Hello i have problem with overclocking my ram. I have ASRock X570s PG RIPTIDE motherboard ryzen 5 5600 and RAM G.Skill Ripjaws DDR4 32GB 4000MHz CL18. Ikn ow that this cpu cant handle 4000mhz but only think i can do is enable xmp and change the mhz to 3200- but this only works with 1.4V. Is that safe? or what can i do to get 3600mhz?
any help?
This is expensive RAM. I tried it, it sucks on AMD.
I can only recommend returning it for refund if possible and getting AMD compatible RAM.
The 5600s don’t really tend to be /particularly/ well-binned on the IMC, but 3600 should be achievable.
However, the slots you put them in matter. A lot. Basic reasoning would say that the slots closest to the CPU would be the “fast” ones as they’d theoretically have the shortest traces, but that has not been the case on any AM4 board I’ve tested. With the various kits I have on hand, putting them in A1/B1 as opposed to A2/B2 results in struggling to run 3200mhz XMP profiles, while I can run at least one kit as far as 4000 without memory errors, though it didn’t consistently perform.
1.4V is safe for the RAM as it’s actually GSkill’s tested voltage. However, check the slots. If you have them in A1/B1, A2/B2 should allow you to run somewhere in the neighborhood of 3600 at least. 3733 maybe. 3800 tends to be a mission with most chips.
bro it worked but are these settings ok? i just load xmp and instead of 4000mhz i just select 3600 and save. the voltage is on 1.4 bcs its xmp default.
I would use a mem checker like occt or anything equivalent to ensure it’s stable. If it passes after a hour then you can slowly tweak and retest again. You maybe able to get 18-19-19-39 at 1.35 volts at that speed.
1.4 should be fine however
Ram overclocking is a very slow process but I find enjoyable to do.
on my old Threadripper system with ddr4 i used ryzen-dram-calculator to find some good timings and settings to start with and then used a few different stability tests tools, (AIDA, OCCT etc.) i think i ran stuff at around 1.45v or a little higher, but more power, creates more heat and more heat makes the ram unstable. i think my ddr4 started becoming unstable at 55c so i ended up watercooling my sticks.
if you want to learn how these tings works i would recommend Buildzoids videos (7 videos) on the topic, they are very technical but if you get through them all you will (hopefully) understand how the timings are related and why just frequency alone does not necessary means you get better performance.
i think these are the timings i used, they probably wont work for you as there are lots of differences in memory controller quality , quality of memory kits, etc. but i think you can do better timings.
i dont know what kind of die your kit got, how many sticks, if its single or dual rank, but i think something like this might be a good starting point (i would recommend to keep the termination block and bus block stuff on auto unless you know for sure your die revision, rank etc)
Now that it’s functional, you have several things you /can/ try, if internet bragging rights mean more to you than your time.
One would be stepping the RAM speed up one increment at a time along with the IF clock. 3733 is probably doable, 3800 can be. When it stops booting successfully, go back down a notch. Mind you, past 3600 is a fairly sharp increase in difficulty and the performance benefits aren’t /great/.
The next is dependant on above, but I’ve found it roughly works (your mileage may vary), but you can try then trimming the timings down. Given this kit’s XMP profile is 4000mt/s, something like 16-20-20-38 should work for 3600. To “recalculate” timings, take each timing (i.e. CL 18), divide by memory frequency, and then multiply by the “new” frequency. This can to some extent work both directions, but trying for a higher frequency than the XMP profile of a given set of memory is a lot more likely to be unstable than tightening the timings at a lower than rated speed.
Finally, and this is the part that’s actually important, memory testing! Even if you don’t try for a higher speed or tighter timings, using various memory testing tools is a big help. OCCT is one that’s common for people to use, and the Ryzen DRAM Calculator includes MemTest with the ability to automatically launch multiple instances of it. Doing both (or more) for a couple hours a piece is strongly recommended before you start using the system heavily.
However, there’s a third type of instability I’ve seen from Ryzen chips that won’t necessarily show up in memory testing. If the Infinity Fabric isn’t happy, you won’t necessarily get crashes, but you can get intermittent slight “hangs” and other weird operation. If you’re way past the point of stability, it’ll crash, but if you’re only a /little/ too far it can be just little oddities.
This option has a deadline and should be considered if OP simply wants to set it and forget it.
Unless that is on the mobo’s qvl. No one looks at those, and I ain’t going to be the first.
I think I have a set similar to OPs at a lower bin if primaries mean anything. Latency was never as good as b die, but everything liked less voltage across the board to run them. I don’t think that translates to much at high frequencies.
The dude that created the DRAM calc now has something called hydra which is more of the same with more options is the only way I can think to describe it. Tried 1.6 through 1.7C if the folders on my desktop are to be believed. I figured a few bills was at least worth the use I got out of the DRAM calc lest I become a commie.
It’s not just the length of the trace that matters. To dumb it down, having the empty slot after the DIMM creates more “noise” on the signal compared to having the empty slot before. This is why the best memory overclockers are 2-DIMM boards like on ITX.
Also those timing are definitely “loose” for a 4000C18 kit, but it generally goes highest clock and then tighten timings for memory OC. I only casually watch Buildzoid though, YMMV.