2 Sets of ram going bad?

[Josie Wales]

Flawed chip designs. Most of the chips simply deteriorated from the heat and loads found during normal operation. Every chip mfg has been through this at one time or another.

[larrymoencurly]

Originally Posted by Josie Wales

larrymoencurly has stated at HardForum that his personal failure rate is based on "my own experience with branded vs. unbranded RAM chips, based well over 300 samples bought from probably at least 10 sources"(link) so it isn't merely a couple of bad sticks. He has even reported multiple problems with Kingston, which is the most reliable brand of memory sold. Very suspicious.

I don't see how Kingston can be the most reliable brand because some chip companies also make modules for the retail market. Also Kingston doesn't normally buy finished brand-marked chips but instead uses whole finished silicon DRAM wafers and slices and packages them in-house. According to a Kingston VP, some of their modules were tested just with PCs, others with the same expensive machines used by chip manufacturers, depending on the customer's specifications (some customers have their own such machines). I'll admit Kingston is a lot better than it was 5-7 years ago, but I haven't found their modules to be as overclockable as modules made with brand-marked chips, not that I overclock often.

Originally Posted by Josie Wales

...Having said that, I am very concerned about your excessively high failure rates as they may be indicative of poor hardware selection, incorrect system settings, and improper handling on your part. When your personal experience doesn't track with published results or the experience of the enthusiast community as a whole you know you are doing something wrong. As you saw in the Toronto study the majority of memory errors are system specific and continued after installing replacement sticks. What were you doing wrong?

Probably testing more thoroughly than most enthusiasts do.

I use the default timings and voltages, except when the BIOS won't set them right -- often voltage won't be raised, but some BIOSes choose timings faster than recommended by the motherboard manual when multiple modules are installed.

I'm fairly careful about not shorting or zapping chips. I usually don't wear an anti-static wrist strap, but when I don't I almost always work on a desk covered with anti-static wrap. Or maybe I'm subconsciously much more careful handling modules that lack heatsinks and have brand-labelled chips?


Additional Comment:

Originally Posted by Mr.Scott

Define 'unidentified' chips. Corsair, G.Skill, Etc. all use completely identifiable OEM chips made by Samsung, Micron, Hynix, Elpida, etc. The only difference is each manufacturers particular binning process. No manufacturer is going to accept sub standard OEM chips and put their name on them. That would be business suicide.

Unidentified = no part number clearly legible by looking at the module (without prying off the heatsink) that can be Googled to bring up the chip's data sheet at Samsung, Micron, Hynix, Elpida, etc.

It's the binning process some major companies use that worries me, especially when so many modules with XMP profiles fail even at their fastest JEDEC profiles. At PC17000 speed, what's more likely to work reliably, a PC17000 rated module with brand-marked PC10666 chips under its heatsinks, or a PC12800 module made of brand-marked PC12800 chips?

[kow_ciller]

Originally Posted by larrymoencurly

Unidentified = no part number clearly legible by looking at the module (without prying off the heatsink) that can be Googled to bring up the chip's data sheet at Samsung, Micron, Hynix, Elpida, etc.

That makes no sense.

So what you're saying is that if there were two sets of ram with the same IC. One of them has no heatspreaders, and the other one has heatspreaders. The one without them would somehow be better? This makes no sense.

Worst-case you can hop on a forum and see which IC is on a given set of ram. Or check the product code to see which IC is used. Lots of reviewers will pop off the heatspreaders on their review samples to let everyone know which IC is used (I know I do in my memory reviews).

[larrymoencurly]

Originally Posted by larrymoencurly

Unidentified = no part number clearly legible by looking at the module (without prying off the heatsink) that can be Googled to bring up the chip's data sheet at Samsung, Micron, Hynix, Elpida, etc.

Originally Posted by kow_ciller

That makes no sense.

So what you're saying is that if there were two sets of ram with the same IC. One of them has no heatspreaders, and the other one has heatspreaders. The one without them would somehow be better? This makes no sense.

Worst-case you can hop on a forum and see which IC is on a given set of ram. Or check the product code to see which IC is used. Lots of reviewers will pop off the heatspreaders on their review samples to let everyone know which IC is used (I know I do in my memory reviews).

OK, it makes no sense if the heatsinked module is rated no faster than it's brand-marked chips, but I've seen a some such modules that were rated faster than their chips, while I've never seen a no-heatsink module like that.

I wish more reviewers would do what you do.

[Josie Wales]

Face it larrymoencurly, you are worried about everything related to computer memory. One minute it is the test capability you are worried about, the next minute it is the test standards, the next it is grey market chips, then it is the markings on the chips, and next it is a suspicion that firms are using failed chips. In your mind everyone else is to blame for your abnormally high memory failure rates. Get real. When your personal experience doesn't track with published results or the experience of the enthusiast community as a whole you know you are doing something wrong. My advice - Stop it.

It's a lot more fun when you do things right. In the overclocking community we test, torture, overclock, and benchmark memory without difficulty or abnormal failure rates (except when on purpose ). It's fun this way. Forsake the dark side and join us. Then when you do fry a stick it's a well earned trophy.

[kow_ciller]

Originally Posted by larrymoencurly

OK, it makes no sense if the heatsinked module is rated no faster than it's brand-marked chips, but I've seen a some such modules that were rated faster than their chips, while I've never seen a no-heatsink module like that.

I wish more reviewers would do what you do.

Well, the thing is that most of the times the memory is under-rated in case there is a flaw in a given set of IC's.

Usually though they can run at a much higher frequency with added voltage and tightened timings. But this is where makers binning IC's comes into play. Some of them for example can fail at 1500 9-9-9 so it gets put at 1333 9-9-9 without heatspreaders as bargain ram, while some of it gets binned as top of the line ram and runs 2200 8-9-8 out of the box. Sure there might be a higher death rate but that is due to the binning process not being perfect. But with the way memory warranties are the worst thing that could happen is that you send them in for RMA and get a new set while running a cheaper set of ram for a week or so.

[larrymoencurly]

Originally Posted by Josie Wales

Face it larrymoencurly, you are worried about everything related to computer memory. One minute it is the test capability you are worried about, the next minute it is the test standards, the next it is grey market chips, then it is the markings on the chips, and next it is a suspicion that firms are using failed chips. In your mind everyone else is to blame for your abnormally high memory failure rates. Get real. When your personal experience doesn't track with published results or the experience of the enthusiast community as a whole you know you are doing something wrong. My advice - Stop it.

I think you're drawing that out too much. All I said was junk or overclocked RAM chips weren't reliable and weren't tested very thoroughly.

I can't find reliable information about RAM failure rates, except from Google's study (probably included only prime quality chips) and a person at HardOCP who mentioned an overall dud rate of roughly 10% for tens of thousands of modules his company bought (did include unknown chips). I don't know the failure numbers from the computer decorating/enthusiast community.

Originally Posted by Josie Wales

It's a lot more fun when you do things right.

That's why I reinforced the ceiling before installing the stripper poles.

[Josie Wales]

Originally Posted by larrymoencurly

I think you're drawing that out too much.

I don't. Let me explain why.

If you look closely at the Toronto study you will notice that the vast majority of errors ( over 90%) were concentrated in a small group of machines (20%) within each platform type. Although machines that exhibited uncorrectable errors had their memory replaced they continue to produce elevated levels of both correctable and uncorrectable errors. This is indicative of the role that system hardware, system management, and handling plays. Systemic issues lead to a higher percentage of transmission errors and hard errors. The uneven distribution of errors both within and across platforms along with the minimal impact of memory replacement show that faulty memory was the source of only a small fraction of the total reported data errors in the study. System generated errors were the primary source of the data errors. Garbage in means garbage out.

If the hardware in a home system is not built with high quality parts, known to be compatible, properly configured, operated correctly, properly located, given clean power, and properly handled, it can exhibit elevated data errors or premature failure. This is why we sometimes see individuals with freakishly high failure rates for parts like hard drives or motherboards and in your case a freakishly high error rate for memory. It seems very reasonable to sing that song from Saturday Night Live, "Oooh baby, what up with that? What up with that?" Give it some serious thought.

P.S. - Can I come over? I'll bring a twelve pack if I can watch your strippers.

[larrymoencurly]

Originally Posted by Josie Wales

If you look closely at the Toronto study you will notice that the vast majority of errors ( over 90%) were concentrated in a small group of machines (20%) within each platform type. Although machines that exhibited uncorrectable errors had their memory replaced they continue to produce elevated levels of both correctable and uncorrectable errors. This is indicative of the role that system hardware, system management, and handling plays. Systemic issues lead to a higher percentage of transmission errors and hard errors. .

But Google used only prime quality chips, while I had both prime and lower quality chips, and my memory test error rate always went to zero when I replaced the bad memory with good. So what systemic issues do you think I had? I swear I didn't rub the DIMMs against the cat for good luck.

[Josie Wales]

Come on now, don't be modest. You didn't try "well over 300 samples" in a single computer at one sitting. There is much more to your claims than you let on. How many generations of ram are we talking about (ddr, ddr2, ddr3)? Did you only buy ram that was validated by the motherboard manufacturers? How many times did you have repeat problems with the same ram SKU ("a dozen 512MB PC3200 Kingstons")? How many different hardware configurations? How do you make your hardware choices (PSU, motherboard, CPU, etc)? What is your overall hardware failure and replacement rate? Could your reckless ESD protocol have caused intermittent hardware issues? How did you differentiate data errors from ram failures? Etc. Usually there is an underlying behavioral cause for freakishly excessive failures. That's how you become an outlier.

[eatdirt40]

In case anyone is curious it seems the new slots are working well. No bluescreens since.

[Mr.Scott]

Originally Posted by eatdirt40

In case anyone is curious it seems the new slots are working well. No bluescreens since.

Nice. Are you going to RMA the board? I would.

[eatdirt40]

Originally Posted by Mr.Scott

Nice. Are you going to RMA the board? I would.

Yes I will once I do a new build.

[larrymoencurly]

Originally Posted by Josie Wales

1. Come on now, don't be modest. You didn't try "well over 300 samples" in a single computer at one sitting.

2. There is much more to your claims than you let on. How many generations of ram are we talking about (ddr, ddr2, ddr3)? Did you only buy ram that was validated by the motherboard manufacturers? How many times did you have repeat problems with the same ram SKU ("a dozen 512MB PC3200 Kingstons")?

3. How many different hardware configurations?

4. How do you make your hardware choices (PSU, motherboard, CPU, etc)?

5. What is your overall hardware failure and replacement rate?

6. Could your reckless ESD protocol have caused intermittent hardware issues?

7. How did you differentiate data errors from ram failures? Etc. Usually there is an underlying behavioral cause for freakishly excessive failures. That's how you become an outlier.

1. Yes, 300 modules, in a single computer, all at once, and it didn't help that world chess champion Gary Kasparov was pressuring us to hurry up. Yeah, that's the ticket!

2. DDR through DDR3, mostly DDR2. I excluded SDR and FPM because I never had problems with it (chips were always name brand, some ECC) and didn't try many samples. I didn't check officially validated lists much, at least not before purchasing the RAM, but some QVL modules did fail, while modules with brand name chips never did, whether or not they were on the QVL. But how can QVLs really be trusted when module companies change to different chips?

3. Probably a few dozen, mostly office Dell and Gateway machines and the budget DIY stuff from Fry's.

4. I buy the cheapest unused non-obscure brands, like motherboards by ECS, BioStar, MSI, and some Asrock. But friends had higher priced models from those companies and from Asus, Gigabyte, etc.

5. Probably around 3% - 5% a year, overall, and the stuff isn't updated much.

6. What wreckless ESD procedures? When I work on the hardware I'm barefoot, my table is covered with anti-static that's in nearly constant contact with my forearms, and I maintain correct hand-off continuity. Garfield The Lucky ESD cat is kept away.

7. By changing the RAM and noticing the errors don't reappear in that machine and sometimes putting the failed RAM in an identical motherboard and finding that errors crop up in it.