1x8 3000 or 2x4 2400 in (2018)

[FJOCK94]

So, Ive read a lot of stuff about faster memory speeds being better for Ryzen CPUs. However, more recently I've heard that the newer Ryzen lineup doesn't scale as dramatically with faster memory. Ive got a R5 2400g on a gigabyte ab350 gaming 3. Ive got a 2x4 kit of skHynix at 2400. I was wondering if grabbing a single 8gig module would be beneficial for me?

 

[Barty1884]

Ryzen still benefits from faster memory, but the 2nd Gen it's not *quite* as important - still important though.

BUT, it does benefit heavily from dual-channel also.

Ideally you'd strike a balance between dual-channel & speed. I'd be inclined to attempt a bit of an OC on your current 2x4GB kit - hopefully you can hit 2666-2800MHz on it.

 

[electro_neanderthal]

Since you're using an APU, I'd suggest the 1x8 @ 3000. It's better for compatibility down the road and I think the faster speeds are better since the graphics part shares the RAM with the processing part.

 

[TJ Hooker]

If you're going to be using the integrated graphics for gaming, you want high speed (~3000 MHz) RAM, with two sticks for dual channel operation. Between those two options, I'm honestly not sure which would be better. The best choice might be to save up for 2 x 8GB 3000 MHz RAM.

 

[FJOCK94]

I'm going to go ahead and try the 1x8 @3000 and see where I'm at. If performance decreases, then I'll trade it out for another 2x4 @2400(unless I get 3000 for cheap) until I have enough for a 2x8 @3000

 

[RememberThe5th]

Sorry i didnt mean to downvote...

 

[FJOCK94]

So, I bought a stick of Corsair Vengeance LPX 1x8GB @ 3000MHz. I think it was a worthwhile investment. I tried all 16GB of total memory @ 2400 and my Cinebench score dropped 200 points!! But after running the 1x8 by itself at its rated 3000 speed, I saw a 14 point increase over my original score with the 2x4 @ 2400. Fortnite and TF2 saw an average 5 FPS increase at 1920x1080 high-very high settings. So I would say as an overall, "SPEED OVER DUAL CHANNEL" is the best solution for my 2400G.

 

[Barty1884]

That would go completely against benchmarks but, if it works fine for you: great!
https://www.gamersnexus.net/guides/3244-amd-r5-2400g-memory-kit-benchmarks-and-single-vs-dual-channel

 

[electro_neanderthal]

That would go completely against benchmarks but, if it works fine for you: great!
https://www.gamersnexus.net/guides/3244-amd-r5-2400g-memory-kit-benchmarks-and-single-vs-dual-channel

That's an interesting article (one I have not seen before), but seeing that dual channel 2400 also beat dual channel 3200 in at least three of the games/tests indicates that more is going on than direct dual channel vs. raw speed. I could make a guess at why, but I'd probably be incorrect. I'm just glad that the advice the OP went with helped them out.

 

[TJ Hooker]

@electro_neanderthal

The only thing I could think of is that maybe the author of that article is just taking out one of the sticks from the 2x8 GB pair to test single channel, meaning the single channel results are for a system with only half as much memory. I would hope he'd know better, but I suppose it's possible.

It sort of looks like that's what Tomshardware did in their own investigation, which shows similar trends for single vs dual channel as the gamersnexus article does.

https://www.tomshardware.com/reviews/raven-ridge-memory-scaling-benchmarks,5489.html

 

[Barty1884]

@electro_neanderthal.

TJ looks to be onto something. The "single channel" mention may well only be a single 8GB module.

Even with that, you're right - some titles do appear to perform better with 2x8gb 2400 vs 2x8gb 3200MHz....
Those are averages across 3 motherboards (i believe)- and as mentioned at the bottom of the article, sub-timings the board implements vary by board vendor.

Just a guess though.

 

[electro_neanderthal]

@TJ Hooker and Barty1884:

Yeah, that very well could be. After some thought the best I could come up with was from a LinusTechTips video where they explained a diminishing return after the 2666 frequency in general... but that was when 3000 was just making it's first debut with a lot of issues on keeping clocks that high long-term, so it probably does boil down to how each motherboard vendor implements sub-timings.

Still just a guess though.