9800X3D RAM buying guide by Buildzoid

JollyJamma said:
The brand itself is not important.

What is important is what memory chips were used - it used to be Samsung B die with DDR4, it’s now Hynix A die or M die with DDR5.

Most RAM modules are just rebranded mass produced modules that come out of the OEM factory and get a brand heatsink and sticker applied.

Just get any 6000CL30 kit as Minis said, doesn’t matter which brand.
Click to expand...
Thanks for this, very helpful!

Is there a difference between the Hynix A or M die?
 
Capitano said:
Thanks for this, very helpful!

Is there a difference between the Hynix A or M die?
Click to expand...
If you're not into RAM overclocking and tweaking, nothing much and either will do.

You can overclocking 6000CL30 to 6400 quite easily since the quality of the DIMM chiplets is very good, hence why Buildzoid is happy to recommend them. Going for A die or M die is not really worth worrying about too much for 99.9% of people since both are good.
 
JollyJamma said:
If you're not into RAM overclocking and tweaking, nothing much and either will do.

You can overclocking 6000CL30 to 6400 quite easily since the quality of the DIMM chiplets is very good, hence why Buildzoid is happy to recommend them. Going for A die or M die is not really worth worrying about too much for 99.9% of people since both are good.
Click to expand...
I used to dabble in some overclocking and tweaking on the RAM but I guess these days it really doesnt matter like how it used to.

Thats great to know, I guess 6000 is the way forward for the 9000 series unless of course the new X3D chips show something different
 
The Corsair 6000Mhz CL30 kit I got runs M-Die I believe.

Tried a Lexar 6800Mhz kit and that was Hynix too but there was no mention if A-Die or M-Die. Works fine on my X870 board, but ofc runs 2:1 ratio at 6800Mhz.

Haven't tried downclocking the Lexar kit yet. Maybe I should?
 
Got a 9800x3D up and running earlier on an x870 Tomahawk.

I used some DDR5 I already had which is some G.skill DDR5-6400 CL32-39-39-102

I think it's running successfully at 6400 in 1:1 mode at CAS 30, but how do I

A) Double check it hasn't changed to 2:1
B) Accurately Test the memory latency
C) Test the stability

I downloaded Ryzen Master and this is what it's showing:

XMZCXlJ.png
 
Last edited:
VincentHanna said:
3Got a 9800x3D up and running earlier on an x870 Tomahawk.

I used some DDR5 I already had which is some G.skill DDR5-6400 CL32-39-39-102

I think it's running successfully at 6400 in 1:1 mode at CAS 30, but how do I

A) Double check it hasn't changed to 2:1
B) Accurately Test the memory latency
C) Test the stability

I downloaded Ryzen Master and this is what it's showing:
Click to expand...
1) that looks like it's running at 2:1 mode and you should really read up on the matter to make sure that you know what you're looking at. Read this.
2) AFAIK from having learnt from Buildzoid, testing latency using an app varies from run to run so you can get wildly different results. You can just use an online calculator to work out the theoretical latency. Looks like yours is running at 10ns as per this calc.
3) You need to run a combination of Testmem5 with the ANTA777 Extreme preset and Y-Cruncher for several hours to make sure it's stable and won't result in data corruption.

I'd suggest you try running 6200CL32 instead and then trying 6200CL30 bbbuuuut I think you need to read up on DDR overclocking to really understand what you're looking at and understand what's good vs what's better for you. Start here.

6400CL32 on paper looks great but it might not work as well for you as 6200CL32 or 6000CL30 which good RAM should be able to hit.
 
I have that GSkill kit. Overclocked them to CL28 but performance was a bit worse, so set them back to stock expo 2
 
That Buildzoid video is pretty great, it really helped me get the maximum out of my ram timings
I know that the latency in AIDA64 probably isn't as good as it should be but that seems to be an issue with the current AGESA as far as i can tell.
If anyone has any suggestions they would be welcome.

G6fsqTH.png
 
Last edited:
ToOo said:
That Buildzoid video is pretty great, it really helped me get the maximum out of my ram timings
I know that the latency in AIDA64 probably isn't as good as it should be but that seems to be an issue with the current AGESA as far as i can tell.
If anyone has any suggestions they would be welcome.

G6fsqTH.png
Click to expand...
Looking at my timings, compared with yours - Is there any obvious reason that people can see that I'm getting 81 ns latency in AIDA 64? What could I change to bring it down?

imbIMwM.png
 
My timings are just a lot tighter across the board, i'm running Hynix A-die so i know that i can push really hard on some of these timings. I think your kit is also Hynix but i'm not certain.

The big ones are tCL 30 vs 32,My TRFC is 390 vs 884, and FCLK 2133 vs 2000, I'm also running DDR5 Nitro at 1,2,0. The rest of the lesser timings are pretty minor but do add up.

You've bought a 6400 kit so you could try to run MCLK & UCKL at 3200, or alternatively you can watch the Buildzoid video which basically hand out the timings on a platter at DDR 6000
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom