• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

AM5 6000MHZ Memory Timings Thread "SK Hynix"

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

SF101

The Printer Guru
Joined
Feb 9, 2002
So I have been fiddling with 6000-MHZ timings for a bit ASROCK does not always share the same naming scheme as other brands.

But it should be ok to follow , I don't see many others showing their timings but these are at my current the best configuration i have found for 6000-Mhz "30 CAS" timings.
I am using Kingston KF560C36BBEAK2-32 (FURY Beast 32GB DDR5 RGB 6000MHz CL36 AMD EXPO Certified Dual Channel Kit (2 x 16GB), Black)
Nothing super impressive with these DDR5 SK Hynix Memory sticks I can imagine most sticks doing the same lower grade or higher.

I should State right now I think these can be improved on by someone more experienced.

I will link 2 Pictures with my Bios timings listed because this is an ASROCK motherboard I'm not 100% that all the configuration names are exactly the same.

These sub timings improved my FPS by at-least 10FPS just by changing them.

And I've seen in some reviews such as the Gigabyte motherboards they are setting some of these timings automatically while the others are not .. the memory modules should accept them however.

And I have been running the modules at 1.4v instead of 1.35 for good measure.

It would be nice if this is a thread we can build off of , its been a while since we had one.

If anyone has any more recommendations then just add to this thread I will try to keep it somewhat updated.
 

Attachments

  • 20230114_175541[1].jpg
    20230114_175541[1].jpg
    1.3 MB · Views: 64
  • 20230114_175530[1].jpg
    20230114_175530[1].jpg
    1.3 MB · Views: 61
Would be better to post RAM-related threads in the RAM section. For some reason, I missed this thread.

I see that all brands have a bit different naming for some timings. I saw that ASUS corrected some timing names because users asked for that on ROG forums. On MSI Z790I Edge I have some confusing names and if not the additional description then I wouldn't know what is it. ASUS Mem TweakIt also has some different namings than ASRock or MSI software.

I wasn't playing AMD much as there are barely any motherboard review samples and I was focusing more on Intel. I bought ASUS Strix B650E-I Gaming last week and there is a chance that the next front page review will have something for AMD with EXPO profiles, so I will try to spend some more time with AMD in the next weeks.

So far I can tell you that I see very nice scaling up to 6000 but not so much above that. It doesn't really matter if you set 6200 or 6400, as results are similar in everything, even AIDA64 is not showing significantly better bandwidth/latency. 6400 doesn't work on every CPU/motherboard, as I see. 6600 works only on maybe 1% of setups.

- Going back to settings. I would set 6200 as the timings are about the same as at 6000. Since it's Hynix, then you can try CL28 or even CL26, but 26 will need 1.5V+. I will check other timings a bit later and add something here.
- Command Rate 1N typically works up to 6200, and 2N enables at 6400 or more (or it won't work at 1N).
- Check how high you can set fclk. It usually works at 2133-2200. On the Strix B650E-I mobo, it works at 2167MHz. At 2200 has problems with booting, while I remember it was working at 2200 on Crosshair X670E Gene. It improves RAM performance more than most RAM settings.
- Check how high tRef/tRefi (or similar) works. It's usually at 2-10k at auto, set it to the limit as higher = faster. Stability problems are sometimes starting at 20k+, sometimes it works at max.

The memory kit which I'm using right now works at 6200 CL26-36-36-36 1N, but so far I was testing it like that only at 1.55V. Max temps were 73°C without additional airflow and it was passing the AIDA64 stability test.


Maybe not perfect, but here are settings with fclk 2133 that passed 30 mins of the stability test. As I mentioned, this is at 1.55V. The same with CL30-36-36 /30-37-37 should work at much lower voltages.

6200c26.jpg
 
Last edited:
You think most of these DDR5 are ok to 1.55v? or is that the highest?
 
It all depends on temps. The higher the frequency, the lowest tolerance to high temps. I have no problems with stability at ~70°C and ~6000-6200, but the same temps will cause errors in stability tests at 7000+. For 7000+, I'm using additional airflow. At higher voltages is even recommended water cooling. You can typically keep RAM stable without additional airflow (unless you have it in a tight ITX box) at 1.40-1.45V and up to 7000-7200. This is why barely any brand sells DDR5 kits at more than 1.40V.
How high the temperature can be, depends highly on the modules. I had kits with the same IC and +/-5°C stability limit. I don't remember what clocks I set back then (something around 7000), but it was like one kit could be stable at up to 62°C and another at up to 67°C. So it's not like you find a guide that tells you exactly where the limit is. Most things you have to test on your own to find out.

Officially the maximum temp for commercial DDR5 is 85°C. Generally, DDR5 has a max temp of ~100°C (I don't remember exactly and have no time now to browse documents). Since all RAM chips are about the same, no matter if they're in server or home/office series, then nothing should happen even during long work at 85°C+, but it's hard to reach that. As I said, I have 72-73°C at 1.55V in stability tests on ITX mobo where modules are nearly touching each other. In a closed case with limited airflow, it would pass 80°C, but it will generate errors faster than break. A simple way, and annoying, to protect the hardware are blue screens in Windows ;)
 
Had some free time today got home early to test a few of your suggestions.

Ran at both 6000mhz and 6200mhz - 28-37-37-37-37-CR1 1.5v
I was getting better read and copy & Latency on 6000mhz - better write speeds on 6200mhz but as i said the latency is quite a bit higher.
Obviously when setting 6200mhz the Bridge clock drops with the /2 multiplier this might be one reason for these differences here Going to see if this is fixable Northbridge clock is doing 466.7mhz when using 6200mhz.

- Check how high tRef/tRefi (or similar) works. It's usually at 2-10k at auto, set it to the limit as higher = faster. Stability problems are sometimes starting at 20k+, sometimes it works at max.
ASROCK X670E Steel Series - Does not seem to have tRef/tRefi anywhere. so this one seems out. perhaps its another name in asrock but none of my settings can go over 2k in this bios i definitly dont see anything up to 10k.
- Command Rate 1N typically works up to 6200, and 2N enables at 6400 or more (or it won't work at 1N).
Oddly no CMD Rate either to choose from so I guess I should be glad its defaulting at 1T or at-least shows that in aida64.

- Check how high you can set fclk. It usually works at 2133-2200. On the Strix B650E-I mobo, it works at 2167MHz. At 2200 has problems with booting, while I remember it was working at 2200 on Crosshair X670E Gene. It improves RAM performance more than most RAM settings.
So far I've set @ 2167mhz its working fine as I can tell this did indeed increase performance.

I will say I'm not getting anywhere near your read / write/ copy /ms performance though so far.

ill revisit this tomorrow going to game tonight and see if its even stable.

Also not sure you noticed but did you catch what expo was setting my TRAS(80) and TRC(96) @ above.. You know its really weird , I bought a Z77 OC Formula from Asrock way back when.. And that board was amazing it also had some of the best timing config profiles I have used ever, atleast great starting profiles you just would choose the memory type and work from there (nick Shi profiles) and that really did influence me buying this motherboard.. what a giant step back in configuration options I'm finding on this Asrock Steel Series though and it was not what i consider cheap in price.
I'm working around it since i already own one and its in the chassis I dont plan on removing it all at this point and changing.. however I really cant recommend Asrock atm maybe the Taiichi is better but I wont take that risk in the future.
 
Last edited:
Had some free time today got home early to test a few of your suggestions.

Ran at both 6000mhz and 6200mhz - 28-37-37-37-37-CR1 1.5v
I was getting better read and copy & Latency on 6000mhz - better write speeds on 6200mhz but as i said the latency is quite a bit higher.
Obviously when setting 6200mhz the Bridge clock drops with the /2 multiplier this might be one reason for these differences here Going to see if this is fixable Northbridge clock is doing 466.7mhz when using 6200mhz.
You can set ratios manually. If not in the tab with OC settings, then it should be somewhere in the general AMD overclocking (in another tab, usually at the bottom of the list where you have to accept OC terms).

ASROCK X670E Steel Series - Does not seem to have tRef/tRefi anywhere. so this one seems out. perhaps its another name in asrock but none of my settings can go over 2k in this bios i definitly dont see anything up to 10k.

Oddly no CMD Rate either to choose from so I guess I should be glad its defaulting at 1T or at-least shows that in aida64.
It's not available on every BIOS. Command rate can be unavailable too. However, it works as I said at auto. It doesn't change much. On older platforms, it was visible, while on DDR5 can't really see it, also in synthetic tests.

So far I've set @ 2167mhz its working fine as I can tell this did indeed increase performance.

I will say I'm not getting anywhere near your read / write/ copy /ms performance though so far.

ill revisit this tomorrow going to game tonight and see if its even stable.

Also not sure you noticed but did you catch what expo was setting my TRAS(80) and TRC(96) @ above.. You know its really weird , I bought a Z77 OC Formula from Asrock way back when.. And that board was amazing it also had some of the best timing config profiles I have used ever, atleast great starting profiles you just would choose the memory type and work from there (nick Shi profiles) and that really did influence me buying this motherboard.. what a giant step back in configuration options I'm finding on this Asrock Steel Series though and it was not what i consider cheap in price.
I'm working around it since i already own one and its in the chassis I dont plan on removing it all at this point and changing.. however I really cant recommend Asrock atm maybe the Taiichi is better but I wont take that risk in the future.
ASRock is nowhere near a higher ASUS or MSI right now. I don't know why but they are not tuning BIOS well, and most of their motherboards have significantly lower RAM support than the competition. I mean, even tested by me, Z790 motherboards could work at a max of 6800 when the average Gigabyte (used to be known from RAM support issues) works at 7200-7600.
ASRock is still stable, and for AMD it shouldn't matter much as it still reaches 6200-6400. I just don't like where they are going with their ideas.
If I'm right, then Taichi is not much better.
 
Truly Bart. I've had some cold boot issues @ 6200mhz its not very forgiving I can bench stable on it but if I shut down sometimes a configuration will just not boot and when it does it will all be defaulted "+mem voltage doesn't seem to change this issue either"

Oddly i did have it cold posting stable 6200mhz - 28-36-36-36-36-1t but then I made a nothing change in the bios to the PBO or CPU OC or something (I forget what it was it wasn't memory related) and it just wouldn't cold boot at that anymore and when it did it was defaulting mem settings Basically I had to start over before it would work again flakey would be the word to describe what happened.

6000-28-36-36-36-36 works though (this also took a lot of playing to get it stable) atm posting and booting fine benching ok on it.

The latency is about 62.3ms +- on testing nothing in the 50ms like your getting.

Perhaps the mem controller is just weaker on this 7600X or its more board related things.
 
I noticed it has problems when the RAM is hot. Like when you test it for some time and restart the PC, then it won't boot on the first try at too low CL. I had no problems with CL28, but it was acting like that at CL26. Maybe it's because of a different RAM or motherboard (usually, it's a +/-1 ratio difference between various kits).

Tref on the screenshot is what helps a lot in synthetic benchmarks. Some other sub-timings too. On the other hand, I wouldn't care much about latency if it's for daily gaming.

1CCD/fewer core chips usually have lower bandwidth, but not latency. I don't know if you have 1 CCD or 2 CCD 7600X. Maybe Ryzen Master will show that. I know there are both versions around, but only for 7600X. 7800X and non-X chips are only 1CCD. All others are 2 CCD. As I mentioned in another thread, even if you find it somewhere, don't install BIOS with 1.0.0.4 AGESA as the motherboard may not recognize the CPU (issue related to 7600X with 2 CCD only).
 
Yes I've noticed that too.

if I boot the next day it boots fine.

If I shut down and then start it sometimes it wouldn't boot and I had to hit reset and then it would start fine.

The odd part is I wouldn't even call the memory hot mines running like 35-40C.

Your right I am mostly gaming on this but after timings improved FPS it sent me down the rabbit hole and I figured some people might appreciate another 20fps or so with improved timings its something we should be figuring out on overclockers.com anyways this is old hat for us here.

To be honest its been quite a few years since I've seen memory speed/timings improve FPS at all this much it just shows our platforms can now take advantage of it a lot more.

I think this just has 1CCD Ryzen master shows CCD0 This bios is 1.11.AS03
 
Then it looks like your CPU has 1 CCD.

The cold boot problem can be related to BIOS too. It's not always the same thing. I just see it at too tight timings. Maybe your motherboard is not tuned so well for tight timings. In general, there are no memory kits below CL30 at 6000+ and considering how high voltages have to be used, I'm not sure if it's even worth it for 24/7 work. If you have Hynix A-die kit, then it can do the same at maybe 0.05V less, so then you can go down from CL30 to CL28 at the same temps. There is still BIOS support and other things. Generally, there is more significant difference from sub-timings than CL only, and sub-timings won't make you to bump voltages so much.

I can't see so significant FPS differences. At 1080p I see like 10-15 FPS gain, but when there is already something around 150FPS+. However, I mainly use Ubisoft games as benchmarks (they have built-in benchmarks) and some other titles, but there are so many games on the market that hard to say how everything will act. Generally, games that use the CPU more, react to RAM speed better.
 
Yes true.. but I'm pushing 240+FPS 1440P on a Alienware AW2721D 240hz monitor.
Feels pretty buttery , and 20FPS doesn't seem much true but its a gain that's basically free FPS on the high end.

Every gain is meaningful otherwise I would have just stuck with my OC'd I9-7900X /x299 platform.

Between CPU/Mem etc Its gained me quite a lot of fps overall on the same GPU.
 
Recently I was testing some settings and at 1080p I saw even 50FPS gain between standard 6000 CL40 RAM and 6400 at tight timings. This was in the Tomb Raider with RTX4080. You can see these results in the latest RAM reviews on the front page.

I also started some longer stability tests with Ryzen 7600 and I see that memory read is about 15-20GB/s lower and memory copy is about 10GB/s lower than with Ryzen 7950X. Memory write is similar. Latency is about 4ns worse, so at 6400 CL30 with some timings already quite tight, I get 58ns+. Just to let you know that if you saw lower results on your 7600X then it's normal.
 
Thank you for this. SF101.
I had some 7200MHz rated (XMP) Ram and couldn't get it to work with my 7700x system! This was very helpful and I'm finally running it at 6000Hz with the timings/settings from your picture (original post)

I also have an ASRock Motherboard (but the Pro RS).

What have you found is the highest/better setting @SF101? Are you able to post some screen shots so I can copy? I am an absolutely n00b when it comes to OC... (hence why I bought 7200MHz XMP ram thinking it would be fine in an AM5 system).
 
Last edited:
7200 kits have better IC than lower series. It still uses Hynix IC but can work at lower voltages on AMD. I don't know what memory kit do you have, but Kingston 7200 works fine on AMD, just with lower frequencies. I mean, the motherboard uses an XMP profile, and you have to set the frequency manually.

I noticed that the 6200 is still stable on every motherboard that I tested. 6400 causes weird problems from time to time (like passes stability tests but crashes in games). I'm testing one ITX rig for full stability right now (the one in post #2, but with Ryzen 7600).
7200 kit should work at 6200 CL30-38-38-38 ~1.35-1.40V mem voltage (whatever it's called on ASRock). tWR 48, tRFC/tRFC2/tRFCSb around 490/460/390, tRRD_L 6 tRRD_S 4 (if it won't be stable, then 7 and 5). Most other timings are not really important and can leave them at auto.
 
Thank you for this. SF101.
I had some 7200MHz rated (XMP) Ram and couldn't get it to work with my 7700x system! This was very helpful and I'm finally running it at 6000Hz with the timings/settings from your picture (original post)

I also have an ASRock Motherboard (but the Pro RS).

What have you found is the highest/better setting @SF101? Are you able to post some screen shots so I can copy? I am an absolutely n00b when it comes to OC... (hence why I bought 7200MHz XMP ram thinking it would be fine in an AM5 system).
On my Asrock mobo the memory voltage is DRAM VDD - this is what Woomack is telling you to run at 1.4v , I have my VDD @ 1.43v.
 
These are my working sub timings these work @ 6200mhz for me.

Also if you plan to increase your infinity Fabric Frequency to 2167mhz I did find SOC/Uncore Voltave (VDD_SOC) increased to 1.415 to help with its stability on my system.

I was conflicted to post these but since ASROCK is so odd i will.
 

Attachments

  • 20230222_143812.jpg
    20230222_143812.jpg
    1.4 MB · Views: 30
  • 20230222_143822.jpg
    20230222_143822.jpg
    1.4 MB · Views: 28
Back