Patriot Viper Steel 4000 19-19-19, 4400 19-19-19, and 4000 16-16-16 are all "b-die" sticks
They also behave fairly differently;
----
The 4000 19-19-19 ones have a lower maximum frequency than the other two, and potentially the weakest secondary (RRD/WTR/tFAW) throughput performance.
They seem to like RTP 7 / tWR 14 or RTP 8 / tWR 14, though some people also run RTP 7 / tWR 10
Unfortunately then I don't have hands-on experience with them, so I can't tell many funny details, but they are by far the "weakest" of the three.
preferred RTT's, something like Nom /6 (40), Wr /3 (80), Park /6 (40) (the divisor is for ""RZQ"", which is 240 ohms)
they have some wiggle-room in what RTTs they can run-run with (not all boards will do Park/6), but those seem to be their "best" performing ones
===
----
The 4400 19-19-19 ones are closer to the 4000 16-16-16, as they prefer to run at insanely low procODTs and raised DRAM Input Voltage (CPU 1P8 / CPU 1.8v / CPU Standby 1.8v[?aSUS?])
They differ from the 4000 16-16-16 and 19-19-19 ones by having strange odd delay requirements,
(MSI LGA1700)
(AddrCmdSetup 56, AM4)
Both the 4400 19-19-19 sticks and the 4000 16-16-16 do very good secondaries at high frequencies (>= 4200 MT/s), the 4400's prefer running either RTP 7 / tWR 14 or RTP 6 / tWR 12
on Intel they have high RDWR requirements (17 at 2100 MCLK, +/- 1 each step of 100 [200 MT/s]) and require RDWR_dd to be +1 of _sg/dg/dr (e.g. 17/17/17/18)
preferred RTT's, pretty much Nom /6 (40), Wr /3 (80), Park /5 (48) or /4 (60)
these sticks are not friendly if you don't run those exact RTT's and a very low procODT
====
----
The 4000 16-16-16 sticks, on the other hand, seem to be the "simplest" to work with of the three.
On AM4 they effortlessly did 3800 14-14-14-28-42, (tRFC 252-188-116), 1T GDM Off, @ 1.50vDIMM / procODT 28.2 with no funny delays, CADBus 40-20-30-20
something like that, didn't take screenshots (🤡), didn't experiment too much:
[Official] AMD Ryzen DDR4 24/7 Memory Stability Thread
Unlike the previous two sticks, they do prefer RTP 5 / tWR 10 rather than 6/12 or 7/14.
They seem to have a lower CPU 1.8v-increase requirement - currently running procODT 24 [/15] @ 1.88v instead of 1.91v like the 4400's (guy with 4000 19-19-19's runs 1.93v for procODT 30) -
but I'm not sure how consistent this observation is.
preferred RTT's, still seem to be the Nom /6 (40), Wr /3 (80), Park /6 (40) of the 4000 19-19-19's - as those were the only ones to even boot 4600 16-16-16
but unlike the 4400 19-19-19's, they don't seem to be very temperamental about it - though they might still be cranky about higher procODT impedances, DIDN'T TEST DIDN'T CARE LOL 🤡
and unlike the 4400's, you require no weird Delays on AM4 nor LGA1700 - the entire "ODT Delay Duration Configuration" section is set to 0, and runs stable unlike the 4400's.
On Intel, they seem to run a lower RDWR than the 4400's, as well as
require (it does not negotiate) RDWR _(dr?)/dd to be -1 of _sg/_dg
RDWR still sort of scales +/- 1 each 100 MCLK
===
tl;dr
(zentimings format of rtt Nom/Wr/Park)
4000 19-19-19: rtt 6/3/6ish, procODT 30, highest cpu 1.8v(?) + rtp/twr of 7/14, 8/14, or 7/10 + flirts with CAD setup timings, unless super specific CADBus 60-20-40-20
"meh" bin (best I've seen is 4200 15-15-15-34-49 @ 1.66v)
idk about daisy good vs. bad slot preference
idk about intel RDWR behaviour
4400 19-19-19: rtt 6/3/5 or 6/3/4 or **** YOU, procODT 28.2 or lower, high cpu 1.8v(?) + rtp/twr of 7/14 or 6/12 + delays/CAD setup timings mandatory(*), CADBus 40-20-30-20(ish?)
decent bin, hits 4200 15-15-15-30-45 @ 1.62v +/- silicon lottery
works great in daisy-bad slots in a 4x8
INTEL RDWR: _dd must be +1 of sg/dg/dr
edit: (* potentially CAD 60-20-40-20 works without AddrCmdSetup 56 ?)
4000 16-16-16: rtt 6/3/6ish, procODT 28.2 or lower, lowest cpu 1.8v(?) + rtp/twr of 5/10 + no delays/cad setup timings, CADBus 40-20-30-20
actually pretty good bin (you can get better for 2-3x the price but who cares), my single reference is 4200 15-15-15-30-45 @ ~1.56v
idk about daisy good vs. bad slot preference
INTEL RDWR: _dr/_dd must be -1 of sg/dg
they're all within 20 euro of eachother, so turns out the newest of them all, 4000 16-16-16, is the one to buy buy