Hi!There would be a problem I can not figure out why the processor's L3 cache measurement results vary greatly.
Update to the latest build of Aida64 and things improve, but unfortunately the results are always rather inconsistent with this benchmark.
actually mine always is same with this benchmark.. here is what i have discovered. since around 2933 - 3500 is in ryzen mapping hole it seems to stres out the L3 cache in this area. also pushing the limits like 4000mhz creates same issue. but if you have an odvddp setting like asrock boards you can up this voltage to close the erroring map gap. i have seen this over and over same thing every time on the around about 3000mhz ram clock speeds. dont think this is just ryzen , every cpu even intel all have a ram error mapping hole it just shows up on these benches easier due to L3 showing and how ryzen uses it with the ram. i will drop some pics and show.
3600mhz here is over the map hole.
4000mhz here stress on L3... not related to mappping error or is it?
i have a 3333mhz pic i will post later i need to dig it up...definatley ram related things going on when you see L3 acting low or goofy..
ok , i couldnt find the original 3333mhz bench. but i have made a discovery this is getting to the bottom of things here. since this 4000 bench is all stock voltage and auto settings i am sure there is some things needed tweaked so i rule this out of this equasion . it seems if you stay in xfr mode by not using the bclk clock and not moving the multiplier past 36.25 to go into overclock mode this causes a multi ripple effect of things to happen L3 being one of those. as you see in this pic 3333 looks good on L3 why? well its because its not in xfr mode. also on top of this once you go into oc mode and move bclk up the north bridge pci-e lanes are being overclocked aswell. so as we all know bclk past 118.2 pci-e drops to gen 2 but at same time its over clocked so around 144 bclk your setting back at original gen 3 speeds on pci-e lanes .
ok so basically this is a forked issue but not a problem. wish i could find that non bclk 3333 pic sorry. to make this shorter basically it seems the instabilty revolves around the xfr. bclk and in general the northbridge speed.
if your having your L3 jump around try this if your board allows it. jump out of xfr mode and into oc mode by using the multipliers anything on bclk 100 is good but you need to put the cpu ratio past 36.25 to go into oc mode . ok now your set once you clock the cpu up it moves up northbridge which will then be fast enough to carry that L3 correctly and smooth rather than jumping around but remember the cpu ratio or ghz of the cpu runs northbridge higher so then you will see your L3 cache speeds increase and stabilize also L1 L2 ... one last thing makes sure before testing voltages are correct to carry this stable before thinking its not working as i have shown.. give this a try and watch the difference.
if this dosent fully fix the issue on L3. the only other thing i can think of would be basically poorly made ram sticks and or just not running smooth with ryzen for whatever reason of the ram sticks being not on par for ryzen.
as a final note: just because your in oc mode does not mean your overclocking. xfr mode will and can go higher than oc mode just not on every core at same time. and you can stay well below xfr turbo. so this is not overclocking unless you go past recommended speeds which i dont recommend. thank you.!
cache memory is static and its much faster than dynamic DDR4
some benchmarks need to be updated to cope with Ryzen
i suggest using the rig and update the BIOS and chipset etc as needed and carry on with business as usual
Retrieving data ...