NewMaxx

Temperature is not too high. Check IOPS in another benchmark like CDM.

SSDs have static (e.g. controller) and dynamic (e.g. NAND) costs, with the latter also having elements of scale (i.e. there is a "sweet spot"). A drive, whether at 250GB or 2TB, may have the same controller for example. This is one reason SSD controller costs are expected to go up 10-15% this year even as NAND remains in decline due to oversupply. So, as an example, you have the dual-core ARM Cortex-R5 Marvell 88SS1074 controller in many SATA SSDs (e.g. WD Blue 3D) while the SMI SM2263 NVMe controller is also dual-core Cortex-R5, albeit clocks vary. The point being that controller costs are a

It's not all that great, strictly an extreme budget option.

That seq. write result is normal if it's hitting TLC at that capacity, which is probably is since the drive is fuller with a diminished SLC cache.

It's normal.

Yes, reviewers have the S70 in-hand. More info soon.

Samsung's 128L flash and 8nm controller are superior to 96L Kioxia and a 16/12nm controller, although I do not know how many cores WD is using on the SN850 yet (the SN750 was tri-core). Hardware doesn't lie. Samsung's software and arguably support are better, too. But if we're talking performance outside hardware, the SLC cache designs are a bit different too - the 980 PRO is using TurboWrite 2.0 which is a fancy way of saying hybrid (static + dynamic) although the dynamic portion is larger here than with TW 1.0 (e.g. 970 EVO Plus). The SN750 had static-only making it very consistent, however

In performance, 980 PRO > SN850 > E18-based (Sabrent Rocket 4 Plus, Addlink S95, etc). The SM2264 is also out but not yet reviewed in the ADATA S70, although I expect it to be quite good for consumer usage. However everything is using 96L flash except the 980 PRO for now. ETA on higher-layered flash is end of Q1 2021 in these other drives.

X570 boards support PCIe bifurcation in the GPU slots at least, yes. That means direct CPU lanes that take from the GPU. You wouldn't be bifurcating a x4 chipset slot, that makes no sense. Again, no performance loss on a RTX 3080 here...just runs a bit hotter so fans need to ramp up more. Don't be using a Hyper in a chipset slot...just get individual NVMe adapters.

The KC600 is an excellent SATA SSD.

No DRAM on the Patriot Burst (Phison S11). It states it has 32MB of SDRAM (which is DRAM, not SRAM) but it's embedded in the controller and AFAIK not utilized for mapping but rather write caching. It's effectively DRAM-less.

I run my RTX 3080 with a Hyper. You don't lose performance. In fact, FPS drop is not huge even at x4 PCIe 4.0. Admittedly it does impact the card's cooling such that I have to run the fans at a higher RPM for same temp. If the board is doing bifurcation, it has to be from the GPU slots generally and x4/x4. I think there may be ONE board that can bifurcate over PCH but I have not confirmed that. In any case, if you have bifurcation on the adapter then you have more flexibility for sure but they are way more expensive. Specifically your comment: "2x 3.0 that should work in a 4x 4.0 s

Caching/tiering SSD with HDDs.

Generally yes they can be switched...although check to make sure the boot order remains the same.

Yep, my EX950 for example had lower sequentials over the X570 chipset. Not sure if that's your issue but it's possible - easy to test by swapping drives of course.