Need help with 5600non-x my 3600 is faster?

Hello. I have been having issues with my 5600 not running as good as my 3600 system and i cant figure it out. 3600 is on b450 tomahawk. 5600 is on b550 tomahawk. tried the 5600 on the b450 nothing changed. dint really touch memory for these test. i know i can run cl 13 on the 5600 but it dont change much


5600 newest bench
thanks for any suggestions. i assume the big difference is the fclk? but should it be that different?
system specs
Report.txt (815.1 KB)

Have you tried running R5-5600 with DDR4-3200 and DDR4-3600?

was worth a shot. no change tho
3600 ddr4
3200

XMP enabled on both? Same number of ranks per channel (and preferably same number of DIMMs per channel) on both?

It probably just comes down to training. It’s kind of a crap shoot on AM4 unless you do a full tune a la this guide. If one board is prioritizing stability over performance and training more conservative values of tREFI, tRFC, etc. that can easily explain the difference. Or one kit is giving additional timings in the XMP. Or one kit is running hotter. If you’re not doing an apples-to-apples comparison with the same board and RAM kit then you simply have too many variables.

If you didn’t also disable Gear Down Mode it will get rounded up to CL14.

the 5600 4 dimms


the 3600 2 dimms

i think the team group stuff is faster but not by much? You are likely right about the timings. Tried 1t and gear down it did nothing. i looked at some old bench results on oc3d and their overclocked 3600 was sligly faster then the not overclocked 5600x they tested. maybe i have nothing to worry about? the highest i have gotten the 5600 with mem oc is just over 51k read in aida. seen others getting 56k reads tho. i might try 2 crucial dimms in each system see if anything changes. thanks for the assist.

Are these on fresh installs with the internet connection cut? Same update version?
Is the f clock and uclock running 1:1
Sorry haven’t read all the posts yet
Overall amount of ranks you have the same on each platform just one has 2 per dimm

Idk if it’s better to have all your ranks on just 2 dimms or to spread them out to 4

they are both fresh installs with updates as of dec 2022. i have uclk set to uclk=memclock but aida reports 56:3 on the 5600 and 57:3 on the 3600/ i looked this up earlier and seen that its normal. i dont know the math but it ='s 1:1 from what i read. i could be totally wrong tho. i cant find infinity fabric benchmarks but i’l keep looking, someone must have that much free time right?
i know buildzoid goes on about how big fclk is but thats a huge performance difference. i’l have to do some more test tomorrow.
i literally set up both systems the same tho. b450 and b550 tomahawk have same bios mostly if not identical.
i’l try the teamgroup dimms on the 5600 tmro its all i can think to do. the 5600 has stronger memory latency but i duno lol it could just be memory is not good. micron e die is not known for being amazing.
i just thought. i can lower the 3600 to 3733 easily and test. i’l do it now
thanks for the help.

1 Like

You’d be surprised at how much tertiary timings matters and how different they are sku to sku brand to brand

We have a 4000 kit that loses to 3600 kits

2 Likes

ok i lowered the 3600 to 3733/1866fclk. its still crushing. maybe i just have a golden egg 3600?


my tfaw and stuff is not set properly but neither is the 3600. thing is on fire. i get better benches with it then most tech review sites did. frustrating tho i want the 5600 crushing too, should just take my 3600 back from the wife and give her the 5600 lol she wont notice

Could be, I mean 3800 pretty high for a 3000 series

i’l swap memory tomorrow, thanks for the help

Also check out curve optimizer and pbo on the 5500
I think it could net you some extra CPU mhz

is that on ryzen master? i usually just manual. maybe it can tell me more tho. it will post 4.675 but temps become an issue. nh d15. havent got into single ccd oc either just dont have time lately. cheers

It’s under the advanced bios tab, AMD overclocking, precision boost overdrive set to manual

Try curved optimizer set to negative and 10 and play with it from there
You want to test single and few core loads
Greater magnitude values means less voltage which means less heater per frequency and more power headroom

Setting the frequency override to positive values means you cant get as much from curve optimizer but higher single core freq

Setting frequency override to negative values means you can get more out of curve optimizer and higher all core load frequency

i reset bios default and messed around with every option it did nothing sadly but i discovered i have the 1900 fclk hole. it just posted 1933!!! gonna keep pushing it.

Be careful of “tolerable” invisible errors
It can clog up your performance

Maybe one stick is bad? Usually men perf increases when you add more ranks not decrease

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.