Overclocking bad quality B(?)-die

[dimi33]

Hello all,

 

I'm trying to see how much i can overclock this ram kit i have which to be honest it probably didnt have much potential from the start since it is a 3200C16 - 32GB. 

Dram calculator(only used it to check bin quality - not using its output for setting the timings) shows a 81% bin quality so im not sure if i am to take this with a grain of salt.I just think this is one of the low end b-dies.

However, in order to see if i can hit 3600, i set docp to manual, set all timings to auto and put 1.45v on dram and soc at 1.15v. That would result in no boot.

However, changing main timings to 22-22-22-25 and at 1.46v it booted but then crashed at windows. This didnt change with either 1.47v/1.48v/1.49v so i decided to stop there since i guess i would have to overkill it with voltage just to do 3600 with ridiculously high timings.3533mhz is achievable with quite high timings and stable results at windows so i went further down to 3466 and this is where i am now trying to tighten timings.

Dram voltage is at 1.3950 and SoC at 1.125.

Changing down trdrdscl and twrwrscl to 3(instead of current 4) results in either crash/reboot or bluescreen/freeze. changing trtp down to 8 from 10 system hangs and have to clear cmos.

For TRFC ive only tried 512(current value 560) but pc doesnt boot and have to clear cmos so i;m guessing i must be quite close to its limit if not already there.

Trp can go down to 18(current is 19) but i get errors on stress test(same voltages).

Tras lower than 62 crashes/reboots pc(same voltages)

 

I'm not sure if theres any point trying for tighter timings.

Any advice/recommendations would be much appreciated

Edited April 5, 2021 by dimi33
clarification

[Alex@ro]

Those might be c-die, hard to believe there are some b-die of that low quality, might be interesting to remove hs and check, software reporting is not to be trusted all times

[ZFeSS]

Value B-Die overclocks similar to C-Die and don't like high voltage at all. Mine stops scaling above 1.35v. Anyway it's highly unlikely to get B-Die in 3200C16 G.Skill kit.

[dimi33]

1 hour ago, Alex@ro said:

Those might be c-die, hard to believe there are some b-die of that low quality, might be interesting to remove hs and check, software reporting is not to be trusted all times

 

33 minutes ago, ZFeSS said:

Value B-Die overclocks similar to C-Die and don't like high voltage at all. Mine stops scaling above 1.35v. Anyway it's highly unlikely to get B-Die in 3200C16 G.Skill kit.

Well that was my thought as well since from what i have gathered around it seems these results are comparable to c-die and not to b-die, yet, the sotware reports it as b-die(even the module part number corresponds to b-die[K4A8G085WB BCPB].Bear in mind though that this is a B1 PCB (dual rank) and i think these overclock way worse than A1/A0.

And you are right about higher voltages. 3533mhz was giving errors at 1.42v, yet it was stable at 1.385/1.395.Could this really be a C-die  being reported as B-die in software?

 

[dimi33]

It all looks more and more that this is actually a C-die  as you guys said (or value B-die if they can be named like that). Haven't removed the HS to check the dimms but based on the tests i did yesterday, and the day before, it looks like they are prolly C-die since using voltages between 1.325 and 1.365 actually worked to hit 3733 and 3800 with loose timings but had to make adjustments to procodt/SoC and Cadbus.

Raising the voltage past 1.38-1.39 at these frequencies immediately made things unstable. Also,It seems higher proCodt has to be set with higher SoC and vice-versa(lower proCodt allows for lower SoC).

Atm i am testing frequency stability for 3800mhz with below settings and then gonna try how low timings can go:

vDimm:1.345v

SoC: 1.0937v

CAS: 20-20-20-22

procOdt: 68.6

CadBus: 30-20-20-20

Rtt_Nom: disabled

Everything else set on auto for now