Rank of RAM Benchmarks (exclusive to PCS Laptops)

FerrariVie

MOST VALUED CONTRIBUTOR
Hi all,

I started to notice in the forum some discussions going on recently about RAM speeds: some people trying to compare speeds from their builds against others with the same specs, while others are just trying to check their clock speeds. So I thought about the idea of creating a rank of memory benchmarks for PCS laptops, even though I know that some people (like me) might have upgraded the RAM along the way (and that's fine, as it will help people that are also willing to upgrade). With this rank, we might also be able to compare different brands of RAM and the difference that clock speeds, ranks and banks might do, as well as spot people that might have either a faulty RAM or a misconfigured system.

Rules:
  1. Download and install AIDA64 Extreme (https://www.guru3d.com/files-details/aida64-download.html). I know it is a paid software, but I don't want people to pay for it (since it's quite expensive) and I haven't bought it myself, so we're all sticking to the memory figures available during the free/trial version, which are: Read speeds and Latency;
  2. Run AIDA64, then click on the "Tools" menu and select "Cache and Memory Benchmark". After the benchmark pop-up comes up, just click on start benchmark and wait for it to finish (it's quite quick, takes just a couple of minutes);
  3. Then go back to the main AIDA64 window, this time selecting Motherboard --> SPD on the left panel (see my own example on the following post);
  4. Create a new post on this thread with a screenshot of your benchmark results and the SPD specs page. Posts without both screens (or with data missing) won't be added to the rank;
  5. Please also mention what is your PCS laptop name (Recoil, Valeon, Destian, Ionico, Nova, etc) and screen size, just to help other people to identify similar builds;
  6. The order of the Rank will be based on the benchmark's read speeds in MB/s.
Any doubts, feel free to ask here or send me a PM.

Rank last updated on 2/January/22

(*)
Desktop Replacement Laptop that uses a desktop CPU, which in some cases allows memory overclock (real > base)

PCS LaptopRAM
Brand
Single/Dual
Channel
Rank+
Banks
Base
Freq.+CL
Real
Freq.+CL
AIDA64
Read(MB/s)
AIDA64
Latency(ns)
User
* Nova 15"CrucialDual (2x16Gb)1 x 163200Mhz CL223600Mhz CL205031276.6@FerrariVie
Ionico 17"CorsairDual (2x16Gb)1 x 163200Mhz CL223200Mhz CL2246823105.0@Illmatc-
* Recoil 17"CorsairDual (2x16Gb)2 x 162933Mhz CL192933Mhz CL194273164.3@Tron1982
 
Last edited:

FerrariVie

MOST VALUED CONTRIBUTOR
My results:

1639521526418.png1639521539929.png
 
Last edited:

Tron1982

Rising Star
Hey
I don't know if this is the right screenshot, but it may help ^^
(i don't really understood what the numbers means, could you explain it to me ?)

edit : my laptop is the 17" recoil DTR with a intel i7-11700k
cachemem.png
ramtest.png
 

FerrariVie

MOST VALUED CONTRIBUTOR
Hey @Tron1982 , thanks a lot for sharing your results. Let's go to the explanations about how to interpret the data (I'll try my best to not be very technical here :LOL: ):

RAM performance cannot be measured by looking into one single data point, but the closest you can get to that is by analysing raw speed (like the reading/write speeds in MB/s) and latency (in nanoseconds, not the CAS) at the same time. So comparing your results with mine, what we can see is that your RAM is quicker to respond to a CPU request (lower latency), but is only able to transfer a smaller amount of data each time (lower speed).

AIDA64
Read(MB/s)
AIDA64
Latency(ns)
User
5031276.6@FerrariVie
4273164.3@Tron1982

Other examples (got them from Crucial's website) that might make it easier for people to understand:
  • Example 1: if the speed rating (frequency) of a standard module and a gaming module is the same (i.e. DDR4-2666) but the CAS latencies are different (i.e. CL16 vs. CL19), then the lower CAS latency will provide better performance (lower overall latency in ns);
  • Example 2: because the latency in nanoseconds for DDR4-2400 CL17 and DDR4-2666 CL19 is roughly the same, the higher speed DDR4-2666 RAM will provide better performance.
So, in general terms, we should always aim to get the biggest amount of data transfer speed at the lowest latency possible and that can be achieved by 2 main factors: frequency and CAS latency (aka CL). However, it's not that simple and RAM modules with the same frequency and CAS can have different performance, as some other factors like ranks (single or dual) and banks (8 or 16) will also have an impact on both transfer speeds and latency. Don't think I'll get into those 2 for now as it can get a bit more confusing, but let me know in case any of you are interested, as I can try to explain that as well.

I would always recommend looking into frequencies first (higher is better) and then CAS latency (lower is better) when buying RAM, but looking at example 2 above we can see that a 2400 CL17 will be worse than 2666 with CL19, but it could potentially be better than a 2666 with CL22. So always keep in mind that both of them are important.
 
Last edited:

Tron1982

Rising Star
Hello,
You're welcome and thank you for the explanation ^^
So, yes, please do explain me, i'm always willing to learn more ^^
(and if you can use my ram test as example, it would be perfect, as i can knoc what i have inside my laptop ^^)

And, to conclude about our ram rank, as i saw that you are in 1st position, does it mean that the read is more important than the latency ?
 

FerrariVie

MOST VALUED CONTRIBUTOR
Let's start with the Rank explanation. Below is the best image that I could find to show what it means (taken from https://www.cgdirector.com/single-rank-vs-dual-rank-ram/):

1641376188704.png


So it means that dual-rank memory can output twice as many data (72b x 144b)? Not exactly, as the memory controller can only handle 1 rank at a time, but dual rank still gives about 3-5% more performance (according to Hardware Unboxed tests):
1641376574341.png

So there are no advantages to single rank memory? Well, they're cheaper and there's also a theory which says that single-rank memory will have slightly lower latency (in ns) than dual-ranked, as well as running a bit cooler (so better for RAM overclock). But I have not yet found any benchmark evidence to confirm this.

Next will be the explanation about Banks, so stay tuned :)

And, to conclude about our ram rank, as i saw that you are in 1st position, does it mean that the read is more important than the latency ?
This is a great question, and the answer is that the rank can only be based on one data field, so bandwidth it is :LOL: . If you look into DDR5 RAMs (talking about desktop RAM now), they're starting with 4800Mhz frequency and 40 CAS, so it seems that bandwidth is indeed been given more importance overall.

PCMag has done a great comparison between DDR4 and DDR5 (https://www.pcmag.com/news/ddr5-vs-ddr4-how-much-performance-will-you-gain-from-todays-newest-ram), which will help you understand that. The best bits are below:

1641377734324.png1641377772429.png1641377802287.png

So even though the latency is higher, the much higher bandwidth actually ends up giving better performance. They're both very important, but bandwidth is usually a bit more than latency (from what I can see).
 
Last edited:

Illmatc-

Member
Hello mate, this is from Ionico 17.3 build. 11800H, 2x 16gb 3200mhz RAM

1642687496210.png



1642687693726.png

If you know of anything that could I could do to improve any of this let me know :)
 

FerrariVie

MOST VALUED CONTRIBUTOR
Hello mate, this is from Ionico 17.3 build. 11800H, 2x 16gb 3200mhz RAM

View attachment 31943


View attachment 31944
If you know of anything that could I could do to improve any of this let me know :)
According to AIDA, it doesn't seem to be any problem with your RAM there, it's reporting a memory bus of 1569 (close to 1600), which then doubles as close to 3200Mhz. Your memory read speeds on the benchmark are also typical of 3200Mhz with high CAS (CL22). Regarding your 3dMark results on the other thread, it might be a bug with 3DMark reporting wrong frequencies.
 
Top