Keep in mind that I am talking about two separate computers here:

• My “daily driver”, a dual-processor HT Intel Xeon s604 system.
• My “heavy lifter”, a dual-processor, dual-core AMD Opteron 285 s940 system.

My daily driver is the one in the OP, and which suddenly started acting all nice and proper once I took her down and put her on the dissection table. Now my heavy lifter is starting to act up, but mainly in the USB ports. When I shut her down this morning and put her on straight mains power (no UPS), most of the USB problems cleared up (most, but not all).

I am assuming a UPS issue until I do further tests on each machine. I am also switching back to my IBM Model M keyboard in order to see if the disconnection/reconnection issues with my keyboard and mouse (through my KVM) is related to going USB-only for my peripherals. My old Model M keyboard (a PS/2 device that required a PS/2 to USB dongle) required constant polling from whatever system the KVM had it connected to (making any BIOS work a PITA, since it gave serious latency to any actions outside of Windows), so I’m guessing it’s this polling which prevented any HID peripherals from dropping out unexpectedly when I was using the Model M.

With this said, most stress tests don’t reveal anything in of themselves. I have been doing Mersenne prime tests both within Windows and from a boot USB (UBCD). Same for the memory -- MemTest 86+ with ECC testing turned on (MemTest 86 does not have ECC testing). Every problem I have come across so far has been without any associated errors being thrown in a testing program.

With my daily driver, it started with a problem between the RAM channels -- all of a sudden only one channel could be occupied at a time; if more than one channel was occupied the BIOS would permanently halt in a very strange way just before it switched over to loading Windows. Then two weeks later it was the IDE channels that started slowing down to the point where no data was going across it, eventually leading to a “hang” of the entire O/S or testing program (no error would be thrown in the testing program, it would just grind to a halt). I would only occasionally get error messages in the Windows logs about I/O errors involving the IDE channels or about IDE channels not having the latest drivers (they did).

So yeah. This is a set of problems that needs a lot more research to figure out just what is going on.

Apparently it is NOT a UPS issue, and is indeed a power issue.

Over the last few days, I’ve had one or more of my drives dropping clear out of the RAID array (power disconnect error) and immediately bouncing back in (power reconnected). As well, my USB devices (all of them) kept disconnecting every 2-5 minutes. Finally, every once in a while during a particularly bad USB disconnect (when even switching channels on my KVM would fail to fix the problem), a manual USB disconnect-reconnect would spontaneously bluescreen the computer.

Plus, no storage device attached to my USB ports was able to transfer more than about 500MB of data before it ceased responding to the computer (and the data became corrupted, sometimes to the point where a reformat of the media was necessary). These are USB ports that are both off of pin headers on the motherboard as well as pin headers on a third-party PCI USB/Firewire combo card (one pair to each source), so I doubt that both sources are failing at the exact same time and in the exact same way.

I now have three 650W power supplies attached to various parts of the computer (one handling the RAID5 array, another handling the RAID0 array, the boot drive and the AGP card, the final one handling the motherboard and anything else attached to it), and it finally appears to be stable.

<sarcasm>
So, judging from what is needed for this system to remain stable, does anyone have a recommended source for a 2,000W SSI-class ATX power supply? Apparently these things currently top out at 1,200W, which is a little less than two-thirds as much as my system seems to need.
</sarcasm>

Rekabis,

I don't know what kind of computer you have, but it sounds intense.

For some reason, this is what I imagine:

I was thinking like this.

Really?

I was picturing something more like this.........