This is when the problems started
![Überglücklich :D](./images/smilies/icon_e_biggrin.gif)
First, you will see tons of results showing 2K Cas12+ by Pro and elite ocers, voltage given 1,8v or a bit higher. Don´t expect this to happen to you when you buy a random kit, chances on retails are low to achieve even 3866 or 4000 12-12-12-28 tight at CR1, most results you will see are by handpicked special samples given by vendors to ocers as support and promo. If you know the oc scene, you will remember that this already started at late DDR3 times, if you didn´t knew, now you do. You can judge this the way you like to do, I state this here to show relations, so if you buy a retail kit that gives you a solid 3866 Cas 12 tight below 1,9v, you are not unluckiest guy in the world but can be happy the kit is solid.
Second, and much more important for me personally, is the issue that brought me the kit together with two others that I tested today. I bought around 20 kits, all G.Skill ripjaw5 and TridentZ, since December, not because I have too much cash or because I am even more crazy than I am for real, but because I constantly needed replacement. Out of the 20 kits I got brand new from german online shops, I had three DOAs and four kits that failed specs out of the box. You will easily determine a DOA because the sticks simply don´t boot, but I can only advise you to test each kit you get sealed for specs with memtest first. I cannot determine if it is the IC itself or sloppy control, but it is an annoying fact which was also confirmed by a lot of other binners and ocers who faced similar problems. To be fair, this also hit other vendors on B-die, but I see no reason to stay quiet about this when it might help people and avoid problems for them by simply telling them the way it is and recommend additional testing.
After I now became popular with some vendors, let´s simply show what my kit could do
![Überglücklich :D](./images/smilies/icon_e_biggrin.gif)
Test system
Intel Pentium G4400
Asus Maximus 8 Impact Bios 0014
G.Skill TridentZ DDR4-3600 C17-18-18-38 1,35v
Dell 7800GTX
Mushkin AP800 PSU
OS Windows7 X64 Ultimate at Maxmem set 4000MB
After checking specs with memtest I decided to start the tests at settings relevant for benchmarks and skip the usual review pattern, so I started at 3600 C12-12-12-28 1T 220TRFC and worked my way upwards, each divider had to pass XTU, superPi32m, Geekbench and several other benchmarks that stress the Ram to see if it is useful for benching, if you want 24/7 settings, you are wrong at this thread^^The Impact undervolts a bit, so shown voltage is 0,01v above real volts
The test and my results
DDR4-3600 Cas12-12-12-28 1t 1,68v
![Image](http://abload.de/thumb/3600c12168v7dsl2.jpg)
Next divider worked as well, I needed 1,765v for 3733 Cas 12-12-12-28
![Image](http://abload.de/thumb/3733c12176vttscf.jpg)
Moving up to 3866, things became harder, the IMC started to show first signs of weakness, I managed to run DDR4-3866 12-12-12-28 1T at 1,84v
![Image](http://abload.de/thumb/3866c121.84vjmsi3.jpg)
Next stop was DDR4-4000, now things started to get really difficult, but not the mems but for my IMC – I needed 1,35v sa to boot cas12 at 1T, it also killed scaling a bit, which is one of the reasons I already needed 1,96v for DDR4-4000 Cas 12-11-11-28 1T 240 Spi32m, XTU should have worked at a bit less volts
![Image](http://abload.de/thumb/2000c12-11196v32m7auwi.jpg)
![Image](http://abload.de/thumb/2000c12-11196vgeekbeny5ueo.jpg)
Time for a short conclusion. The tested kit is clearly above average on B-die kits, it worked flawless at specs and gave me excellent results, especially when I take in that 3600C17 is no tight highend bin. If you wonder why I set maxmem for my benchmark tests, this is a platform problem that occurs at all B-die based kits, at tight timings and high voltage at 3600 and above, they lose stability at x64 OS. This is neither vendor specific nor is it a reason to complain, benchers like us use hardware outside of its specifications, so we have to find workarounds if this causes issues. B-die is the way to go for benchmarks at the moment, it is the dominant IC at the moment, and especially if vendors find a solution to improve the ratio of DOAs and kits failing specs, there will be no serious alternative to these as they offer performance and timings no other ic can do even on cold. I hope you liked my short review, feedback is welcome, I hope I can answer these before modul houses delid me^^
Best regards
websmile