Memory performance on low end hardware: Single channel VS dual channel

yRaz

Posts: 3,573   +3,444
So, I was playing around today and noticed that not only was my memory not running at it's rated speed, it was also in single channel. It has been like this since I built it in 2017....

My specs are:
AMD 1700X
Asrock AB350 Gaming K4
2X8gb G.Skill ripjaws DDR4-2400
Asus 1070ti

So I first noticed the memory speed was at 2133 so I upped it to 2400, it made an immediate difference in boot time. I never timed it before but it then went from post to desktop in under 10 seconds. I checked CPU-z to check the speed and I noticed it was in single channel mode. So, shut down, move a stick of ram to another slot and it's in dual channel mode. It's now basically instant on. It's posts, flashes the OS logo(tested in both linux mint and windows 10) for what seems like a few frames and bam, I'm at the desktop.

I got curious and tested a few games. It seems like 1)micro stuttering in games has stopped and other ones have nearly doubled in a few cases, so have seeming doubled in performance. The biggest difference I saw was in Elderscrolls Online where I went from 42-50 FPS to a max pegging 97 FPS and "drops" in the mid 80's.

I started googling around and found this article on techspot. It only shows almost no performance gain going from single channel to dual channel, but it only uses the best of the best hardware. This got me wondering, does memory speed or the number of channels become more important on eliminating bottlenecks on lower end hardware? Is the 1700x somehow limited by memory speed?

The benchmarks shown in the article above show a fraction of a percent in performance at best. However, on my hardware going from single channel to dual channel gave me a 20-30% increase in gaming performance on the low end to a near 100% increase on the highend. This is aswell as the overall system feeling quicker and much more responsive overall. Applications and webpages load much faster.

Anyway, I found this fascinating and wanted to get some opinions on the matter. Is memory performance more important on lower end hardware? Am I just an ***** and accidentally did something to increase system performance?
 

neeyik

Posts: 1,788   +2,061
Staff member
The TechSpot test wasn't a single vs dual channel examination; it was 2 vs 4 DIMMs, both in dual channel configurations.

Zen processors benefit enormously from higher RAM speed (up to a certain point) and again with dual channel memory (you're doubling the available bandwidth).

Edit: I should add that the performance of any system, irrespective of the hardware used, is only as fast as its weakest point. In your case, it just so happened to be the RAM: 2133 MHz is slower than the maximum, non-overclocked speed that your CPU and motherboard support (which is 2666 MHz), hence why increasing it to 2400 MHz made such a difference. However, that's still down from 2666 MHz, so when you switched to dual channel mode, you helped ease the situation by upping the available bandwidth.
 
Last edited:

yRaz

Posts: 3,573   +3,444
The TechSpot test wasn't a single vs dual channel examination; it was 2 vs 4 DIMMs, both in dual channel configurations.

Zen processors benefit enormously from higher RAM speed (up to a certain point) and again with dual channel memory (you're doubling the available bandwidth).
Ah, well I feel like an ***** now, I kind of skimmed over it after looking for benchmarks in google. This was oddly hard to find information on. That, or googles search results were sending me information I wasn't looking for.

Anyway, I had no idea that could have such a large impact on general performance. Everything from gaming to boot times. I was under the impression that memory speeds had gotten to the point that they only make a difference in specific use cases. But, hey, I'll take a free performance increase even if it was due to user error.
 

neeyik

Posts: 1,788   +2,061
Staff member
Don't feel bad about it - after all, one BIOS setting and an extra RAM stick, and you've got a brand new computer! I've had to do the opposite with mine: I've been running with 4 x 8GB for ages now, but my system has always been on the edge of instability because it. Took 2 out and it's a much happier machine.

You're right about memory speed only making a difference in specific cases. It's as before: if the RAM is the weakest link in the performance chain, then any improvements in this area will have a noticeable difference. Running a Cinebench test? That's really down to your CPU and its core count. Running game at 4K with Ultra settings? That's down to your GPU.

In your case, it was simply all round general use, and the fact that you gained so much performance shows that, for specific hardware combinations and usage scenarios, RAM speed can make a huge difference. If you changed the RAM to faster 2666 MHz, again in dual channel mode, you'd see another improvement - albeit quite not as good as before.

2133 > 2400 = 12.5% data throughput improvement
2133 single > 2400 dual = 25% improvement
2400 dual > 2666 dual = 11.1%

After this and you then need to enable XMP profiles in the motherboard BIOS, to get to 3000 or higher. The gains will still be there but now much smaller, because the RAM weakness has now gone - the performance limitation will be elsewhere in the system.

You can see this in this test:


ACO_Ultra.png


Notice how overclocking the RAM and adjusting its timings has a noticeable outcome in the game test - but only with the 2080 Ti. Assassin's Creed Odyssey is very GPU limited, even at 1080p, so the RX 5700 and RX 580 are the performance limiters, not the RAM.
 

martinsmith

Posts: 49   +3
A single stick of RAM will operate on a single 64-bit data channel, meaning it can push data down a single pipe that is 64-bits in total width In the case of dual-channel configurations, we've now got 2x64-bit channels available to the memory.