Question RAM For AMD And 2 sticks vs 4

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
D

DefinitelyNotTom

Respectable
Jul 20, 2017
1,053
5
2,295
I have a 3200 x2 kit of Samsung b die, but the latency is the odd 15, whereas about everything on the msi ace qvl is 14 or 16 latency.

There is a 3600 x4 kit with latency of 16 for only a couple hundred dollars.

Is it worth the trouble and losses to buy the 3600 kit and try selling my other kit on ebay?

Someone said you shouldn't use a 4 stick kit, you should only use a 2 stick one. is this true? I forget what his reaosnw as, but seems like he said it's slower when using a 4 kit one.
 
Sort by date Sort by votes
TJ Hooker said:
There's a very similar kit on the MSI ACE QVL: F4-3200C14D-32GTZKW. Looks to be identical (also dual rank Samsung B die) other than that all the primary timings are one lower. Given that tighter timings are harder to run, if it can run that kit it seems like it ought to have a good chance of running yours.

Pretty sure it'll just default to 16 rather than 15 when you enable the XMP profile. Not a big deal.
Click to expand...
It may be worth a try to contact msi and see if they'll test it with that exact kit, although I doubt they'll do that. I may keep it, but not sure. Any little difference, including the timings, can make a big difference in what works properly on a mobo. That guy from gamersnexus said g.skill kit worked fine and another with the exact same model number EXCEPT rgb in it had issues on some board. So even a slight change can affect things.

Also, people kept making a big deal about how every slight change in CL would affect productivity in a big way and they'd go on and on about how CL16 kits sucked blah blah blah. I really do think some people exaggerate on how important hings are, though. Some people act like a 5% difference can determine whether something flat out sucks or flat out is great. lol. Either way, so much of that is in my head making me unsure what to do, though.
 
Upvote 0 Downvote
There's a lot of info here, but I also see focus, not the big picture.
Yes, 3600MHz is the sweet spot for Ryzen 3000's. But to get the big picture, you have to look at what 3600. Namely it's the lower end, cheaper, higher Cas 3600, because most ppl don't see any value in spending 1.5x to get the good stuff.

For Ryzen 1k and 2k series, it's 3200MHz, because that was what ran best, was the most compatible, gave the least amount of issues, was the highest speeds on many boards.. And ppl bought the hell out of Cas 16 3200MHz. Very few bumped the price up considerably for the far better Cas 14 sticks.

That's Samsung B-die. More than likely it'll take a manual setting at 3200 Cas 14, or bump upto 3433 Cas 15. Or maybe as high as 3600 Cas 16 timings.

If I can get Patriot Intel Masters 1866MHz Cas 10 1.5v to hit stable at 2400MHz Cas 11 timings 1.505v, anything is possible.

Question is, is a 3600 kit really worth the extra cash over what you own now.
 
Upvote 0 Downvote
Does Op realize that just in the gskill Trident-Z series, there's over 3000 individual models? Then add in the RipJaws, value, Aries, Snipers etc and gskill alone has a list of over 10,000 individual models. Then add Patriot, Corsair, Kingston, Adata, Crucial, and all the others and there's well over 250,000 individual model numbers.

And about 100 different IC chips. Same B-die in Trident-Z series are used in Patriot Elite, Adata XPG, and multiple others. 1Gb ic's, 2Gb ic's, 4Gb ic's. All the same. Different name brand paint job on the heatsink.

The QVL is Qualified Vendor List. Not the Qualified Ram list. All it is is the manufacturer has tested several brands, speeds, Cas, kits. What it owns, what it is donated, what it has on hand. Just a selection. There's absolutely no guarantee that an identical model will work on a different mobo, only that those tested were compatible with that board.

And red heatsink ram has a different model number to black or blue or rgb, but is the exact same ram underneath.

There's not a single mobo vendor willing to spend hundreds of thousands of dollars in purchases, man-hours testing and compiling on the multiple different mobo models it makes to test 250k different ram setups and print out a QVL listing thousands of pages long. That'd be company suicide. Especially when there's really not that many different ram ic's to begin with.

QVL is a reference tool. It's not Gospel.
 
  • Like
Reactions: TJ Hooker
Upvote 0 Downvote
"all it is" is you still have to follow it or if you have a problem no company will do anything about it. you can't go to them and say "there were way too many kits not tested". You have a problem, FIRST thing they ask is whether or not it's on the qvl. If not, they say bye bye and do nothing. Go read some reviews on Newegg where manufacturers respond to them and immediately ask is it on the qvl.

of course I know there are a ton of kits, which is stupid and ridiculous for them to make that many to cause such a headache like this. But it's irrelevant. You have to buy from the qvl or play lottery.

And even when it's ON the list, it still may not work. Look at the reviews on newegg for the 64GB g.skill trident z neo 3600 kit, in other words the 64gb version of the 32gb kit I was thinking about. One person says once you update your firmware for your RAM, it stops working on the meg ace.

As fort it being shown that 3600 isn't much better than 3200, then why do people keep pushing getting 3600 for ryzen 3? If increased MHz doesn't matter, then what is the point of selling it and people buying it? Depending on which programs you use, it would matter. heck, look right in this thread a couple posts up where he's implying even 16 latency 3200 was bad compared to 14 latency 3200. If a 2 difference in latency is enough to be a big difference then 400 extra MHz is.

In fact that is basically equal... 200 MHz speed increase is equal to about a 1 difference in latency at the lower speed. (I mean latency would be about equal, so then you're getting the extra speed at no latency loss).
 
Upvote 0 Downvote
I'm really going off of what other people say. Look even on these forums, over and over people say get high RAM speeds, ryzen loves high ram speed blah blah blah, so important. Either it's important or it's not.

Someone told me a couple years ago that anything above 2667 is such diminishing returns that you might as well not go much above that. then person after person saying it's soooooooooo important to get 3200 for ryzen 2xxx.
 
Upvote 0 Downvote
Actually, I was right before I started editing that. 200MHz difference at same latency is the same as 1 difference at latency at the same MHz. So if it was really such a big difference between 3200 CL14 and 3200 CL16, then it's just as much difference between 3200 CL16 and 3600 CL16. I mean BASICALLY the same. 8.888 ns vs. 8.75ns.

And the speed of my kit would be 9.,375 ns per byte transfer, whereas the kit I was considering would be 8.888. I see opinions on both sides as to whether that would be noticeable, but someone points out that since your pc is constantly doing something that needs memory it would be building up over time how much the difference matters.

Same person who was posting how to figure that up did mention, though, that even 3200 CL16 is only around 1% from being the best you'd get on ryzen 3. So more or less backed you up. But I am not surre if they're basing that only on gaming or not. And it still doesn't solve my not on qvl issue. It may mean if I buy another I could buy around my same speed, though... like a 3200 cl13 and sell this one and then be guaranteed due to it on the qvl.

edit: sigh. even then people don't agree. someone said infinity fabric with ryzen 3 means higher sped still matters and someone said the closest to 3733 is always going to be the best.
 
Last edited:
Upvote 0 Downvote
@dnt,
I personally recommend G.Skill RAM because in my experience, they honor their lifetime warranty on RAM.
I have only RMA'd G.Skill RAM twice in the last 20-25 years and both times they replaced the faulty RAM without any hassles.
I have had maybe 4 or 5 other RAM failures in that time.
I won't name those other companies, but I will say that they all found some reason to not replace my RAM.
Which is why I buy and recommend G.Skill RAM to anyone looking for quality RAM at a fair price.
So if you are worried about the warranty, I would definitely consider G.Skill.
 
  • Like
Reactions: Crosslhs82x2
Upvote 0 Downvote
DMAN999 said:
@dnt,
I personally recommend G.Skill RAM because in my experience, they honor their lifetime warranty on RAM.
I have only RMA'd G.Skill RAM twice in the last 20-25 years and both times they replaced the faulty RAM without any hassles.
I have had maybe 4 or 5 other RAM failures in that time.
I won't name those other companies, but I will say that they all found some reason to not replace my RAM.
Which is why I buy and recommend G.Skill RAM to anyone looking for quality RAM at a fair price.
So if you are worried about the warranty, I would definitely consider G.Skill.
Click to expand...
I am definitely using g.skill. The kit I have and the one I was considering are both g.skill. I did temporarily consider corsair, but it's even more impossible to figure out if that is even on a qvl because it says version numbers! Like we know what version number we'll end up with when ordering online. Also, I am not so sure any of theirs are Samsung b die because even some that "say" they are b die have timings where they aren't the same... such as 14-16-16. Normally you can tell b die by the timings being the same, such as 14-14-14. I'm sure that isn't ALWAYS the case, though.

Btw, when I do start using a pc, how will I know if the RAM is performing properly?
 
Upvote 0 Downvote
DefinitelyNotTom said:
As fort it being shown that 3600 isn't much better than 3200, then why do people keep pushing getting 3600 for ryzen 3?
Click to expand...
a) People just repeat what they've heard, without checking the evidence (like the benchmark results that were previously linked in this thread)
b) This forum is full of 'enthusiasts', which can often mean having more money than sense and pushing for the 'best' even though it doesn't really get you any noticeable improvement

DefinitelyNotTom said:
If increased MHz doesn't matter, then what is the point of selling it and people buying it?
Click to expand...
Why do they make $100k cars when a $20k car gets you from point A to B just as well?

DefinitelyNotTom said:
Someone told me a couple years ago that anything above 2667 is such diminishing returns that you might as well not go much above that. then person after person saying it's soooooooooo important to get 3200 for ryzen 2xxx.
Click to expand...
Things change over the course of years...
 
Last edited:
Upvote 0 Downvote
DefinitelyNotTom said:
edit: sigh. even then people don't agree. someone said infinity fabric with ryzen 3 means higher sped still matters and someone said the closest to 3733 is always going to be the best.
Click to expand...
We provided you with hard numbers showing the differences between various speeds. You don't need to listen to the evidence-less opinion of every chucklehead on the internet. And of course "best" can mean different things. Best performance? Yes. Best value? No.
 
Last edited:
Upvote 0 Downvote
With Intels, speeds don't matter much, how they work is different to Ryzens. With Ryzen, the infinity fabric (communication between cores) is based on ram speeds. So technically the faster the ram, the higher the cpu performance. Upto a point. After that point, there's diminishing returns. With 1k and 2j series, that speed was 3200MHz. There's @ 20% increase between 2133 and 3200, and pricing was pretty equitable. The price for 3433 or 3600 was far higher, and the performance gap far smaller the faster you go. Yes, performance generally was better, but not really cost effective.

And that's just the ram. With the motherboard and cpu, after 3200MHz the bios started really having issues with speed compatability, next to impossible to get 4k stable.

With Ryzen 3, that whole thing has moved from 3200MHz to 3600MHz, and prices are equitable to the performance gains. 3200 is still good, 3600 is better. Infinity fabric makes better use of the faster ram speeds.
 
Upvote 0 Downvote
True. But scroll down to pricing. 16Gb of 2666 is less than $100. 16Gb of 4400 is $330. For @ 10% or less performance gain. Just in gaming at higher resolutions, that got even worse difference. Even with a 10% difference at 4k, that's still only averaging 4-8fps. Is 4-8fps honestly worth spending over $300 for ram? If the game even benefits in the first place. Many really don't. A few do.
 
  • Like
Reactions: TJ Hooker
Upvote 0 Downvote
Thanks for the info. As far as hard numbers, we could probably find some other person's tests that show a bigger gap, though, and they didn't test every program ever made, either.

Either way, if I am set on having t get something from the qvl, it's like what was said, price is close to the same between 3200 and 3600, so might as well get 3600. The main thing now is just whether to chance this kit or not. It goes down in value a bit when opened. I wish g.skill had a type of warranty where they'd swap it out for a totally different model kit of the same value if it ahs issues in a mobo. Not like it costs them anything, as they'd just send out a refurbished one, anyway.

Also, I know things change over time, that was my point, though, is it has also changed to be 3600 recommended.

In actuality, I have already gone BEYOND insanity in as good of parts I have bought already. Most of the time I'm only going to freakin use the internet or use simple programs and I have gone overboard just to improve things I'll rarely do.

Eh and like I said I am somewhat tempted to get 64gb ram just to be that much longer before I need more, but even 32 is overkill.
 
Upvote 0 Downvote
In the article you linked (by Tom's 😊) the RipJaws V 2666MHz 16Gb is $95. The same kit in 3200 is $175. Almost double the price. Gaming at 1080p (same article) puts the 2666 at 108.4% and the 3200 at 113.7%. That's a 5.3% improvement for double the cost. And that's if it translated directly to fps, which it doesn't. You'd be looking at @ 2-3fps increase in games which have fps bounce all over the place anyways. Not static or even close. For that pricing difference, the faster ram is seriously not worth the difference in speeds as the aggregate difference is basically not visible at all except to a benchmark.

The differences with those speeds to a Ryzen can see 10-20 or more fps, by comparison, depending on the game.

But. While I agree there is some benefit to Intel and faster ram, most times it's moot, either so far above refresh it really doesn't count, or so small you can't see it anyways, so doesn't justify the expense.
 
Upvote -1 Downvote
Karadjgne said:
In the article you linked (by Tom's 😊) the RipJaws V 2666MHz 16Gb is $95. The same kit in 3200 is $175. Almost double the price.
Click to expand...
I don't know why you'd use the pricing from the article, given that Tom's price links are often buggy (the price link for that kit is broken) and the fact that there's no reason to get that exact kit specifically from Newegg.
G.Skill Ripjaws V 2666 MHz CL19 is $63
G.Skill Ripjaws V 3200 MHz CL16 is $68
If you want to be picky about using the same CL as in the article then:
G.Skill Ripjaws V 2666 MHz CL15 is $70
G.Skill Ripjaws V 3400 MHz CL16 is $83
All prices from Newegg.

Karadjgne said:
Gaming at 1080p (same article) puts the 2666 at 108.4% and the 3200 at 113.7%. That's a 5.3% improvement for double the cost. And that's if it translated directly to fps, which it doesn't. You'd be looking at @ 2-3fps increase in games which have fps bounce all over the place anyways. Not static or even close. For that pricing difference, the faster ram is seriously not worth the difference in speeds as the aggregate difference is basically not visible at all except to a benchmark.

The differences with those speeds to a Ryzen can see 10-20 or more fps, by comparison, depending on the game.
Click to expand...
Uh, what? I provided you with empirical evidence showing you that RAM speed can affect FPS with Intel as much as AMD (look at the performance summary pages in each of those articles). Yes, they're testing with a 2080 Ti to exaggerate differences, so you probably won't see that kind of improvement in real world usage, but that applies equally to AMD or Intel. I simply don't understand how you say that RAM speed is insignificant for Intel but important for AMD in light of the results in those articles unless you're being deliberately obtuse.
 
Last edited:
Upvote 0 Downvote
Karadjgne said:
In the article you linked (by Tom's 😊) the RipJaws V 2666MHz 16Gb is $95. The same kit in 3200 is $175. Almost double the price. Gaming at 1080p (same article) puts the 2666 at 108.4% and the 3200 at 113.7%. That's a 5.3% improvement for double the cost. And that's if it translated directly to fps, which it doesn't. You'd be looking at @ 2-3fps increase in games which have fps bounce all over the place anyways. Not static or even close. For that pricing difference, the faster ram is seriously not worth the difference in speeds as the aggregate difference is basically not visible at all except to a benchmark.

The differences with those speeds to a Ryzen can see 10-20 or more fps, by comparison, depending on the game.

But. While I agree there is some benefit to Intel and faster ram, most times it's moot, either so far above refresh it really doesn't count, or so small you can't see it anyways, so doesn't justify the expense.
Click to expand...
you're doing like most and using only gaming as the basis, though. I care about productivity.

(I know you're talking about what tj said, but point still remains... gaming isn't the only decider.)
 
Upvote 0 Downvote
DefinitelyNotTom said:
you're doing like most and using only gaming as the basis, though. I care about productivity.

(I know you're talking about what tj said, but point still remains... gaming isn't the only decider.)
Click to expand...
The article linked on the first page provides results for non-gaming tasks. Are any of the benchmarks included relevant to your use case? From their test suite, the only non-gaming benchmark that showed any real difference with RAM speed was file compression.

I think you have all the info you could need at this point. If you're willing to try non-QVL RAM (which will most likely work fine) use what you already have, if not than buy something off the QVL list (knowing that spending extra for 3600 MHz probably won't get you much in the way of improved performance).
 
Last edited:
Upvote 0 Downvote
TJ Hooker said:
The article linked on the first page provides results for non-gaming tasks. Are any of the benchmarks included relevant to your use case? From their test suite, the only non-gaming benchmark that showed any real difference with RAM speed was file compression.

I think you have all the info you could need at this point. If you're willing to try non-QVL RAM (which will most likely work fine) use what you already have, if not than buy something off the QVL list (knowing that spending extra for 3600 MHz probably won't get you much in the way of improved performance).
Click to expand...
The 3600 is likely not going to cost much above the 3200. only thing, though, is for Samsung b diw 3600 it's $350 vs. the $250 for non-b-die. And also I don't think the b die 3600 was even on the qvl.

I'd do file compression when I upload apps I make, but they are so small it wouldn't make much difference there. But also one never knows for sure what all of his future needs will be. Programs become more and more relied on high levels of resources. On my current pc I have been on sites where it locks the pc up where it takes literally MINUTES to close the window or get task manager opened. But just a year earlier there weren't sites causing that. Out of nowhere websites or software developers can start doing something that requires far more resources than were needed the day before.
 
Upvote 0 Downvote
Well. Hmm.

That would work, if it was actually the gskill 3200MHz at Cas 16 that was tested and price linked. It wasn't. It was the Gskill Cas14 instead.

Memory: G.Skill Ripjaws V Series 16 GB (2 x 8 GB) DDR4-3200 Memory ($67.99 @ Newegg)
Memory: G.Skill Ripjaws V Series 16 GB (2 x 8 GB) DDR4-3200 Memory ($123.98 @ Newegg)
The 3200 Cas14 is considerably more expensive than the 3200 Cas16. Almost double the price. It's also somewhat faster.

With Intel memory controllers, % performance increase doesn't translate as fps in a 1:1 ratio. The tests performed showed only the performance of the ram, not the performance of the cpu to increase max fps. It doesn't scale that way. A 5% performance increase in ram doesn't mean a 5% increase in fps, it's closer to half that. And that's best results at 1080p. Go up in resolution, and the differences go down, at 4k there's almost no difference at all. Anyone spending $300+ on uber fast ram sure isn't gaming at 1080p.

Productivity is a whole different ball of wax. Ram speeds/timings can and do have a sizable impact on time in use. But that's only because ram is a data conduit. If a cpu can use or needs the data, the faster the ram can get it to the cpu, the faster the whole process is. Gaming is different. The individual files are small, a few kb or even less. The scripted code for a gun or a leaf or blade of grass is only a few lines. Productivity files can be massive, 10Gb or more.

With gaming, it's a matter of shaving off a few nanoseconds having faster ram move the files to the cpu for pre-rendering, with productivity it can easily be minutes, 10's of minutes, an hour that gets saved.
 
Upvote 0 Downvote
The fact that it can take a lot of time to create a file with video editing/creation sofwtare is why I also went as high as 3900x on cpu (although many programs don't use the cpu as much as gpu….. I did get a 1070 ti too in case I use a gpu-heavy editor... I don't pc game much, so sure wasn't going above that gpu.). But still not sure what I'll do RAM-wise. It's harder to find benches on all the different editing software. Usually they'll show how the cpus perform on a few of them, but I don't recall seeing RAM comparisons.
 
Upvote 0 Downvote
Another problem is I have $300 in Newegg credit, as I expected to get a case and/or mobo there.... then I bought them both from amazon. Now I am kind of screwed because there's not a lot I'd want from Newegg if I don't buy RAM. And people keep insisting I should keep my rm550x, despite pcpartpicker showing it incompatible with msi x570 ace because so many people say the second 8-pin cpu connector is optional. So if I have no psu or ram to buy, about all that's left is a monitor at some point. for now was going to use my old lcd monitor via an adaptor cable.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom