Just picked up the new 9950X3D and looking for optimal DDR5 recommendations. I've seen conflicting info about whether to go with DDR5-6000 CL30[1][7] or push for DDR5-6400[8] speeds. Planning on 32GB capacity for gaming and content creation. Using 4 sticks seems problematic above 5200MHz[5], so probably going with 2x16GB. Any real-world experience with this chip? What kits are you running and how's the stability? Thanks for any insights!
For the 9950X3D, you'll want to focus on DDR5-6000 with tight timings rather than chasing the highest speeds. AMD's sweet spot is still around 6000-6400 MT/s due to the FCLK limitations. I'd recommend G.Skill Trident Z5 Neo DDR5-6000 CL30 or Corsair Dominator Platinum RGB DDR5-6000 CL30. The Neo series is specifically optimized for AMD and tends to have better compatibility. Make sure whatever you get is on your motherboard's QVL list.
You can't go wrong with DDR5-6000 with a CAS Latency of 30, such as VENGEANCE DDR5, VENGEANCE RGB DDR5, or DOMINATOR TITANIUM DDR5. Since you're using the 9950X3D, either 32GB or 64GB of DDR5-6000 RAM is great, with a CAS latency of 30 being ideal (CL30). You'll be in a good spot performance-wise with that configuration. The Corsair Vengeance RGB DDR5-6000 32 GB memory kit is what we recommend for value seekers. Just install the memory, access your UEFI/BIOS, and enable the EXPO profile for optimal performance.
> I've seen conflicting info about whether to go with DDR5-6000 CL30 or push for DDR5-6400 speeds. tbh I'm still kinda learning the ropes with the AM5 platform, but I'd be really careful about pushing for 6400 MT/s. Like, I've been reading that the memory controller on these high-end chips can get super stressed when you go that high, especially if you're doing content creation where you need it to be 100% stable for long renders. I'm honestly a bit worried that chasing the highest speeds might lead to some silent data corruption or random BSODs later on. iirc, anything over the official AMD spec is technically an overclock, so you're basically at the mercy of the silicon lottery. Wouldn't it be safer to stay right at the official limits if reliability is the main goal? Has anyone actually run long-term stress tests like Prime95 at 6400 without errors? Basically, it seems like a lot of extra heat and SOC voltage for a tiny performance gain, ngl.
Bump - same question here
Saving this whole thread. So much good info here you guys are awesome.