DDR4 H-Die

[水水水(BANNED)]

I recently got some new H-Die ddr4 memory chip from samsung this seems very promising. Like earlier Samsung 8Gb die it is very voltage tolerant and like over 2.1v, unlike earlier die it is ok with high VTTDDR voltage up around 1V. This is higher then Samsung B-Die and E-Die and D-Die which like around 0.85 - 0.95V in my experience. This memory chip like cold temperature and scale to around -110C, more cold allow for more temperature scale and less tRCD timing. tRP go below tRCD on viable platforms which allows for much better performance in less consistency workload.

After a while of toying, on A2 G.Skill PCB I was viable to get to 5.2GT/s 12-13-13-28 2T. With a better board and cpu, 1T is possible since memory chip able to run 1T in memory down configuration at 5100 using 10980HK ES cpu (this cpu using R0 cpu die). This memory chip seems like might be best for tight timing bench-marking, only problem with this chip is it is very bad termination and signal integrity wise compared to hynix and micron options. With this chip, VREF like around 1.2V and go to around 1.32V under cold temperature with 8700K and Apex XI motherboard. This is quite low for VREF comparison to other memory chip which makes it bad for out of the box compatibility, though if you are able to tune termination to a good value IC has much capability. Due to bad signal integrity and dirty DQ output from memory chip buffer which is mainly architectural issue, this memory chip not very good for high memory clock speed, though that is not much problem since Samsung 8Gb E-Die is going to be released, which much like D-Die clocks high while being light on IMC. Even though it clocks high and is light on IMC it still is not good for signal integrity and has dirty output, this is just trait of samsung memory chip.

For use such as server this will be very good memory chip, since it is least powerful samsung memory chip to date with IDD4R and IDD4W at 81mA and 73mA when operation at 2666V, due to this it is very efficient.

[Leeghoofd]

You got any screens to back them statements sir? 

[Alex@ro]

Your screen showed oem samsung sticks, now you say a2 g.skill pcb....

[Leeghoofd]

well either we have fun with him or just kick em... I vote option 2

[zeropluszero]

I don't know how you can stand this place Albrecht

[Leeghoofd]

4 minutes ago, zeropluszero said:

I don't know how you can stand this place Albrecht

All the cash that flows in makes it all worthwhile I heard

[Guest alatron978]

1.2V VREF isn't even possible at 2.1V vdimm on mainstream S-line intel chips, these chips only support as low as 60% of VDDQ for VREF, 60% of 2.1V is 1.26V, not 1.2V.

[Leeghoofd]

9 minutes ago, 4n0nym0u5b3nch3r said:

And neither does H-die exist.

 

not sure bout that  but it ain't this good...

[safedisk]

 

Photoshop failed

[zeropluszero]

1 hour ago, Leeghoofd said:

All the cash that flows in makes it all worthwhile I heard

Is roman actually paying you to run this chickenshit outfit?

 

[Splave]

Your sub said helium on the memory as well not -110c but close enough

[SparkysAdventure]

I would like to say one think about this post that has basis in reality.

Running a R0 Stepping 10980HK ES on Z370 and older is viable and has been done by a handful of people, and might be possible to run on z390.

However, because of the interposer used to put mobile chips on desktop, achieving even 3200mhz will be difficult.

The R0 ES has not made it into the Asian market afaik, and is still in testing. Skylake and Kaby Lake mobile chips are available currently.

Edited August 14, 2020 by Sparky's__Adventure

[Leeghoofd]

I don't even care if it was Angelina Jolie making the statement... the screenshot was as false as it could be... though I would buy her a drink ... or two ?