Samsung VS. Corsair memory

[Tullphan]

Currently in my laptop (Latitude E6510) I have 1x4GB stick of Samsung ram model #m471b5273ch0-ch9.
I can get a really good deal (basically 2 sticks for the price of one) right now from a seller for 2x4GB kit of Corsair ram.
Quality-wise are they about the same?
Thanks.

 

[modulus]

well keep in mind samsung manufactures memory while corsair purchases memory from a supplier, dresses it up and rebrands it.

i'm not saying that makes it worse in any way, just food for thought.


you won't notice a difference between the two brands unless speeds are dramatically difference and then you'd probably only notice on benchmarks

quality and durability-wise they're both fine. if one dies it's just random bad luck. i wouldn't say you'd be more or less likely to have one brand die vs. the other.

 

[larrymoencurly]

What brand chips do the Samsung modules have?

What brand chips do the Corsair modules have?

 

[Aehs01]

I have never had any issues with all the Corsair memory I have purchased over the years.

 

[bigddybn]

RAM is RAM is RAM.

Buy for warranty and customer support.

 

[silent-circuit]

RAM is RAM is RAM.

Buy for warranty and customer support.

This. It's all the same. Binning may be different, but still.

 

[larrymoencurly]

well keep in mind samsung manufactures memory while corsair purchases memory from a supplier, dresses it up and rebrands it.

i'm not saying that makes it worse in any way, just food for thought.
quality and durability-wise they're both fine. if one dies it's just random bad luck. i wouldn't say you'd be more or less likely to have one brand die vs. the other.

I'm saying it does make the Corsair worse, except by luck. After all, it's not as if RAM is RAM is RAM and the quality is all the same. Modules made with RAM chips rated prime quality by actual chip makers, which passed testing done by a $1M machine, are far, far less likely to give errors than the typical major brand modules with decorative heatsinks stuck over the chips, which were likely tested only with regular motherboards, and I've had a lot of bad Corsair in the past 3 years but not before.

 

[silent-circuit]

I'm saying it does make the Corsair worse, except by luck. After all, it's not as if RAM is RAM is RAM and the quality is all the same. Modules made with RAM chips rated prime quality by actual chip makers, which passed testing done by a $1M machine, are far, far less likely to give errors than the typical major brand modules with decorative heatsinks stuck over the chips, which were likely tested only with regular motherboards, and I've had a lot of bad Corsair in the past 3 years but not before.

Lies.

 

[iconiK]

What brand chips do the Samsung modules have?

Samsung.

What brand chips do the Corsair modules have?

Usually Hynix or Elpida, maybe Micron.

 

[Mike Clements]

Currently in my laptop (Latitude E6510) I have 1x4GB stick of Samsung ram model #m471b5273ch0-ch9.
I can get a really good deal (basically 2 sticks for the price of one) right now from a seller for 2x4GB kit of Corsair ram.
Quality-wise are they about the same?
Thanks.

FWIW, you do get a lifetime warranty with the Corsair.

 

[larrymoencurly]

What brand chips do the Samsung modules have?

Samsung.

What brand chips do the Corsair modules have?

Usually Hynix or Elpida, maybe Micron.

That's about all but one Korean chip maker and the Taiwanese makers. With Samsung modules the chips will be prime quality Samsungs that passed all factory testing, but with Corsair modules you can't be sure that the Hynix, Elpida, or Micron chips on them will be prime quality, and if they're covered with a heatsink or don't have chip company logos or full part numbers printed on them, there's a good chance -- maybe even a virtual guarantee -- that they failed some testing.

 

[larrymoencurly]

I'm saying it does make the Corsair worse, except by luck. After all, it's not as if RAM is RAM is RAM and the quality is all the same. Modules made with RAM chips rated prime quality by actual chip makers, which passed testing done by a $1M machine, are far, far less likely to give errors than the typical major brand modules with decorative heatsinks stuck over the chips, which were likely tested only with regular motherboards, and I've had a lot of bad Corsair in the past 3 years but not before.

Lies.

What specific evidence do you have?

Why did the Corsair factory tours reported by several websites, including HardOCP, show no $$$ memory testers, just ovens and motherboards (maybe equipped with PHD RST cards) running modules in them? I think I saw a high priced tester only in a report of a tour for a Taiwan company, probably ProMOS or PowerChip. The lack of good testing could explain why 1 out of the last 3 sets of Corsair I bought was defective.

 

[Mike Clements]

That's about all but one Korean chip maker and the Taiwanese makers. With Samsung modules the chips will be prime quality Samsungs that passed all factory testing, but with Corsair modules you can't be sure that the Hynix, Elpida, or Micron chips on them will be prime quality, and if they're covered with a heatsink or don't have chip company logos or full part numbers printed on them, there's a good chance -- maybe even a virtual guarantee -- that they failed some testing.

This is not correct. We don't buy "failed" ICs. And, we have our own internal procedures for testing and validating ICs before they are mounted on a module and also for testing modules after they are built and before they are shipped. Please try to be a bit more informed when handing out advice.

 

[djsi]

I am sure that quality wise they are about the same.Either brand will fit your needs perfectly.

 

[larrymoencurly]

That's about all but one Korean chip maker and the Taiwanese makers. With Samsung modules the chips will be prime quality Samsungs that passed all factory testing, but with Corsair modules you can't be sure that the Hynix, Elpida, or Micron chips on them will be prime quality, and if they're covered with a heatsink or don't have chip company logos or full part numbers printed on them, there's a good chance -- maybe even a virtual guarantee -- that they failed some testing.

This is not correct. We don't buy "failed" ICs. And, we have our own internal procedures for testing and validating ICs before they are mounted on a module and also for testing modules after they are built and before they are shipped. Please try to be a bit more informed when handing out advice.

Then why have my Corsair modules failed far, far more than my Samsungs? (from Crucial but whole modules made by Samsung with Samsung chips.) How do Corsair and Samsung testing differ?

What brands and models testing machines do you use? The photos from factory tour reports haven't shown anything like the testers offered used for well over $100K US on eBay. All I saw were automatic DIMM handlers that were way smaller (probably based on motherboards) than those machines and DIMMs manually plugged into ordinary motherboards (some in ovens) that maybe ran PHD RST cards. (I don't remember) That's not exactly first-rate testing.

Since when are prime quality DDR3 chips rated for > 1.5V recommended operating voltage? (not absolute maximum) Please try to inform me a bit better by providing a data sheet from Samsung, Micron, Hynix, etc. indicating that 1.65V is normal for any DDR3 chip.

 

[Mike Clements]

It's simple. Anything with a semiconductor can fail. That's why we offer a lifetime warranty. If you have a set of memory that won't run spec or has failed, RMA them. We'll gladly replace them.

We don't do everything under one roof so don't try to draw too many conclusions from pictures on the internet.

I can't give a lot of details on our or anyone else's testing procedures because I don't know. I don't care to know or need to know these things to do my job. It's pointless for me to try to debate procedures and machines on the manufacturing end of things. What I DO know is that we don't buy failed ICs. I would appreciate it if you would stop telling people that as it is completely untrue.

As for voltages, you can easily see for yourself that ICs that meet the DDR3 spec are spec'd to run their rated speed at 1.5v. OCing them requires more voltage. Every DDR kit we sell will run JEDEC spec at 1.5v.

 

[Teenyman45]

It's simple. Anything with a semiconductor can fail. That's why we offer a lifetime warranty. If you have a set of memory that won't run spec or has failed, RMA them. We'll gladly replace them.

It is quite true that Corsair reps will readily replace failed parts... and replace them, and replace them, and replace them, and replace them, and it looks like I will need the kit replaced again unless 1.68v is suddenly considered spec.

 

[larrymoencurly]

Then why have my Corsair modules failed far, far more than my Samsungs? (from Crucial but whole modules made by Samsung with Samsung chips.) How do Corsair and Samsung testing differ?

What brands and models testing machines do you use? The photos from factory tour reports haven't shown anything like the testers offered used for well over $100K US on eBay. All I saw were automatic DIMM handlers that were way smaller (probably based on motherboards) than those machines and DIMMs manually plugged into ordinary motherboards (some in ovens) that maybe ran PHD RST cards. (I don't remember) That's not exactly first-rate testing.

Since when are prime quality DDR3 chips rated for > 1.5V recommended operating voltage? (not absolute maximum) Please try to inform me a bit better by providing a data sheet from Samsung, Micron, Hynix, etc. indicating that 1.65V is normal for any DDR3 chip.

1. It's simple. Anything with a semiconductor can fail.

2. That's why we offer a lifetime warranty. If you have a set of memory that won't run spec or has failed, RMA them. We'll gladly replace them.

3. We don't do everything under one roof so don't try to draw too many conclusions from pictures on the internet.

I can't give a lot of details on our or anyone else's testing procedures because I don't know. I don't care to know or need to know these things to do my job. It's pointless for me to try to debate procedures and machines on the manufacturing end of things. What I DO know is that we don't buy failed ICs. I would appreciate it if you would stop telling people that as it is completely untrue.

4. As for voltages, you can easily see for yourself that ICs that meet the DDR3 spec are spec'd to run their rated speed at 1.5v. OCing them requires more voltage. Every DDR[3] kit we sell will run JEDEC spec at 1.5v.

1. Then why have so many Corsair DIMMs been defective right out of the box? Not to single out Corsair on this because I've experienced the same with Kingston, Mushkin, OCZ, and K-byte but only once with with modules made with prime quality chips, i.e., chips with the actual manufacturers identified. How can that be a coincidence?

2. Almost all retail DIMMs are warranted for life, regardless of quality.

3. Do you honestly believe that chips sold without identifying markings are as good as chips that carry the logo and full part numbers? There has to be a reason why module makers buy unmarked chips, and it's not just to print their own logos on them. I don't mind if they do, so long as they screen them as thoroughly as the chip manufacturers do for the chips sold with their logo and part numbers printed on them, but I've found no evidence that module makers do, except in the case of one Taiwan company that also makes chips. With every one of the factory tours shown, it seemed the module makers were trying to brag about themselves, so if they had an uber-expensive testing machine, why would they keep it hidden but instead show just simple DIMM loaders that fed mobos?

4. I have no way of knowing if the chips meet JEDEC specs at standard voltage and timings, nor did I know that JEDEC had established reliability standards for DRAM. So why do Corsair and other companies specify 1.65V for some of their DDR3 modules or overclock them when chips made for the proper timings are available?

 

[Jeremy C]

Here's my way of thinking when it comes to whose part to buy. Look at this thread:

For Samsung, you have a regular member here trying his best to slander Corsair to make the Samsung stuff sound better.

For Corsair, you have one of their public relations reps right here in case you run into any issues.

I think I know which product I would want (and have - 4 years and counting of JEDEC spec voltage on these old parts and no problems whatsoever). I've been running Corsair ram exclusively for about 8 years now with a dozen sticks in current use and have had 1 RMA in that time. Considering each module has been OC'd to its stable limit for most its life, I consider that pretty good reliability.

Unless you can prove that Corsair uses inferior products, all you're doing is slandering the name of a trusted company. If you had a bad time of it with your parts, then so be it. I won't touch another AMD chip due to the problems I had with a handful of them, but you won't see my saying they make bad parts or use inferior processes compared to Intel.

 

[larrymoencurly]

Here's my way of thinking when it comes to whose part to buy. Look at this thread:

For Samsung, you have a regular member here trying his best to slander Corsair to make the Samsung stuff sound better.

I'm not slandering Corsair -- notice I said that the problems with unmarked memory chips was NOT limited to Corsair. I also didn't say Samsung was the best, only that memory made with marked chips was better. I mentioned those Corsair and Samsung only because the OP did, and I didn't mean to single them out as opposite extremes of quality.

Unless you can prove that Corsair uses inferior products, all you're doing is slandering the name of a trusted company. If you had a bad time of it with your parts, then so be it.

"So be it" = defects are acceptable. Defects are unacceptable, especially when memory is so important and defects are easily avoidable.

I won't touch another AMD chip due to the problems I had with a handful of them, but you won't see my saying they make bad parts or use inferior processes compared to Intel.

As I said, I've had bad luck with about every brand of RAM made with unmarked chips, but RAM isn't like CPUs, where the vast majority of chips are prime quality. Instead most retail modules are made from chips bought on the UTT market, and Corsair was one of the companies that told me they use such chips (Kingston said they never did, but that was back when I was having much worse luck with Kingston than Corsair).

 

[Mike Clements]

1. Then why have so many Corsair DIMMs been defective right out of the box? Not to single out Corsair on this because I've experienced the same with Kingston, Mushkin, OCZ, and K-byte but only once with with modules made with prime quality chips, i.e., chips with the actual manufacturers identified. How can that be a coincidence?

2. Almost all retail DIMMs are warranted for life, regardless of quality.

3. Do you honestly believe that chips sold without identifying markings are as good as chips that carry the logo and full part numbers? There has to be a reason why module makers buy unmarked chips, and it's not just to print their own logos on them. I don't mind if they do, so long as they screen them as thoroughly as the chip manufacturers do for the chips sold with their logo and part numbers printed on them, but I've found no evidence that module makers do, except in the case of one Taiwan company that also makes chips. With every one of the factory tours shown, it seemed the module makers were trying to brag about themselves, so if they had an uber-expensive testing machine, why would they keep it hidden but instead show just simple DIMM loaders that fed mobos?

4. I have no way of knowing if the chips meet JEDEC specs at standard voltage and timings, nor did I know that JEDEC had established reliability standards for DRAM. So why do Corsair and other companies specify 1.65V for some of their DDR3 modules or overclock them when chips made for the proper timings are available?

FWIW, "UTT" was Samsung ICs. They were untested TCCDs which was DDR2 and that was about......5+ years ago IIRC. And, that meant "UNTESTED", not "FAILED". And, we tested them ourselves before we made modules with them.

As for why there are "so many defective Corsair DIMMs right out of the box", that is such a vague unsupported reference, I have no idea how to really answer it. We sell a LOT of RAM world wide and I mean a LOT. So, if you are pulling numbers out of what you see in enthusiast forums, you don't have any solid verifiable numbers to work with.

I can't comment on what other companies do or how they do it. As for JEDEC, we make JEDEC compliant memory but we also make OCing memory. To OC, you have to add voltage hence the 1.65v rating on some modules. As an example, no one is actually manufacturing an 1866MHz IC. You have to OC JEDEC compliant ICs.

I'm not really sure what you are driving at here. We don't buy or sell "failed modules".

 

[larrymoencurly]

I thought UTT chips came from virtually all RAM chip manufacturers, but the manufacturers didn't want their identities associated with them. I can understand chip manufacturers selling whole wafers untested, but why would they let sliced and packaged chips leave the factories untested, after spending so much money on their production, especially when chips that test 100% good can be sold at much higher prices? Wouldn't that give manufacturers a strong incentive to sort the good chips from the duds? But how can they sort them, except by testing all the chips? This is why I don't believe real UTT packaged chips exist, only duds that may be good enough if extreme conditions are avoided.

Bad "right out of the box" means the computer indicates errors the first time it runs the module, either by its speaker beeping a memory problem at boot or by a diagnostic reporting bad bits in anywhere from a few seconds to several hours. I used to find no defects with Corsair modules, but 3-4 years ago my luck changed, and ~30% of the newer Corsairs showed defects. This is the opposite of my experience with Kingston, which went from junk (the majority of my 512MB PC3200 DDR modules were bad) to very good.

I'm not including defects that occur when memory is run out of spec, and I don't overclock, undervolt, or overvolt, except to see if it causes errors or eliminates them. OTOH when the recommended voltage is 1.65V, I'd use 1.65V even when running at the JEDEC timings.

I said Corsair used failed chips (didn't meet all factory specs), not that it sold modules that had failed its own testing. OTOH that testing seems to be grossly inadequate (not just with Corsair but most retail brands). Otherwise how do you explain 5-10% initial failure rates reported by careful, knowledgeable customers who run only quality computers that are in good condition? It's as if the module companies are testing their memory with just ordinary motherboards, rather than with the much more expensive machines used by the chip manufacturers, or they're much more lax with their standards. As I said before, I've never found a defect in any module made with name brand chips, including minor module brands like Centon and K-byte, except with a PNY/CompUSA/Infineon with bad soldering.

Why is so much retail brand RAM overclocked even when run just at advertised specs? Why not instead build them with faster chips that meet those specs without overclocking?

 

[Mike Clements]

Otherwise how do you explain 5-10% initial failure rates reported by careful, knowledgeable customers who run only quality computers that are in good condition?

If you can show me any reliable documentation of this, I'd like to see it. And before you ask, NO, I am not going to reveal our internal numbers for failure rates but I will tell you that it is nowhere close to even 5%, much less your 5% - 10% you seem to be pulling out of thin air.

You are doing quite a bit of speculating here while offering zero proof of anything. You even said yourself that you had 1 out of 3 Corsair kits fail. This hardly qualifies as what you referred to as "a lot" of Corsair memory failing. 3 kits is hardly any type of proper sampling to determine a meaningful failure rate.

In my time as a Corsair customer then as an employee, I have had less than 5 modules (not kits) fail since about April 2002. And I have handled hundreds of our modules in that time. I've done all sorts of extreme benching and had servers and workstations with thousands of hours of use on them.

You have made assumptions about what you think might be under our heat spreaders. You also only presented speculation and conjecture about our internal testing procedures based on some pictures you saw somewhere. If you have an axe to grind, at least use some meaningful facts to back it up.

If you don't like Corsair, fine. Move on. But please don't come into a thread with a customer asking a legitimate question and piece together a lot of random numbers and words to support your obviously slanted and misinformed opinion.

We've obviously gone WAY beyond what the OP asked for here. I'm done with this thread unless the OP needs something else.

 

[larrymoencurly]

1. If you can show me any reliable documentation of this, I'd like to see it. And before you ask, NO, I am not going to reveal our internal numbers for failure rates but I will tell you that it is nowhere close to even 5%, much less your 5% - 10% you seem to be pulling out of thin air.

2. You are doing quite a bit of speculating here while offering zero proof of anything. You even said yourself that you had 1 out of 3 Corsair kits fail. This hardly qualifies as what you referred to as "a lot" of Corsair memory failing. 3 kits is hardly any type of proper sampling to determine a meaningful failure rate.

3. In my time as a Corsair customer then as an employee, I have had less than 5 modules (not kits) fail since about April 2002. And I have handled hundreds of our modules in that time. I've done all sorts of extreme benching and had servers and workstations with thousands of hours of use on them.

4. You have made assumptions about what you think might be under our heat spreaders. You also only presented speculation and conjecture about our internal testing procedures based on some pictures you saw somewhere. If you have an axe to grind, at least use some meaningful facts to back it up.

5. If you don't like Corsair, fine. Move on. But please don't come into a thread with a customer asking a legitimate question and piece together a lot of random numbers and words to support your obviously slanted and misinformed opinion.

6. We've obviously gone WAY beyond what the OP asked for here. I'm done with this thread unless the OP needs something else.

1. You want me to show mine but you won't show yours??? Yet you're complaining about me not providing proof. Isn't that a double standard? I'm counting sales receipts. Even Google said 8% of their DIMMs gave at least one error annually, and I assume they've used only brand name RAM chips. OTOH considering that you have much more access to RAM failure statistics than I do, not just for your company but also for the industry overall, what are the industry's numbers, both for no-name chips and branded chips?

2. That's a rate of 1 out of 3 Corsairs failing, not just 1 failure out of 3 tested. Why would I say a lot of them failed if I had tested just 3 modules? OTOH I did mention trying only 2 Mushkin modules and seeing both of them fail, remember?

3. What tests did you run on those hundreds of modules that resulted in only 5 modules showing failures, and what hardware did you use? Some diagnostics can run day and night and not detect any problems (DocMem), while a properly designed diagnostic may report bad bits within a few minutes or hours (Memtest86/86+, Gold Memory). Also how did you test for internal pattern sensitivities that require the diagnostic to know the layout of the data bits on each chip?

4. So why do modules with name brand chips and no heat spreaders test better for reliabilitly than modules with no-name chips and heat spreaders?

5. I have nothing against Corsair in particular, only the common practice of most of the retail RAM industry of using no-name chips but not screen them adequately.

6. I simply told OP, who asked about Corsair vs. Samsung, to prefer Samsung because the Samsung modules were made with branded chips while the Corsair ones were probably not. From that, some people falsely accused me of being a fanboy of Samsung and as having a vendetta against Corsair.

 

[iconiK]

1. You want me to show mine but you won't show yours???

Company secret likely, might need authorization to reveal.

Yet you're complaining about me not providing proof.

Last time I looked, the plaintiff (aka the one making accusations) provides proof. You don't go accusing people or companies and except them to do everything to defend themselves, when you have ZERO reliable sources. I trust Corsair's employees a lot more than some random guy on a forum.

I'm counting sales receipts.

Whoose? From where? A retailer? Link to statistics please. If yours, how many samples? If it's less than 100 it's totally meaningless.

Even Google said 8% of their DIMMs gave at least one error annually, and I assume they've used only brand name RAM chips.

Where? I can't find that.

2. That's a rate of 1 out of 3 Corsairs failing, not just 1 failure out of 3 tested. Why would I say a lot of them failed if I had tested just 3 modules?

How many have you tested then? And did you source from multiple retailers to avoid bad batches?

3. What tests did you run on those hundreds of modules that resulted in only 5 modules showing failures, and what hardware did you use? Some diagnostics can run day and night and not detect any problems (DocMem), while a properly designed diagnostic may report bad bits within a few minutes or hours (Memtest86/86+, Gold Memory).

I expect a memory manufacturer to have industrial devices that handle this. Maybe running batch memtest, but it could be some specialized software. It's a company secret and will likely not be revealed. They may not be as through as something done by Samsung or Micron or Hynix, but it's certainly adequate.

4. So why do modules with name brand chips and no heat spreaders test better for reliabilitly than modules with no-name chips and heat spreaders?

Where is it said that they do test better?

5. I have nothing against Corsair in particular, only the common practice of most of the retail RAM industry of using no-name chips but not screen them adequately.

Prove that is it common practice to ship products without adequate testing. You're pulling that out of your ass.

6. I simply told OP, who asked about Corsair vs. Samsung, to prefer Samsung because the Samsung modules were made with branded chips while the Corsair ones were probably not.

A label on a chip means little. It matters who made it and Corsair likely sources Samsung for most of their chips, with Hynix, Micron and Elpida for their high-end stuff. Just speculating though.

 

[larrymoencurly]

I'm not accusing anybody, just citing my own experience with branded vs. unbranded RAM chips, based well over 300 samples bought from probably at least 10 sources, including the major national store chains, NewEgg, Dell, and Crucial. There was nothing noticeably different about any particular batches, except in the case of about a dozen 512MB PC3200 Kingstons, where the country of origin and chip markings were good predictors of defect counts.

Google's study of ram failure is a widely cited paper, and Googleing "ram failure rates" will return several hits, even on the very first page. Here's the link:

http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf

I don't see how anybody with Internet access could fail in an attempt to find this. Exactly what did you search for?

I expect a memory manufacturer to have industrial devices that handle this. Maybe running batch memtest, but it could be some specialized software. It's a company secret and will likely not be revealed. They may not be as through as something done by Samsung or Micron or Hynix, but it's certainly adequate.

In the case of Corsair, it's just an RST card -- a $700 company secret:

http://teamau.net/5

Please cite proof that the testing done by module makers is "certainly" adequate. The main reason I say it's not is because I've had over 30x as many bad modules made from unidentifiable chips as from prime chips. Furthermore, why would chip manufacturers go to the trouble of buying extremely expensive test machines if the 99.9% cheaper ones used by the module makers were "certainly" adequate?

Yes, I am saying that the label on the chip matters a great deal, more than do the circuit boards, soldering, or handling, probably because the best and worst modules are all made from the same parts (except for RAM chips) and on the same production lines. And if chip quality didn't matter, I'd always just buy the cheapest memory from major module makers that offered lifetime warranties and good service.

 

[Spooony]

RAM is so common it is actually traded as a commodity.

 

[iconiK]

I'm not accusing anybody, just citing my own experience with branded vs. unbranded RAM chips, based well over 300 samples bought from probably at least 10 sources, including the major national store chains, NewEgg, Dell, and Crucial.

That's good. I assumed you were just another wiseass who bought ten DIMMs and has bad luck, because that's what happens in most cases. Thanks for proving me wrong, that is indeed a pretty good sample.

Google's study of ram failure is a widely cited paper, and Googleing "ram failure rates" will return several hits, even on the very first page.

I found that study, but it is kinda old - DDR2 tops, thus it may or may not be relevant. Probably the latter.

In the case of Corsair, it's just an RST card -- a $700 company secret:

Still, I doubt the employees are allowed to say on a public forum, no matter the material value of the company secret.

Please cite proof that the testing done by module makers is "certainly" adequate.

I haven't found anything solidly conclusive so far. Mostly because it's fine for average Joe, which is why everyone sells untested DIMMs and do their own testing on them.

Furthermore, why would chip manufacturers go to the trouble of buying extremely expensive test machines if the 99.9% cheaper ones used by the module makers were "certainly" adequate?

Cause they aren't adequate for their larger market - the enterprise space. That is where manufacturer-tested chips are absolutely required. It's why buy the exact same DIMM directly from Micron instead of their Crucial retail brand shoots to a two to three times increase in cost (judged by looking at Supermicro Micron DIMM prices vs Crucial DIMM prices). While 48 hours memtest might be adequate for Joe's Facebook browsing sessions, it certainly isn't for a five grand server.

Yes, I am saying that the label on the chip matters a great deal... And if chip quality didn't matter, I'd always just buy the cheapest memory from major module makers that offered lifetime warranties and good service.

Sure, it matters, but for whom? For Joe? Probably not. For businesses? Hell yeah! For you? It depends on how much you care.

 

[larrymoencurly]

It would be one thing if only 1 in 1,000 DIMMs was bad and testing showed no errors for 48 hours, but it seems that more like 1 in 10 or 20 DIMMs is defective, and the errors usually show up in 2 seconds to 2 hours of testing. That can't possibly be good enough for even the average Joe. Also in most cases I've seen where a computer randomly locked up, rebooted, BSODed, etc., even at a rate of just every several weeks, and nothing was found wrong with connectors, power, capacitors, or software, the memory failed testing (always memory with no-name chips or heatsinks over the chips -- coincidence?), replacing it cleared the problem for good. (coincidence?) I have a feeling that many memory problems are instead blamed on malware or Windows.