Just wanted to add for the record, that on my MSI B650 Edge WiFi board with v17 bios, booting a 32GB kit at 4800 or 5200 & with manual primary timing changes that I entered did not require memory training. The system booted up like a normal system should. Perhaps this is also to do with pushing memory bandwidth beyond AMD officially supported spec for Zen 4 IMC which is 5200MT/S?Is there any evidence that memory training is needed on every boot up?
Here’s what Crucial has to say about it:
DDR5 systems may require training
A small number of DDR5 systems and motherboards require a period of "training" newly installed memory to work with the system and allow access to the UEFI, or otherwise to complete the system's power-on self-test (POST).www.crucial.com
”A small number of DDR5 systems and motherboards require a period of "training" newly installed memory to work with the system and allow access to the UEFI, or otherwise to complete the system's power-on self-test (POST).
[…]
After this process is complete, future reboots will not require this wait [training], unless hardware is changed again in the future or other modifications are made in the system's BIOS or UEFI, in which case another delayed startup for this process may be seen.”
The SOC voltage issue from some time ago now is certainly fixed limiting it to 1.30v max only. Yes, If I am to continue using this board for OC I'll have to get one of those eternal switches to clear the CMOS. But one should not have to do that when the Bios recovery screen can enter into the failed boot process & prompt the user to reconfigure their system like GIgabyte have done on my B650 board from them. Don't know why MSI can't do this. I only had to manually clear the CMOS on that board once within a couple months of experimenting with RAM OC. Completely different situation with this MSI board.
Why MSI?
I've had more trouble on my daughter's Gigabyte B650 board than I have with my MSI B650 board.
I haven't had any trouble with USB devices on either AM5 platform at my house. I believe with the x670e boards you get 10gb lan over 2.5 with b650. When I made my purchase I couldn't see spending the extra money on x670E.
Could have been early BIOS issues. Are you still having issues with it? From what I've read, Gigabyte AM5 mobos so far have the best DDR5 support and shortest boot times. And @biostud is pretty pleased with his ASROCK mobo. I would have remembered if someone here had sung the praises for his MSI AM5 mobo.
Because I narrowed it down to those 2 motherboard. However, I will consider the Gigabyte B650 Aorus Pro.
Does this really matter if I plan on using DDR5-5600 JEDEC memory without XMP/EXPO and run it at the JEDEC 5200 profile setting? I plan on getting the 64GB Crucial Pro DDR5-5600 kit, which should have a JEDEC 5200 profile and I don't intend to use anything other than the JEDEC 5200 profile as that's what is the fastest memory officially supported by the Ryzen 7800X3D. I've read some user reviews that some people with Ryzen 7000 series CPUs can't run DDR5-6000 at their rated 6000 speed without issues and that companies consider XMP an overclock and warranty will be voided if the CPU manufacturer finds out that XMP was used such as if you tell them that you had XMP enabled in the BIOS.
No, B650 is just as reliable as X670. I’d recommend Asrock for any AM5 motherboard but MSI is fine as well.
Why leave performance on the table? I believe the latest AGESA is happy with DDR5-6400. Go with that.
No it's been working fine lately. Had issues with memory on first hoot. Took a few hours to get things working. She was also having issues with the Ethernet port randomly not working on boot up. Didn't even show up in device manager. She would have to reboot it. I don't remember what fixed it. It's possible it was a BIOS update.
I just don't want to take my chances. If I end up with a 7800X3D that can't run at 6000 memory speeds stable because the memory controller on it wasn't up to it and it wasn't actually defective memory, I don't want to fiddle around with memory controller/memory voltages, timings, or even lowering the speed settings for the memory. I wouldn't be satisfied if I bought 6000 speed memory but at the end I could only get it stable at 5800 and even then I would still question if it's going to remain stable at 5800 speed. I might take my chances buying a DDR5-6000 kit if there is one that actually has a JEDEC 5200 profile with native 5200 speed chips, but I won't take my chances buying a DDR5-6000 kit that only goes up to 4800 for it's JEDEC profiles and only has 4800 native speed chips. I want to fall back on the fastest officially support memory speed of the CPU including the timings and memory voltage if it's the memory controller on my particular CPU that has issues with 6000 speed. This reminds me that when I used to own a Pentium 166MMX, some people said at that time that they were running at at an 83MHz bus speed on that CPU for a performance increase without overclock the CPU. I tried that but I ended up getting frequent illegal operation messages in Internet Explorer and starting up Flight Simulator 98, until I went back to the officially supported bus speed of 66MHz for that chip, despite owning a motherboard that supported the 83MHz bus speed (SuperMicro P5MMS with the Intel 430TX chipset) and SDRAM with 10ns latency chips. I also had a Maxtor 3.5GB hard drive that was flaky at 83MHz bus speed.
Well, next year JEDEC rated DDR5-6400 kits willl become commonplace so maybe you should wait.
LinpackXtreme v1.1.5 &/or Prime95 small FFTs. With Linpack at least 10 runs with 10GB model (anything higher is experimental according to the author) , check consistency of Gigaflops output for all runs, this will show if the FCLK is stable with the CPU overclock. With p95, at least a few hrs with no failed cores.
Evidently Nitro was introduced after v15, because I don't have those options. I also don't use the Settings menu to set up my OC and RAM settings. I use the OC menu.
Interesting point regarding when DDR5 memory training is needed!
1. DDR5-5200 will still trigger a memory training if memory context restore is off on a full power cycle reboot. If you just shut down windows this doesn’t happen. If you’re running lower memory speeds the training session won’t be as long compared to DDR5-6000.
ditto. This whole memory training thing is all blown out of proportion. If its that big a deal, don't power it down. If its not, wait the extra few seconds.
How come Hotrod2go did not have to perform any memory training on his Zen4 system then?
If standard JEDEC DDR5-5200/5600 modules are used and they default to 5200 speed in the BIOS with Ryzen 7000 series, does memory training still occur? I'm not referring to DDR5-5200 modules that require XMP or manual settings to run at 5200.
He either had MCR off or was running lower memory speeds where the training only took a few seconds.
Yes, it's a mandatory requirement (same goes for Intel, it's just a part of modern memory spec). The memory training is just faster and doesn't need as tight of tolerances when running lower memory speeds.
Facebook
Twitter