Crosshair VIII 2501's for testing

[Baio73]

2311 was first BIOS with changes added in for 5000 series. Why are you asking this ?
I can run the 2702 BIOS, and we have similar systems, anyway 1302 was very stable BIOS for 3000 series but PBO really didn't work back then `imo.



Some Generalised Information:-
B-die is best for tinkering and overclockers tend to prefer it because it is more tolerant and able to adapt with predictable results and generally it's expensive (I think worth it). You can use SK Hynix, Crucial / Micron or Samsung (the only real manufacturers of RAM dies) and *generally any die for set and forget installations and maybe able to tighten timings a little to achieve better than average speeds using tools such as Ryzen Dram calculator.

*4x8 is sometimes harder to run but a kit specifically approved for use on AMD such as g.skill NEO branded would work out of the box. 4x8 will give you dual rank (it's fairly rare to see 8gb stick dual rank to start with).

*2x16 will most likely give you dual rank RAM (although should verify) and g.skill NEO branding are all approved for all Ryzen CPU's

*RAM can be verified for use, in three ways.
(1) Motherboard Manufacturer RAM QVL list.
(2) RAM Manufacturer QVL list.
(3) AMD RAM QVL list.
https://www.amd.com/en/products/ryze...-gen-amd-ryzen – not sure they have done 5000 series list but given the memory controller is identical fairly safe bet.

All lists are a valid way to have approval but not all RAM will be listed on all three. Ideally, as a popular song goes, 2 out of 3, aint bad. But even one is better than none and I would avoid not listed.

If your heart is fixed on Samsung b-die (newegg or amazion is likely source for a decent price) and this tool https://benzhaomin.github.io/bdiefinder/ in advanced mode can provide answers about if it is or isn't.

Specifics:-
I would be looking at 3600 kit CL16 2x16.
You can run DDR4 3200-3600 CL16 as an ideal in 2x8, 2x16, 4x8 (see note)

Note:- For any 4xRAM config be sure it has verification to run and be aware mixed kits are currently having more than usual troubles for the 5000 series and even 3000 series on latest BIOS.

Some known examples below,
2x16
https://www.newegg.com/global/au-en/...2E16820232860? – bdie kit
https://www.newegg.com/global/au-en/...2E16820232861? – Hynix
4x8
https://www.newegg.com/global/au-en/...2E16820232895? – bdie kit

Last thing, only listed g.skill Neo as become little rusted on it and there are many many many kits which work. Once upon a time, I was rusted on Corsair (with Intel prior to problems with Ryzen) and before that really liked and used to use OCZ & Mushkin. I've used Patriot Viper memory (4133 b-die), Team Group RAM and Crucial / Micron memory, all with Ryzen's without issues.

Hope you find this helpful.

Thanks for this 'little guide', they are all known piece of information, but having all of them in row it's very usefull.

My experience can unfortunately confirm all you wrote... I've passed from 2 2x8Gb different kits to 2 2x8Gb identical but still having no luck... I can push a single kit to 3800 CAS 14 but I0m having great troubles in running the 2 kits togheter at 3800 CAS 16. They all are G.Skill but not the Neo version.

What hurts me is that untill 2206 I could run the different kits smoothly, with later releases it's a drama.
I'm an old Asus mobos owner, but must say that with this generation they did not work well at all.

Baio

[pelaaja]

I havent seen any of the issues ppl are having. Im running a 5900x in my VIII Hero with 32 GB G.skill 3600mhz CL16. Im currently running the following without issues:
4800 / 4700 @ 1,33v
3800mhz CL16 @ 1,42v
FLCK 1900mhz
I have been loving my new 5900x.

[78Staff]

Honestly, 3600 CL16 is good enough. CL14 is extra few FPS and might shave a bit more off time dependant applications like rendering. CL18 kits we need to move away from CL xx gauge and look at the real latancy NS response time ... confused yeah me to... trying to say CL18 or even CL22 kits can be great as well.

This article explains why it bad measure, but we all use CL xx.
https://www.crucial.com/articles/abo...ed-and-latency

This site often has RAM kits actual NS rated speed
https://pcpartpicker.com/

2x16 is simply way to get rank2 advantage (as most are rank2 not rank1). While I run 4x8 not in habit recommending what is not work to keep running BIOS to BIOS. I know, I tinker and like it but I know not everyone likes doing this and would not do this on work type machine.

I will read up on the article - I am ready to replace my 4x8 with the Neo/Neo-Z 2 dimm kit, just need to decide on 2x16 or 2x32. 2x32 seems to be the safer option (over 4x16) if I want 64GB, while possibly being a titch slower, although maybe not noticeable in real world use... I just have to decide how much I really need 64GB lol... not all the time, but would be nice for running multiiple VM's or working on large RAW files I would think.

That was the desire of going with the 2x16's, I could jump to 4x16 by adding 2 more dimms later if desired. But that doesn't seem to be the case with Ryzen's IMC, at least currently.

[Jackalito]

Guys, I've started tinkering with my RAM 2x8GB F4-3600C15D-16GTZ (Samsung B-die, Single Rank), and I cannot make it stable beyond 3666/1833 with my new Ryzen 7 5800X. I'm currently using BIOS 2702. Is there any previous BIOS without the dreaded WHEA errors that could help going higher with the OC?

For example, checking this out, I can see 2 different people with CH8 currently enjoying 3800/1900, although they're using BIOS revisions 2311 and 2502
https://docs.google.com/spreadsheets...#gid=321590489


Thanks!

[78Staff]

I notice some sites are starting to list the new lower CAS GSKILL sets for the 5000 series - but they are 14-15-15-35 - I though Ryzen didn't do odd timings and just bumped it up 1 - so thus 14-15-15-35 basically becoming 14-16-16-36? Or is that for older gen chips?

[darkmessias94]

I notice some sites are starting to list the new lower CAS GSKILL sets for the 5000 series - but they are 14-15-15-35 - I though Ryzen didn't do odd timings and just bumped it up 1 - so thus 14-15-15-35 basically becoming 14-16-16-36? Or is that for older gen chips?

What you are refering to is the GearDownMode. If it is enabled it will force the CAS Latency (tCL) to be rounded up to an even number (tCL is the first number, so in your example 14-15-15-35, the other latencies are not affected by it. So the G.Skill kit will run at the advertised 14-15-15-35 with Ryzen.
Btw: You can also disable the GearDownMode in the BIOS settings. If you do so you can also run uneven timings like tCL 13.
Why would AMD implement such a feature you might ask? Well, you might have heard of the "Command Rate" before. Long story short: A lot of high-end RAM is not able to operate at 1T with higher frequency, therefore normally you would put it at 2T to get it stable. GearDownMode is kinda T1.5 - it offers higher compatibility than 1T while performing better than 2T - or vice versa being slower than 1T and less compatible than 2T. If your RAM can handle it, you can deactivate the GearDownMode and put it manually to 1T for the best performance.

[1ah1ksa]

Hi

I just build a new system and decided to stress test my Ryzen 7 5800x so i want to check whether cooling worked properly. during an CinebenchR20 stress test, the temps reached up to 83 degrees in ryzen master and hwinfo64 but today i opened AIDA64 Engineer ver 6.30.5500 i want to test my memory by using system stability test i found that it has cpu temperature So for idle it was the same temperature with all the ryzen master and hwinfo64

but when i benchmark my cpu using CinebenchR20 the temperature differences appears

HWinfo64 80.5c

ryzen master 79.77c

AIDA64 Engineer 69c

when full bench load https://imgur.com/gF9Q8Ki

here with Idle https://imgur.com/a/asLLRK4


Can anyone confirm that? Only in full load the temperature differences appears.

I am using ASUS x570 viii hero 2702 bios

ARCTIC Liquid Freezer II 360

Thank you

[RedSector73]

Polling rate of each is different, they are all right and wrong. Ryzen Master is the best choice of the bunch.

[Silent Scone@ROG]

Either Ryzen Master or HWInfo depending on what sensors you wish to monitor.

[wyliec2]

I primarily use Aida64 - there are several different CPU values - I always monitor CPU Package which is usually 10-12 degrees C hotter than CPU.