3700x all core overclock to 4.3ghz on 1.2875v, is that too much? Max temp 81C on very rare spikes, but average around 75-77C.

Haven't tried prime 95 yet but loaded worldcommunitygrid to try folding on all core workload. Seems stable for 24 hours.

i'd say u can follow micron b die which is inferior than e die in dram calculator & see how it goes. if stable then go tightened it.

i wouldn't manually oc ryzen 3k, might degrade ya chip in long run. cool it as much as u can then let pbo does its job is d safer way to go. i could be wrong thou.

https://www.bilibili.com/read/mobile/6393757

hmm looks like it doesnt like to be tight. found some reviews on china sites that talks about trfc. i guess i wil start from 600.

after some google sessions i saw these chips being used on 32gb modules. ya 2x 16GB must be single sided perhaps?

that was initially my plan, but when i see it go 1.47-1.48v at 4-4.1ghz, i kinda feel something off, seriously don't know what the heck going to happen on this kind of voltage in longer term. The temperature was similar to my manual settings, so...i don't know....i'll do more observations between my manual settings and letting pbo to do the job. Thanks for the tip regardless

hopefully, they can better manage the parameters on the 4000 series or maybe with AM5 when can implement a whole new platform.

try to use overrid vid (vcore) + offset, i remember reading somewhere saying manual oc will degrade d chip in long run. i think i had d same issue on my 3500x. now on 3600 i dare not to run manual anymore. with offset i managed to keep d load vcore at 135v ish, which is pretty fair i would say.

Mind elaborating more on what you observed with your previous 3500x?

yeah.. they're high density 2GB chips so it's a single rank. Dunno why team used them for the 16GB. Maybe temp advantage?

currently on memtest on the 3rd hour with trfc at 600.. 32GB is taking a bloody long time but thanks to that aida64 bench mark is 46GB read and 74ns latency. Pretty decent stats now.

why not running try them at 3733/3800 Bonchi?

Im just lazy because it's build in an ITX case and to reset the CMOS jumper is super troublesome if i encounter boot stability with overclocking. Im happy if it can run at a full potential benchmark on 3200mhz CL16. Maybe will try pushing to 3600 when i take the board out to do cleaning in the future.

lazy but wanting to have low latency

well to assist in that case, you could try these on 3733 or 3800 (if you know your IF can do that)
micron b-dies are not that bad.

Anyone got experience with "Kingston" C-die?

Feel like is bootleg SK Hynix C-die to me because DRAM calculator settings doesn't work and manual overclock only get 3533 16-19-19-37@1.4V.

Dunno is bad silicon lottery or is bootleg.

Even kingston don't want put full information when reading with Thaiphoon Burner.

HAHA that's why im just looking for the common 3200mhz average stock settings because you know... xmp dont fully work with amd and those "auto" settings will go crazy like my trfc becomes 880 on the xmp

These are not the B-die but D9XPF ..... which are still pretty much undocumented and seems to appear only on high capacity server rams or the corsair 32GB sticks. So it's quite a surprised for me to see it on these team vulcanZ... which i was expecting to see hynix.

Well, i got this at RM390 so i didnt bother to do much research about it. After that trfc tweak, it seems to be up to standard now and im finally getting back those FPS (upgraded from a hynix 16GB 3200mhz). Maybe one day when i have time, I will push it a little to 3600mhz and post my outcome here

3600 ish @ 16 19 19 (or 16 20 20) sounds pretty much like cjr/cfr.

D9XPF is just the fbga code for their 16Gb density dies. And those are micron revision b.

exactly sir.

https://www.micron.com/products/dram/ddr4-s...mt40a2g8va-062e

degraded.... used to run 4.2Ghz manually with 1.325v, but after 2 months not prime stable anymore.

Results from hynix manufactured c-die seem to have tighter timings, usually see 3600 16-18-18@1.35.

Therefore I'm questioning isit due to lottery or lower manufacturing standards by Kingston but it could be my MB limits too.

I can get my secondary timings pretty tight but the biggest let down is tRFC, best i could do is 315ns.


Attached thumbnail(s)

1.325 vcore b4 vdroop? whaaa that's seriously rapid degradation for this voltage.
you sure this is not down to temp or other variables?


for myself I always check what the lowest voltage (on stock) is for some specific prime95 runs. that voltage is then my safe voltage for all core oc.
but so far I haven't had a chip where I wouldn't lose single core performance if I applied a static (ccx/ccd) oc.
however with some of those more recent chips I am pretty sure (especially the 3100-3600) might see quite a few of them being able to profit from it.

yup sir, b4. always take bios vcore as std. temp has always been normal, albeit on slightly high side (small fft at lower 80s), i hardly prime nowadays so i doubt it's temp degraded d chip.

used to be a hardcore static oc practitioner but given how hot it's when primed (i'm not talking about crazy oc) i've totally give in to pbo setup nowadays. gettin old i guess...

16 18 18 is abit pushing it, higher volts & perhaps a better imc & board needed.

i'd say gkskill sniper series is decent yet their XMP is rated 3600 @ 19 19 19 iirc, using cfr too (correct me if i'm wrong). cfr/cjr is capable for high frequency with loose timing, yup, i'm talking about above 4k. in term of personal preference, i would rate:

1. micron e die (high frequency, low volt, medium timing)
2. samsung b die (high frequency, high volt, best timing)
3. hynix cjr/cfr (medium frequency, medium volt, loose timing)



This post has been edited by ah_khoo: Aug 4 2020, 04:51 PM