Neo 3800 CL14-16-16-36 @ 1.50V kit turns out to be a Samsung D-die, not B-die as I thought earlier Source: Newegg

This post has been edited by owikh84: Dec 1 2019, 01:47 PM

HT off is faster on HyperPi, not sure bug or what
BTW isn't that 3500X only available in China?

Ya 3500X is a good choice for budget users.
However, gaming nowadays needs HT otherwise frame rates especially min fps will suffer.

Crucial = Micron, there's a Micron Memory factory located near my workplace here in Penang mainland. Their SSD factory is located at Batu Kawan, which is also not far from there.

Not bad for 4 sticks to run at 3600 on old gen, in your case 1700 and X370. Very nice. Did you run any stress test to confirm stability?

The Crazy Record appears to have one error there so not really stable. From what I heard Micron E has better thermal tolerance compared to Samsung B-die. The latter prefers cooler condition not more than 42C to achieve OC stability. Only downside on Micron E is that tRCD cannot be tightened as much as B-die.

MCO Day3 benching with 3700X and Micron-E (D9VPP).
Using 1usmus' DRAM Calculator 3800 Fast profile for primaries and Safe profile for subtimings.
Full Fast profile thrown me an HCl error at 100-150%, that's why Safe profile being used for the subs.
I know it's not efficient but I'm definitely working on it.



owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | Crucial Ballistix Elite DDR4-4000 CL18 (Micron-E) 2x8GB | DDR4-3800 CL16-19-16-36-1T @ 1.45V | vSOC @ 1.100v


Update: Today MCO Day4, I managed to tighten the subtimings a bit. tWTRS=5, tWR=18, both SCL=5 are pretty much hitting the wall. So far 1.45v is compulsory otherwise HCl will throw an error as soon as 150%.


Update 2: tWRWRSD & tWRWRDD reduced from 7 to 6, resulting in higher read bw 55k.


Update 3: Increasing the ClkDrvStren to 120 has allowed me to reduce primary timings to 16-19-14-14-34, as well as both SCL and tRDRD down to 4, but a slight voltage bump of 1.46v is required. AIDA64 scores are now improved a bit.


Update 4: GDM off is free of error as well.


This post has been edited by owikh84: Mar 29 2020, 03:18 PM

Thanks for the tips bro.
The bold part already tried earlier, not possible at 1.45-1.46v.
Maybe need to pump more juices to cross the HCl 150%.

Quick test 3800 CL16 @ 1.35v.
Manual primaries, subtimings auto, manual terminal block 34.3, off, off, RZQ/5 (<= auto will give error at HCl 120%).


XMP @ 4000 CL18-19-19-39-1T @ 1.35v


This post has been edited by owikh84: Mar 28 2020, 12:57 PM

My 3700X at stock clocks auto voltage registers at ~74C Tdie even under high-end custom loop. This is just a single loop of CB R20 test, let alone long-term stability tests such as p95, AIDA64, OCCT etc. Let's see how it goes with manual voltage later. I need that der8auer custom bracket badly.




Can't imagine how did others get with AIO or even air cooling.

This post has been edited by owikh84: Mar 29 2020, 02:20 PM

Moving to my regular B-die now, it was a 5-minute job because I just reused most of the settings from my previous run with 3600X. Now 1:1 is possible with this new 3700X.

owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | G.Skill TridentZ RGB DDR4-4266 CL19 (Samsung B-die) 2x8GB | DDR4-3800 CL14-16-14-28-1T @ 1.48V | vSOC @ 1.100v


This post has been edited by owikh84: Mar 29 2020, 03:18 PM

Not really sure but I believe up to 1.50v should be fine, provided it's properly cooled with fan/water block.
So far there are two kits rated at 1.50v XMP:

https://www.gskill.com/specification/165/32...N-Specification

https://www.corsair.com/us/en/Categories/Pr...16GX4M2Z5000C18

CL14-15-14-28-1T is also possible at the same voltage

owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | G.Skill TridentZ RGB DDR4-4266 CL19 (Samsung B-die) 2x8GB | DDR4-3800 CL14-15-14-28-1T @ 1.48V | vSOC @ 1.100v


Will continue tightening the subs

Update 2/4/20: Tried the other B-dies including my older TridentZ RGB 4266C19, D80 3600C17, NightHawk 3600C18, none of them is able to achieve stability using the above settings. Can't even stable at 14-16-16 at 1.50v.

This post has been edited by owikh84: Apr 2 2020, 08:46 AM

Well done bro.
At least your copy speed is always above mine, in which I don't know why haha.

I believe there are still some other locals who are into RAM OC just that LYN is not a medium of sharing their OC results. Hey this morning I met one local OC enthusiast in shopee, who has already delidded his 3800X chip, benching quite a number of B-die, E-die RAMs and resell them.

I used to write reviews for OCDrift.com website and most recently in LYN review section, no youtube though. It was fun but now no more due to time constraint. Btw, don't bother reading as those are noob reviews.

Ahh my bad, I swear I didn't notice that your E-die is DR haha.. mah eyes

or better don't delid it cos not worth the effort and time.
Definitely need to try that der8auer offset bracket, which he claims to offer 5-10c improvement on the temps.
Too bad Caseking doesn't ship to Malaysia.

This post has been edited by owikh84: Apr 2 2020, 07:13 PM

I think as long as micro-fins of our waterblock cover the whole chiplets we should be fine.
On the other hand, I need to spend time dedusting the rads, it's been a while.. haha

This post has been edited by owikh84: Apr 2 2020, 09:51 PM

Continue benching with B-die
As mentioned earlier, my older TridentZ RGB 4266C19 is weaker than the newer kit.
Only able to settle at 3800 CL16 instead of C14.

owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | G.Skill TridentZ RGB DDR4-4266 CL19 (Samsung B-die) 2x8GB | DDR4-3800 CL16-16-16-32-1T @ 1.42V | vSOC @ 1.100v


Mixing strong and weaker sticks, my copy speed is now 55k after messing around with some settings...

owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | G.Skill TridentZ RGB DDR4-4266 CL19 (Samsung B-die) 4x8GB | DDR4-3800 CL16-16-16-32-1T @ 1.43V | vSOC @ 1.100v




This post has been edited by owikh84: Apr 2 2020, 10:26 PM

Weekend benching with Hynix AFR... Took me quite a while to figure out that tRC=65 and tRFC=530 are minimum values that I could set, otherwise system will auto reboot in less than 10 minutes.

owikh84 | ASUS ROG Crosshair VIII Hero-WiFi | G.Skill TridentZ RGB DDR4-3200 CL16 (Hynix AFR) 2x8GB | DDR4-3533 CL16-19-19-40-1T @ 1.45V | vSOC @ 1.025v


As a comparison, here's AIDA64 score at stock RAM speed of DDR4-3200 CL16-18-18-38-1T @ 1.35v:


2x8GB Samsung B-die + 2x8GB Hynix AFR @ DDR4-3200 CL16-18-18-38-1T @ 1.35v

BIOS auto set without manual tweaking, AIDA64 lower read, higher copy, higher latency due to some loosen auto subs.

This post has been edited by owikh84: Apr 7 2020, 10:33 PM

Taking a break from RAM OC, now let's talk about CPU OC

PBO auto: - CB R20: 4674 cb


PBO auto, vCore offset -0.125v: - CB R20: 4816 cb


PBO manual, 200MHz OC, PPT 395w, TDC/EDC 255a: - CB R20: 4840 cb



Manual OC 4.2GHz @ 1.200v - CB R20: 5042 cb


Manual OC 4.25GHz @ 1.232v - CB R20: 5101 cb


Manual OC 4.3GHz @ 1.25v - PC auto reboot with CPU overtemp error

Edit 26/4:
Manual OC 4.3GHz @ 1.275v - Finally able to complete CB R20 with XMP off (auto RAM)


Need at least 4.4GHz to match my stock 9900KS: - CB R20: 5277 cb




This post has been edited by owikh84: Apr 26 2020, 12:14 PM

1080p Gaming Performance of Max OCed Samsung B-die vs Micron E-die vs Hynix AFR

Overall no significant difference but in terms of total fps B-die > AFR > E-die.

Finally able to do 4.3 without overtemp protection error.
I reduced the vDIMM down to 1.3v, vSOC 0.975v, VDDG 0.9v, VDDP 0.7v.
CPU phase control set to default instead of extreme.
Put case fans above DIMM slot and VRM, as well as a small fan under the CPU socket.

3700X 4.3GHz all core @ 1.275v




This post has been edited by owikh84: Apr 28 2020, 02:06 PM

Good news for B-die lovers, ADATA 3600C14 @ 1.45v is coming to our local market soon.



For record top binned B-die was G.Skill Neo 3600 CL14 @ 1.40v, even better than 3800 CL14 @ 1.50v kit.
However, this kit is now EOL and replaced by 1.45v model.