Why matched pair of RAM

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
It has been recommended here and everywhere else to use matched pair of ram but it still puzzles me. Why this need if specs of the ram are same and are of the same brand or for that matter of different brands. Heck, why do brands sell matched pairs like say 2X4GB. If I m correct they are of the same batch so more likely to be identical; but what is the parameter that even manufactures do not have direct control on and how it may effect other tech specs.
Where in tech specs can ram modules differ if not in voltage, speed,timings,capacity,and not sure but may be no. of chips. I would like to know the real reason behind and not experiences you may have had. Thank you.
 

Toxxyc

Distinguished
Jan 6, 2011
969
0
19,160
Matched RAM sticks are used in order to allow the motherboard to run the RAM in dual or triple channel (if supported). This effectively means that you gain a little speed between the CPU and the RAM due to a higher 'bandwidth' so that more data can travel faster. It sounds a lot more impressive than it really is though, and even though the increase in performance is there, sometimes you can't even notice it. I'd still stick to it though, it's helped me a little from unmatched to matched sticks and the PC just feels a little more alive.
 



The reason is the exact same reason not all CPUs that come from the same piece of silicon wafer won't run the same speed. They are all tested, or "binned" before they are sold as the final product depending on how that particular item performs in testing, even though they all were made at exactly the same time, with the same process, to be identical. There are microscopic differences that affect how the particular circuit may perform alone, or paired. Obviously, if the memory comes from different manufactures, it can only compound the issue.

Motherboard manufactures even go to the length of testing memory on their boards, and supplying a list of approved memory modules that will work on their boards, as even not all motherboards will "like" just any type or brand of memory you slap into them, even if the memory is "specd" the same as a brand that has been proven to work. Just call ASUS and ask them why your new memory won't work, the first thing they will ask you is "Is this memory you have installed on our approved list of memory you can use on this board?" If it is not, they will simply say you need buy memory that has been approved to work on the board.
 
TBH, the best reason to buy matched pairs (and indeed say 1 set of 4 sticks instead of 2 sets of 2 sticks of the exact same kind) is just that it helps to maximally rule out RAM incompatibility as the source of potential problems.

There have been times when two different RAM types have been tested separately to work perfectly and together they just don't for no reason as far as anyone can tell. Buying 1 pack with all the RAM you want to have in it is the best way to try and avoid this sort of problem.

Unmatched pairs may work just fine 95% of the time. Indeed many years ago when I was still chained to DELLs I was offered $100 for 1 stick of 1GB RAM from DELLs online store and I was able to find a set of 2x 1GB sticks of PQi with exactly the same specs for $80. I ended up taking a chance with the PQi and it worked out just fine.

However, it could have also worked out horribly as well. I looked up reviews of people saying they tried that exact RAM in my exact DELL model and they all said it worked and none said it didn't so I took the risk and it paid off.

On the other hand, I could have been one of the dozen people every day coming in here asking why mismatched RAM doesn't work too.

That is just the nature of the beast.

Pretend, for the sake of argument that matched pairs work 99% of the time and unmatched pairs work 95% of the time and the difference is a RAM problem that will probably take at least 8, 10, 12 man hours to resolve. Is it worth it to pay the lower cost and go with the 95% vs paying more for 99%? Only the individual's budget can really tell you the answer.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510



@jitpublisher: Thanks for the analogy. However, CPUs tested to run at given frequency,will run at same frequency in every system(with identical base clock & multiplier on all mobos) and give same benchmark scores. Even if they are not identical(some r capable of higher overclocks) they can be used in servers together. So why cant RAM modules which also has transistors and is much less complex than a CPU also share this feature.
To be illustrative: lets say one module is used at a time, specs are same, brands not. Now say one can be overclocked to 2133Mhz and another to 1800Mhz individually, but still they don't have a good chance to work together at 1600Mhz than a matched pair each module of which, theoretically may not be capable of an overclock at all.


Another thing which ur post reminded me is why do individual mobo manufacturer tests memory, if practically every CPU has a memory controller. I mean what business do they have with memory nowadays, except for providing dimm slots and a data path and most probably providing power.
 
Motherboard makers need to care about the RAM used in their boards. They are required to warranty their own boards. If you put bad RAM in the board and blew up the motherboard, you might ask them to replace the motherboard which costs them money whether they say yes or no.

It probably costs them less to say no than to say yes, so having a QVL allows them to say no the maximum out of times for a defensible reason.

RAM production is one of those things where there are not super super high barriers to entry in the market. If someone can make a stick of RAM that works at some given quality level and they can price it really low then chances are they can find somebody to buy it even if the quality sucks. There are a lot of people out there on an extremely tight budget willing to chance it or enough people who aren't smart enough to choose good RAM who only care about the price.

Those people will take bad RAM if it is cheap if they can possibly do so.

For a lot of people, they view RAM as a commodity product where everything is visibly the same and the only thing to do is maximize your bang for your buck by minimizing cost.

In some ways it is and in some ways it isn't.

PSUs are the same way. Many view them as commodities and don't understand why they should pay more for the same wattage. There is a good reason, they just don't know it.

Yes, RAM internals are a lot the same regardless of maker, but different chips do have different properties. A 512 mb chip on a RAM stick that is made in Japan is different than a 512 mb chip on a RAM stick that is made in China. The chips made in Japan are pretty widely known to be better quality and higher price than the ones made in China are. That is just something people have come to accept as a fact of life for the most part.

Therefore the RAM cards that use chips from Japan will cost more and potentially be higher quality than those which use chips from China.

This matters because it applies most to the lowest common denominator, the chips from China. The people who are using unmatched RAM are maximally likely to be in the ultra low budget range.

If they were in a higher budget range, they could just buy 4 x 4 GBs of some kind and just throw away the 2x 4GBs of a different kind they already have and they wouldn't care about it.

The ultra low budget people (I am one of those), however, tend to be the most incentivized to get maximal usage out of what they already have and maximally minimize the cost of new expenditures in order to achieve a given set of goals. This often means they want to buy cheap RAM that doesn't match what they already have because what they already have is more expensive than something else that is available.

So you are really talking about a low quality chip from China being stuck in with another low quality chip from China made by some other company.

Whereas good brand + other good brand is maximally likely to work out of unmatched types, bad brand + bad brand is maximally unlikely to work. You get all the drawbacks and none of the advantages.

It just takes one tiny defect on one RAM stick to prevent a system from booting completely. That is maximally likely to occur with bad brand + bad brand.

Mind you its still quite possible that bad brand + bad brand will give you a set of RAM that has no defects. From the point of view of the manufacturer of a motherboard, though, they don't want to service calls like this if they can possibly avoid it.

All the motherboard maker can really tell you is to send it back to them, they will test it at their own expense, and they will ship it back to you at their own expense. It is in their interest to limit this to the greatest extent possible.

Quite often, if not most of the time, they just test things, figure out that there is nothing wrong with it, and they send it back with a note saying that. A lot of people just RMA things that aren't broken because they don't know what is broken and what is.

Your motherboard maker could probably think that your bad ram is the cause of the problem, but they don't really have any desire to try and test your bad ram. They would rather just tell you to buy something else if at all possible so they don't have to test it. Especially since bad RAM has a habit of working when you don't want it to and only not working when you do want it to work. They could test the sticks for 8 hours, find no errors, and send it back with a note that it is good when it isn't.

Really, having a QVL is the most pragmatic way that motherboard manufacturers can limit the amount of money that people on ultra low budgets can damage their bottom line, that is about all there is to it.
 
Ram is sold in kits for a reason.
Ram from the same vendor and part number can be made up of differing manufacturing components over time.
Some motherboards can be very sensitive to this.
Although, I think the problem has lessened with the newer Intel chipsets. Still,
it is safer to get what you need in one kit.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
@Raiddinn: For PSUs there are various protections, higher continuous power at "elevated temperatures" and higher overall power output which raises cost and is differentiator. Likewise for every single component in a PC. Hard disks do not come in matched pairs to be used as RAID 0, it will suffice if they are same model, and i haven't tried but maybe even different models. Even crossfiring is possible with slightly different GPUs.

This issue is with running dual (or triple) channel memory, and not only about generic ram makers, every reputed makers market twin ram modules. Why this is need? I mean is there something wrong with JEDEC? Each and every ram tested to work properly in single channel should work with same type in dual channel or should not be allowed to sell.
However the question still remains, Chinese make poor quality rams; what exactly is "poor" about them. if u ask about overclockability, definitely a yes. they may even fail early, but if working and benchmarked a DDR3 1600Mhz 1.5V with timings of 9-9-9-24 in single channel will fair equal to a Corsair or a Gskill with same specs. So what makes their use in dual channel a matter of luck, but not only for generic rams, even branded ones have uncertainty, to a lower extent of course, if not "matched".
What exactly differs in ram modules if they have identical tech specs? What is in "quality" that is not is tech specs, and what could go wrong that makes manufactures to ship matched pairs. I know this happens, I just dont know why?
 
There are not very many companies that make RAM from start to finish.

Micron is one of the companies that does.

If you go to Crucial's website www.crucial.com you can look down there at the bottom and see "Copyright 2011 Micron Technology Inc".

What happens is the Micron company makes say like 1 million of those little black chips on the RAM sticks. There are quite a few of them on one RAM stick so its less RAM sticks than it might sound like.

So Micron makes these things and Micron is one of the companies that tests every single part they make in QC.

Some makers test a quarter of them and then, if the error rate is within a certain rate (say, 99% are good) they will just ship the whole thing and pay to replace bad sticks the consumers receive with other ones. They figure the cost of replacing the sticks is less than the cost of trying to manufacture RAM chips that are 99.9% of the time within specs instead of 99%.

Those 1%s stack, so if you have a manufacturing run with a 99% pass rate and there are 8 chips on the RAM then each of them is 99% likely to work, for a RAM stick that is about 92% likely to work.

Crucial/Micron, though, tests absolutely everything. Every part that isn't within specs gets tossed into the "fail" bin. The ones that pass get sent to the next stage in the manufacturing process.

Also, I want to point out here that the number and quality of tests performed does matter. Some manufacturers like Micron run a dozen tests on each chip. Other manufacturers may only run 2 or 3 tests on each chip. Whereas Micron may test RAM chips at super high temperatures because this causes the chips to fail some % of the time, other makers may skip that test completely and just opt to replace RAM that makes it to the end user and fails.

Also, if given specifications say anything below 10 and above 20 is a fail in some metric, a maker can opt to fail anything below 12 or above 18. The reading of 10 could have been a 9.9 on the metric if the chip had been tested for 100,000 more hours and averaged out. If the maker puts the minimum at 12 they can be pretty confident that it would average above 10 if the chip were tested for 10 times as long or 100 times as long.

So what happens to all the chips that are in the "bad" bin?

Micron sells them to somebody else that wants them. The buyer doesn't have the capability to manufacture these black chips, but they do have the capability to put them on a board and use them for something if they work. It is easier to put them on a board than it is to make them.

The buyer knows that all of the chips they are buying are "fail" chips, however, they also know that some % of these "fail" chips they just bought from Micron are chips that were in that gap between the worst acceptable quality and the higher standard that Micron has set. Going back to the prior example, if Micron sets the cutoff at 12 but anything above 10 is OK, then there are chips in the fail bin that are rated at both 10 and 11 which are acceptable for use, even though they are below Micron's tolerance level.

The buyer may buy a million of these fails for pennies on the dollar. To get a million fails would require many different production runs because a lot of what Micron makes isn't a fail. Say, for the sake of argument, that it takes 3 runs of 1 million chips to generate 1 million fails (33% fall outside tolerance levels) that means that the buyer will be getting chips from 3 different manufacturing batches.

The buyer gets these little black chips at pennies on the dollar because a lot of them are actually bad just completely worthless, they have to buy those too just like they are buying the ones that got 11 (good but not within tolerance).

The buyer has to try to test each of these parts to determine if they are a "good bad" one or a "bad bad" one. They throw the "bad bad" ones in a fail bin and sell those again to someone else if they can and they keep the 10s and 11s. With those 10s and 11s, they put them on a board and sell them under their own RAM brand name.

Some % of the time the ones they tested at 10 that Micron also tested at 10 really are only 9.5 or 9.9 instead of 10 if they would have been tested 100x longer. Those chips will cause the RAM to fail, but not as often as if the RAM would have been constructed with "bad bad" chips.

The distributors who these companies sell to know that they didn't make the RAM chips themselves and they know that all the RAM chips this company uses were failed by Micron, so they won't pay the rates for the RAM from brand x that they would have paid for RAM from the Crucial brand. They know the quality is worse and they pay less for it in order to subsidize the losses they are going to take on replacements.

They may offer to buy the "good bad" RAM for half of what they would pay Micron for "good good" RAM.

The distributor then sells the RAM chips it has, some from Crucial and some from other companies using "good bad" RAM at a markup and the retail stores that get it sell it at a further markup.

The end result is that the "good good" RAM in the stores just plain costs more than the "good bad" RAM does. Micron charged a premium and so did everyone else in the chain all the way up to and including the retail store. At the same time, the rates were cut along the way for the "good bad" RAM because it was known all along that these were made completely with reject chips and the likelihood of the end product needing to be replaced is much higher.

We figured earlier that if 99% of the chips pass, then on an 8 chip RAM stick the pass rate would be 92% approx when it was used (assuming the board its on doesn't also bring its own failure rates). If Micron's QC sets the bar high for its chips (12 - 18 instead of 10 - 20) then maybe 99.9% of these black chips will work. That means the end result product is 99.2% likely to work (compared to just 92%).

BTW, did you wonder what happened to the "bad bad" RAM earlier? Do you think it is possible that a 3rd company might have tried to do tests on it and see if it could find any 9.1s to 9.9s and try to pass it off on someone else as a 10? If they were getting "bad bad" ram at pennies on the dollar compared to what Micron sold the rail ram to the second company for, do you think there is a potential to try and salvage some of those chips and sell them even knowing the failure rates would be astronomical in practice?

I hope now that you are starting to understand the differences in RAM after this explanation.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
Thanks Raiddinn for the detailed post.
What i got from above was:
1. There is higher probability of a generic RAM failing.
2. They are OK to replace their bad RAMs.
3. Those RAMs are not guaranteed to run a million cycles(say) and may fail earlier.
4. They may or may not work at elevated temperatures.
But :
1. Still a high probability of getting an OK RAM ( by ur parameters between 10 & 20)
2. As long as they are OK, they will fulfill their specs.

So,
If 2 bad ram ---will not work even in single channel
if 2 sticks found good---will in single channel.

If they are OK in single channel, why not in "dual channel"? Does dual channel have more stringent requirement. If that is the case, I dont know of any.

Two rams are tested to be OK in single channel, why matched pair ought to have higher probability of working together.
Or is it like, reputed makers have higher quality chips so they marked RAMs as "matched" and are willing to replace both sticks if only one is faulty so that one has not to hunt for which stick is bad. Due to higher overall QC they will still get less returns. They can sell it for slightly higher cost and for end users the advantage is less troubleshooting hours. If that is the case then there is no technical issues, but i am inclined to believe otherwise.

Why work in single channel with no problem and not work in dual channel?
 
There is no guarantee that dual channel WONT work with mismatched sticks.

Said another way...

Mismatched sticks might or might not work in dual channel mode.

When you buy a dual channel kit of 2 sticks, if it is from a high end manufacturer, that means they probably tested them in dual channel and proved they work in dual channel together before they packaged the sticks together and sold them as a dual channel kit.

Again, the brand matters, no name brands are less likely to do this sort of testing and more likely to just shove 2 sticks in a pack and call it a day.

If you get two sticks that are from different brands with the exact same statistics then it just might work in dual channel.

The problems mostly come when you try to match pairs that aren't exactly the same, like 9-9-9-24 and 9-10-9-27 or something like that. You are almost guaranteed to 1) have to run the RAM at a performance handicap, or 2) have to accept a certain amount of crashes as OK.

Basically, as long as both sticks are good and if both sticks are exactly the same in specifications, then it is somewhat likely they will work in dual channel mode.

Your computer might even enable dual channel mode with wildly different sticks and just crash a lot because of it.

It really depends a lot on the BIOS and the motherboard and stuff.

 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
@Raiddin: "The problems mostly come when you try to match pairs that aren't exactly the same, like 9-9-9-24 and 9-10-9-27 or something like that."
I thought so too. But then again "as long as both sticks are good and if both sticks are exactly the same in specifications, then it is somewhat likely they will work in dual channel mode." Why this "somewhat likely"?
1st stick----works OK in singe channel. All possible specs and details noted. (From CPU-Z or BIOS or whatever)
2nd stick---works OK in single channel. All possible specs and details noted.
Brand is not not known. All specs and other details identical.
Is this detail necessary and sufficient condition for a deterministic and not a probabilistic answer for the question:
Will 1st and 2nd stick work together in dual channel mode at their previous(identical) specs?
If not why not? If yes then why matched pairs are in market?

Appreciate ur effort and hope this is not getting annoying.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
I was missing the details mentioned here: http://en.wikipedia.org/wiki/Multi-channel_memory_architecture. Phew... so much of unnecessary effort.
From that link: Modules rated at different speeds can be run in dual-channel mode, although the motherboard will then run all memory modules at the speed of the slowest module. Some motherboards, however, have compatibility issues with certain brands or models of memory when attempting to use them in dual-channel mode....Several motherboard manufacturers only support configurations where a "matched pair" of modules are used. A matching pair needs to match in:
Capacity (e.g. 1024 MB). Certain Intel chipsets support different capacity chips in what they call Flex Mode: the capacity that can be matched is run in dual-channel, while the remainder runs in single-channel.
Speed (e.g. PC5300). If speed is not the same, the lower speed of the two modules will be used. Likewise, the higher latency of the two modules will be used.
Number of chips and sides (e.g. 2 sides with 4 chips on each side).
Matching size of rows and columns.
.......Dual-channel architecture is a technology implemented on motherboards by the motherboard manufacturer and does not apply to memory modules. Theoretically any matched pair of memory modules may be used in either single- or dual-channel operation, provided the motherboard supports this architecture.

So what I conclude from above is that culprit are the motherboard makers; so I have partial answer to my question. I still don't know why, as mobos don't have memory controllers. Can someone throw some light on this?
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
Here's some more after search. http://www.dslreports.com/faq/9231

Q: What is Matched or Certified Dual Channel Memory?

Companies like Corsair, Mushkin, OCZ, etc produce what they call "dual channel" memory, or Dual Channel Kits. These are sold in pairs, so for instance you might buy a 512MB or 2x256MB Dual Channel Kit, which consists of 2 sticks of 256MB DDR memory paired together by the manufacturer.

Companies don't just throw two sticks of RAM together to produce these kits, but they don't necessarily produce a totally different batch of RAM either. Testing or qualifying Dual Channel memory might involve something as simple as technicians booting up pairs of RAM in a Dual Channel motherboard and ensuring they work together under a set of conditions, or it could be more complicated, including so called "SPD" optimisation's and even chip selection (we're inclined not to put much trust in any of those claims ). For your purposes, you should assume that Dual Channel memory is qualified through testing as all companies will claim that every pair of Dual Channel memory is tested prior to being packaged.

Q: Will Non-Dual Channel Matched RAM work in my nforce2 motherboard?

It most certainly, will. As long as it fits the requirements of Dual Channel operation (two of the same types of memory, same size modules, speed, etc). Two modules of the same model/brand purchased from the same vendor at the same time is essentially as likely to work properly in a dual channel configuration as is a dual channel kit.

The ONLY thing you can lose by buying "single channel" memory for use in Dual Channel mode is that manufacturers may or may not provide support and replace your memory if it won't work in dual channel mode, whereas if Dual Channel memory fails to work in Dual Channel mode, the manufacturers will help you resolve the problem and possibly replace the memory to ensure proper Dual Channel operation.

There are instances of people using two different types of RAM together and have had no problems. You can't damage your motherboard or RAM just by trying to use two un-identical module in Dual Channel mode. But be advised that if the machine is unstable for any reason, it is entirely possible to corrupt your data upon operation of the machine. Mis-matched memory sticks in a dual channel configuration often produce unstable operation, so as with any new overclocking or upgrading venture make sure you have adequate backups so you can recover from a data loss.
 




Holy cow dude, how many more explanations do you need before you get a clue? It's not always the motherboard, it can be the motherboard, but more often it is the memory its self that is causing the problem.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510


Ya thanks. As if it helped. I questioned ur replies which I couldn't accept as it is. Does that make me dumb? Sorry, but I cannot believe everything that is said.
My question was specific. Raiddinn even wrote very elaborately but I did not get what I wanted, so I looked myself and found something which seemed more logical. Now u r saying that those sources are incorrect. I am not saying they can't be, they just seemed plausible. If u say they r wrong on what grounds do u base ur judgment. Lets re-frame the question for u:
What property or problem in memory makes two identically good memories to (maybe)not work in dual channel, but work flawlessly in single channel?
 
Somehow I don't think you will be satisfied by anything anyone says.

The fact of the matter is that there are a huge number of tiny tiny things that can go wrong and that number is the lowest with 2 sticks of RAM that come in the same package from a high end brand like Crucial.

The only reason to really even have this question arise is if someone tried to cheap out the first time they bought RAM and they want to do it again the second time they buy RAM. If that is the case, then the number of things that can potentially go wrong is quite high.

1) One or both of the RAM sellers lied about the capabilities of the RAM. Perhaps one really is 1333 RAM while the other is actually 1066 RAM that the seller managed to OC to 1333 in their 5 second test boot.
2) Both RAM sticks are probably coming out of bad lots anyway.
3) Environmental damage to chips on one or both cards because of the way they were treated prior to being inserted in the packaging.
4) Whatever other stupid thing you can think of.

All we here really need to know is that x can happen or is more likely to happen in situation y and how to take situation y and derive it to problem x. This isn't really a forum where silicon valley electrical engineers gather to discuss the finer points of 45 nm chip design or anything like that.

We are a practical bunch and I already spent like 5 hours trying to tell you everything there is to know about RAM and its not good enough.

I think you need to take your question to someone else if you want more than I gave you. I just don't think you will be satisfied until someone who actually designs these things for a living teaches you enough about it for you to have your own electrical engineering degree.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
Cool it down people. It was just a question. And just for info, I have an electrical engineering degree(sadly no dual channel architecture course there). And @Raiddinn "The only reason to really even have this question arise is"--- curiosity and search for answers.More than once have I seen some very good questions answered here. So I thought maybe I could get the exact technical details for the problem here.
Its nothing to get pissed off at if one is not satisfied with your answers.
And as a passing note do ponder over this--if two systems take same inputs, perform on them in what way you do not know, and perform same outputs.How will you differentiate between the two systems. You will appreciate my question better when you have an answer to this.
So once again no hard feelings and I cannot select a best answer from these.
 
I am a computer programmer.

I redesign systems all the time to have the same input and the same output except wildly different things that happen in the middle, so I CAN appreciate how different internals can be.

No hard feelings, but I am kinda sorry that I wasted my time in vain trying to educate you.
 

Gh0s7y

Distinguished
Dec 15, 2011
34
0
18,540

Maybe you can merge your posts into one post explaining all this about "Why RAM kits are better than individual parts", then your time won't be wasted as other people can learn from it.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510


It looks like you took it on your ego. You are a veteran here and people 'listen'when you speak. And you are like-"what does this newcomer knows about and who the f--- he is is to question my anwers". But its nothing personal.
And btw you mistook me again, I didn't want to appreciative the difference, I want you to find the difference without looking inside. In your case its possible as changing code can change the complexity of a program by say O(n) to O(logn) for same input/output which can be timed in no of cycles taken. So try it for something that is different from inside but exactly similar from outside over all allowed input range. PM me when you have the answer, I have a gut feeling, this is going to be interesting and heated(sic). I can give you my gmail id and we can chat it over.
Thanks everyone.
 
What have we learned here today? A lot of very good information that can be applied directly to the OP's original question.
Then we must also consider a theory as a side note, first proposed by Ron White in 2006, "........, But let me tell you something, folks: You can't fix stupid. There's not a pill you can take; there's not a class you can go to. Stupid is forever."
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510


And I know now how he looks likes-no how he sounds.
 


It doesn't have to do with ego it has to do with how many other people I could have helped in the same time that I was futilely trying to help you.

I explained many times now that this cheapy thing + that cheapy thing can have problems for this, that, and the other reasons and I explained why a lot of those things might happen (defects in manufacturing, environmental damage, etc) and I said that there is no guarantee they won't work in dual channel mode if indeed the sticks are both of good quality and neither manufacturer is doing shady things with labels and such.

You, as an electrical engineer, should be explaining all this stuff to me instead of the other way around.

Electrical Engineers need to know about Six Sigma and how it applies to manufacturing and transportation and all that stuff.

With manufacturing specifications in nanometers even a tiny tiny tiny flaw in a chip can throw off the internal workings considerably. The least flaws occur with the premium brands and with the ones that go through the greatest amounts of QC testing.

If you are talking about premium brand x working with premium brand y then there is no argument really. If you have enough money to buy 2x premium brand RAM from different makers then you have enough money to buy 2x premium brand RAM from the same maker too and there is no good reason not to do it.

If you are talking about subpar brand x working with subpar brand y then you probably don't have enough money to buy RAM that is guaranteed to work like it is supposed to and if that is true then you open yourself up to all the sorts of problems I previously described.

My wife refuses to allow more than the bare minimum to be spent on her computer and therefore she has off brand RAM unlike me. Her RAM sticks say they are 800 mhz and both sticks are from the same maker, same model, same exact part number and everything. They don't run at 800, though, they run at 667 mhz. Both individually and in combination her motherboard recognizes them at 667. If I wanted to get the 800 it says on the pack I would have to try to OC them.

Mind you she is fine with the RAM at 667, but that isn't the point. The point is that the package clearly says one thing and the BIOS clearly says something else. These are just the sorts of problems you face when you are trying to use sub par sticks together even if they are exactly the same brand and everything. You can expect more problems like these if you go as far as buying different brands while you are at it. One may really be 800 like it says and one may lie and say 800 when it really means 667, then the brands will have trouble working together in dual channel mode.

To me, it doesn't really seem like there needs to be an exhaustive listing of every possible thing that can go wrong at every step in the process. I think it is just enough to understand the main sorts of problems that could be experienced when trying to use mismatched parts.

Risk management is also something that I would expect an electrical engineer to be familiar with. In risk management it is important to decrease the number of variables in play if at all possible. This leads to more consistent results every time. The fewest variables exist when you get 2 sticks in the same package. I think most risk managers should be able to see problems coming when they are thinking about using two highly complex parts that have to be manufactured to specifications in nanometers that have "Made in China" written all over them.

I really don't think there is that much more that can be said about all of these things.

I would really like to give you what you are looking for but I just don't think it can be done. I also really hate leaving people with questions still unanswered and also the idea that I might have to spend 20 or 100 more hours to answer all of yours to the ultra exacting standards you are looking for.

On the one hand I just want to call all this bad and give up on it and on the other hand it feels like you aren't understanding what I am saying so maybe if I say it differently you will understand it.

There is just so much that can be different about RAM that you can't see looking at the label. If you had a microscope that could easily see inside every part at the nanometer level and you could compare every square millimeter of both parts and verify they are the same, then yes its probably going to work in dual channel mode, but I think if you had such a microscope and you could examine the sticks at that level then you would find that these things aren't exactly the same on every square millimeter and therein lies the answer to your questions.
 

raghwendra123

Distinguished
Mar 2, 2010
19
0
18,510
I had the same problem as your wife; with premium brands. Earlier I had Kingston hyperX ram 1600Mhz single stick. It ran at 1333Mhz. I had to overclock to make it run at 1600Mhz. Same case when I upgraded to 2X 4GB Corsair Vengeance 1600Mhz.But I knew that my CPU had support for upto 1333Mhz only, so either the memory controller or the main clock itself needed an overclock.I speculate that your wife's pc might not be overclock friendly. I don't know if that might be the case or the memories are just bad. Anyways, if they are bad, the point you made is what I said. If the memory is bad, it is bad even in single channel too(as you stated), it does not have to wait to be paired to become bad(if it does).

So far whatever you have specified can go wrong, will make that RAM bad(no single or dual channel, it is bad everywhere). And everything you have mentioned about QC, Risk Manangement, Six Sigma and what not applies to a lot. If one does not have high QC, more of its products can fail. If something is working within the specified parameters, it is "good", never mind that all other products from the same batch failed or was bad. This "good" product can be used wherever the same parameters are required, if may (or may not) fail if parameters are made more extant.
What you talk about being same at microscopic level, it cannot happen and is not expected to happen, that is why we have "permissible errors and we have standards". (Theoretically, the probability of finding two exact same sticks is zero. However if you specify a range,howsoever small, the probability becomes finite). As long as a product is within the range specified, it is guaranteed to work, though some may be more closer to the specified parameters and hence temptation to call them "better". But its either a pass or a fail.
A product that passes tougher parameter check (higher speeds and latency here) will work if the parameters are relaxed. A product which passes a relaxed parameter check is good to work at that parameter set. Thus both are guaranteed to work at relaxed parameter level. What you said earlier about good good ram and bad good ram gets covered here. If a bad good ram passes the parameters,it is good enough,if not, is faulty and needs replacement. What I know is the BIOS reads the SPD and assumes all info to be correct and run it at the same frequency, timings and voltage. If the RAM module and its SPD contents are different, problems can occur(even in single channel) and it is liable to be exchanged for a "good" stick. So if two RAM have same SPD contents and are "good" they will work good in single channel.And as I searched and posted earlier :
"Two modules of the same model/brand purchased from the same vendor at the same time is essentially as likely to work properly in a dual channel configuration as is a dual channel kit."

At this point, what I can guess as a reason for failure exclusively in dual channel may have to do with no. of ranks, or no. of rows and columns which are the parameters which are not specified and can change from period to period even for a premium seller( like as higher density chips arrives rank is reduced to 1 keeping capacity constant; just a guess). Generic ones might sell different combinations of those together, thus higher probability of them not working in dual channel properly.


And for Six sigma, etc try a mechanical,not an electrical engineer :p