A few random 4800 c20 32GB units from V Color
A few random 4800 c20 32GB units from V Color
I was casually checking the eshop app for high-speed DDR4 modules and their costs. Since I had nothing better to do at 2 a.m., I stumbled upon a kit that caught my attention. The timing specs were pretty rough—20-30-30-50 cycles—but that’s normal for 1.55V operation. Was it just a micron revision or are there other 16Gb modules that can match such speeds? Also, since 8Gb modules already have several options delivering impressive performance (like RevE/J, DJR, B Die), I’d like to know if there are more 16Gb variants beyond RevB that can reach this level.
They offer dual rank DJR or a 16Gb DRAM? Everyone says 16Gb is just MJR. If it can reach that speed, it might be similar to an 8Gb DRAM and handle those high-end chips like the 5200+ 1.7V+ without any issues. I don’t think even a powerful APU with over 8GB in 4x8 would be a problem.
It seems to be DJR, though I’d need to research further... The links provided cover various Hynix memory specifications. There’s also a reference to SkyLake Intel DIMM details and a general note about DDR4 performance expectations. Most of the information points toward DDR3/DDR4 configurations rather than Samsung B-Die. It’s likely you’re looking for specific data sheets or benchmarks to confirm the exact model.
Nope. For them too, at least most of the time. The issue lies with the infinity fabric. Memory data must pass through this fabric, which, even at its maximum capacity, increases its processing speed—either by raising the divisor or decreasing the multiplier. English isn’t helping here. If you exceed 3600 or 3777, the speed drops significantly; it halves from 1:1 to 1:2. The fabric can be overclocked, but only up to this point. This is how zen3 manages to run at 4000mhz memory. It generally handles the necessary boosts. If you push it high enough to support 4800mhz at full speed, it should work well. But that seems unlikely.
laughs in Cezanne highlights the advantage of bandwidth, with more bandwidth outperforming latency. The idea of beating FCLK penalties seems to favor bandwidth above latency, especially for weak CPUs that can't reach 1900FCLK like multi-CCD 3950x/5950x or some Ryzen 3000 chips. Exceeding FCLK is unnecessary on an APU, though very fast clocks (5200+) might still cause issues even with overpowered Cezanne.
Sure. With a sufficiently fast memory, you can handle it without much trouble. The 4800 isn't quite 5200, though my own stats show it's just barely above the threshold. It's still deep in the dip on the upward side, but not as fast as expected. I doubt a 3600 cas 16 would outperform it or at least not fall behind. That would change my thoughts, but currently there doesn't seem to be any real examples. If you have something different in mind, let me know.
Best setup for slow runs sits around 3733-4000 c14/13 1.6-1.75v, offering a genuine match since 5400+ brute speed feels like wild high-bin djrs zooming at about 1.8v, which isn't really a fair comparison when you look at top-end raw speed versus mid-to-low speeds.