Discussion Zen 5 Builders thread

[igor_kavinski]

Well, I have 7200 cas 34. @6600 I should be able to run cas 32 or 30.

From my rudimentary experience in overclocking RAM, every 200 MT/s less lets you decrease the latency by 2. So you could get 7000 CL32. 6800 CL30 and even 6600 CL28. But you would need to pump in some extra voltage for stability.

But let's say you are very unlucky. I still see 6600 CL32 being possible for you.

 

[Det0x]

Well, I have 7200 cas 34. @6600 I should be able to run cas 32 or 30. @Det0x ???

6600 CL30 is very easy for hynix a-die, i guess that's what you have ?

But sadly 6600 1:1 is not so easy for the rest of the system..
I'm pretty sure you wont be able to just enable EXPO/XMP and select 6600MT/s and have a stable system, without any other finetuning

 

[igor_kavinski]

without any other finetuning

Let's say we don't get into subtimings and just want a straightforward 6600 CL32 (so even crappy 7200 CL36 can manage that). Which component voltages would need to be increased? (Appreciate if no ASUS specific terminology is used )

 

[Markfw]

6600 CL30 is very easy for hynix a-die, i guess that's what you have ?

But sadly 6600 1:1 is not so easy for the rest of the system..
I'm pretty sure you wont be able to just enable EXPO/XMP and select 6600MT/s and have a stable system, without any other finetuning

Its on a 670E Taichi. Thats the best 670E out there IMO

 

[Det0x]

Let's say we don't get into subtimings and just want a straightforward 6600 CL32 (so even crappy 7200 CL36 can manage that). Which component voltages would need to be increased? (Appreciate if no ASUS specific terminology is used )

The same voltages that would need to be adjusted for Zen4 to run stable 6600MT/s 1:1
I cant give any universal voltages that will work for every cpu, even motherboards have different voltage sweetspots (VDDIO is a prime example for this)

The main voltages you would need to find the range for is: (they all affect each other)

VSOC
CLDO_VDDP
VDDIO (which need to be more or less "matched" with VDD and VDDQ on latest agesa)
Proc and RTT memory training values
Nitro delays
Sometimes CCD and IOD
And not least, your IMC must be capable of this speed in the first place.. 6600MT/s 1:1 if much harder to run than 8000MT/s 2:1 if you have capable motherboard (read not 2DPC)

 

[Markfw]

The same voltages that would need to be adjusted for Zen4 to run stable 6600MT/s 1:1
I cant give any universal voltages that will work for every cpu, even motherboards have different voltage sweetspots (VDDIO is a prime example for this)

The main voltages you would need to find the range for is: (they all affect each other)

VSOC
CLDO_VDDP
VDDIO (which need to be more or less "matched" with VDD and VDDQ on latest agesa)
Proc and RTT memory training values
Nitro delays
Sometimes CCD and IOD
And not least, your IMC must be capable of this speed in the first place.. 6600MT/s 1:1 if much harder to run than 8000MT/s 2:1 if you have capable motherboard (read not 2DPC)

So, where does the Taichi fall in this ? and what values do you recommend for the above voltages ?

 

[IEC]

So, where does the Taichi fall in this ? and what values do you recommend for the above voltages ?

Any board with 4 DIMM slots is going to underperform in memory OC versus 2 DIMM ITX boards.

I strongly doubt you will get 6600 speeds working 1:1 at anything resembling out of the box settings/plug and play. Even if your CPU's IMC is capable.

6400 CL32 is more likely to be doable without much fiddling. But still TBD.

 

[B-Riz]

I think I may run up to MC and grab a 9600X tonight to play with, but am really waiting for 9800X3D.

 

[igor_kavinski]

I think I may run up to MC and grab a 9600X tonight to play with, but am really waiting for 9800X3D.

Buy the one whose serial number adds up to 8. It will either be crap or a golden sample.

 

[lixlax]

I'm a bit disappointed in the performance as most here, but I'm still going to upgrade to 9700x as was the plan. For some reason I want to do it even more because of the hate it gets.

 

[B-Riz]

I'm a bit disappointed in the performance as most here, but I'm still going to upgrade to 9700x as was the plan. For some reason I want to do it even more because of the hate it gets.

Intel's stuff is eating chips inside out, but AMD is the bad guy for this release 🤣

 

[Hotrod2go]

Just ordered my 9700X, interesting OC days ahead!

 

[Thunder 57]

Since the lack of advantment in games seems to be the primiary disappointment, may I ask something of you brave new 9700X owners? Try some game benches at 105W TDP with SMT off. I'm wondering if that 2x4 decoder is hurting client computing rather than if they had gone say 6 wide. I've got nothing to back that up at all other than it looked like an easy way to keep decoders similar to previous designs where Zen 5 is still closer to 4 wide than 6, let alone 8.

 

[igor_kavinski]

Games will eventually run faster on Zen 5 when their game engines are updated to use the AVX-512 version of memcpy and other related functions.

 

[MS_AT]

Games will eventually run faster on Zen 5 when their game engines are updated to use the AVX-512 version of memcpy and other related functions.

Problem is Zen5 might be already obsoleted before AVX-512 reaches sensible availability that devs will consider using it. Either AVX10/256 will win the market if Intel has its way or AMD will release newer replacements given the pace of adoption. [Every CPU since Haswell had AVX2, yet only now some games are requiring it to run, probably because latest console gen supports AVX2]. Oh and memcpy and friends are a special case, as they might be statically compiled in or linked dynamically from system libraries.

 

[B-Riz]

Zen5 shenanigans mode in full force! 🤣

I did not make it to MC last night, the wife had a commitment I had forgotten about and then I had to be a responsible adult and do stuff around the house :|

BUT, we have l00t incoming, for science! Sadly, I am out of town all next week on last hurrah vacation before school starts for the offspring.

It worked out better in the end, I had Amazon discount points and my state of residence (Ohio) was having sales tax holiday for anything $499 or less, so win win?

The board is open box "good" condition, I have usually had good luck with these, but this is the first AM5 open box from Amazon I have tried, hopefully all the pins are looking good.

Memory was just the cheapest 5600 according to PCPartPicker, and Amazon reviews say it is an easy OC to 6000.

Coolers are in my ~$20 range, I have wasted too much money on coolers over the years that never got many socket updates, so, that is why I picked those two, which ever one I like the best in person goes on the 9600X.

Video card is tricky, my highest end is a 7900XT, so, no 4090 full send results

I have 6700XT, 3070, 3080, 3080 Ti I can try out when time allows, but I will not be able to do big chungus testing like the dedicated people, too much other life going on

 

[B-Riz]

I'm a bit disappointed in the performance as most here, but I'm still going to upgrade to 9700x as was the plan. For some reason I want to do it even more because of the hate it gets.

After watching the Tech YES City 9600X review, I feel most big channels missed the .1% lows highlight, Zen5 is ground up, and is a beast. 9800X3D is going to actually be an upgrade over 7800X3D 😂😂😂

 

[igor_kavinski]

After watching the Tech YES City 9600X review, I feel most big channels missed the .1% lows highlight, Zen5 is ground up, and is a beast. 9800X3D is going to actually be an upgrade over 7800X3D 😂😂😂

View attachment 104905
View attachment 104904
View attachment 104903View attachment 104902

EXCELLENT FIND!

That's the best entry level non-X3D 6 core chip then. The BG3 result is staggering!

So it's definitely not a cache issue there coz the 7800X3D is also faltering. Something about that game is really punishing and the fact that 5700X3D is doing better than the 5800X3D leads me to believe that it's heating up the chip too quickly and not letting the 5800X3D boost as high and as frequently.

 

[Hail The Brain Slug]

EXCELLENT FIND!

That's the best entry level non-X3D 6 core chip then. The BG3 result is staggering!

I'd like to know the relevance of game results using a $1000 graphics card at 1080p low.

I'm not dismissing the results, but they have very little value being completely unrealistic configuration. Maybe proving that this improvement in lows happens on an actual configuration someone in real life would use would have more value.

 

[igor_kavinski]

I'm not dismissing the results, but they have very little value being completely unrealistic configuration. Maybe proving that this improvement in lows happens on an actual configuration someone in real life would use would have more value.

I'm sure @B-Riz plans to shed more light on that

 

[In2Photos]

I'd like to know the relevance of game results using a $1000 graphics card at 1080p low.

I'm not dismissing the results, but they have very little value being completely unrealistic configuration. Maybe proving that this improvement in lows happens on an actual configuration someone in real life would use would have more value.

The idea is to remove the GPU as a potential bottleneck. Since they are testing CPU performance they run resolution at a lower setting, putting more emphasis on the CPU rather than the GPU.

I agree, in real world use this pairing would never be used together.

 

[Hail The Brain Slug]

The idea is to remove the GPU as a potential bottleneck. Since they are testing CPU performance they run resolution at a lower setting, putting more emphasis on the CPU rather than the GPU.

I agree, in real world use this pairing would never be used together.

Changing graphics settings alters the load on the CPU from the engine, in addition game engines are more complicated than that - interplay of several threads all waiting on work from one another. Running an unrealistic config like this doesn't simply "remove the GPU bottleneck" but fundamentally alters the performance behavior altogether.

 

[In2Photos]

Changing graphics settings alters the load on the CPU from the engine, in addition game engines are more complicated than that - interplay of several threads all waiting on work from one another. Running an unrealistic config like this doesn't simply "remove the GPU bottleneck" but fundamentally alters the performance behavior altogether.

So how do you suggest the reviewers do it then? What GPU should they all use?

 

[B-Riz]

I'm sure @B-Riz plans to shed more light on that

I can, lol, I have a 5700XT 8GB, 1080 Ti 11GB, 5500 XT 8GB and a 480 8GB and 480 4GB, will be happy to futz around in 2 weeks with them

 

[CakeMonster]

Its not perfect, but its certainly preferable to benchmark the CPU's impact on game performance when you take the GPU out of the equation. 1080p is also increasingly quite reasonable, as the input resolution for various upscalers are in that area.