RAM Config for Ryzen 3000 series - 2x16 MB or 4 x 8 MB configs

[Midnitewolf]

I have run into a curious question. Which has the best performance, a 2x16 MB configuration or 4x8 MB configuration.  I ask because I have seen several video where it seems like 4 modules out performs 2 modules by a fair amount, however none of the videos are comparing apples to apples (same brand and timings for both configuations) so it is hard to say how accurate they are.

[asand1]

Some people have had issues running 4 sticks at advertised speeds. I do have 4x8g running at the advertised 3000Mhz though. Theoretically 4 sticks should run faster due to bandwidth. Ymmv

[Bombastinator]

37 minutes ago, Midnitewolf said:

I have run into a curious question. Which has the best performance, a 2x16 MB configuration or 4x8 MB configuration.  I ask because I have seen several video where it seems like 4 modules out performs 2 modules by a fair amount, however none of the videos are comparing apples to apples (same brand and timings for both configuations) so it is hard to say how accurate they are.

Depends on the trace on your motherboard.  If it’s daisy chain you get nothing from 4 sticks except additional PITA 

[Jurrunio]

GB, not MB

 

1 hour ago, Midnitewolf said:

I ask because I have seen several video where it seems like 4 modules out performs 2 modules by a fair amount,

have you made sure their frequencies and timings are the same?

[Midnitewolf]

13 minutes ago, Jurrunio said:

GB, not MB

 

have you made sure their frequencies and timings are the same?

That is the problem.  I haven't found any comparisons that definitively compared apples to apples however every one of the video I found showed a distinct benefit to having all 4 slots populated vs just two.  

 

49 minutes ago, Bombastinator said:

Depends on the trace on your motherboard.  If it’s daisy chain you get nothing from 4 sticks except additional PITA 

Yeah I was watching some videos about this as well stating that theoretically this would be correct but EVERY video I found comparing 4 vs 2 modules showed a benefit for the 4 modules and from what I have seen, most of the AMD MBs seem to be Dasiy Chain.

 

Honestly I guess to further confuse the issue there is this quote from this article: https://www.tomshardware.com/reviews/amd-ryzen-3000-best-memory-timings,6310-2.html

 

"The best practical configuration is four ranks of DDR4-3600 at the lowest stable latencies. Reminder: Four ranks can be achieved from either four single-rank or two dual-rank DIMMs." 

 

Basically I am wondering if this might have something to do with things but since I can't seem to find anything referring to whether the memory tested is single or dual rank, it is hard to say.  I have also found that it is damn hard to determine if the RAM your trying to purchase is single or dual rank as it doesn't even seem the mfg websites list this information.  

 

 

 

 

[Bombastinator]

8 minutes ago, Midnitewolf said:

That is the problem.  I haven't found any comparisons that definitively compared apples to apples however every one of the video I found showed a distinct benefit to having all 4 slots populated vs just two.  

 

Yeah I was watching some videos about this as well stating that theoretically this would be correct but EVERY video I found comparing 4 vs 2 modules showed a benefit for the 4 modules and from what I have seen, most of the AMD MBs seem to be Dasiy Chain.

 

Honestly I guess to further confuse the issue there is this quote from this article: https://www.tomshardware.com/reviews/amd-ryzen-3000-best-memory-timings,6310-2.html

 

"The best practical configuration is four ranks of DDR4-3600 at the lowest stable latencies. Reminder: Four ranks can be achieved from either four single-rank or two dual-rank DIMMs." 

 

Basically I am wondering if this might have something to do with things but since I can't seem to find anything referring to whether the memory tested is single or dual rank, it is hard to say.  I have also found that it is damn hard to determine if the RAM your trying to purchase is single or dual rank as it doesn't even seem the mfg websites list this information.  

 

 

 

 

So 16b is almost certainly dual rank but 8gb only might be dual rank.  2x16.  More space for later, easier to set, more known config.

[Jurrunio]

9 minutes ago, Midnitewolf said:

That is the problem.  I haven't found any comparisons that definitively compared apples to apples however every one of the video I found showed a distinct benefit to having all 4 slots populated vs just two.  

link?

[Midnitewolf]

19 minutes ago, Jurrunio said:

link?

Here is a good example:

 

 

4x4gb appears to be be significantly better but they are using two different brands of memory which I know can sometimes make a huge difference.

 

Here is another one with the same brand but with an older 2700x Process

 

 

Again the 4x4 configuration is significantly better.

 

But what I can't find is 32GB tests which I know could make a difference and under what conditions does 4 modules make sense, i.e. does it depend on the ranks of each module?.  The problem is if these videos are anything to go by there might be as much as a 10% increase in performance when using 4 modules which is pretty huge considering there is very little cost difference between 2 x 16 GB modules and 4 x 8 GB modules.  Also I know I am not ever going to need more than 32 MB of RAM so whether I go 2 modules or 4 modules isn't a factor.  

 

I guess I could actually just buy a set of 2 x 16 MB and a set of 4 x 8 MB, test them both and return the set that performs the worst but I hate to do that if someone can give me a definitive answer beforehand.

 

[Jurrunio]

4 minutes ago, Midnitewolf said:

Here is a good example:

Those you quoted didnt match with what I know though. More ranks are better for a set frequency and timings indeed, but 1) that's a difference you'll see in benchmarks, not games and 2) it's easier to overclock memory further with less sticks and less ranks.

[Bombastinator]

I can’t visually see a bit of difference between either image myself.  The only visible difference is some very slight number shifts.  What I’m seeing there is very very small though unequal use of GPU in percent of use.  Less than 5%.  Frame rate change of less than 1.

I don’t know much about this kind of testing, but the impression I have received watching testing videos is this kind of thing is within test accuracy ranges.  If they’re not equal, they’re very very close.  I’m not sure even a pro FPS gamer would worry about a change that small if it is even there, which I am not convinced of.  

[Midnitewolf]

14 hours ago, Bombastinator said:

I can’t visually see a bit of difference between either image myself.  The only visible difference is some very slight number shifts.  What I’m seeing there is very very small though unequal use of GPU in percent of use.  Less than 5%.  Frame rate change of less than 1.

I don’t know much about this kind of testing, but the impression I have received watching testing videos is this kind of thing is within test accuracy ranges.  If they’re not equal, they’re very very close.  I’m not sure even a pro FPS gamer would worry about a change that small if it is even there, which I am not convinced of.  

To be honest in most case scenarios I agree and it is not a huge worry for me, however if you can get an extra 5-10 fps just by buying 4 x 8 MB instead of 2 x 16 MB, why not do it.  That is why I was hoping someone might have some experience with this.

[Bombastinator]

2 minutes ago, Midnitewolf said:

To be honest in most case scenarios I agree and it is not a huge worry for me, however if you can get an extra 5-10 fps just by buying 4 x 8 MB instead of 2 x 16 MB, why not do it.  That is why I was hoping someone might have some experience with this.

Ah.  Given your examples it seems to be a difference of 0-1 fps.  1-2 at the outside. And then only if the frame rate is very high.  For this you trade complication.