QVL - Myth, Legend, Marketing/Advertising, what is your take?

[Contra]

I'll add that people return "non-working" memory under warranty a lot, especially TG and GSkill. And very often it turns out to be the inability of the motherboard or processor to run XMP... well, that's also why more responsible manufacturers write more than one profile

 

[Bill_Bright]

Apologies for replying after I said I was done. But I was quoted so feel I must reply to the inaccuracies that followed.

The problem is, the QVL list has been proven to be useless. (Correct?)

No, not correct.

As is now typical, you only listen to those with the same opinion as you, and can't tolerate and therefore ignore, and/or disparage everyone else.

Nearly 80% of the poll responders have indicated the QVLs help. That clearly indicates they are not useless to the overwhelming majority.

Shit in bold has nothing to do with QVL lists.

Sure it does. I even explained how. You just don't get it - or as I said above, can't tolerate and choose to ignore it because it does not align with your own opinion.

Even your own QVL list is a joke.

Huh? And as is also now typical, you throw out nonsensical stuff to obfuscate the issue. No one, including me, has created our own QVL.

Why can't you simply say that, for you, QVLs are not useful? Why must you try to force your opinion on everyone else as the only possible option?

Why can't you accept that for some (most it seems) QVLs are useful? I have totally accepted, as I said multiple times before, that for many advanced, experienced, enthusiast builders/upgraders, QVLs may not be useful.

It really is not hard, or should not be hard to accept that opinions opposite ours can still be right too.

Now - to stop ShrimpBrime from degrading sneekypeet's thread further, he can post his inevitable, must-get the last word in. I am moving on for good now.

Have a good day.

 

[thestryker6]

I think QVLs have both gotten better and worse over time. Better in terms of being updated more often and generally providing a little more detailed information (how many modules the speeds work with etc). The place where they're worse is that all the manufacturers are clearly using top bins to get the results. If you don't know what a CPU is generally capable of running it would be easy to fall into the trap of buying something that will never work at its rated speed.

For my own builds I think my current one was the first time I've used a QVL to pick a kit. That's because I just wanted a kit that just drop in worked while I waited for CUDIMMs to drop in price. I did look at the QVL for my W680 setup because I wanted to make sure the ECC memory was compatible, but none of the modules listed even existed on the open market.

When giving someone else build advice I tend to stick to QVL unless they're knowledgeable about memory tweaking. One other place I've found it can be beneficial is running XMP kits on AM5 as sometimes it seems like some boards just aren't properly designed to.

Overall though I mostly just view it as a general guide as to what may work on any given motherboard.

 

[TheLostSwede]

I can sort of answer this, as I've written about this before, but a lot of people don't seem to believe me, despite the fact that I have lived in Taiwan for 15 years of my life and been to the head quarters of all the motherboard makers except ASRock. Yes, that includes DFI, ECS, EVGA, Biostar, Jetway and more. I have also asked some of them about this in the past.

The QVL isn't marketing from the motherboard makers side for sure, unless they make their own DRAM of course, which I guess only Gigabyte has done and it wasn't widely available afaik.

However, it most definetly is marketing from the memory module makers, as they pick the kits they send in to the motherboard makers, especially during the early days of a new platform.
In fact, the motherboard makers buy very few of the kits in the QVL, as this is only done in rare cases when they have to test a specific kit they don't have, due to complaints from users to support.

Another thing to keep in mind here is that not all memory kit makers have a working relationship with all of the motherboard makers, so some brands would be missing entirely from a QVL, but might work just fine at the end of the day. Some of this is based on the fact that there are memory makers that do branded kits for one or the other motherboard maker, then the others are less inclined to work with them.

Yes, the QVL can be an ok guide, but ultimately, it's just the memory kits at hand at the motherboard maker and it doesn't mean other modules won't work or work better than the ones in the QVL. It's important to note that the QVL only applies to the same testing conditions as the board makers used as well, i.e. same CPU model, UEFI release etc.

Also keep in mind that the motherboard makers generally only tests with the default settings or the various profiles available and won't try any kind of tuning/tweaking. If they have an in-house overclocker, that person might do such things, but those results are not posted on the QVL.

I should also add that some memory makers have their own QVL's where they test different motherboards with their RAM, which might be different from what the motherboard makers QVL for the same board.

 

[freeagent]

good to know

 

[TheLostSwede]

I tell everyone who is a new builder that it is basically Legos and really hard to mess things up now. No longer do you have to read a manually or move dip switches.

Dip switches? What kind of luxury hardware did you grow up with.
Moving jumpers on pin-headers was the way to do it back in the day...
Sometimes it still didn't do what it was supposed to do. No, I do not miss those days.

 

[Caring1]

That board was akin to an LGA 775 AsRock board which had agp and pcie.

That reminded me of the Asrock Twins, that could run DDR2 or DDR3, haven't thought of that board for ages.

 

[Waldorf]

not sure why anyone cares to go thru QVL for perf.
at least for any ryzen build the MB page list all that on the bottom, not gonna dive into pages of kits what i can find in seconds.

@eidairaman1
maybe something from Patriot. always find most of their kits have decent perf and dont look like designed with "gaming" in mind.

 

[bug]

Missing from the poll is the real reason QVLs exist: cost cutting. Mobo manufacturers can drop many problematic service request simply invoking QVL "we're sorry, you didn't buy the right kind of RAM". Between JEDEC which says pretty clear how RAM is supposed to work and XMP/EXPO which further define that, QVLs should actually be illegal, imho, as they only state "our mobo doesn't work according to standards, but we believe it works with these sticks".

 

[Waldorf]

@bug
QVL means qualified vendor list.
not sure where this mentions anything regarding what standards/believe is involved.

ignoring you have other things doing virtually the same, like cars being listed with just a few wheel/tire (size) combos. doesn't mean you cant install other sizes.
fact is, XMP/AMP/EXPO is ocing, and not guaranteed, even IF the kit is on the list, so they can just say stop overclocking ram, without even blaming it on unverified ram.

 

[lilhasselhoffer]

I wonder if people think computers are like Lego.

Put the rectangle in the rectangle, match that up with that

That's a lot of pins I hope it goes in this way

You...might be surprised. Had youtube on in the background the other day, and it fell to a "my first PC build" video. They described it as electronic legos that are more fun if you build them yourself. They had issues with the CPU bracket, issues with getting the SSD loaded into the slot with the spacers in the box, had issues with reading about how the front panel worked, and topped things off by powering the system on with the HDMI port plugged into the motherboard while complaining about such poor performance. Comments ripped them a new one over choosing the RAM with RGB lights that was not cheaper than one with better timings, the ability to open the manual to read memory placements but not to go into BIOS and set them, and absolutely not reading about the video output. This was professional video games journalist levels of bad, and somebody put it onto youtube as a point of pride.


With that in mind, I think the QVL is a fantastic trouble shooting tool. If this were early AM4, I'd call it something that could mean the difference between booting and blue screening for no apparent reason...and definitely the difference between stable and screwy behavior. That said, both AMD and Intel worked out the memory issues a long time ago, and I've only seen failures on non-jedec memory configurations in the last several years. I will adamantly suggest that if you have to touch 1xxx to 3xxx processors you double check the QVL...but anything more recent is more likely to be operator error.

 

[Waldorf]

@lilhasselhoffer
seen to many times where i build 5xxx on x570, and the recommended gskill kits didnt work at AMP, some not even at all, while the cheap/rgb corsair, most were loathing and saying to stay away from, worked as expected. all on qvl.

the other way around, any quality kits (e.g. known to have b-die/equivalent) all worked fine, most better than AMP profile, and at least for the MSI boards i used, even auto (AMP) was stable.

 

[lilhasselhoffer]

@lilhasselhoffer
seen to many times where i build 5xxx on x570, and the recommended gskill kits didnt work at AMP, some not even at all, while the cheap/rgb corsair, most were loathing and saying to stay away from, worked as expected. all on qvl.

the other way around, any quality kits (e.g. known to have b-die/equivalent) all worked fine, most better than AMP profile, and at least for the MSI boards i used, even auto (AMP) was stable.

Anecdote.

My anecdote counters yours (plenty of DDR4 that did not play well with AM4 back in the day, but no problems after the 5xxx series). This is why anecdote sucks...yours is as good as mine. Neither is solid truth, and both is.

What sucks worse is that memory doesn't remain with the same chips, so I've heard dark rumors about an ancient QVL having sticks that decided not to work despite being on the list...because the manufacturer decided to use a batch of chips that were cheap as...well, chips...but subsequently refused to work with some systems.

Ain't life just grand? Confusing, contradictory, but always worth having a laugh over. RNGesus and Lady Luck are a fickle lot.

 

[Waldorf]

not really, unless i misunderstood the meaning (short (funny) story about person/place).
for those that i build it was a fact, and nowhere did i state this was the rule or statically relevant, and only in response to you claiming operator error, if it didnt work on 5xxx chips.

 

[lilhasselhoffer]

not really, unless i misunderstood the meaning (short (funny) story about person/place).
for those that i build it was a fact, and nowhere did i state this was the rule or statically relevant, and only in response to you claiming operator error, if it didnt work on 5xxx chips.

Maybe a reminder on what an anecdote is. An anecdote is personal experience, not constituting a significant enough quantity or without supporting evidence to support a conclusion. It is generally not an acceptable argument, because without significant enough sampling to rule out biases, your data is not acceptable as factual.

Read: It's a fact to me, but I cannot prove it.

Read also: 10% drop rates in video games. 10% of people get it round one. 0.9^100 or 2.66*10^-5, or 1 in about 37,649 do not get that 10% drop in 100 draws. Anecdotally the former person sees a 100% drop rate, and anecdotally the later person has a 0% drop rate, but because neither of the two add to a significant sampling size their experience is anecdote. Note I said "...I've only seen failures on non-jedec memory configurations in the last several years..." It's like I'm aware that your facts and my facts aren't more or less valid, because both are anecdotes and thus not statistically significant.



If instead you listed a few dozen specific instance we could maybe start somewhere (I offer that to you, because I've got only about a dozen, and do not meet better than anecdotal requirements)...but that would be assuming some controls. I...don't think we should go down that path. Also note I said both of our facts are not inherently facts, but anecdotal evidence. That is reality...not something I've used to decide one or the other is better.


As a side note, I'm usually that person who looks at 10% of tickets as winning and can buy 100 without ever seeing a winner. Having lived most of my life without luck I've found laughing at ones self is the only way to stay sane. Also, it's why I secretly enjoy seeing the silicon lottery hit others as hard as it does me...which is why I personally prefer the QVL and things like it. My first 1700x was...about as stable as a bipolar, one legged, drunken man inside a bee filled costume, with severe sleep deprivation. Everything I've gotten in the 5xxx series, multiple 5600x, 5700x, and a 5500 have just worked. Crappy no-name memory clocks to detected jedec, and the only crashes are software. That's maybe 8 CPUs...or so statistically insignificant it's funny. On the other hand, a 3930k was replaced twice, barely ran at stock memory frequencies on one board, and only on the second did I ever get the thing to run (under water) at anything less than nuclear reactor levels of temperature (100C reported at a junction I could never find, and after 9 year was probably the death of the system...with almost no overclock (did have to goose memory voltage to populate all 8 slots). Anecdotally, my Intel experience sucks...but I'd never state it as fact because 4 CPUs in 18 years is a silly basis to judge anything on.

 

[ShrimpBrime]

Maybe a reminder on what an anecdote is. An anecdote is personal experience, not constituting a significant enough quantity or without supporting evidence to support a conclusion. It is generally not an acceptable argument, because without significant enough sampling to rule out biases, your data is not acceptable as factual.

Read: It's a fact to me, but I cannot prove it.

Read also: 10% drop rates in video games. 10% of people get it round one. 0.9^100 or 2.66*10^-5, or 1 in about 37,649 do not get that 10% drop in 100 draws. Anecdotally the former person sees a 100% drop rate, and anecdotally the later person has a 0% drop rate, but because neither of the two add to a significant sampling size their experience is anecdote. Note I said "...I've only seen failures on non-jedec memory configurations in the last several years..." It's like I'm aware that your facts and my facts aren't more or less valid, because both are anecdotes and thus not statistically significant.



If instead you listed a few dozen specific instance we could maybe start somewhere (I offer that to you, because I've got only about a dozen, and do not meet better than anecdotal requirements)...but that would be assuming some controls. I...don't think we should go down that path. Also note I said both of our facts are not inherently facts, but anecdotal evidence. That is reality...not something I've used to decide one or the other is better.


As a side note, I'm usually that person who looks at 10% of tickets as winning and can buy 100 without ever seeing a winner. Having lived most of my life without luck I've found laughing at ones self is the only way to stay sane. Also, it's why I secretly enjoy seeing the silicon lottery hit others as hard as it does me...which is why I personally prefer the QVL and things like it. My first 1700x was...about as stable as a bipolar, one legged, drunken man inside a bee filled costume, with severe sleep deprivation. Everything I've gotten in the 5xxx series, multiple 5600x, 5700x, and a 5500 have just worked. Crappy no-name memory clocks to detected jedec, and the only crashes are software. That's maybe 8 CPUs...or so statistically insignificant it's funny. On the other hand, a 3930k was replaced twice, barely ran at stock memory frequencies on one board, and only on the second did I ever get the thing to run (under water) at anything less than nuclear reactor levels of temperature (100C reported at a junction I could never find, and after 9 year was probably the death of the system...with almost no overclock (did have to goose memory voltage to populate all 8 slots). Anecdotally, my Intel experience sucks...but I'd never state it as fact because 4 CPUs in 18 years is a silly basis to judge anything on.

Lol. It's funny because when you sum up all the anecdotes, through generations of Double Data Rate memory, there are conclusions made.

So lets go back to DDR1, work our way up.

DDR1, top modules would be TCCD and BH5 kits.
DDR2, some of the best kits where D9's.
DDR3, many would argue PSC or Hypers..
DDR4, B-Die kits / DJR - (this is not back in the day quite yet)
DDR5, A-Die / M-Die kits

And this isn't just some QVL picked theory, these where all tested tried and proven.
By people that took time for testing. To have an experience and share the same with others.

Since XMP is overclocking, you want the best memory chips. QVL or Not, that is enthusiast?? No that's a gamer that wants high speed memory. And it may not be on the list, but if it has certain memory die, then the chances are better for a successful overclock. The QVL does not give you this information, that is learned through human trial and error processes that extends beyond an incomplete, possibly non-updated list of venders that are qualified to be sold as a bundle in a pre-built.

Sure the list helped someone. So does a fork so you don't starve. Not that you need it really, you won't starve, promise.

 

[lilhasselhoffer]

Lol. It's funny because when you sum up all the anecdotes, through generations of Double Data Rate memory, there are conclusions made.

So lets go back to DDR1, work our way up.

DDR1, top modules would be TCCD and BH5 kits.
DDR2, some of the best kits where D9's.
DDR3, many would argue PSC or Hypers..
DDR4, B-Die kits / DJR - (this is not back in the day quite yet)
DDR5, A-Die / M-Die kits

And this isn't just some QVL picked theory, these where all tested tried and proven.
By people that took time for testing. To have an experience and share the same with others.

Since XMP is overclocking, you want the best memory chips. QVL or Not, that is enthusiast?? No that's a gamer that wants high speed memory. And it may not be on the list, but if it has certain memory die, then the chances are better for a successful overclock. The QVL does not give you this information, that is learned through human trial and error processes that extends beyond an incomplete, possibly non-updated list of venders that are qualified to be sold as a bundle in a pre-built.

Sure the list helped someone. So does a fork so you don't starve. Not that you need it really, you won't starve, promise.

?

Your conclusion from two people talking about anecdotes is that if you dramatically widen the net the anecdotes stop being anecdotal.


?

Bro, and I say that knowing nothing about you, use your words. You've told me a mole hill stops being a mole hill after you pile a few million mole hills onto it. I said that anecdotes are fine...but not facts, because they don't have enough examples...and you said if you increase the examples they cease to be anecdotes. Please, use the words. This was a thread about QVLs. I said my piece that they are a great tool, and anecdotally in the past I've found them to help but not so very recently. Anecdotally, because I am one person sharing an opinion.

Maybe the finer point is lost...but I think I'm done here. Your anecdote works for you...if you want to go out and claim QC and good processes make good product, and the market will find said good product, that's fine. Never said it wasn't the case. I don't know why I have to defend that...but here we are.

 

[eidairaman1]

not sure why anyone cares to go thru QVL for perf.
at least for any ryzen build the MB page list all that on the bottom, not gonna dive into pages of kits what i can find in seconds.

@eidairaman1
maybe something from Patriot. always find most of their kits have decent perf and dont look like designed with "gaming" in mind.

Yeah the Crucial Ballistix Gamer had a Red HS, reminded me of the days of Mushkin Redline, but that Crucial is good ram inspite the gamer name, at least it didnt have rainbow vomit.

 

[Waldorf]

@lilhasselhoffer
my problem is, while you do seem to understand the relevance of numbers (to be more detailed: basically anything below ~2500 units "tested", is statistically irrelevant),
you still say "..but anything more recent is more likely to be operator error...".
which would require you to have "fixed/helped" to get +2000 non-working ryzen setups to work, by adjusting settings (e.g. not defective ram), to make the conclusion
that its caused by operator error.

Im using mainly amd cpus since 2001, and know at least since 2007 that they were more picky about ram than intel.
and while fixing over 1000 computers in the past 10y is still not statistically relevant, short of defective parts, i never had a single time where swapping a non working kit (jedec and/or XMP/AMP),
with a different die, had the same issues (both QVL btw).
so my educated guess is, its very unlikely related to be an operator error (as in bios settings), and much more likely to be caused by certain cpu/mb/die combo,
and what i have "heard" from others troubleshooting/tweaking ram on ryzen rigs (in the past 5y), seems to supports that.

 

[Frick]

As a side note, I'm usually that person who looks at 10% of tickets as winning and can buy 100 without ever seeing a winner.

Everything has a 50% chance to happen: either it does or it doesn't.

 

[Waldorf]

@Frick
unless you have enough money to buy all tickets

 

[DirtyDingusMcgee]

It kind of depends on what you are building. If you are building a gaming system, QVL can be a useful guide as to what chips have been tested at what speeds......go nuts with hot rod ram, but it's good info.

If you are building a mission-critical system, you will probably want to use QVL ram so that you can give and receive proper support or warranty if needed.

 

[lilhasselhoffer]

@lilhasselhoffer
my problem is, while you do seem to understand the relevance of numbers (to be more detailed: basically anything below ~2500 units "tested", is statistically irrelevant),
you still say "..but anything more recent is more likely to be operator error...".
which would require you to have "fixed/helped" to get +2000 non-working ryzen setups to work, by adjusting settings (e.g. not defective ram), to make the conclusion
that its caused by operator error.

Im using mainly amd cpus since 2001, and know at least since 2007 that they were more picky about ram than intel.
and while fixing over 1000 computers in the past 10y is still not statistically relevant, short of defective parts, i never had a single time where swapping a non working kit (jedec and/or XMP/AMP),
with a different die, had the same issues (both QVL btw).
so my educated guess is, its very unlikely related to be an operator error (as in bios settings), and much more likely to be caused by certain cpu/mb/die combo,
and what i have "heard" from others troubleshooting/tweaking ram on ryzen rigs (in the past 5y), seems to supports that.

No....stop. READ THE WORDS.

"...I will adamantly suggest that if you have to touch 1xxx to 3xxx processors you double check the QVL...but anything more recent is more likely to be operator error...."

Now that I've copied and pasted, maybe you can pay attention. I will adamantly suggest. Do you get it? Anecdote. Bolded for your convenience of reading.

If your kink is somehow that I'm suggesting it's operator error, when literally saying it's my anecdotal experience, then I cannot help you. You refuse to see the words...and I'd appreciate it if you could stop talking to me. Have a little bit of self value, and read between the lines. In my experience, it's operator error. Hell, I didn't even say that as a requirement, only it's more likely to be.



Now that we've gotten off topic, because you've decided to drag this out until I have to copy and paste words to show you that your magical interpretations were already wrong, can you please let it die? I will not pretend that I can understand your logic...but I've done everything to not claim things as facts, and you seem Hell bent on thinking I did...for no apparent benefit. Please, stop.

Everything has a 50% chance to happen: either it does or it doesn't.

Can I buy a lottery ticket in your world? I might have a chance there.

Also, the retort to this is the simple "outcome does not equate to the probability of outcome." That's a lot less snappy though.



Trying to make this educational, this is the three doors problem personified. IE, you have three doors. Two lead to a trap, and one to a reward. You choose a door. The host then opens one of the remaining doors. You are offered to switch doors, with the other one that remains. Do you switch?
-Math, yes. The probability of your pick is 33%, whether shown a failure or not. Choose to switch.
-Human brain, no. I've got a 50-50 chance, and I feel lucky.
-Outcome, no matter what you can lose. Cool. Probability is not really fun.

 

[Waldorf]

@lilhasselhoffer
i dont care to read between the lines, type something or dont.
otherwise its assumptions i have to make, likely to be incorrect.

@DirtyDingusMcgee
but then you usually dont care much about speed (above sweetspot), and more about things like EEC.

 

[unwind-protect]

I ignore them and have no problem with many builds.

But I usually use ECC RAM, so obeying to them would be easier said than done even if I wanted to.

Many QVLs don't even have sets on there to reach is stated max capacity.