Return to

Kingston 2400 CL18 @ 2933 CL16 - stability testing

So I’ve gotten Windows 10 and Fedora 28 to boot with this overclock, which seems a little too good to be true. I’ve successfully transplanted a Ryzen 7 1700 and 32GB of the aforementioned Kingston RAM from a Dell tower into an MSI B450 Tomahawk. I’m a little dubious about the stability of this OC, but Windows seems OK and the RAM checks out with a couple of passes from the Windows memory checker.

I’d check with memtest86 but apparently it’s not UEFI compatible, so I’m not sure what to test with. SiSoft Sandra reports memory bandwidth of 35.7GB/s and latency of 67.1ns, but I’m not sure if it uses any integrity checks. I did get one warning from SiSoft about insufficient voltage so I might need to bump that up a bit (I think it’s at 1.36v currently.)

Any suggestions for testing? I’m open to Linux or Windows tests, but phoronix at least seems to be unable to phone home at present.

In case anyone ever has a similar need, the best torture/stress test for Linux I found was namd (Nanoscale Molecular Dynamics), by way of the Phoronix test suite. It locked up several times while trying to find a stable OC, where other tests passed. It seems particularly good at generating lots of heat, though I didn’t try comparing it to an AVX-enabled prime number cruncher. To run, download and install the Phoronix test suite, then run:

$ phoronix-test-suite run namd

It’ll install needed components, though you may need to feed it your root password. You may need to tweak the namd xml config file to get it to do more than 3 passes at a time.

My final stable overclock was [email protected] with the RAM going at 3200MHz CL18 @ 1.4v. I’d try for a higher OC on my Ryzen 7 1700, but I’d have to feed it more voltage which I’m pretty sure would overwhelm my cooling (Corsair H60.)

I will try this out.
But what are the tests that passed? mprime?