It's been a while since my last submission
For starters, here's my early stage of OCing adventure with the 2nd Gen Ryzen and X470 chipset.



Max Prime95 stable OC for this particular Ryzen 7 2700X chip of mine is 4.25GHz @ 1.44v. Load Tdie temp reached 74.4C under custom loop.



Continuing with RAM OC, I managed to get the Patriot Viper LED DDR4-3600 CL16 (Samsung B-die) kit running HCl Memtest Pro 300% stable at DDR4-3466 CL14-15-14-30-1T @ 1.40V, vSOC @ 1.125v.

owikh84 | ASUS ROG Crosshair VII Hero-WiFi | Patriot Viper LED DDR4-3600 CL16 (Samsung B-die) | DDR4-3466 CL14-15-14-30-1T @ 1.40V | vSOC @ 1.125v



Now still figuring out how to stabilize the RAM at DDR4-3533. Not an easy task as it's not a higher binned B-die kit.

DDR4-3600 is far from being stable but still able to run a quick test like AIDA64


This post has been edited by owikh84: Aug 9 2018, 08:45 PM


Attached thumbnail(s)

For B-die, you can get one of these awesome kits:

G.Skill Trident Z RGB DDR4-3200 CL14 (F4-3200C14D-16GTZRX)

or

Team Dark Pro DDR4-3200 CL14 (TDPGD416G3200HC14ADC01)

Today decided to swap the RAM with the higher-grade B-die coming from TridentZ RGB DDR-4266 CL19 kit.



I managed to get DDR4-3533 CL14 @ 1.44v with the help of DRAM Calculator 1.3.1 Safe preset. HyperPi 32m stable but just a little bit disappointed with the memory latency, will need further tweaking later...



====================================================================

owikh84 | ASUS ROG Crosshair VII Hero-WiFi | G.Skill Trident Z RGB DDR4-4266 CL19 (Samsung B-die) | DDR4-3533 CL14-15-15-30-1T @ 1.44V | vSOC @ 1.175v

HCl Memtest 400% passed after a bit of tweaking: using Alt2, bumping the SOC up to 1.175v, and RTT_NOM disabled:


Mem latency improved now


This post has been edited by owikh84: Aug 9 2018, 08:43 PM

owikh84 | ASUS ROG Crosshair VII Hero-WiFi | G.Skill Trident Z RGB DDR4-4266 CL19 (Samsung B-die) 4x8GB | DDR4-3333 CL14-14-14-30-1T @ 1.400V | vSOC @ 1.100v



That's 1 hour, for 400% I will need to run at least 5 hours.

Update: I managed to reduce the voltages by adjusting the Termination and CAD_BUS Blocks as displayed in the DRAM Calculator's Rec fields. As a result, HCl Memtest 400% passed for 5 hours without any error.

owikh84 | ASUS ROG Crosshair VII Hero-WiFi | G.Skill Trident Z RGB DDR4-4266 CL19 (Samsung B-die) 4x8GB | DDR4-3333 CL14-14-14-30-1T @ 1.390V | vSOC @ 1.025v





This post has been edited by owikh84: Aug 12 2018, 06:43 PM

Nice try bro
For more efficient performance and reduced latency, I highly recommended using the Ryzen DRAM Calculator.
This is Hynix CFR 3200 Fast preset:


Use DangWang HCl Memtest Pro: https://drive.google.com/file/d/1m8Qg6HECXV...ZRM0omM-O4/view

Nice bro.
Those subtimings, did you refer to DRAM Calculator or let the mobo auto-set?

This post has been edited by owikh84: Sep 10 2018, 11:41 AM

Just want to share another RAM testing tool from 1usmus @ OCN, not sure about its consistency though.

TM 5 0.12 1usmus config v2 (memory test)



Why this test/config?
Most of the tests work on a strict template. I think it's all a waste of time. In this configurator, the pattern is random. There is a re-allocation of memory for tests. Also, temperature-related tests for errors are added, tests in which the write and read offset occurs.
For a dual rank of 5 cycles, for a single - 10, (recommendation).

Download test:
http://testmem.tz.ru/tm5.rar

Download config:
https://drive.google.com/file/d/1oN5MPOxJvi...iew?usp=sharing

How to instal:

In the bin folder replace MT.cfg

Source

This post has been edited by owikh84: Sep 10 2018, 12:00 PM

3600, nice!
That's the advantage of a 2-DIMM mobo with short signal trace.
At what voltage for that OC? How about the RAM latency now?

For 3600, 4-DIMM mobo would need more than that or worse, not possible at all with average IMC and RAM chips.

I've recently sold ditched my 2700X + C7H and went for 1400 + Prime B450M-K in my chick's rig. If got free time I will try playing RAM OC with the latter to see how it clocks.

Oh yeah, your RAM latency is quite high most probably because of that tRFC being too loose. I'd say tightening it up would shave off a good amount of ns.

This post has been edited by owikh84: Nov 14 2018, 11:14 PM

It will work and your mobo is C6H, should be easy to run at the rated speed even with populated DIMMs.
I have 2 kits of non-QVL B-die TridentZ RGB 4266 CL19 running stability at 3333 CL14 4x8GB @ 1.39v with the 2700X + C7H.
With 1700 + C6H, it was able to do 3200 CL14 4x8GB @ 1.40v.
MemTest HCl 400%+ of course, don't use other methods unless Prime95 custom blend 1344K FFT with 90%+ RAM capacity stressed for 3 hours minimum.

Not only Corsair, actually any other RAM vendor will advice end users to purchase kit rather than mixing sticks. Your case was with X99 chipset, which had tendency to have RAM incompatibility issue.
Yes, it is safe to mix between different production as long as they are both based on the same RAM chip AKA both are B-die's. My kits were from initial batch of 2017 (RGB just died and pending for RMA) and 2018 (a much better kit with improved RGB quality according to G.Skill).

Your current kit is GTR or GTRX (AMD certified)? And what edition are you looking to buy?

The GTRX has been in local market since quite a while ago, you can find it at RM1099-1199.
Good question... but so far i've never seen any people mixing the GTR and GTRX. But I think they will mix well cos both are also B-die. Only difference is that the latter has been tested and certified to work well with the Ryzen platform.

Great! Good to know that it actually works perfectly on Ryzen as well.
4x8GB B-die, should be able to OC up to DDR4-3333 CL14-14-14-30-1T @ 1.390V

As mentioned by abam khoo, an average Ryzen 5 1600 should be able to overclock to 3.75-3.8GHz on stock cooler.
I'd say 1.35v max at 100% load before throttling.
Keep the CPU core temps (Tctl/Tdie measured through HWiNFO64) below 75c if possible.
I also noticed that your mobo is AB350, might be a good idea to actively cool the VRM with a fan pointing toward the heatsinks surrounding the CPU socket.

Performance gain in CPU intensive tests according to TPU is about 8%:


However, in gaming environment based on G3D results the improvement is almost negligible:


This post has been edited by owikh84: Feb 20 2019, 11:20 PM

abam khoo bought new mobo again...
But where's your AB350M?

B-die tRCDWR can be set as low as 8 and still maintains stability!

TridentZ RGB for AMD 3600 CL18-22-22 1.35v kit is MFR, remember?

RAM IC doesn't matter, be it MFR/CFR/AFR/B-die. They are comparable to each other at the same specs.
Among these chips, Samsung B-die is the best overclocker as it scales well with the voltage.

Ryzen CPU architecture uses AMD's Infinity Fabric, which directly linked to RAM speed.
So faster memory will give a better CPU performance.

RAM clock speed (higher is better) and timing (lower is better) are more important.
Best way to know which RAM kit is faster is by calculating its true latency (lower is better):

3600 CL18 = 2000/3600*18 = 10.00
3600 CL16 = 2000/3600*16 = 8.89 (faster than 3600 CL18)
3200 CL16 = 2000/3200*16 = 10.00 (same as 3600 CL18)
3200 CL14 = 2000/3200*14 = 8.75 (fastest)

New toy in a small footprint.. Pairing it with 2600X