Reduce RAM speed using an XMP profile setting
Reduce RAM speed using an XMP profile setting
I understand. Thank you for the video. My main concern is whether two high-end RAM models with identical Hynix M chips, such as the g.skill Trident Z5, perform similarly. If the only variation is in the motherboard settings, would adjusting to lower speeds like 7000mhz or 6800mhz improve stability? Your insights are appreciated @RONOTHAN##
Here are some alternative options to consider:
- Invest in four RAM sticks, each with 16GB of Hynix A-die memory, aiming for a maximum speed of 6000 MHz.
- Alternatively, opt for a dual-rank 6800 MHz kit with two 32GB modules.
Regarding your question about dual-rank configurations: not all 64GB kits are guaranteed dual-rank, so it's best to verify the specifications before purchasing.
More or less. There is some argument to be made about binning in the higher rated kits, where a theoretical 8200 CL38 kit is likely to need less voltage for running the same settings as the same brand's 6800 CL34 kit, though in practice getting a high binned kit in a lower rated kit is not unheard of (the 6000 CL30 kit I own behaves very similar to 7600 CL36 rated kits from what I've seen in terms of required voltages for given settings, my best kit of DDR4 is in a 3600 CL17 bin, arguably the 2nd or 3rd worst Samsung B die bin, though my 6400 CL32 kit is noticeably worse than a 7200 CL34 kit). Yes. You can't necessarily go the other way just in case your kit requires more voltage or looser timings at specific settings, but going down and just copying the settings of a lower rated kit pretty much always works. 6400MT/s never fails on 13th/14th gen, 6800 almost always works, 7200 usually works, and anything above that requires a fair bit of luck. I'd try for 7200 and run Y-Cruncher VT3 to check that it's stable, though if you're dead set on not running a stress test, drop down to 6400MT/s. Avoid going 4 DIMMs at all cost with DDR5, it's hit or miss at best, and while you can get it to work at decent frequencies, it will fight you all the way there. If you really need the capacity, go 6800 max, probably sticking to 6400 or 6000 to give yourself some stability headroom. Do know that dual rank is a fair bit harder to run, where I'd expect that if your CPU can do 7200 max, it likely maxes out at 6400-6800MT/s on dual rank, and if you have a chip that can't do 7200, it might max out at just 6000MT/s. As of today, 32GB and 48GB DIMMs are dual rank. This will change as the 64GB sticks get introduced later this year, because then it will mean that 32GB DIMMs can physically be single rank, though as with DDR4 that transition won't likely happen for ~6-12 months after 64GB DIMMs are publicly available. Dual rank doesn't really have a performance uplift on DDR5 though, so only really go dual rank if you need the capacity (which it sounds like you don't), otherwise just try to go for a higher frequency instead.
It’s common to dive into this forum more than just confirming if a certain frequency holds up. Usually, an hour of testing with Y-Cruncher VT3 is enough to verify stability for a specific frequency—most kits run about ten minutes before reaching an hour, and longer runs tend to keep going. The key factors are timing, frequency, and voltage, all of which should appear on the product specifications. In practice, you can stick with your current setup at 1.35V and adjust the lower timings and frequencies used for other kits. The 24Gb M die is fairly resistant to voltage changes, so that’s a plus. The main thing to watch is memory temperatures; since these chips get hot, especially with higher voltages, removing the heat spreaders can actually lower temps during testing. Frequency isn’t usually affected by memory voltage, but it can help with timing accuracy.
I understand. I’ll stick with what I have and follow your instructions. Could it be that removing the heat spreaders lowered the temperatures after taking out the RGB? I was planning to eliminate the heat spreaders just to achieve that effect. Do you know any aftermarket heat spreaders that would work better? *Edit* Just noticed you can’t remove the RGB in the same way as with Corsair parts.
I owned a kit without RGB lighting, and from what I've observed, RGB options tend to be even less reliable. Even taking off the heat spreaders doesn’t eliminate the issue—the LEDs are directly bonded to the memory stick, which causes problems when using third-party heat spreaders. The main problem is that the PMIC isn’t given a thermal pad, so it has to channel heat through the memory chips, leading to higher temperatures. Some brands like Bitspower offer better alternatives. My approach was to remove the heat spreaders from an older DDR3 kit I had, which helped significantly compared to leaving them in place. Most aftermarket heat spreaders are made for non-RGB models. In reality, it’s usually not worth modifying your kit’s heat spreaders due to warranty concerns. The 24Gb M die kit I own performed much worse in temperature than my A die kit—running around 60°C versus about 45°C—but the only noticeable impact was on a few memory timings and the inability to push voltages beyond 1.45V in summer, which limits low CAS latency. For XMP profiles, these details don’t matter much. If I install custom heat spreaders even with full custom settings, the performance gain is minimal—only a slight boost in memory usage efficiency. Taking them off would also void any warranty, especially since I purchased that kit used and didn’t have coverage.