Hello, I’m struggling with an old ram and a new one with very similar specs but of different brands and I need help enabling dual channel for these Kingston and Ballistix of 8GB, 3200MHZ, 1.35v, it seems i can’t do it on my own. I’ll give details in the pictures (XMP is disabled as of now): ram — ImgBB
I benchmark my rams and i get 8427.17 MiB/sec which in dual channel it’s supposed to be much higher right? My motherboard is a B550 M K gigabyte and my CPU ryzen 5 5500gt. Bios is from march 2024. Using slots A2 and B2
In just about every situation I have ever seen dual-channel will happen automatically as long as its possible - ie: the memory modules match and are in the correct seats.
I dont know about your rig but … maybe these should be A1+A2 or B1+B2 ?
Unclear. I might expect higher in most cases … but some see a measly 5-7% increase in data transfer speed and we dont know what the single-channel benchmark was.
A2+B2 is correct for two modules on that motherboard (and most B550 motherboards).
How to interpret dmidecode output is… contentious
One simple way to be absolutely sure is to run memtest from the grub menu and it should be 128-bit Mode for dual channel - https://superuser.com/a/1621323
The suggestion by @MrLavender is a good one to be super sure.
I still dont think the speed is ridiculous though.
On a laptop I just checked with dual-channel (also not overclocked or anything special) it came out to 7798.92 MiB/sec … but only after plugging it in. Before it was about half of that. Which reminds me to tell you to double check any power saving features or similar that may be interfering.
Lately there have been more than a few threads related to power-profiles-daemon being left in power-saving mode in particular. Just a possibility if it really should be higher.
You may need to install the memtest86+-efi package, I can’t remember if it’s installed by default. Then update grub.
Having said that I just tried it myself and it seems that memtest86+-efi (which is different from the original memtest86) doesn’t show that Mode information…
As @cscs said earlier dual-channel will be enabled automatically as long as you put the dimms in the correct slots, which you have. Any further problems after that are likely caused by mismatched memory.
Just enabling XMP profile 1 should be all you need to do because the XMP1 profile is the same for both. But that does not guarantee that two sticks from different manufacturers that were manufactured 4 years apart will happily work together.
If XMP is disabled then the memory is not running at 3200 (unless you set all the timings manually) so that would explain your benchmark numbers. What are you using to benchmark?
I tried enabling XMP profile 1 which is only different on tRC (one is 72 and the other one 74, check images I uploaded), and manually, I set System memory multiplier to 32.00 and it’s currently set at that.
I run “inxi -m” and it shows that it’s at “speed: 3200 MT/s”, which is what CPU-X also says.
I do my benchmarks with sysbench
Is there a simpler way to know if dual channel is activated or not? Despite the fact that I really doubt it.
As far as I am aware (no expert here) if the memory modules aren’t identical, you have to mess with timings etc. in BIOS to match the lowest common denominator (i.e. both match the slowest of the two) for this to work properly.
There is nothing I know of on Linux that will say “your ram is running in dual channel mode”. But it is because the dimms are on different channels.
What speed are you expecting to see and why? This is sysbench memory run on my B550 system which has better CPU and RAM than yours (5800X + 2x 16GB 3600 C16)
89217.01 MiB transferred (8921.24 MiB/sec)
But I can get much higher numbers with different parameters;
sysbench memory run on it’s own is not a great test because it uses 1K block size. Needs to be much larger for realistic testing. Compare before and after with at least 64M block size instead - sysbench memory --memory-block-size=64M run.
8361.35 MiB/sec