I am new in this forum and I’ve decided to get in this insane adventure as well. I already have a 3950X and 5950X systems and I’ve build a brand new 7950X on an ROG CROSSHAIR X670E EXTREME. I am a 3D artist so 128GB are a must for me.
At 3600MHz the system is very stable and had no issues at all till now, I used the workstation for only 2 days but no crashes or other crazy things.
The only thing I did was to set manually the speed at 3600 (recommended by AMD if 4 DDR5 banks /128GB are used) and the rest is set to auto.
I tried 4600Mhz but my system was blocked on the Memory Test step and I had to pull out 2 RAM banks to be able to get in the BIOS and go back to 3600.
Now, I imagine that setting the ram speed at 4600 and leave all the rest on AUTO the system won’t work so well, do you suggest any other settings to put manually so at least it can start? As we can’t use any XMP or EXPO profiles the manual mode is a bit new to me. I am not still very expert on RAM OC yet.
I see some ZEN Timmings tables from you here but is it enought to insert those values or I need to do something else as well?
I know, my questions may seem a bit basic and hard to answer so no worries if you can’t help. I’ll follow you in any case to see how things will evolve.
Just to let you know another thing, I wrote to GIGABYTE and ASUS support a month ago and both of them confirmed (naturally), that if we need a 128GB configuration, the solution is to buy modules of 4800 speed and let the system downgrade them to 3600.
I am not very positive on a future radical change on this, 128GB is a lot for the memory controller so I gues is a bit normal having difficulties to go over it.
Hi I had never really changed memory timings before this so I have a very basic understanding too but yes you can just copy settings from the Zentimings screenshots on this thread and stick with whichever one is stable for you.
I am really sure you will get higher than 3600 stable, that is the absolute guaranteed minimum speed that will run on the worst example of motherboard, ram, and cpu, so higher must be possible in 99% of the other cases.
I would first try to set your memory speed to 4000, Cl40-40-40, set your memory voltage to 1.25v and then leave everything else on auto. Run a short memory test with TestMem 5, it’s important to use test mem 5 because it will show errors incredibly quickly compared to normal memory tests. Seconds vs minutes or even minutes vs hours/days.
If you pass TM5 with those settings then move up slowly until you start getting errors then decide if you want to keep pushing for higher numbers or if you’re happy with whatever one you found to be stable.
You can save your bios configuration and if you get stuck with memory settings that don’t allow you to boot you can press the reset cmos button and be able to boot again, then reload the old bios settings and then try a different memory speed.
It saves a lot of time if you’ve already tweaked other settings in there so you don’t have to start from scratch each time. Especially if you’ve manually set the memory timings and just wanted to try 4600 instead of 4400 memory speed
Initially, when I first turned on the Workstation, all the dimms where install and everything was fine. I got in the BIOS, lowered down the speed manually at 3600 and start installing Win 11.
Once this was done and I did some tests to verify the speed of the build, I went to the BIOS again and brought the dimms speed at 4400 and all the rest was set on AUTO. This was my initial error, the build was always blocked at the memory training step for a lot of minutes, more than 20 if I remember well.
The strange thing is that clearing the CMOS didn’t help at all! Strange no?
At that point I took off 2 of the 4 dimms and everything came back to normal.
Last night I finished installing all my software, all the dimms are installed but they work for safety reasons at 3600.
I’ll do a backup of the C: drive first and then I’ll start playing a bit with the values you guys reported here, trying to find my build’s limits.
After setting up three different machines with 128GB same Corsair Vengeance SKU on DDR5 and AMD 7950X, I can tell you that the RAM seem to be very different in terms of stablity. One machine seems to be stable at 4600 but another fails with blue screens at 3600. If you have Gigabyte motherboard I do not recommend latest bios as it seems to have a bug for ram stability, I had to downgrade to make it work. Most important thing I’ve seen is to set SoC voltage to 1250 (seems to make pretty much any 128GB capacity work at minimum 4200 Mhz). Increasing this would let you higher frequency, if you are unstable at 3600, downgrade the bios to a earlier version, test first with first version of bios that the motherboard comes with.
Just made an account specifically for this topic, as I’m interested in building a server array of 7950X’s, all with 128GB RAM. I’m particularly interested in ECC RAM, since my workloads will be focused on data applications (different types of databases, to be more specific).
I’m curious if anyone here tried any ASUS boards with ECC UDIMMs. I was looking at Kingston, specifically this - KSM48E40BD8KM-32HM
There’s a list of 16/32G UDIMMs I found on another forum, I’ll cross post it here, just in case anyone has tried any of them and can report back.
Some motherboards have enabled ECC RAM in recent bios updates. Just don’t spend too much on high frequency and save yourself some money there 3600 is safe and every 200Mhz you risk not getting what you pay for a little more.
Isn’t DDR5 3600 @ CL40 very slow compared to DDR4 3200 @ CL20? I’m not sure if there’s gonna be much performance benefit for going AM5 vs cheaper AM4 at this point.
Then I guess we’ll stick with a dozen Zen3 chips with ECC DRR4 for now, maybe upgrade end of Zen4 cycle in a couple of years. Seems to risky to go 128gb on Zen4 right now, judging by this thread. Especially with no known DDR5 ECC compatibility on consumer motherboards.
I took these AIDA64 screenshots from a while ago directly comparing two of my 128GB workstations, one with a 5900x DDR4, and the new one with 7950x and DDR5.
This was before I went down from DDR5-4800 to DDR5-4600 (to avoid memory errors), so the 7950x results are very slightly higher than what I would get now.
When comparing to Zen3 I don’t see an issue with not being able to hit high DDR5 speeds on Zen4. Yes, I’d like it to be faster, and hope it gets better in time, but it still is faster with 4x32GB than the previous platform!
Also don’t forget the architectural differences of DDR5 (two channels per DIMM); you can’t just look at timings and understand how it’ll affect any given application since so much has changed.
