Discussion Zen 5 Builders thread

[DaaQ]

Anyone want to weigh in on memory modules, would 32x2 or 24x2 be one's option? I know it depends on speed and latency also price. but if you were going for a kit what would you choose for 9800x3d?

Or even 2x64GB kit? NM all sold out

 

[Hail The Brain Slug]

Anyone want to weigh in on memory modules, would 32x2 or 24x2 be one's option? I know it depends on speed and latency also price. but if you were going for a kit what would you choose for 9800x3d?

Or even 2x64GB kit? NM all sold out

6000C30 EXPO in the biggest capacity you care to pay for (or need). 2x32 and 2x48 will have slightly better performance for being dual rank dimms.

 

[utahraptor]

I don't get the new CPU until Monday, but I got the RAM up and working. What a pain:



I got it running at the EXPO I config, 6000/CL 30. Going to do some more testing through gaming. I had to remove the RGB headers on the RAM sticks as they would not physically fit in my case against the radiator fans :*(

After 2 hours of Marvel Rivals no crashes. It seems sable at these settings.

 

[Makaveli]

6000C30 EXPO in the biggest capacity you care to pay for (or need). 2x32 and 2x48 will have slightly better performance for being dual rank dimms.

Single ranks kits overclock abit better 2x16 or 2x24

2x32 2x48 kits are all dual rank.

But doing the secondary timings is where the most gains are.

I went with 32x2 DR kit for my build.

 

[StefanR5R]

Stable memory := a suitable memchecker does not detect errors during a >24h run.

 

[Makaveli]

Stable memory := a suitable memchecker does not detect errors during a >24h run.

Not only does it need to not error in those tools. It needs to pass a month of my usual workflow on the machine before I rule it fully stable.

 

[utahraptor]

ya as long as it gets me through the work day and then gaming at night with no crash good enough for me heh

 

[StefanR5R]

Some of the differences between testing and everyday observation are:
– When an application software crashes, the cause may not be obvious. Software bug? Hardware defect?
– Data corruption in files may not be noticed when they are introduced, but days/ weeks/ months/ years later.
Tests of course aren't perfect; they need a proper test design. I deliberately put the ">" into the ">24h" suggestion.

edit: grammar​

 

[Hail The Brain Slug]

Some of the differences between testing and everyday observation are:
– When an application software crashes, the cause may not be obvious. Software bug? Hardware defect?
– Data corruption in files may not be noticed when they are introduced, but days/ weeks/ months/ years later.
Tests of course aren't perfect; they need a proper test design. I deliberately to put the ">" into the ">24h" suggestion.

I like to run 33 Passmark Memtest86 passes. 32 Passes to run every test pattern plus one pass to make up for the first pass being truncated and not a complete pass.

It also adds the element of long uptime which helps reveal instability.

I am also insane and on some systems it takes an entire week. On my AM5 system with 96GB it takes around 60 hours.

 

[IEC]

Latest AM5 build:

9600X on a Gigabyte B650M Gaming Plus Wifi with an AMD Wraith Prism RGB cooler, 32GB of DDR4-6400 CL36 with XMP profile loaded, and 2200 FCLK, 1:1 UCLK:MCLK.

Just to see if even bargain bin boards can do it.

 

[Sonikku]

Good news: new parts are here.

Bad news. The cd rom tray doesn't allow enough clearance for the motherboard by a couple mm.

Blasted proprietary cases... How do I remove these drive bays? They couldn't just use normal screws...

 

[Markfw]

See those 2 buttons and light using a tie to the screw hole in the case ? Thats because this damn case has some very odd proprietary cables going to the on/off/reset and lights. I bought a generic cable that has these at the end of it. "Blasted proprietary cases" indeed !!!!! seconded. and it is Zen 5, Turin is Zen 5.

 

[DaaQ]

6000C30 EXPO in the biggest capacity you care to pay for (or need). 2x32 and 2x48 will have slightly better performance for being dual rank dimms.

Single ranks kits overclock abit better 2x16 or 2x24

2x32 2x48 kits are all dual rank.

But doing the secondary timings is where the most gains are.

I went with 32x2 DR kit for my build.

Latest AM5 build:
View attachment 123525
9600X on a Gigabyte B650M Gaming Plus Wifi with an AMD Wraith Prism RGB cooler, 32GB of DDR4-6400 CL36 with XMP profile loaded, and 2200 FCLK, 1:1 UCLK:MCLK.

Just to see if even bargain bin boards can do it.

How much difference is there really between 26/36 30/36 vs like 30/40 (G.Skill puts the 1st two timings in their model numbers.

EDIT: Versus like 6400 at 32/39? That 400mhz difference will not be as large as 28/36 or 26/36, which should be able to go up higher in speed anyway seeing as it is a better bin correct?

Mainly asking because 26/36 cas + is $100 more than the 30/36 this is 6000mhz 2x32 DIMMs

What was the better IC chips the A-die or M-die?

 

[Sonikku]

I got the board to fit. My reward was Windows 11 not being as much of a nightmare to install this time as it was the last time. But while it was able to download updates and drivers without issue during the windows install process and I didn't need to use some obscure command prompt work-around to get it to do its fucking job, it still infuriates me that it DEMANDS to be done mid-install when I have few options and can't do anything but stare at the screen while it works when Windows 10 would save all that shit until after the base install. Ya know, when I can actually DO other stuff while that driver/windows update download and install process runs in the background. Why in blazes did Microsoft do this? No wonder nobody wants to leave Windows 10. I've hated 11 since day one and only use it out of necessity.

 

[Hail The Brain Slug]

How much difference is there really between 26/36 30/36 vs like 30/40 (G.Skill puts the 1st two timings in their model numbers.

EDIT: Versus like 6400 at 32/39? That 400mhz difference will not be as large as 28/36 or 26/36, which should be able to go up higher in speed anyway seeing as it is a better bin correct?

Mainly asking because 26/36 cas + is $100 more than the 30/36 this is 6000mhz 2x32 DIMMs

What was the better IC chips the A-die or M-die?

A few points:

• Dual rank dimms are much less likely to work at 6400 1:1 due to board signaling and IMC variance. You tend to only see single rank being used at 6400. If you want turn-key, practically guaranteed to work without hassle, 6000 is the highest I would buy in dual rank.
• 16 Gbit M-die is not the same as 24 Gbit M-die. While 16 Gbit A-die is typically considered superior to 16 Gbit M-die, 24 Gbit M-die has different characteristics that make it not as easy as "which one is a better IC?"
• If you're using an X3D CPU and youre gaming, the timing difference should be minimal to actual performance. Unless you want to spend the extra money to "know you bought the best" I'd save the $100 and get the 6000C30/36 kit.

 

[CakeMonster]

I've had 2 AM5 CPU's, its just not worth trying to see if higher than 6000 actually works. If you're OC hobbyist, instead tweak timings running at 6000. I now have a 96GB 6400 kit running at 6000 30 + various tweaked subtimings.

 

[DaaQ]

A few points:

• Dual rank dimms are much less likely to work at 6400 1:1 due to board signaling and IMC variance. You tend to only see single rank being used at 6400. If you want turn-key, practically guaranteed to work without hassle, 6000 is the highest I would buy in dual rank.
• 16 Gbit M-die is not the same as 24 Gbit M-die. While 16 Gbit A-die is typically considered superior to 16 Gbit M-die, 24 Gbit M-die has different characteristics that make it not as easy as "which one is a better IC?"
• If you're using an X3D CPU and youre gaming, the timing difference should be minimal to actual performance. Unless you want to spend the extra money to "know you bought the best" I'd save the $100 and get the 6000C30/36 kit.

Actually it is 100 cheaper for the 28/26 kit.
EDIT:
Think I will see if I can get these to lower timings at 6000 1st.

 

[Hail The Brain Slug]

Actually it is 100 cheaper for the 28/26 kit.

If the better timings kit is cheaper, what is even the question? What are the exact kits youre considering?

 

[DaaQ]

If the better timings kit is cheaper, what is even the question? What are the exact kits youre considering?

These are the 2. The make an offer wasn't available earlier.



Bottom is little more than. 100 more

 

[Hail The Brain Slug]

These are the 2. The make an offer wasn't available earlier.

View attachment 123641
View attachment 123642
Bottom is little more than. 100 more

Oh, yeah, okay. CL 26 is in no way worth that much of a premium. I'd just pick the CL28 kit.

Maybe if you were trying to buy the absolute best bin kit for some memoc benchmarking shennanigans.

 

[DaaQ]

Oh, yeah, okay. CL 26 is in no way worth that much of a premium. I'd just pick the CL28 kit.

Maybe if you were trying to buy the absolute best bin kit for some memoc benchmarking shennanigans.

No no, no OC epeen for me. II bought this lexar kit early on, and it WAS on the QVL for ASRock, but has siince came off, II thiink the 7400 or 7000 one is on there. I have has memory issues, using EXPO then backing down to 6200 resolves mostly, but I can't do 2200 Fclk 2133 is stable though.

But the biggest issue is D4 has again opened a memory leak. Which at 32GB takes about an hour for the crash dump.

 

[IEC]

How much difference is there really between 26/36 30/36 vs like 30/40 (G.Skill puts the 1st two timings in their model numbers.

EDIT: Versus like 6400 at 32/39? That 400mhz difference will not be as large as 28/36 or 26/36, which should be able to go up higher in speed anyway seeing as it is a better bin correct?

Mainly asking because 26/36 cas + is $100 more than the 30/36 this is 6000mhz 2x32 DIMMs

What was the better IC chips the A-die or M-die?

Ideally you will have a dual rank kit of 64GB or 96GB and just run it at 6000 CL28 or similar. It's a lot easier to run 6000-6100 and takes low vSOC and runs cool.

You can always take a faster kit and downclock it. I have a 6400 CL32 kit in my main rig and run it at 6000 CL28 (really 6030 because of 100.5 BCLK).