Bclk, Cache Speed, L2 >= L1? (Ryzen)

Nobu

Nobu

[H]F Junkie
Joined
Jun 7, 2007
Messages
10,023
After updating to latest bios, I was trying different OC settings, and while testing for mem stability I saw this:
_20170827_103049.JPG

L1 Cache: 58.67GB/s
L2 Cache: 58.98GB/s

Now, I know it was probably a fluke/error in the speed detection, but is it even possible for the L2 cache to be faster than the L1 cache?
For reference, was running 104bclk, 35.75 multi. Got errors shortly after this photo.

Lowered it to 103.50bclk and got a more reasonable 40-somethingGB/s, but still got errors. Reduced multi to 35.25 and loosened timings to 17, and my L2 is back above my L1 again. Still a bit unstable.
Edit:Hmm, thinking about it, maybe I should have posted in overclocking...though I'm more interested in the architecture than overclocking specifically.
 
Last edited:
Hmm, maybe it wasn't a fluke...AIDA64 shows L2 running darn near as fast as my L1. Need to go look at other's results to see if that's typical of Ryzen CPUs before I start freaking out. lol
(Edit: Yeah, seems typical, holy shit that L2 flies)
cachemem.png
 
Last edited:
Could it somehow be related to the fact that the memtest code itself fits in L1 and artificially lowers the speed reading?
 
I would think as long as it's not being accessed it wouldn't be a big deal. Especially, they could fill the L1 cache and all that would happen is the memtest code would be pushed to L2 or L3 and still directly accessible from any core (at least in the case of L2), with a slight latency bump.

Pieter3dnow said:
Supposedly this can have a beneficial effect as well :
tighter ram timings :
http://www.overclock.net/t/1624603/rog-crosshair-vi-overclocking-thread/20660#post_26178558
Click to expand...
Thanks, though I'm running hynix m-die, which seems a bit more difficult to pin down timings on. Doesn't help that I don't well understand their relations.
 
You must log in or register to reply here.
Back
Top