Optimal selection for a 3900x configuration
Optimal selection for a 3900x configuration
I’m exploring my latest Ryzen 3900x configuration and am facing a decision on RAM. I need a minimum of 32GB, preferably 64GB, because of my work commitments. For gaming, I also want to enjoy smooth performance. I’m using a Gigabyte X570 Aorus Master, and I need the system to remain functional for several years—my replacement cycles are quite long (7-8 years). Right now, budget isn’t my main concern.
For running 3600Mhz on Infinity Fabric at a 1:1 ratio, I’m limited by DIMM size and timing options. So far, I’ve evaluated these choices to get the best performance:
1) GSkill Trident Z RGB 32Gb (2x16) DDR4-3600 CL16-16-16-36 – €300
2) GSkill Trident Z NEO 32Gb (4x8) DDR4-3600 CL14-15-15-35 – €550
3) GSkill Trident Z RGB 64Gb (4x16) DDR4-3600 CL16-19-19-39 – €365
4) GSkill Trident Z RGB 64Gb (4x16) DDR4-3600 CL16-16-16-36 – €630 (no dual channel)
I’m weighing the pros and cons of each, especially considering cost differences and technical trade-offs. I’m open to alternatives if they seem reasonable.
One option stands out: GSkill Trident Z RGB 32Gb offers dual channel and the lowest timings.
Another: GSkill Trident Z RGB 64Gb is also a strong pick, though it’s pricier and lacks dual channel.
Option 4 is cheaper but sacrifices dual channel and has higher timings.
I’m also thinking about other frequencies—maybe upgrading later if timings improve. Right now, I’m leaning toward the options that balance speed, capacity, and longevity.
Feel free to share any insights you might have from other builders. Thank you in advance for your help!
And how would this compare to the GSkill Trident Z RGB 32Gb (2x16) DDR4-3200 Cl-14-14-14-34? It’s priced at 261 EUR. So far, the lower 3200 MHz didn’t really affect performance much. You can also find 64Gb kits with identical specifications...But the clock speed might be more important for production builds. The testing mainly focused on gaming environments, correct?
it would likely run slightly slower than the first option, but I think cl16 3600 would be better than cl14 3200 on Zen 2. The exact app matters, so testing is necessary. I'm not certain if four sticks can handle 3600, but it seems safer to stick with 4 sticks at 3200. The X570 Taichi is designed for four sticks and has a better heatsink overall. So I’d consider: 64GB 4x16 DDR3200 with Taichi, or 32GB 2x16 DDR3600 with Master 4x16 DDR3600. It could work with Taichi, but it's not a certainty.
The Taichi supports four sticks due to its flexible expansion capabilities. For the X570 board, it can accommodate a 4x16 DDR3600 configuration. If you install two 16-bit modules, the upgrade path would involve adding more sticks or using higher-speed modules to maximize performance.
it's a T-top board designed for four sticks, whereas a daisy chain setup is more suitable for higher frequencies and two sticks. My quick search on Google suggests that most users with 4x16 3200 units perform well, though there isn't enough data for 3600. If you need a board that supports 4x16 3600, the Taichi is a possibility, but it's not a certainty.
Based on the tests I've reviewed, the optimal choice is DDR4 3600 with the most restrictive timing. DDR4 3600 CL16 and DDR4 3200 CL14 appear to offer similar performance, though the 3600 edges slightly ahead. For module capacity, 32 GB units generally have more flexible timing than 16 GB ones. While I don’t have direct experience, it seems a configuration of four 16 GB modules with tighter timings could outperform two 32 GB modules with looser timing, though this is just an assumption.
I also believe this applies, though it affects the selection of the Mobo. As xg32 mentioned, choosing a t-topology board is ideal if you fill all four slots. I just noticed my confusion about Dual Channel Mode. If I fill all four slots with identical DIMMs, should I still operate in Dual Channel Mode? I found this article that confirms yes. So, is there any drawback to filling all slots on a t-topology? Probably not—just consider whether the board supports the required frequency. If not, sticking with 4x8 or 4x16 should be fine. The main trade-off would be lower frequency, which could be addressed with tighter timings.