DDR5 Memory Performance Scaling with AMD Zen 5

[Dr. Dro]

Yeah, the pre-binned kits are about 50% more than 6000/30. You probably could get there on a regular set of A-Die with some FAFO.

View attachment 365090

Some of the price is also because it's Z5 Royal Neo (looks), but not as bad as I'd expect given that tightening is harder than just giving it a few extra MTs

 

[Deleted member 223428]

Some of the price is also because it's Z5 Royal Neo (looks), but not as bad as I'd expect given that tightening is harder than just giving it a few extra MTs

Yeah, there's also a 6400/30 kit for $149, but you're right - a good portion of that is bling.

 

[ir_cow]

Some of the price is also because it's Z5 Royal Neo (looks), but not as bad as I'd expect given that tightening is harder than just giving it a few extra MTs

Yep, you are paying for the Royal aesthetics. I got a DDR5-5600 JEDEC KLEVV (naked DIMM) kit that does 8000 CL36 with 1.45V. If you know how to change the timings manually or flash a new SPD profile, you can do better for half the price.

 

[catulitechup]

@W1zzard

Good review but can test with ryzen 5 8600G for see how much impact have 1:1 and 1:2 modes using igp aka 760M

Thanks

 

[Rowsol]

I would like to see a test where you test similar speeds with different timings, to see which has more effect on fps.
Example would be 6000,28 vs 6000,48 and 4800,40 vs 4800,20.

 

[x4it3n]

Thanks for that great review W1zzard!
Now we know that buying some DDR5 8000 kits for ZEN 5 is useless, unless we want to downclock them to 6400MHz and get crazy with low timings haha

 

[csendesmark]

Thank you @W1zzard @ir_cow !
Extremely useful for me, and thank Ir_cow for the advice in the other topic.
Just bought my Kingston Fury Renegade 6000MHz 96GB (KF560C32RSK2-96) instead of the other models I were eyeing between 6400~6800MHz for a lot higher price!

 

[ir_cow]

Thank you @W1zzard @ir_cow !
Extremely useful for me, and thank Ir_cow for the advice in the other topic.
Just bought my Kingston Fury Renegade 6000MHz 96GB (KF560C32RSK2-96) instead of the other models I were eyeing between 6400~6800MHz for a lot higher price!

That KF560C32RSK2-96 is a nice memory kit. Enjoy! If you decide to start changing some of the subtimings, I'm sure people on the forums here can help with that.

 

[jxdking]

It is tested with win11 22H2 with VBS on. And without inter CCD latency bios patch.
I don’t think the results are correct.
7200 should be much slower than 6000.

 

[sector-z]

Excellent Review waited that since a long time ! But 2 important Ram setup missing. Like I said in anothers Ram review @W1zzard

@ir_cow I know you tested 6000MHz cl 28 but next time if you can it will be interesting to have some OC like Buildzoid of is 6200MHz CL26 he have make it on couple kit board like Here again. Because we never see number about what it give in gaming

+ Many on Overclock.net are in the 8000MHz CL36CL to 32. 38 is just stock wirst possible. Having a test at the middle CL34 would have been great or if you can add it later . 8200MHz have been acheived to. With that 3 Ram config I think you tested everythings


Edit: ah just see you tested with a 360mm AIO ... so yeah some number should have been better if you have a custome loop or 420mm AIO at end on config where you was capped by the 95° . Having a Liquid Freezer II 420 + 3 Arctic Cooling P14 Max fan swaped is really cheap and a good setup for not be temp limited as cheap/performance solution.

 

[ir_cow]

Excellent Review waited that since a long time ! But 2 important Ram setup missing. Like I said in anothers Ram review @W1zzard

What is missing?

 

[sector-z]

What is missing?

The 6200MHz CL 26 and 8000MHz CL34. Because the 6400MHz 1:1 was nice but timing are very very basic

 

[ir_cow]

The 6200MHz CL 26 and 8000MHz CL34. Because the 6400MHz 1:1 was nice but timing are very very basic

I think the reason is Wiz was sticking to retail memory kits and no adjustments to the subtimings, like 99% of users would do. Those two memory kits don't exist yet either.

 

[Pepamami]

I think the reason is Wiz was sticking to retail memory kits and no adjustments to the subtimings, like 99% of users would do. Those two memory kits don't exist yet either.

Many DDR5 kits have really awful default subtimings, some DDR4 now too. Its just a good extra too see how it gonna work if set correctly.

For example this vid has like 227k+ views

Easy memory timings for Hynix DDR5 with Ryzen 7000

This vid also

AMD Ryzen 7000 Series [Zen 4] RAM/Memory Scaling feat. Tuned Memory

So I don't think its 99%, I think its way more, since DDR5 kits today have a lot of marketing garbage inside, instead of directly specifying Dies they are using or even number of ranks.
So you have to do everything manually nowdays.

 

[Launcestonian]

What's with the bios version for this board? its not even listed on the support page for the motherboard!
If you ran with the latest AGESA (bios version 2403), you'll see different results; I saw that in my own system with a 9700X & the Asrock X670E board so now achieving 2200FCLK with modest 1.21v SOC for 1:1 @ 6400MT/S! Could never do that with the previous AGESA no matter what timings, voltages or nitro training applied. Even trying to run 6400 1:1 with any FLCK value in the previous AGESA was impossible to stabilise under stress testing, now it's just easy...

Not trying to be off topic but here is the evidence of my claim for 2200FCLK ,1:1, 6400MT/S, CL30 with the new AGESA.

I think if you ran these tests again with the latest official non beta bios, you'll have different results. I don't know why TPU does this with Zen 5 testing, do they dislike AMD or something?

 

[ir_cow]

I think if you ran these tests again with the latest official non beta bios, you'll have different results. I don't know why TPU does this with Zen 5 testing, do they dislike AMD or something?

Why does TPU do a memory scaling article? Idk maybe it's because people would like to know what to expect when buying memory for AMD as not everyone knows the "fastest" isn't the best for the majority of users.

 

[render_guy]

This are the official AMD ram spec for 9950x:
Max Memory Speed
>2x1R DDR5-5600
>2x2R DDR5-5600
>4x1R DDR5-3600
>4x2R DDR5-3600

If I get the Corsair 192GB (4x48GB) DDR-5 5200 kit so I can only use 3600 speed?

 

[low]

What can we learn from this article?

Users go for AMD and Cache CPUs to save money from buying expensive DDR5 kits.
Now no one will go for DDR5-8000 and prices drop further. I saw some 48gb DDR5-8200 kits for under 130€.

9800x3D will be expensive due to high demand.
Biggest flaw on AMD CPUs is higher IDLE power consumption.

 

[danc]

The power consumption of IOD is weird. Normally with DDR8000, you can get away with 1.1-1.15v vsoc and it always have lower IOD power than 6400 1:1.

I think if you tweak DDR8000 a bit more, such as lowering VSOC and tighter CL/TRCD/TRP/TRAS/TRTP to 34/46/46/62/16 at 1.5-1.55v VDD, you will beat 6400 1:1 in more scenario and still have more leftover CPU package power.

 

[sector-z]

I think the reason is Wiz was sticking to retail memory kits and no adjustments to the subtimings, like 99% of users would do. Those two memory kits don't exist yet either.

Yes but if nobody test we will never know if the gain is good enough for take the time to tune . Having that 2 Ram setup I talked, can help many to decide if they will start for the first time to OC their memory or not. But without knowing if it's worth it .... yeah few people will start do it.

What's with the bios version for this board? its not even listed on the support page for the motherboard!
If you ran with the latest AGESA (bios version 2403), you'll see different results; I saw that in my own system with a 9700X & the Asrock X670E board so now achieving 2200FCLK with modest 1.21v SOC for 1:1 @ 6400MT/S! Could never do that with the previous AGESA no matter what timings, voltages or nitro training applied. Even trying to run 6400 1:1 with any FLCK value in the previous AGESA was impossible to stabilise under stress testing, now it's just easy...

Not trying to be off topic but here is the evidence of my claim for 2200FCLK ,1:1, 6400MT/S, CL30 with the new AGESA.
View attachment 365123

I think if you ran these tests again with the latest official non beta bios, you'll have different results. I don't know why TPU does this with Zen 5 testing, do they dislike AMD or something?

Have you tested 28-36-36-72 6400MHz ?

 

[W1zzard]

amazing job @W1zzard, thank you.

i have a question, is the controller in the 8000X processors any different from the 9000 ones? i've seen different memory behavior from 7000 / 8000 / 9000 series, from 7000 to 9000 it doenst appeared to have any improvements, but 8000 series APUs seem to have a better Memory controller, anyone knows why?

You mean Ryzen 7000 vs Ryzen 9000 I guess? Zen 4 vs Zen 5? The IO die is physically identical, but it's running slightly different firmware for Zen 5

I wonder if the 9950X3D and 9900X3D with an X870E board make any significant difference in the comparisons. I think might finally upgrade from my X99 workstation.

I don't think so, depends on your definition of significant though, but I'll definitely explore it, once all the pieces of the puzzle are together: X870E, Agesa, 24H2, X3D

 

[Nhonho]

6000 CL28 is the new "gold standard" for Zen 5, it's a very low latency kit. Not a budget-friendly option, nor the most common, but as DDR5 advances, should become more commonplace

So can you show us tests to see if this DDR5 6000 CL 28 memory, which must be quite expensive, really makes that much of a practical difference compared to other DDR5 6000 CL 32 to 36?

 

[kondamin]

OK, stick to +-6000 +-32 and you are good

must have been a very tedious test, tnx for the effort

 

[RootinTootinPootin]

and here's me dailying 8533MT's CL38..

 

[bug]

You can shrug off that idle power draw on the desktop, but I hope they can do something about it on their mobile parts.