Have AMD got the BIOS optimized to handle higher RAM clock speeds for R7 CPUs or is the 2017-18 recommendation still true?
I've heard that Ryzen 7 don't really like RAM speeds above 3000mhz & running above this speed long-term can affect system stability because of the way the memory controllers work.
I have 3200mhz RAM & it passed MemTestx86 & the stressapptest with no errors using it's XMP timings.
So I'm wondering whether recommendations to stay to < 2933mhz is still best or whether the latest BIOS updates for AMD ComboPI 1.0.0.3abba have any impact on R7 RAM handling.
Its a question stability & performance between
•XMP Profile #1: DDR4-3200 CL16-18-18 @1.35V
•XMP Profile #2: DDR4-3000 CL15-17-17 @1.35V
My rig is :-
850w PSU & Watercooling.
the ryzen 1000 series are designed for DDR4-2133 and the 2000 series are designed for DDR4-2933, the 3000 series bump it up to DDR4-3200 speeds
Thanks for the reply.
So there is no truth in the idea that the microcode for the older CPUs has been updated to improve what they can reliably handle now?
unless your having stability issues. the BIOS settings should be at default
I don't think you would have a stability problem with running at higher memory speeds if the memory you are running is rated for that speed and/or you're using XMP settings. Running Memtest should prove this by not having any errors no matter how long you run the test. Now, performance could be a problem at higher speeds. I'm not familiar enough with the 1st gen Ryzen, but for 3rd gen Ryzen (Zen 2), memory operating above 3600 typically performs worse due to the speed of the infinity fabric.
I don't think running 3200 speed on your memory should be a problem. However, my unprofessional guess is you should run XMP profile #1. This profile has a CAS of 16, which matches the design of the memory. XMP profile #2 may run fine, but I would suggest testing the memory to be sure it can handle the tighter timings. Back to the performance of the memory, you could always run a benchmark test (i.e., Timespy) to see how the PC performs with different memory speeds. My guess is the faster the memory the better the performance.
Thanks guys.
I had a chance to run a benchmark (below).
BTW I've found that Memtest is great for assessing the stability outside the OS but with the OS running the only way that seems to be foolproof is with stressapptest. This became really clear when I was running Handbrake & couldn't understand why I'd get BSODs when the tests I'd run had passed.
Maybe there is a more recent version of Memtest which addresses this but last year I could go for hours & see zero Memtest flags & get errors with stressapptest in the OS within seconds.
It is a useful tool if you're having problems.
MHz // MB/s Read Write Copy Latency
2933 - 44064 43450 39308 84.3ns
3200 - 48154 47431 42551 78.4ns
Both of these speeds pass stressapptest.
Guess I'll just play around & see if Handbrake will still cause a BSOD, on 1 or 2 hours of work its' been OK so far.
I'm curious, could you explain more how stressapptest is foolproof compared to Memtest? I'm not sure I understand. What did you change to avoid the BSODs? Are you running this memory with XMP settings, or are you adjusting settings?
Regardless, glad to see that the memory runs great at 3200!
I’m using XMP but I’ve found that it’s dependent on the BIOS & BIOS version as to whether the values match 100%. This BIOS version is the best so far, the earlier ones could adjust a lot of values if they weren’t set/forced manually.
I found the Asus mobo I had was worst for doing that. Hopefully Asus have got past that now, they are pushing updates.
I double check the XMP values vs what the BIOS has set & adjust where necessary. Usually it is the tRFC values that get tweaked by the BIOS. So I set them to the listed values & test.
I’d managed to figure out that the RAM settings were the cause of the BSODs but couldn’t figure out why.
I was pointed to stressapptest & that showed the memtest results were ‘flawed’ once you got the OS involved….
These RAM settings are way more granular than my previous FX CPU rig was, so I had to start going through everything until I got it to pass the test in Win10. That took several days but taught me a fair bit.
Hence now when I find this BIOS version so much more friendly/stable I wonder if AMD had retroactively stabilised the code for the infinity fabric on this gen of CPUs.
Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows 10
I found some RAM with XMP to be unstable with AMD Ryzen, depends on the CL and other timings