[SOLVED] Does RAM brand matter?

[fivefiveopen]

Im building a pc and I don't know which RAM brand to choose. ADATA? Corsair? Kingston? If the rams have the same specs, does ram brand matter?

 

[Deleted member 2720853]

Look at the RAM frequency, CAS latency and price.

Check your motherboard's QVL list before buying.

G.Skill Trident Z is generally good for Ryzens.

 

[egda23]

Im building a pc and I don't know which RAM brand to choose. ADATA? Corsair? Kingston? If the rams have the same specs, does ram brand matter?

If you are not familiar with choices of RAM, stay with the QVL for you motherboard. This is the RAM that has been tested by the motherboard manufacturer (all motherboards have a web page with this listing).
Otherwise you can go here for example:
https://www.gskill.com/configurator
this allows you to choose RAM compatible with your MoBo.
The other brands: Kingston/xFurry, Crucial, Corsair have a similar configurator

 

[Endre]

Im building a pc and I don't know which RAM brand to choose. ADATA? Corsair? Kingston? If the rams have the same specs, does ram brand matter?

Yes, the brand does matter.
Brands that are “legit” can afford to test the DIMMs properly.
I usually recommend one of these: Kingston HyperX, Corsair, or Crucial.

Most people having problems with RAM, own G. Skill modules. (I’m not a hater, it’s just a reality).

 

[fivefiveopen]

Yes, the brand does matter.
Brands that are “legit” can afford to test the DIMMs properly.
I usually recommend one of these: Kingston HyperX, Corsair, or Crucial.

Most people having problems with RAM, own G. Skill modules. (I’m not a hater, it’s just a reality).

Thanks for your advice bro.

 

[egda23]

Yes, the brand does matter.
Brands that are “legit” can afford to test the DIMMs properly.
I usually recommend one of these: Kingston HyperX, Corsair, or Crucial.

Most people having problems with RAM, own G. Skill modules. (I’m not a hater, it’s just a reality).

Very interesting, but could you back up this claim with data.
That would be very interesting for everybody

 

[Endre]

Very interesting, but could you back up this claim with data.
That would be very interesting for everybody

Sure, just type “RAM XMP instability issues” on Google and you’ll see that most people that had RAM related issues, own G. Skill modules.

 

[egda23]

Sure, just type “RAM XMP instability issues” on Google and you’ll see that most people that had RAM related issues, own G. Skill modules.

Just for the fun of it, I slightly modified your search by specifying the brand name :
ex GSkill RAM XMP instability issues

and left all Google search specs by default
what came out is a certain number of references for each brand

ADATA 125
GSkill 117
Crucial 103
Ballistix 100
(which makes Crucial + Ballistix = 203)
Corsair 95
X furry 94
Kingston 82
(which makes X furry + kingston = 176)

which all means nothing as we don't know the market share for each brand.
But it's fun and clearly, all brands have problems

 

[Karadjgne]

Sure, just type “RAM XMP instability issues” on Google and you’ll see that most people that had RAM related issues, own G. Skill modules.

Doesn't work that way. Gskill and Corsair have the worldwide best selling ram when it comes to Ryzens, for 1st release the Only recommended ram Was Gskill, so you'll find gskill in most recommended builds. When you have 40% of the market, a 1% failure rate is massive compared to a company that has 5% of the market and a 1% failure rate.

If: 1,000,000 kits sold.
400,000 gskill
400,000 corsair
50,000 kingston
50,000 crucial
Rest noname
That's 4000 issues for corsair, gskill, 500 issues for Kingston, crucial. If 10M kits were sold, that discrepancy gets much larger. So shows up a lot more on websites, forums etc.

So you will find a lot more listed issues with gskill than is warranted, simply because there's that many more purchases. Gskill actively fights that by being one of the few companies that tests their own ram on other vendors mobo's, but even then that's not a perfect solution since much depends on the actual pc usage by the owner, what software, what bios, what revisions, what drivers etc.

Also take into consideration that many of the posts are actual user errors. 'Windows says my Gskill ram is only 1066, but I bought 2133' or 'my gskill ram is 3200 but is only running at 2133' etc.

Gskill, like crucial and Kingston will also usually bend over backwards to make you happy. If that means full kit replacements, upgrades, changes, that's what they'll do. I've had several conversations with both Kingston and Gskill reps, here at Tom's, where the rep has actively stepped into a post to help solve an issue.

I don't see Adata doing that.

 

[Endre]

Doesn't work that way. Gskill and Corsair have the worldwide best selling ram when it comes to Ryzens, for 1st release the Only recommended ram Was Gskill, so you'll find gskill in most recommended builds. When you have 40% of the market, a 1% failure rate is massive compared to a company that has 5% of the market and a 1% failure rate.

If: 1,000,000 kits sold.
400,000 gskill
400,000 corsair
50,000 kingston
50,000 crucial
Rest noname
That's 4000 issues for corsair, gskill, 500 issues for Kingston, crucial. If 10M kits were sold, that discrepancy gets much larger. So shows up a lot more on websites, forums etc.

So you will find a lot more listed issues with gskill than is warranted, simply because there's that many more purchases. Gskill actively fights that by being one of the few companies that tests their own ram on other vendors mobo's, but even then that's not a perfect solution since much depends on the actual pc usage by the owner, what software, what bios, what revisions, what drivers etc.

Also take into consideration that many of the posts are actual user errors. 'Windows says my Gskill ram is only 1066, but I bought 2133' or 'my gskill ram is 3200 but is only running at 2133' etc.

Gskill, like crucial and Kingston will also usually bend over backwards to make you happy. If that means full kit replacements, upgrades, changes, that's what they'll do. I've had several conversations with both Kingston and Gskill reps, here at Tom's, where the rep has actively stepped into a post to help solve an issue.

I don't see Adata doing that.

Yeah, market share is very important.

 

[egda23]

Karadjgne, when I answered Endre's post with my "statistical analysis" I was just joking.
I do statistical analysis professionally, so I have some vague idea how to proceed (although in a completely different field). but Tomshardware is a family friendly site and i didn't want to be aggressive (and you do not know if you answer to a professor at MIT or a 15 year old kid that you don't want to hurt).
So, I was not going to come with "here comes the scientist who certainly prepares DT press conference about COVID"
That said until a few years ago (they stopped in 2017) a french website reported 2 times per year the incidence of RMA for all computer parts that they obtained from the largest French retailer (it was LDLC.fr).
Their last publication about RAM can be accessed here:
https://www.hardware.fr/articles/962-4/memoires.html
it's in french, but the numbers are easily understood. So the RMA percentage for RAM sticks was:
Kingston 0,30%
Crucial 0,35%
G.Skill 0,63%
Corsair 0,77%
And I think that this leads us as close as we can to some sort of estimation of brand reliability

 

[geofelt]

Ram brand does not matter unless you have a problem.
Then it is important to know about the support system for the brand.
All of the name brands are sufficiently reliable, so in that respect, brand does not matter.

 

[Karadjgne]

And I think that this leads us as close as we can to some sort of estimation of brand reliability

Not reliability. Rma is a simple return, can be for any reason. Bought wrong kind, wrong voltage, doesn't fit under heatsink, got better deal elsewhere with rgb, etc. Ppl rma for stuff as simple as Amazon sent the wrong color.

Very little is an actual doa or reliability issue. Reliability is more normally an age thing, ram died suddenly after 3 years etc.

And while I can understand your intent, I'm well informed enough to make some determinations, unless explained then other readers of this post might not come to the same conclusion and take more meaning from your answer. Many understand a simple joke, but satire is something far deeper.

Take first release Ryzen for example. There was only one ram that any real success at reaching 3200MHz. Samsung B-die Gskill at Cas 16. The biggest failure rate to be even compatible with those cpus was the SkHynix Corsair LPX. With its popularity, it was a huge seller, but then many rma's later, ppl bought gskill instead. Absolutely nothing wrong or unreliable about the LPX, just didn't work with Ryzen really at all beyond 2133MHz. Thousands of rma's for no reason other than user error, they didn't do their homework, just picked what they liked.

Many also went cheap and bought the gskill Cas18, Samsung D-die, and had issues with 2933MHz, few got to 3200MHz, so there were multiple rma's of gskill too.

Nothing unreliable about the ram, everything wrong with the way the cpu talked to the ram.

As to actual brand, that's immaterial. There's only a handful of OEMs for ram, they slap on a chunk of aluminium and a paint job, but you'll find the exact same SkHynix chips in Corsair, Patriot, Adata, Crucial, Kingston etc. Brand is just a name of who bought it from the OEM.

 

[egda23]

Not reliability. Rma is a simple return, can be for any reason. Bought wrong kind, wrong voltage, doesn't fit under heatsink, got better deal elsewhere with rgb, etc. Ppl rma for stuff as simple as Amazon sent the wrong color.

Very little is an actual doa or reliability issue. Reliability is more normally an age thing, ram died suddenly after 3 years etc.

And while I can understand your intent, I'm well informed enough to make some determinations, unless explained then other readers of this post might not come to the same conclusion and take more meaning from your answer. Many understand a simple joke, but satire is something far deeper.

Take first release Ryzen for example. There was only one ram that any real success at reaching 3200MHz. Samsung B-die Gskill at Cas 16. The biggest failure rate to be even compatible with those cpus was the SkHynix Corsair LPX. With its popularity, it was a huge seller, but then many rma's later, ppl bought gskill instead. Absolutely nothing wrong or unreliable about the LPX, just didn't work with Ryzen really at all beyond 2133MHz. Thousands of rma's for no reason other than user error, they didn't do their homework, just picked what they liked.

Many also went cheap and bought the gskill Cas18, Samsung D-die, and had issues with 2933MHz, few got to 3200MHz, so there were multiple rma's of gskill too.

Nothing unreliable about the ram, everything wrong with the way the cpu talked to the ram.

As to actual brand, that's immaterial. There's only a handful of OEMs for ram, they slap on a chunk of aluminium and a paint job, but you'll find the exact same SkHynix chips in Corsair, Patriot, Adata, Crucial, Kingston etc. Brand is just a name of who bought it from the OEM.

I agree with you. In conclusion, I think that the best answer to the original question by the OP, would probably be: "if your are not familiar with all the intricacies of RAM, stick with the QVL for you board". There are no "good" or "bad" brands (at least if you remain with the majors), there are good or bad combinations with your motherboard that you can minimize by sticking with the QVL or configurators as the one from GSkill.

 

[Karadjgne]

Almost. Stick with what's on the QVL. Not what's in it. Take for instance the gskill Trident-Z series (which btw is on the QVL). That series alone compromises of close to 5000 individual model numbers. The model number changes for Cas, speed, voltage, color, kit, heatsink style etc. But is the same ram underneath. A blue 3200 Cas 18 is the same as the red or black or gray 3200 Cas 18.
But only 1 was tested. Might not have tested 3000, or 3600, but did test on 3800.

So going by the QVL is ok, but in 40 years± of building pc's I've never actually used any actual listed model. Close, different color or kit, but never the exact thing.

 

[Endre]

Almost. Stick with what's on the QVL. Not what's in it. Take for instance the gskill Trident-Z series (which btw is on the QVL). That series alone compromises of close to 5000 individual model numbers. The model number changes for Cas, speed, voltage, color, kit, heatsink style etc. But is the same ram underneath. A blue 3200 Cas 18 is the same as the red or black or gray 3200 Cas 18.
But only 1 was tested. Might not have tested 3000, or 3600, but did test on 3800.

So going by the QVL is ok, but in 40 years± of building pc's I've never actually used any actual listed model. Close, different color or kit, but never the exact thing.

I’d say that memory instability related issues started with the “XMP era”.
Going beyond the JEDEC standards involves risks, though the vendors advertize it as totally safe & working.
It’s a lot of marketing involved, because we, as people, love high numbers.

 

[egda23]

I’d say that memory instability related issues started with the “XMP era”.
Going beyond the JEDEC standards involves risks, though the vendors advertize it as totally safe & working.
It’s a lot of marketing involved, because we, as people, love high numbers.

You could almost say the same thing for CPU overclocking as many people take a single number and fixate on it, not understanding that it is almost a lottery.
But I come to think that instability and reliability are not entirely the same problem. And this is compounded by the fact that with present generations of computers the memory controller is in the CPU, which brinks yet another variable in the equation.