DDR5 performance and timings on Zen 4, need help

[neurotix]

 

First of all, what kind of memory do I have? What die? It doesn't say in Thaiphoon Burner.

 

Second of all:

 

 

 

This is my current bandwidth and latency, which seems low to me for running DDR5-6000 in synchronous mode. I've seen similar pics on Zen 4 with DDR5 at this speed and they were like 90GB/sec for both read and write, and had like 62ns latency.

 

I've never overclocked DDR5 before, if anyone with a similar kit can post their Zen Timings or someone could tell me what dies these are I can look on reddit or something for a screenshot of timings to use with this kit. Also wondering about upping the voltage, it's at 1.35v, and from what I gather 1.4v is the limit for DDR5?

 

Any help appreciated. Thank you.

 

@Fluxmaven since I know you have a Zen 4 setup, maybe you can help. And yes, I used the latest version of Thaiphoon Burner and that's what it says.

 

EDIT: I am also open and willing to try to overclock it to 6200 or 6400MHz and OC the fclk higher. I know it's stupid and doesn't matter, and I have no idea if it's good or not but my SP rating is like 115, if that helps.

Edited March 11, 2023 by neurotix

[Nikado7]

Cpuz will tell you under the spd tab.  Will say hynix or whatever you have.  

[Nikado7]

Actually it says Samsung in your typhoon shot.

[neurotix]

13 minutes ago, Nikado7 said:

Actually it says Samsung in your typhoon shot.

Yeah but it says nothing about what die it actually is and so I can't look up timings from other people using this memory on Zen 4...

[Nikado7]

1 minute ago, neurotix said:

Yeah but it says nothing about what die it actually is and so I can't look up timings from other people using this memory on Zen 4...

 

Ah.  No idea.  

[Nikado7]

According to this you have B-Die.  

G.Skill Trident Z5 RGB DDR5-6000 C36 Review: DDR5 Performance King | Tom's Hardware

WWW.TOMSHARDWARE.COM

---

Trident Z5 is raising the bar once again

 

 

Thing is, the ? on yours (K4R number..) is the determining factor I'm pretty sure, there should be a B there.  

 

Edited March 12, 2023 by Nikado7

[neurotix]

So I found someone's timings for this memory including bios screenshots of what voltage he was using etc. at OCN. I managed to do this but I have no idea if it's stable or not (about to test now)

 

 

He was actually running these timings at 6400MHz. Anyway, this is a huge improvement over the one I posted previously and if I have to drop back to 6000MHz, I will.

 

I really hope it's stable, going to run the memtest feature in the old Ryzen DRAM Calculator to 3200% coverage and see if it crashes, first I'm going to try Horizon 5 benchmark mode though to see how it handles games.

[neurotix]

Memtest passed to 3200%!

 

 

Though since this is getting benched for hwbot eventually, I'll know for sure whether it's stable or not after that.

 

I'm over the moon though, look at the difference:

 

 

[Nikado7]

6 hours ago, neurotix said:

Memtest passed to 3200%!

 

 

Though since this is getting benched for hwbot eventually, I'll know for sure whether it's stable or not after that.

 

I'm over the moon though, look at the difference:

 

 

 

So this is what I have to beat.  Why can't it be something easy.  

[neurotix]

Is one pass of memtest86 enough? I need my rig for other stuff atm.

 

 

[The Pook]

1 hour ago, neurotix said:

I need my rig for other stuff atm.

 

I stopped using MemTest all together, I've ran it overnight and passed and then failed in GSAT within an hour. GSAT (or TM5) are what most people are using nowadays anyway and you can use either within the OS. 

[tictoc]

20 minutes ago, The Pook said:

 

I stopped using MemTest all together, I've ran it overnight and passed and then failed in GSAT within an hour. GSAT (or TM5) are what most people are using nowadays anyway and you can use either within the OS. 

 

Longtime stressapptest user.  It almost always uncovers stability issues quicker and more reliably than memtest. 

 

Additionally, there are some other things you can do to hammer the memory and cache to look for instability.  Feeding pbzip2 a large (100GB+) mix of compressible and incompressible data (text files, pdfs, jpgs, etc.), has also helped me find edge case instability when dealing with large amounts of fast memory on DDR4 platforms.  I haven't played around with DDR5 yet, but the same principles should hold true. 

[neurotix]

2 hours ago, tictoc said:

 

Longtime stressapptest user.  It almost always uncovers stability issues quicker and more reliably than memtest. 

 

Additionally, there are some other things you can do to hammer the memory and cache to look for instability.  Feeding pbzip2 a large (100GB+) mix of compressible and incompressible data (text files, pdfs, jpgs, etc.), has also helped me find edge case instability when dealing with large amounts of fast memory on DDR4 platforms.  I haven't played around with DDR5 yet, but the same principles should hold true. 

I totally forgot about stressapptest! I haven't heard that in like 10 years. Yeah, I'll try that, thank you. How long should I run it?

[tictoc]

10 minutes ago, neurotix said:

I totally forgot about stressapptest! I haven't heard that in like 10 years. Yeah, I'll try that, thank you. How long should I run it?

 

Depending on how close to stable your memory is, it will usually error out in less than 30 minutes, but I usually give it an hour.  If that checks out then I'll usually run through a few other stress tests like mprime (Prime95) large FFTs and Blend.

[neurotix]

Tictoc, what command line arguments do you pass to stressapptest to check your systems? There's a lot and I'm not sure what to use, what size, etc

[The Pook]

depends how you use your system. if you're just gaming then 1 hour is plenty. 

 

I usually test for 30-60 minutes when dialing in the OC and then when I think I've got it dialed in I let it run overnight. 

 

stressapptest -s XXXX -M 29000 --pause_delay 9999999999 -W --max_errors 1

 

XXXX = time in seconds you want to test 

 

 

[tictoc]

1 hour ago, neurotix said:

Tictoc, what command line arguments do you pass to stressapptest to check your systems? There's a lot and I'm not sure what to use, what size, etc

 

Pretty much what Pook posted, although I forgo the --pause-delay, since I am building from source and that workaround has been fixed in the upstream version.

 

*Edit* Depending on how you're running it, you might need/want to allocate a different amount of memory, but what Pook posted (leaving 3GB available for other processes) should be more or less universally good unless you're doing some other crazy memory intensive things while running stressapptest.

Edited March 14, 2023 by tictoc

[neurotix]

So I downloaded stressapptest, it was in the repos (Linux Mint so probably in Ubuntu repos). I passed an hour but that's not enough for me. Linux is my main OS and I only use Win10 for benching and gaming (which doesn't happen often).

 

I need 100% uptime for my system for various reasons, so..

 

Hence, I am now running stressapptest for 4 hours. Wish me luck in passing it. Also, thank you both for your help. I remember telling people about stressapptest on OCN and posting screenshots and getting ignored, with people prefering x264 stress tests or IntelBurnTest. Glad to know this is popular now. Thanks to both of you for recommending this. Or rather, reminding me of it.

 

Ever tried byte-unixbench tictoc? That one is like the only benchmark I know of that will tell you your Instructions per Second in MIPs which you can compare against the timeline here: https://en.m.wikipedia.org/wiki/Instructions_per_second

Edited March 16, 2023 by neurotix

[neurotix]

Can anyone recommend changes in timings/other settings for my current OC?

 

 

For example, Samsung liked low ProcODT with my previous DDR4 B-Die kit. It was 34.3. I wonder about the rest of the subtimings, though the guy who posted these settings said that's the lowest he can get tRFC and have it post.

 

Any advice appreciated.