Good value DDR4 RAM option available.
Good value DDR4 RAM option available.
are the double sided e die sticks unavailable? part number is m378a1g43eb1 and how about 4gbit d die sticks? pn m378a1g43db0, all ive heard is that these sticks clock lower than e die but i dont think op is gonna be running >5000 esp not on 4 sticks theres also hynix 4gbit afr apparently somewhat close to e die in performance and it seems to be the only 4gbit ic hynix makes so random ordering wont accidentally get you a shit ic like 4gbit t die pn hma41gu6afr8n im curious as to how 16gbit rev b is the 2nd best ddr4 ic cause i would have suspected something like hynix djr instead, i mean the loose trfc on 16gbit ics isnt gonna help on amd with no adjustable trefi
I discovered those listings and checked "DDR4" on eBay, ordering by lowest price first until I found something acceptable. Those were the initial two options I considered reasonable. It seems a dual rank E die might exist, but since I didn’t spot it right away, I couldn’t recall the part number easily. I also didn’t pay much attention to it, so I assumed they were fine enough. I’m not very familiar with them, but since E die matched the price, I leaned toward that choice. DJR doesn’t offer much tighter tRFC (it’s likely 600 vs. 500, an improvement but still lacking), and for most other timing specs, Rev.B is more precise, especially for tRCD. This means a few Y-Cruncher rankings shifted to Rev.B instead of B die, mainly because it handles higher voltage better due to stricter memory limits. DJR still holds for frequency checks, but not enough to offset the timing drawbacks.
Doesn't the rev limit stop at around 1.7v? Are some components able to go higher than that? From what I've seen, most seem to cap around 1.7v, except for the max frequency case where someone achieved 5400c18 stable. I'm wondering about the rest of the primaries beyond that, especially since the trcd effect on performance seems to be minimal. It feels like cl isn't really providing much benefit in terms of voltage or frequency headroom, and tras doesn't seem to make much difference.
I understand there are certain components that work well at 2V, especially when they're cold. Frequency seems to drop noticeably after a specific voltage threshold. I plan to look into the scores later this evening. tRCD handles it somewhat, tCL a bit more but less effectively, tRP offers minimal improvement compared to tRAS, and tRAS essentially does nothing. The other sub-components aren't much help, but together they provide noticeable gains in performance when optimized.
Generally little is done, yet many manage a solid effort. When I adjust my rams for better performance, timing comes before primaries; I tend to leave cl and tcwl relaxed so voltage needs drop, allowing higher frequencies. If scaling is possible, I direct all power into the TRFC = P. When cl and tcwl don’t consume much voltage at a given frequency, I slightly tighten the settings. If I’m just experimenting with high voltage and the IC handles it, I can increase the voltage. It’s frustrating because the motherboard often limits things, which is annoying since I’d rather tweak my X58A audio if needed. I’m curious how subzero would affect things if I’m not hitting extreme frequencies—I wonder what would happen if I didn’t exceed the limits on frequency and timing. For now, I’m looking at a few components that seem to handle cold better: some green PCB boards, dual-sided Hyundai chips starting to fail around 2V, and a PSU that stays stable at 2800V 2.1V. Only a few things I know will improve with cold are a junkyard board, some PSC models, and a dual-sided chip that starts to fail near 2V. PSU max boot is around 3150V 2.1V, and it can handle 2800V 2.1V at lower voltages (like 1.7V or 1.65V) on my i7. Running subzero might reduce degradation, especially at around 2V, since it’s less likely to start failing compared to other parts. It could also push performance past the limits of my current ICs, but I’m still open to testing a better board if it helps.