Feed my curiosity: Why aren't all RAM modules 100% compatible with every motherboard?

[octoberasian]

So, I'm on the ASRock support forum (via Tweaktown) asking about RAM compatibility and this thought came to my head:

Honestly, it would be great if every memory module would work because with the way memory pricing is nowadays-- I bought those G.Skill Sniper 8GB kits for $43.99 back in 2012-- I'd like to get a brand I like that's also affordable, pop it in the board and have it work for once WITHOUT ISSUE.

Why can't it be that way? It's too much of a hassle having to return incompatible RAM or looking at memory compatibility lists hoping it'll work. It's like playing Russian Roulette with the RAM and hari-kari with your wallet.​

Given JEDEC standards compliance and memory ICs coming from companies like Hynix or Elipida or Samsung or other semiconductor memory manufacturers, you would think that there shouldn't be compatibility issues with motherboards.

But, it doesn't work that way?

Why is it even after JEDEC compliance and similar DRAM technologies we still have compatibility issues with RAM?

(Note: Disregarding buffered vs. unbuffered and ECC vs. non-ECC RAM as only certain boards and processors (with IMC) can accept these types of RAM.)

It sorely limits the kind of RAM you can install in your system and limits what affordable RAM you can buy at the same time. It both perplexes me and angers me at the same time.

 

[Dreamerbydesign]

I almost never use "recommended" or "compatible" RAM with my builds. I stick to standard voltage and decent timing ram. Never had a single issue with compatibility.

 

[drescherjm]

Why is it even after JEDEC compliance and similar DRAM technologies we still have compatibility issues with RAM?

Because a lot of ram that is sold is overclocked and is not JEDEC standard. I mean aren't the chip manufacturers producing 1.5V or lower DDR3 1600 CL11 but the module manufacturers are selling modules using these same chips at non standard speeds, timings and voltages.

 

[drescherjm]

I almost never use "recommended" or "compatible" RAM with my builds. I stick to standard voltage and decent timing ram. Never had a single issue with compatibility.

Same here.

 

[pcjunkie]

We just fried a Dell Precision T1500 at work because the person that ordered the ram thought the standard non-ecc 1.5v sticks would work. After checking back with Kingston it was confirmed that they weren't compatible. If they are Dell/HP machines always check to see if they are compatible. Standard OTC motherboards should be fine with most any ram now days, but not always.

 

[Skripka]

QVLs for RAM are a massive joke.

 

[cyclone3d]

We just fried a Dell Precision T1500 at work because the person that ordered the ram thought the standard non-ecc 1.5v sticks would work. After checking back with Kingston it was confirmed that they weren't compatible. If they are Dell/HP machines always check to see if they are compatible. Standard OTC motherboards should be fine with most any ram now days, but not always.

This story sounds fishy to me.

Were non-ECC sticks added to a system that already had ECC sticks in it?

There is no way that 1.5v sticks of the same type should have "fried" the system. Maybe cause it to be unstable, but fry???? Something else is going on here.

 

[cyclone3d]

QVLs for RAM are a massive joke.

This.

Anybody who thinks any company has the time/manpower/money to test every single brand/model of RAM in each and every single motherboard they make needs to get some common sense.

 

[daglesj]

I think the last time I had incompatible ram was...hmmm...around the turn of the century.

A Packard Bell desktop a friend owned that we got in a 128MB ram stick for needed specific gold edged ram for some reason. I avoided Packard Bell PCs from then on, a good rule for anyone I'm sure you will agree.

 

[octoberasian]

This.

Anybody who thinks any company has the time/manpower/money to test every single brand/model of RAM in each and every single motherboard they make needs to get some common sense.

Oh, I know no company would test every RAM out there. I rarely check QVL lists myself for motherboards and just buy RAM that's good quality and priced right. It just annoys me that even with standards in place that the RAM you just bought simply doesn't work. If I get a 1600 MHz or 1866 MHz RAM from Newegg from Company A, Company B at the same speed and voltage should work without issue. Or, if I get Model A at 1600 MHz and upgrade to Model B At 1866 MHz from the same company, they should still work on the same board.

Unfortunately, a lot of times both scenarios do not work that way.

 

[drescherjm]

This story sounds fishy to me.

Were non-ECC sticks added to a system that already had ECC sticks in it?

There is no way that 1.5v sticks of the same type should have "fried" the system. Maybe cause it to be unstable, but fry???? Something else is going on here.

A mix of REG ECC and UDIMMS? Although I do not see that frying anything.

Edit: Looking at the Crucial memory advisor the T1500 did not support registered or ECC memory so I have no idea how standard DDR3 UDIMMS were not compatible..

 

[drescherjm]

This.

Anybody who thinks any company has the time/manpower/money to test every single brand/model of RAM in each and every single motherboard they make needs to get some common sense.

On top of that what is the testing procedure? How many hours / days of testing before calling the ram good? How many motherboards do they test? How many sets of modules are selected for the test? Why don't they specifically list the RAM that failed the test? Then there is also the fact that QVL lists are usually done during the design of a motherboard and are rarely updated after the board is released when new ram modules are produced.

 

[Skripka]

Oh, I know no company would test every RAM out there. I rarely check QVL lists myself for motherboards and just buy RAM that's good quality and priced right. It just annoys me that even with standards in place that the RAM you just bought simply doesn't work. If I get a 1600 MHz or 1866 MHz RAM from Newegg from Company A, Company B at the same speed and voltage should work without issue. Or, if I get Model A at 1600 MHz and upgrade to Model B At 1866 MHz from the same company, they should still work on the same board.

Unfortunately, a lot of times both scenarios do not work that way.

This is like expecting to be able to mix different sizes of wheels on cars.

 

[octoberasian]

This is like expecting to be able to mix different sizes of wheels on cars.

Technically, it should work-- in my imagination, anyhow.

If I buy 1866MHz RAM that's from Kingston with the same timings and 1.5V as 1866MHz RAM from G.Skill, both RAM should work.

Or, if I buy 1600MHz G.Skill RipJaws RAM and move on up to 1866MHz RipJaws, they should work as well.

Unfortunately, in many cases both scenarios do not work that way. And, I wonder why.

There are standards in place and DRAM technology is the same going from Hynix to Elpida to Samsung ICs yet somewhere along the line one module works while another brand with the exact same speed, timings and voltage does not.

 

[cyclone3d]

Technically, it should work-- in my imagination, anyhow.

If I buy 1866MHz RAM that's from Kingston with the same timings and 1.5V as 1866MHz RAM from G.Skill, both RAM should work.

Or, if I buy 1600MHz G.Skill RipJaws RAM and move on up to 1866MHz RipJaws, they should work as well.

Unfortunately, in many cases both scenarios do not work that way. And, I wonder why.

There are standards in place and DRAM technology is the same going from Hynix to Elpida to Samsung ICs yet somewhere along the line one module works while another brand with the exact same speed, timings and voltage does not.

It is because of the differences in sub-timings.

Good luck getting two different model sticks of RAM that have the same exact sub-timings.

Even the refresh time can totally screw up modules working together and it is one of the higher level timings for DDR3.

You mix sticks with different sub-timings and you are just asking for trouble.

 

[cortexodus]

It is because of the differences in sub-timings.

Good luck getting two different model sticks of RAM that have the same exact sub-timings.

Even the refresh time can totally screw up modules working together and it is one of the higher level timings for DDR3.

You mix sticks with different sub-timings and you are just asking for trouble.

I agree with this assessment. I really try not to mix n' match anymore. Granted, it can be hard to anticipate the likelihood of filling all the slots on a mainboard, but I try damned hard to make sure if I do a later upgrade that I either replace or match with same model (which even then may not be peachy since PCB designs change)

 

[octoberasian]

By the way, not mixing RAM.

I'm more curious why if one brand works and I switch the RAM to another brand, why won't the second brand work as well?

As I said, if two RAM modules are from different manufacturers but same speed, timing and voltage, why would one work and the other wouldn't?

I have 4x G.Skill Sniper 4GB DDR3 sticks so I'm not mixing RAM.

But, let's say in a future system I'm using 1600MHz RAM from one manufacturer and upgrade all of them to another manufacturer that's the same speed, timings, voltage, but higher capacity OR faster speed (i.e.- 1866MHz or 2133MHz) and same capacity, the latter memory module from the other manufacturer should work.

Sadly, it doesn't work that way in every motherboard.

If I were to go from 4GB G.Skill sticks to 8GB G.Skill sticks, but have same speed like 1600MHz, voltage and timings, I would expect the 8GB G.Skill sticks to work without issue.

Or, if I replace them all with Kingston 8GB sticks for 32GB RAM total, the Kingston should still work too.

That's what I'm curious about. We have the JEDEC standards yet RAM incompatibility still exists.

 

[hcforde]

While not ram related this happened to/for me. Spring 2010 I answered an ad on craigslist for 2 Asus Eyefinity6 cards for $600. A company was selling them because the HP rep. told them if they put them in their new HP computer it would void the warranty. Dell & HP have strict guidelines for what ever crazy reason sometimes.

On the RAM issue it would seem like it would work or not on startup of the machine, that would be the first clue and it would not burn up. However, one of my ASUS motherboards(P7P55 workstation) has a button that will adjust the timing of the ram banks on board so any ram can be used in them. So with that thought in mind they understand that not all Ram is created equal.

 

[octoberasian]

So with that thought in mind they understand that not all Ram is created equal.

I was looking at my Newegg purchase history and I've returned a 1GB Kingston RAM and went with a Corsair ValueRAM instead of the same capacity. I believe that was the last time I had incompatible RAM and it was for an old Socket AM3 system.

If I were to strictly follow QVL from motherboard manufacturers, it really severely limits what RAM you can purchase for your system.

There has to be at least one RAM that would seem compatible regardless of platform, socket, and processor. On the ASRock support forums, one of the admins responded saying that compatibility issues started when IMC started coming into play especially after DDR3 came on the market. DDR2 never had this issue, or if it did, was very rare since the memory controller was still on the chipset. He mentioned that Samsung tended to have the best compatibility regardless of platform and processor, but aren't sold anymore.

I fear this is going to be the same issue when DDR4 comes to market when Haswell-E is released.

Maybe it's time to move the memory controller back to the northbridge/southbridge?

I guess, as you said, even with standards in place by JEDEC, not all RAM are created equal.

It would probably be the equivalent of installing an AC Delco alternator into your 1995 Maxima after replacing the Nissan branded alternator and finding out the AC Delco one doesn't work even if it was built and stated to work for your car.

Why bother with standards anyway if no manufacturer follows them strictly?

 

[dandragonrage]

Why is it even after JEDEC compliance and similar DRAM technologies we still have compatibility issues with RAM?

Because engineering is hard, and with each generation it gets harder (higher clockspeeds to support).


^That is a serious answer, btw.

 

[cyclone3d]

I was looking at my Newegg purchase history and I've returned a 1GB Kingston RAM and went with a Corsair ValueRAM instead of the same capacity. I believe that was the last time I had incompatible RAM and it was for an old Socket AM3 system.

If I were to strictly follow QVL from motherboard manufacturers, it really severely limits what RAM you can purchase for your system.

There has to be at least one RAM that would seem compatible regardless of platform, socket, and processor. On the ASRock support forums, one of the admins responded saying that compatibility issues started when IMC started coming into play especially after DDR3 came on the market. DDR2 never had this issue, or if it did, was very rare since the memory controller was still on the chipset. He mentioned that Samsung tended to have the best compatibility regardless of platform and processor, but aren't sold anymore.

I fear this is going to be the same issue when DDR4 comes to market when Haswell-E is released.

Maybe it's time to move the memory controller back to the northbridge/southbridge?

I guess, as you said, even with standards in place by JEDEC, not all RAM are created equal.

It would probably be the equivalent of installing an AC Delco alternator into your 1995 Maxima after replacing the Nissan branded alternator and finding out the AC Delco one doesn't work even if it was built and stated to work for your car.

Why bother with standards anyway if no manufacturer follows them strictly?

Not sure how much experience the ASROCK person that responded to your post has, but RAM compatibility issues have been around pretty much forever, even going back to 30pin SIMM days.

As for those compatibility issues as well as the Samsung LV RAM being very compatible, I am still, years after its implementation, highly suspicious that XMP is basically a pile of trash.

The Samsung LV RAM does not have an XMP profile, and the timings change automatically when running at different speeds.

I used some RAM with an XMP1.3 profile for a while and it never changed the timings when the RAM speed was changed.

I really dislike XMP profiles as it has always been this way, even on DDR2 systems. It makes it much harder to overclock since you have to manually adjust all the timings.

As for moving the RAM controller back to the chipset.... not going to happen. If they did that, you could expect about 1/8 to 1/4 of the throughtput of current systems.

 

[Skripka]

Not sure how much experience the ASROCK person that responded to your post has, but RAM compatibility issues have been around pretty much forever, even going back to 30pin SIMM days.

As for those compatibility issues as well as the Samsung LV RAM being very compatible, I am still, years after its implementation, highly suspicious that XMP is basically a pile of trash.

The Samsung LV RAM does not have an XMP profile, and the timings change automatically when running at different speeds.

I used some RAM with an XMP1.3 profile for a while and it never changed the timings when the RAM speed was changed.

I really dislike XMP profiles as it has always been this way, even on DDR2 systems. It makes it much harder to overclock since you have to manually adjust all the timings.

As for moving the RAM controller back to the chipset.... not going to happen. If they did that, you could expect about 1/8 to 1/4 of the throughtput of current systems.

The very first time I found XMP to work worth a damn at its job was LGA1155. Before that it ducked things up more often than not.

 

[larrymoencurly]

Given JEDEC standards compliance and memory ICs coming from companies like Hynix or Elipida or Samsung or other semiconductor memory manufacturers, you would think that there shouldn't be compatibility issues with motherboards.

But, it doesn't work that way?

Because module companies don't test very thoroughly and some exaggerate even the JEDEC profiles. Most modules are tested only in regular PCs, often in hot chambers, but the max temperature is way below the 85C guaranteed by chip manufacturers, probably only 58C (Corsair factory tour). And until about early 2012, one of the top 3-4 module companies proudly declared that it allowed only 1-2 errors in final testing for its 1600 MHz and slower modules, zero for its 1866 and faster ones (they used MemTest86+). Almost all modules rated 1333 MHz and some rated 1600 MHz aren't made from branded & guaranteed 1333 MHz chips but from rejected or untested chips, again not screened very thoroughly.

If XMP profiles are realistic, why don't the chip makers assign those values to their DRAM chips?

Did you ever wonder why those Samsung 1600 MHz low voltage modules were so highly praised? Because they were made from 1600 MHz chips, just like the better Kingston HyperX 1866 MHz modules and 2933 MHz Corsair Vengeance Pro.

 

[cyclone3d]

Because module companies don't test very thoroughly and some exaggerate even the JEDEC profiles. Most modules are tested only in regular PCs, often in hot chambers, but the max temperature is way below the 85C guaranteed by chip manufacturers, probably only 58C (Corsair factory tour). And until about early 2012, one of the top 3-4 module companies proudly declared that it allowed only 1-2 errors in final testing for its 1600 MHz and slower modules, zero for its 1866 and faster ones (they used MemTest86+). Almost all modules rated 1333 MHz and some rated 1600 MHz aren't made from branded & guaranteed 1333 MHz chips but from rejected or untested chips, again not screened very thoroughly.

If XMP profiles are realistic, why don't the chip makers assign those values to their DRAM chips?

Did you ever wonder why those Samsung 1600 MHz low voltage modules were so highly praised? Because they were made from 1600 MHz chips, just like the better Kingston HyperX 1866 MHz modules and 2933 MHz Corsair Vengeance Pro.

The Samsung LV modules also do not have an XMP profile and the timings change automatically as the speed changes. With XMP, the timings always stay at the XMP profile settings, which is a retarded way of doing it.

The Samsung modules act as all memory modules should act in regards to speed changes.

 

[jon666]

There has to be at least one RAM... Samsung tended to have the best compatibility regardless of platform and processor, but aren't sold anymore.

I believe some of the G.Skill Trident X use Samsung. Been hearing that for a little while now, and after a quick google search I found this http://www.overclock.net/t/1385526/gskill-trident-x-2400mhz-wont-overclock

Samsung is still out there, they are being sold off to other manufactorers now though I think.

Seems to me that a strong IMC is the biggest thing affecting RAM.

 

[lostinseganet]

Because motherboards are waiting for the holy memristor ram?

 

[drescherjm]

Because motherboards are waiting for the holy memristor ram?

I would highly doubt that. And since the memory controllers are on the CPU, new CPUs would have to be created to support this. I do not see this happening any time soon especially for consumer class stuff. I mean do 99.9% of desktop users need 100s of GB of most likely slower ram than we currently have?

 

[larrymoencurly]

I believe some of the G.Skill Trident X use Samsung. Been hearing that for a little while now, and after a quick google search I found this http://www.overclock.net/t/1385526/gskill-trident-x-2400mhz-wont-overclock

Samsung is still out there, they are being sold off to other manufactorers now though I think.

Seems to me that a strong IMC is the biggest thing affecting RAM.

G.Skill using Samsung chips is one thing, but G.Skill using them properly is another, and I've never heard of any chip manufacturer specifying an XMP profile for any of its DRAM chips, unless you count Micron's Crucial division.