Can this output be 3600M/T/s from a 3200 on Samsung D-die? Yes, it is possible.
Can this output be 3600M/T/s from a 3200 on Samsung D-die? Yes, it is possible.
I had those kits in my current configuration on A520M ITX Asrock /AC, 5600X R5.
3200 Cl16, Samsung D-Die... Why couldn't they use B? It's not too far from D.
😆
Any chance someone has the same kits and managed to overclock them to at least 3600 without exceeding the clock speeds?
Keep in mind that achieving an "XMP" OC at this level requires a motherboard with robust VRM and stable voltage. The 'A' chipset isn't the main factor. It's unlikely you'd notice the difference, but performance might suffer if latency is increased to reach it.
I'm not certain about the exact kit, but I know it's similar to other D-die kits. Generally, adjustments in timing and sub-timings are necessary for stability during memory testing. It seems a tolerance for higher-than-normal DRAM voltage settings might be required. In reality, increasing the voltage to around 1.45V first could help determine if it works. DDR4 supports up to 1.5V, though many DRAMs become unstable at higher voltages.
Don't doubt increasing timings, especially CL, since latency decreases significantly with higher clock speeds. Adjusting timing is often essential for stability, but doing it improperly could negatively affect performance.
VRM quality shouldn't be a major concern here because DDR4 memory consumes very little power. While good VRM matters for CPU overclocking, it's not critical for an A520 board. Similarly, X570 or B550 boards are unlikely to improve memory clocks and timings beyond what your A520 already provides.
I absolutely disagree about your thoughts on the VRM on an A series chipset.
I also would mention (there is a chart somewhere) that once you start 'loosening timings' and get past a certain point you are actually slowing the module down. Overall this is going to be such a small part of the performance puzzle, that beyond the curiosity of doing it, probably will not net results worth the risk.
.02
I understand that pushing too hard isn't the way forward. Adjusting overclock settings requires careful attention: stability checks and latency tests are essential to ensure you're not exceeding safe limits. A slight reduction in clock speed can often allow for much lower timings, improving latency. This approach still makes sense if you're interested in experimenting.
I need to test 1.45V again, but I recall something didn't work when I set it that way. Probably due to tight timing settings. I'll try it back home.
I was attempting to keep timings on auto at 1.4V and got around 26.28, so maybe the voltage was a bit too high. 🙈
The value 26-28-28 is excessively high for DDR4 even at 3600, and the overall latency would likely be very poor. It might also cause timing issues and contribute to instability. I should definitely look up overclocking guides for DDR4 with Ryzen. A significant jump like from 3200 to 3600 usually needs checking sub-timings, and tools such as DRAM-Calculator can be useful. If the tool provides stable timings, they’re usually quite loose, but it serves as a good starting point for further adjustments.