After reading about the newest rant from Linus Torvalds about Intel and ECC RAM I was reading a little more about ECC RAM.
It looks like the newer AMD Ryzen processors (Matisse upwards) support ECC RAM and so do many of the the X570 Boards from Gigabyte, Asrock, Asus, etc.
I have a Ryzen 7 3700X which is supporting ECC RAM. Now I was wondering if I could just upgrade my MB + RAM. But I am hesitating. Does anybody here have experience with ECC RAM on the AMD consumer platform? Any arguments why I should or shoudn’t go for ECC RAM?
Work related: memory was increased by 32Gb in my work workstation a few months ago, it was surprisingly expensive.
The amount spent almost equaled the amount I spent on my personal box upgrade, i.e. motherboard + Ryzen 2700 processor + 16 Gb RAM
It is more that I am looking for confirmation that this is actually working. Therefore I was looking for somebody with some experience who could share some lessons learned.
It’s actually a good question that I also posed to myself.
If you don’t mind the rather large performance loss, knowing that Ryzen 3xxx already suffer from limited memory bandwidth, then go for it.
RAM errors are extremely difficult to diagnose.
Everything I’ve read about the “support” on Ryzen and Threadripper is that it’s unofficial, and getting it to actually WORK is often a MAJOR PITA even though it’s “supported”. Most of the time the system recognizes it has ECC, but will not actually enable it.
Note, this is just what I’ve been reading, I haven’t tried it myself.
@mbod
No i don’t have any experience with it using on an AMD consumer platform. But i will tell you this although it may work on an X570 platform it probably does not support the ECC functions. Just my opinion from my experience so you would be better off asking the manufacturer of the particular motherboard.
The last time I had actual RAM problems due to failing chips was back in the 80’s in the age of DRAM. That was back when the RAM chips were about 1Mb each and all were socketed. Back then I had to write my own memory test. Moving pattern blocks of data one byte revealed stuck bits in one address, fixable by replacing a chip. Of course that didn’t auto-correct the corrupted text files…
Since then, my hardware failures have been in the mass storage devices, the power supply, batteries or the display. RAM typically fails quickly (not compatible, infant mortality, static) or almost not at all. On the other hand, I don’t do anything risking life or death on my systems…
I always run ECC RAM on my computers. Most of my computers have Xeon CPUs. I have some APUs from AMD running ECC RAM. They are more for NAS/Server operation.
Ready. I migrated my system to new MB and ECC RAM. Everything is working fine and my first darktable benchmark show that the 2666 MHz EEC RAM is not slower tahn my former 3200 MHz non-ECC RAM.
Physical Memory Array
Location: System Board Or Motherboard
Use: System Memory
Error Correction Type: Multi-bit ECC
Maximum Capacity: 128 GB
Error Information Handle: 0x0009
Number Of Devices: 4
As a consumer, i would ask a simple and precise questions:
Why somebody even want to use RAM, or anything else for that matter, which can lead to error?
Why not use ECC only and close the whole problem?
Ain’t point of any PC to be failure-proof, coz results of it’s calculations matter the most?
