• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Trying to optimize ram Need Help!

Joined
Oct 27, 2022
Messages
150 (0.16/day)
Location
Fort Worth, Texas
System Name The TUF machine shh...
Processor Ryzen 7 5800X3D (4.5Ghz)
Motherboard TUF GAMING X570-PRO (WI-FI) BIOS 5031
Cooling EK 360mm AIO Elite, D-RGB, 2x240mm rads
Memory G.SKILL RIPJAWS V 32GB (2 x 16GB) DDR4 3900mhz cl18-20-21-20-29
Video Card(s) ASUS TUF RX 7900 XTX OC/PTM 7950 (Liquid Cooled)
Storage T-FORCE CARDEA ZERO Z330 1TB , Crucial P3 Plus 1TB
Display(s) Gigabyte G27FC A, G32QC A
Case Thermaltake Tower 500
Audio Device(s) R-120SW, R-100SW, Logitech X-240 2.1 Speakers, Skullcandy PLYR
Power Supply Corsair RM850x
Mouse Logitech G502 Hero
Keyboard Logitech G815
Software Windows 11x64 Home
Benchmark Scores Time Spy 21,266 6700XT(2x)/5800X3D 12,569 6700XT/5800X3D
Over the past few days i started tweaking my memory because of CPU Bus/Interconnect errors and hitting roadblocks with my memory kit.
ZenTimings_Screenshot_28746955.2957632.png
I am new to Ram tuning so don't question the numbers or voltage. This is the current configuration it seems stable enough.

ZenTimings_Screenshot_28746674.9935507.png
ZenTimings_Screenshot_28745184.3882544.png
These are other timings that I've tried. Anything close to 4000 MT/s spits errors like crazy. The 3866 MT/s tune works well but games crash easily.

XMP is 4000 MT/s CL 18-22-22-42-64 at 1.40V I think it is a Hynix CJR die
I found a form that is dealing with the same memory. https://www.techpowerup.com/forums/threads/is-this-b-die.260089/
 
Last edited:
Run MEMTEST86 for a full pass to validate your memory
(burn USB drive, boot to it in UEFI/BIOS..)

Don't worry about performance as you won't gain much especially with a large cache on the CPU. Just go for stability. If MEMTEST86 passes it's maybe 99% certain the memory is fine.

I don't know if you need to tune. I could be wrong, but I thought they'd give you multiple DOCP/XMP profiles to choose from. If there's something closer to 3800MHz I'd use that rather than trying to tune it yourself.

Your CPU settings also matter as the memory controller is in the CPU so memory errors could be the CPU itself. If you're overclocking, then set back to default values. AUTO?

If in doubt, take the CPU and Memory to their defaults, get it stable then go from there. I think Memory JEDEC is 2133MHz. You said 3866 was stable but Memtest86 might not say so. This test reads and writes in different combinations of memory access so it's fairly reliable.
 
I have been running 4000Mhz since I got the kit, it is stable But there is a boatload of CPU bus errors at 2000 FLCK. Just want to fix the errors and tune the lower clock speed, the tCL won't go lower than 18.
 
I need to know what is STABLE before I can help. Your CPU has a memory controller in it. So if you increase the frequency etc to the memory then you stress the memory controller in the CPU.

So SETTINGS related to memory can cause issues with the CPU.

From what I can tell, 2000 for the FCLK isn't always stable. Is this a default value?

Put the CPU and DRAM to their defaults and run MEMTEST86. I'm not certain what's stable, but again I'm guessing something like "AUTO" and find a DOCP/XMP profile that works with MEMTEST86. Select it in the BIOS. Usually there's a few options.

(And once you get it stable, use it for at least a week to confirm. Do what you want, but if I had your system I'd get it to stability and ignore overclocking.)
 
I have been running 4000Mhz since I got the kit, it is stable But there is a boatload of CPU bus errors at 2000 FLCK. Just want to fix the errors and tune the lower clock speed, the tCL won't go lower than 18.
It is not stable. If it was you would not be getting bus errors.
 
It is not stable. If it was you would not be getting bus errors.
Nothing major was happening that i could see but that's why I want/need to change it.
 
You do NOT seem to be listening to what I'm asking.

I want you to run MEMTEST86 with your CPU and system memory at DEFAULT values to confirm stability.

"Nothing major was happening that i could see but that's why I want/need to change it."

?? So if it's stable WHY do you want to change it?

*Anyway, I'm going to leave this thread if you're not interested in running Memtest86. THIS is the proper way to test a computer. By that I mean revert to default values and run tests to confirm stability with software designed for this purpose.

If that's not something you want to do I understand. But I won't help further.
 
I'm not at home so I can't do anything at the moment besides posting on forms.

Done. Thanks for your help even though you left prematurely.
 
Last edited:
Glad it worked. Whatever you did.
For the record I hadn't left yet. By "I won't help further" I meant if you weren't doing what I suggested.

If anybody else sees this, please post any SOLUTIONS to your problems when you fix them as that may help others that stumble across a forum like this.
 
Very few AM4 systems/CPUs can run bus and memory subsystems (FCLK/UCLK/MCLK) past 1800MHz
I suggest you start there and see if you are getting any bus interconnect errors, or any kind of WHEA.

I strongly suggest that while you are trying to find stability on tweaked memory to lower or completely disable any tweaks with curve optimizer. Once you find your FCLK:UCLK:MCLK and timings limits then you can start giving negative values on the V/F curve with Curve Optimizer.

Also you have a 5800X3D and these 3D CPUs do not need fast memory.

You can tweak of course for the fun of it but lower your expectations. There is no magic trick to fix these kind of errors. Most times lower speed fixes them. To start cranking up voltages won’t help either. FCLK likes low voltage. Just saying…

Try Thaiphoon Burner to determine the type of your DRAM ICs

Also do not compare your system with other systems even if the have the exact same DRAM kit.
Every CPU is different and the combination of CPU/Board/DRAM kit makes things way more complicated.
 
Last edited:
If you do some research and you'll see plenty of sites and videos covering memory speed scaling with 5800X3D; There's very little difference between loose-timing DDR4-3000 and tight-timing DDR4-3800.

If you want to learn to tune RAM for fun then don't let me stop you but the 5800X3D is insensitive to RAM bandwidth and you should just pick some nice stable timings unless you're just interested in tinkering for no real gains other than the sake of tinkering.

I personally wouldn't bother tuning DDR4 at this point, especially if you don't know what you're doing as DDR4 is on the way out and if you want to learn how to do memory tuning, learn on a DDR5 platform. At least then it's useful information you can use for the next 5-10 years. Becoming a DDR4 tuning expert is soon going to be a pointless achievement. Your goal with DDR4 and an X3D chip is to get the most rock-stable, never-have-to-mess-with-it-ever-again daily-driver setting, and not worry about raw memory bandwidth.

I use my 5800X3D with a Kingston Renegade DDR4-3600 kit at safe timings that are a good 15% slower than the fastest seemingly-stable timings that the 1usmus calculator recommends. I've tuned my kit to within an inch of instability with a 5950X and yes, it was a bit faster. With the 5800X3D there's absolutely zero point in bothering, and the difference between a good tune and a defaults is within the margin of measurement error and run-to-run variance.
 
Last edited:
Run testmem5 with anta777 extreme profile for one hour. If it shows even 1 errror, its not stable. In which case loosen timings, reduce frequency, or increase voltage. Or some combination of those things.
 
Last edited:
Back
Top