8 PACK MEMORY RANGE GROWING: SAY HELLO TO 8 PACK RIPPED EDITION & 32GB KITS!!!

Guest2 said:
Nice.
Is there any reason for going 4x8GB over 2x16GB? Timings on that are slightly better but it works out more expensive and can't use the other 2 slots in future (i doubt 64GB RAM will be needed any time soon mind)

Does Ryzen like having all 4 slots used?
Click to expand...

The new ryzen 5000 series seems to like all slots occupied, there was a thread on it and here is some testing https://www.youtube.com/watch?v=-UkGu6A-6sQ&feature=youtu.be
at the moment i am using some X4 sticks G-Skill trident Z 3600 which runs 3800 cas 16 but want to try some of this 8 pack team group cas 14 to see how it compares as ryzen likes fast memory. I think gigabyte did a great job with the Aorus extreme X570.
The bios layout isnt as good as the Asus ROG boards but all the settings are there and once you find your way around its fine. The Asus new dark hero also runs the 16 phase VRM's now so i expect that to be a top board also.

The Ryzen 5950x pulls some juice when over clocked and i like a 24/7 stable clock not on the brink of death clock. so 4.5ghz at 1.265v is good 24/7 clock, can get to 4.8ghz but need 1.3v plus and that gets toasty even under water for a 24/7 clock.
so ram speed is the next improvement.
 
vapor matt said:
The new ryzen 5000 series seems to like all slots occupied, there was a thread on it and here is some testing https://www.youtube.com/watch?v=-UkGu6A-6sQ&feature=youtu.be
Click to expand...

This is true but I'm pretty sure its been the case with all Ryzen generations. Also the 2x16gb kits should give the same performance benefits as 4 ram sticks as they are dual rank.

I've gone for the 2x16gb kit myself, just waiting on ocuk to dispatch my order. Ordered on Tuesday, stock appears to have arrived yesterday, still no dispatch notification...:(
 
SpudMaster said:
This is true but I'm pretty sure its been the case with all Ryzen generations. Also the 2x16gb kits should give the same performance benefits as 4 ram sticks as they are dual rank.

I've gone for the 2x16gb kit myself, just waiting on ocuk to dispatch my order. Ordered on Tuesday, stock appears to have arrived yesterday, still no dispatch notification...:(
Click to expand...

No dispatch for my order yet either, seems the focus is on GFX cards not the ram buyers lol although i paid for next day delivery!
 
vapor matt said:
The new ryzen 5000 series seems to like all slots occupied, there was a thread on it and here is some testing https://www.youtube.com/watch?v=-UkGu6A-6sQ&feature=youtu.be
Click to expand...

Classic misinformation by uninformed media. Dual ranks has always been more efficient than single ranks since ever, its nothing to do with Ryzen. The testing is fundamentally flawed as it does not take into account manual tuning of timings or clock speeds. This therefore tilts the testing in favour of multiple ranks of memory.

The only thing that piece proves is that if you plug and play and leave XMP and Motherboards to do auto settings that dual or quad ranks are usually faster than single ranks. Of course auto values for secondaries and tertiaries vary tremendously between different motherboard models and manufacturers, so its not even applicable beyond their own test hardware. The "conclusion" they come to is only applicable to their exact test hardware. Classic case of noobs jumping the gun while lacking full information and pretty bad when the noobs in question are media with a massive youtube following. This leaves people who know better fighting misinformation for years afterwards and getting told they're wrong because some tard who jumped in with both feet published a video on the internet. This kind of stuff makes me a sad panda and undermines everything I try to help people with.

I've bought top bin 2 x 16gb and equivalent top bin 2 x 8gb G.Skill kits and will publish fair test results using both fully fixed timings and fully tuned timings. There will be no game testing, only actual benchmark and stress testing so the raw numbers can be observed in a vacuum as any independent and informed test should be done. You can all come to your own conclusions then. Unfortunately I can only do this on AMD as I don't have a suitable modern Intel system to compare against. £400 out of my own pocket to disprove this crap.
 
Last edited:
@MrPils Can't agree more. You have no idea how many posts I've read on Reddit or here with "I hear 4 sticks is better than 2". That video being up is a disservice to information. It should be taken down and an apology video made for how misinformed they are. I'm not slating normal users for not knowing this but a channel with 1.1 million subscribers should be consulting this far better... I mean jesus Buildzoid does videos for them and not one person on their "staff" looked at this video he had made like 1-2 weeks prior:
 
Dropped in a 5800X in place of a 3600
RAM at stock speeds (right pic) at the moment but notice the different in custom latency

Is that down to the single CCX?

yxy33xI.png
 
Guest2 said:
Dropped in a 5800X in place of a 3600
RAM at stock speeds (right pic) at the moment but notice the different in custom latency

Is that down to the single CCX?
Click to expand...

Pretty sure its not reading correctly for Custom and Inner CCX latency. Inter CCX latency no longer exists as you only have 1 CCX. :)

The reported random latency decrease is what you've got from the 5000 series upgrade, you should now be able to get that down into the 50's with Ryzen 5000. Read Speed and Write speed will see a small gain from that too - less waiting around for data to flow.
 
@MrPils

I just wanted to take the time to thank you for all the information you've shared on the thread. Prior to reading your comments I thought RAM clocking was a voodoo magic and never really understood it. Now, I get it - the penny has dropped as they say. Without you imparting what you have learnt I would not be making the progress I a now with my B-Dies. And I'm certainly now versed to tackle any new sticks I happen to purchase if OcUK Black Friday the 32g sticks.

So, thank you again!
 
MrPils said:
Pretty sure its not reading correctly for Custom and Inner CCX latency. Inter CCX latency no longer exists as you only have 1 CCX. :)

The reported random latency decrease is what you've got from the 5000 series upgrade, you should now be able to get that down into the 50's with Ryzen 5000. Read Speed and Write speed will see a small gain from that too - less waiting around for data to flow.
Click to expand...
Isnt that correct then if its showing 00ns for inner ccx latency?

CAS14 timings

TjaCrl3.png
 
Last edited:
Price on the 32GB kit has creeped up to 240 now. I'm finding it a little hard to justify that considering you can get g.skill trident z for barely much more and it has better default timings. I wonder if anyone has an opinion on options compared to this 32GB kit?
 
Guest2 said:
Isnt that correct then if its showing 00ns for inner ccx latency?
Click to expand...

Inner ccx latency should be latency between cores within the ccx, so im fairly sure that is wrong too yeah. I dont think that should've changed much between 3000 and 5000 series - I can't see how it could be zero anyway. Honestly though I don't know, its expectations not confirmed information. :)
Maybe ask 1usmus on twitter?

Aida looks really nice man, good result there. 56ns is WAY under what 3000 was capable of :D

@Awahwah @Nitefly Thank you both for the kind words, I'm glad I was able to help :)
 
vapor matt said:
is this any good?
Click to expand...

That (sadly) is a brilliant example of how tFAW can be used to completely throttle memory performance. You should be looking to get that time at least 25 seconds lower, that is pretty much around the speed you should get at 2666mhz not 3800mhz.

Go back to 3600mhz, set your tRRDS to 4, tRRDL to 6 and tFAW to 16 then set tWTRS to 4 and tWTRL to 12, Disable Gear Down mode and set Command Rate to 1T. That should get you below 120 seconds for the total test time.
To tweak further after that you can then try lowering tRRDL (aim for 4) and tWTRL (aim for 8) for a little more performance or try pushing back up towards 3800mhz (don't do both at once, get one or the other stable first).

Good luck! :)
 
Last edited:
MrPils said:
That (sadly) is a brilliant example of how tFAW can be used to completely throttle memory performance. You should be looking to get that time at least 25 seconds lower, that is pretty much around the speed you should get at 2666mhz not 3800mhz.

Go back to 3600mhz, set your tRRDS to 4, tRRDL to 6 and tFAW to 16 then set tWTRS to 4 and tWTRL to 12, Disable Gear Down mode and set Command Rate to 1T. That should get you below 120 seconds for the total test time.
To tweak further after that you can then try lowering tRRDL (aim for 4) and tWTRL (aim for 8) for a little more performance or try pushing back up towards 3800mhz (don't do both at once, get one or the other stable first).

Good luck! :)
Click to expand...


thanks for that will try :)
 
Managed to shave the Aida latency down a bit by dropping tFAW
Thanks for all your help too @MrPils
All this info should really be in a RAM timings/general info sticky IMO
I forgot how addictive this stuff is. Thats why I do it. I sit here drinking my xmas present IPA keg trying to lower latency and increase RAM read write speeds :o

o7lP4Mw.png
 
Whats the Time / Best time difference?
Based on Matts scores with read/write speeds i probably should set 3800Mhz and CAS 16 instead of 3600Mhz CAS 14
@vapor matt , what's you Aida latency out of interest?
 
You must log in or register to reply here.
Back
Top Bottom