• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

Do you know if thats the case on all AMD motherboards welshrat, or all ASUS? or just your specific board? :)
Defo on Asus, but I am unsure on others

Just for clarification, on ASUS boards its slots A2 B2 if you have two sticks otherwise you wont get full speed.

EDIT - Just reading around and ASROCK etc recommend the same. I am only going off my experience but DOCP (XMP) would crash in slots A1 B1 but has been stable as a rock in A2 B2
 
Last edited:
Defo on Asus, but I am unsure on others

Just for clarification, on ASUS boards its slots A2 B2 if you have two sticks otherwise you wont get full speed.

EDIT - Just reading around and ASROCK etc recommend the same. I am only going off my experience but DOCP (XMP) would crash in slots A1 B1 but has been stable as a rock in A2 B2

All ASUS boards this gen (470) are daisychain topology, not sure on other vendors. Could be different, although most other vendors always opt for daisy chain.
 
I know on my main rig (intel) cpu consumption in the few games I tested almost halved going from 3000-14-14-41 to 3000-12-13-39 (and optimised secondary tertiary timings). Probably from drop in ram i/o wait.

On ryzen I done almost no game testing but I think it see's at the least similar benefits.

However with that said samsung B die has massive price premiums especially for higher binned stuff so I am not sure if thats worth it.
 
Last edited:
Defo on Asus, but I am unsure on others

Just for clarification, on ASUS boards its slots A2 B2 if you have two sticks otherwise you wont get full speed.

EDIT - Just reading around and ASROCK etc recommend the same. I am only going off my experience but DOCP (XMP) would crash in slots A1 B1 but has been stable as a rock in A2 B2

I am wondering if my next board will be asrock, given my existing issues on my asrock z370 board where the secondary dimms can only reach 3000mhz (compared to advertised 4000+) and that 90% of reports on the net for poor secondary dimm slots are from asrock boards, is making me question that choice. Asrock tech support when i contacted them even told me their secondary dimm slots are only designed to meet jedec spec, no overclocking, in my asrock b450 board, the manual even admits they spec'd lower showing how fast certain dimm kits will run in primary vs secondary slots. Sadly it seems only asus do t-topology dimm slots?
 
Wow! How did that impact actual game performance?

none as none of the games were cpu bottlenecked, but still is nice to see.

Division2 was the closest to been cpu bound, it was before the change averaging 40-50% cpu and spiking to the 80s, now its averaging in the 20s and spikes to about the 50s.
 
none as none of the games were cpu bottlenecked, but still is nice to see.

Division2 was the closest to been cpu bound, it was before the change averaging 40-50% cpu and spiking to the 80s, now its averaging in the 20s and spikes to about the 50s.

Is there any guides you followed? Seems a bit of a mine field from a quick Google. I couldn't move mine from xmp settings without getting crashes in games so gave up. Hours of testing after changing each setting with no stability wore me down
 
Is there any guides you followed? Seems a bit of a mine field from a quick Google. I couldn't move mine from xmp settings without getting crashes in games so gave up. Hours of testing after changing each setting with no stability wore me down

Drop tRFC to about 300
Increase tREFI to 65535

These two should be stable and will give you a noticeable latency and throghput boost, what they do is control how often ddr4 needs to refresh its bits, if they not refreshed then they bit flip and you get of course corruption.

Now basically the hotter ram runs at the more often it needs to be refreshed, so if your ram runs hot, you need to be conservative on these values, if they run cool you more likely to get away with more aggressive values.

So the issue with this is e.g. you might be stable on christmas day, but in the middle of a heatwave in july suddenly you getting blue screens due to the ram been hotter and the infrequent refresh not been good enough.

If you increasing your dram voltage which in turn makes dimms run hotter this will also potentially affect this.

After adjusting this do a 2 hour run of HCI and a 2 hour run of karhu ram tester. This will likely heat up the ram more then you will see in gaming, every day loads etc. If its ok, you should be good, but perhaps wise to check again in the summer.

If its unstable or you just paranoid then reduce tREFI, for my ram a value of about 11k is JEDEC spec for 3000mhz (the value depends on the clock speed of ram), doubling it over jedec space should be pretty safe, if you dont want to risk 65535.

For other ram timings ude the ryzen memory calculator, try it on the middle setting, and see how you get on with that, if its unstable revert to safe. Tuning tREFi and tRFC in my opinion is much easier than CAS etc.

Primary timings primarily are dependent on quality of ram and ram voltage. IMC voltage and quality may have a small affect. B die is probably safe for 24/7 usage up to about 1.45v when passively cooled even with maxed out tREFI.

Secondary timings apart from tRFC will have very limited impact on throughput but can help latency and can help make tuned primary timings stable when they otherwise wouldnt be, these tend to be set by the board manufacturer and not affected by XMP.

Third timings, can have a noticeable affect on throughput performance, especially tREFI, and these seem affected by IMC voltage more than any other timings. They effectively manipulate how efficient ram is at its clockspeed for throughput e.g. they can make 3000mhz ram faster than 3200mhz ram. The people who helped me indicate its motherboard quality and IMC voltage that primarily affects how tight these can go.

For me even on intel I used ryzen calculator, as a guide, someone helped me on reddit, and also I looked at timings posted by other overclockers, as close as possible to the clock speed of my ram and used those as a baseline.

Some people have got paranoid about how high tREFI as it caused them e.g. to have HCI runs fail overnight, but the heat generated by HCI is way in excess of normal system load. So whilst I cannot say for sure it be safe, I reckon it should be. The tREFI+TRFC combo has a bigger impact than primary timings. Also those guys with the failures were already pushing ram to extreme levels such as 4000CL14 1.5+v.

I will post a screenshot of my intel timings but this is intel, however about 70% of the timings were taken from the ryzen calculator so wouldnt be too different.

On my 2600X, I have hynix A die spec'd for 3000CL16, its stable at both 3000CL14 and 3200CL16 using the help of ryzen memory calculator, I ended up using 3000CL14 as it has better latency over 3200CL16.

Also to correct my earlier post its going from 14-14-14-31 to 12-13-13-28 on primaries.

bCjLInc.png


One of the reasons I want to switch my main rig to ryzen is the community is so much better, getting people to help me on intel was really hard, it seems that community only has extreme overclockers and no one else, whilst reddit ryzen users are providing a bunch of help to each other and of course we have the amd calculators.

Also on ryzen use CR1 not CR2. Disable geardown.
 
Last edited:
I know on my main rig (intel) cpu consumption in the few games I tested almost halved going from 3000-14-14-41 to 3000-12-13-39 (and optimised secondary tertiary timings). Probably from drop in ram i/o wait.

On ryzen I done almost no game testing but I think it see's at the least similar benefits.

However with that said samsung B die has massive price premiums especially for higher binned stuff so I am not sure if thats worth it.

So you reduced timings from CL14 to CL12 and your cpu utilisation went down by half.
I'm calling BS right there.
 
So you reduced timings from CL14 to CL12 and your cpu utilisation went down by half.
I'm calling BS right there.

No I did more then that.

Why would I BS? Whats in it for me.

Feel free to do the same yourself (also reread my posts as you didnt read properly think I only adjusted CL) to provide your own data.

I think just coming out calling BS based on absolutely nothing is BS in itself, very rude.
 
Last edited:
Back
Top Bottom