Spread Spectrum...wonder if it's causing STRANGE issues

[drjay]

Ok here's the deal. In this room I have 4 computers, a monitor, a small fan, and two APC's. In the last month I have had 6 gigs of ram fry on me all split between 3 of the 4 computers. I had 4 gigs go out in one, 1 gig go out in the one next to it, then skipping the next one the one after it just fried a gig strip today.

This is getting insane and I can't figure out what it is. It can't be the APC's unless both failed at the same time and are spiking. Each computer is about 1" apart.

I was digging around the bios and found an option for spread spectrum control. In reading about it I really have to wonder if EMI is causing issues in the ram of the computers nearby. Strange to think, but I'm out of ideas otherwise.

Has anyone heard of an issue like this??

 

[chuckshissle]

As far as I know the Spread Spectrum basically spreads the frequency of the pc components. Meaning the frequency is constantly changing different level and back so that it can't or reduce the interference to a certain components with a certain frequency. As far as your problem I don't think the interference would cause another pc to fry up it's ram, from what I am understanding on your delima. Usually the case will reduced EMI from the inside going out. So what's the spread spectrum settings right now? Have your tried spacing the pc apart?

 

[audiovoodoo]

..

I was digging around the bios and found an option for spread spectrum control. In reading about it I really have to wonder if EMI is causing issues in the ram of the computers nearby. Strange to think, but I'm out of ideas otherwise.

Has anyone heard of an issue like this??

Seriously doubt that this has anything to do with it. Spread spectrum is designed to reduce EMI. The one thing most people will tell you is that it can get in the way of a good overclock.

Read all about it here wikipedia

Sounds to me like your PSU's might be on the way out or that you are getting power spikes...

 

[audiovoodoo]

This is getting insane and I can't figure out what it is. It can't be the APC's unless both failed at the same time and are spiking. Each computer is about 1" apart.

Can you monitor the voltage with the APC's? Is it just the computer that is connected to the APC or do you have any other equipment strung off it..

Must admit it does seem a little strange. Is the equipment of a similar age?

 

[Mondoman]

It's not the spread spectrum. Here are some possibilities:
1) APCs no longer providing surge protection. Cheaper APCs don't provide an indication when their surge protection circuitry has sacrificed itself.
2) Power supplies in the computers. Cheaper power supplies are well-known to fry components.
3) Running conditions in the PCs. What are the case/system temps? Are you overclocking? Are you running at higher-than-standard DIMM voltage?
4) How do you know the DIMMs are bad? Did they pass memtest86+ and PRIME95 stress-test before, and now fail one or both? If not, maybe they were bad (or the system was bad) to begin with, and they just got worse?

 

[drjay]

Thanks for the replies so far. I'm wondering about EMI because unless I'm forgetting something the ram seems to be about the most sensitive component in the computer and it's the only thing going out.

I can't space them out much further, the desk nearly perfectly fits them with around 1" between them, 2" for the server.

I do have horrible power here, it flickers a lot. It's certaintly possible I'm getting spikes but I would think at some point more than just ram would go out. I don't have a way to monitor the output of the APC's though. The APC's are probably about 2 years old and may be around the same age but I don't know.

I'm also leaning towards something 'strange' because I usually keep 3 of the 4 running at 100% for weeks on end. Yeah I abuse them. But it saps a LOT of power, I literally doubled my electric bill. Plus they're all loaded with equipment, I have 12 hd's (4 scsi), 15 fans, 9 gigs of ram and the average processor speed is 3.7ghz. I don't think I ran their life out though because collectively since they were new I probably used the equivalent of 2 years of life.

I'm going to buy a better UPS and hope it helps but if anyone has heard of a similar issue I'd love to hear about it.

 

[CompTIA_Rep]

Questions: What are you using to confirm the RAM is bad?
Secondly, if your APC's do not stop Spiking then you need different APCs.
On brownouts, the first thing to go (usualy) is the PSU. On spikes, it can be anything, but I hardly ever see the RAM being the only thing damaged. Are they all the same MOBO? All the same PSU? All the same memory?

 

[GavinLeigh]

If you are having electrical issues I would seriously consider getting a high quality UPS unit, with more than your basic feature set.

You probably need something that conditions your power all the time and modifies the outgoing frequencies too. There's a lot of potential issues with bad power... even before it reaches your PC power supply.

Under the circumstances you need to remove as many variables from the equation and a good stable supply of power is a great idea. A decent double conversion unit like a Liebert unit would be good purchase. They convert the AC ot DC and then back to AC, cleaning the power as it goes.

However if you are overclocking your memory and upping the voltages to achieve this, then it might be as simple as pushing the envelope too high.

 

[BustedSony]

The main way that perfectly good memory can be fried is if the clock stops. In other words if they are just receiving DC but no access or write is running then they can quickly overheat. So maybe look for something that is causing the system's clocks or dividers to shut down. That could be the dirty power mentioned. The other thing is that it IS possible to simply get a run of bad memory. In 2001 I had to go out of business from building small Business systems because almost every single ram I had received, from two suppliers, put in some 50 computers, went bad, and after paying for replacement for those out of supplier warranty and for shipping the PC both ways I was in the hole.

 

[drjay]

Yipes that's unfortunate. I'm just glad mine have a lifetime warranty...wonder when they'll be sick of me though haha.


All 4 comps have different motherboads and PSU's, and of the 3 that had ram fail 2 were the same ram 1 was different. I had 4 gigs of OCZ go out then in two computers at different times 1gig each of Kingston KVR. The 4 gig failure I found by chance while trying to install Windows. It wouldn't install right (would freeze up at random times before this too) so I ran memtest and it listed thousands of errors. Replaced it and all was well. A short time later I flipped to a computer and it was stuck in a reboot loop. Ran memtest with errors, replaced and all is well. This time I flipped to a THIRD computer and it was blank and wouldn't boot. Pulled a stick and it boots fine, put it back in and it complains of bad ram on bootup. This is all within about a month.

I guess it could be a fluke that I just happened to get bad batches from both companies, or maybe I got one bad batch and the other computer had a spike. I just read about EMI and I'm sitting here listening to the hummm of all the stuff in them. Collectively I have around 2,000w packed in a few square feet. But maybe I'm just freaking out.


GavinLeigh - I'll look into that company. Anything to stop this madness

 

[GavinLeigh]

That's a terrible story about the faulty memory. I remember there being a glut of cheap memory a couple of years ago and gossip on the boards was that generic memory companies had given up testing memory at the factory altogether. Mainly because it cost them too much time and money, but also because the cost to replace the items returned by a consumer was relatively cheap. In that case the end user becames the stress test. I don't know what the story is now... the last OCZ DDR800 I bought had a faulty stick and the packet looked already opened... I wasn't pleased about that.

I have also had memory that ran perfectly at a given frequency for say 12 months... and then just all of a sudden start erroring. I throttled the memory down about 10 or 15mhz and the errors went away. And it's been fine since. It sounds like you work you memory pretty darn hard... so maybe you basically work it to death.

But my feeling is that you are creating EMI within your own power circuit, and that is being transmitted to every machine in the room. Maybe through the air as RF interference but more probably as noise on your AC circuit. You need a friend with a good oscilloscope to look at your sine wave on the AC.

Sorry about your woes... I hope you have better luck in the future.

 

[Mondoman]

Gavin's exactly right about getting some power-conditioning equipment (not just UPSs). Also, assuming your PCs all have the normal metal cases, they are shielded from broadcast interference (though potentially not from interference through the power line, as Gavin noted).

 

[drjay]

Wow you guys are really helpful, I sincerely appreciate it.

So far it sounds like I have a few areas to address; clean AC power, heat, new APC. Here's what I'm thinking:

Line Conditioner:
http://www.gutwire.com/conditioner.htm
Totally new area for me, no idea what to look for. Will it go in-line with the UPS?

UPS:
http://www.tigerdirect.com/Applications/SearchTools/item-details.asp?EdpNo=770423&Sku=ULT31502
Do I need two? 3 of them are almost always at 100% and have 450w power supplies each. The 4th (server) only blips one of the processors every so often and is generally near idle.

Cooling:
Kind of at a loss here. All of the memory coolers I read about were not effective at all. The rest of my house is nearly freezing but even with the door open and two fans always on HI this room stays warm, not hot, but warm.

Thanks again for everyones help!

 

[maury73]

Absolutely not. EMI level produced by a PC, even with Spread Spectrum turned off, can't do any damage to any electronic circuit.
You would need at least 10,000 times the RF power generated by a PC for produce some little damage.

 

[GavinLeigh]

Okay let's forget electromagnetically created interference.... EMI or RF.

But could these machines be producing "unclean" power on the AC side? I have been told stories of fluorescent lights affecting systems.... even of transformers on desk lamps affecting things. I have used a computer compact fluorescent case light that produced such a hum on the audio outputs the speakers were unusable.

Brown out's, spikes, frequency drops.... they could all be the culprit. I think that a stable AC power supply is going to make a good starting point (for a full diagnosis). Hell I once had a PC that wouldn't post because it was on a 10 foot extension cord.

 

[CompTIA_Rep]

If fluoresenct lights caused real problems, schools would be in trouble lol. I just wanted to ask again: What did you use to test the memory and determine it was bad?

 

[audiovoodoo]

If your power really is bad it might be worth talking to your landlord / electricity supplier. If its flakey old wiring inside your building then getting a dedicated circuit installed might help matters. The electrical supply company is under an obligation to provide electricity within defined bounds. I have heard of cases where they will leave a monitor at your premesis to check for fluctuations. You might even be able to sue for your time in fixing the problems..

 

[GavinLeigh]

Some UPS units include software to monitor the power. Some even allow you to create log files of the input and output voltage, frequency, surges and the like. If your current units allow that then you might want to seriously look at that issue.

As AudioVoodoo said your power company is under an obligation and if you can take a print out to them and say "here" look at this... it's down to 92 volts here, and upto 118 here, then they might do something.

At my work we had similar issues and all of our UPS units would go into "boost" mode in the afternoon as the voltage dropped. The power company came out and replaced the cable drop from the power pole to the building and things have been fine since.

 

[drjay]

Welp things got even worse for me. Last night there was a pretty bad storm and the power shut off. The battery backups did their job for as long as they could but the systems shut down. That's all well and fine but I woke up to the power flashing on and off in about 2-3 second intervals. I freaked and ran in here to find my F'n server with the power lights flickering. It's the only computer set to come on with the power. The power was back on and steady but the computer was fried. Replaced the power supply but the fans only jitter a little. Even pulled the mobo out with same issue..goner. Mobo is still under warranty though, hope that's all that went. It's a VERY expensive computer so if more stuff is fried...mmmm...The other comps are fine.

Sooo I figurd I had enough of this crap. Soooooo pissed right now. Just went out and got this guy: http://www.compusa.com/products/product_info.asp?pfp=SEARCH&Ntt=apc+1500&N=0&Dx=mode+matchall&Nty=1&D=apc+1500&Ntk=All&product_code=298348&Pn=Back_UPS_1500_UPS

Just for the server. It's the best one CompUSA has.
I also have two of these:
http://www.compusa.com/products/product_info.asp?pfp=SEARCH&Ntt=back%2Dups+ES&Pn=Back%5FUPS%5FES%5F8%5FOutlet%5F350VA%5F120V&N=0&Dx=mode+matchall&Nty=1&D=back%2Dups+ES&Ns=display%5Fprice%7C1&Ntk=All&product_code=318854&Pn=Back_UPS_ES_UPS

That I HAD all of the comps running on. I also called and complained my head off to the electric company, they're sending a guy out. The 1500 APC has monitoring so I'll keep that running to see what happens.

 

[Mondoman]

APC makes good stuff.

You will likely find that the MB warranty doesn't cover damage due to power issues, though.

 

[GavinLeigh]

Well APC makes pretty good equipment, although you have to be careful to check what the unit actually does to the power. Not all UPS units are created equal and some only provide mediocre smoothing of the output power. The automatic voltage regulation (AVR) will help remove spikes and boost brownout conditions, but the unit may not provide frequency correction or clean the waveform up. But then those are premium features. There is a lot of difference between a relatively cheap APC and a serious server room UPS unit.

I really feel for your situation, and I hope that things improve with your new purchases.... it's gotta be a step in the right direction.

One tip... whatever you do don't mention "lightning" when you do that RMA for the motherboard. Lightning is an insidious killer of equipment and if they know the failure is due to the situation you described... well they would be within their rights to not honor the warranty.