Help with Ryzen RAM OC performance/stability choices

Have a couple questions on a Ryzen 3 2200G system I’m building.

Please bear with me, I’m a newbie with OC’ing. I didn’t realize that if I wanted my RAM to run above 2133 on these systems I would be required to OC it to some degree. Since I spent the money on faster RAM and from all I hear Ryzen will excel with speeds 2600+ I’m forced to play around :wink:

Some of my questions here may be nit-picky, but I would like to really understand what is going on here and some of the decisions/math serious OC’ers look at.

I’ve got a hold of two different sets of RAM:

a) G.SKILL - which is 2666 @ 15-15-15-35


b) PATRIOT which is 3000 @ 16-16-16-36

So first, I have some questions about THEORY:

If my math is correct ((T = 1 /f) * CL), based on what I can gather the total latency for these is 11.25ns and 10.67ns respectively.

(Seemingly making the PATRIOT the better choice based on total latency)

The G.SKILL I have tested so far to 2933 @ 14 CL.
This would equate to a total latency of 9.54ns.

The PATRIOT I have tested so far to 3200 @ 18 CL.
This would equate to a total latency of 11.25ns.

Is my “math” right here in that on paper the G.SKILL RAM should perform better (at least in terms of latency)?

As far as real world testing, I got the following results from AIDA64 memory tests:

The G.SKILL @ 2933/14:

42676/43613 /39350/70.6

The PATRIOT @ 3200/18:


It would seem the G.SKILL has slightly better latency, but the PATRIOT has faster read/write speeds. As a data point, the PATRIOT at 2933 has the following numbers:


So at equivalent speeds it seems the read/write speeds are very close, but the G.SKILL has better latency. I’m assuming that this is due to the lower CL that the G.SKILL has ??

Also, I’m interested to know how the latency computed by AIDA compares to the theoretical latency computed above. I computed 11.25ns for the PATRIOT at the given speeds, but AIDA is saying 71.9ns.
However, the latency listed for the L3 cache was about 10.7ns…much closer to the theoretical computed latency. I have no idea if the closesness of those numbers is coincidental - I’m just curious the relationship between computed and perceived latency.

FINALLY - I’m interested in knowing which of the RAMs I should pick…I think the answer is the PATRIOT, and here is why:

  1. The G.SKILL required more tweaking to my settings to get it running above 2400. It did some really weird stuff with the Ryzen GPU without disabling GDM and changing all the timings. OTOH the PATRIOT worked well out of the box up to 3200 with only a small bump in voltage.

  2. The G.SKILL seems like a more “recommended” RAM, in that they seem to use better quality. Plus it has better latency seemingly even at much lower speeds. However, the best I’ve been able to get it was 2933 with the tighter timings and at that I am having some failures in Prime95. I’ve run for over 12 hours without issue but before that (and once since) I had a shut down in under an hour.

  3. I can’t seem to push the PATRIOT past 3200 either without BIOS retraining down to defaults. Even if the latency is a bit slower on the PATRIOT, it seems to be a bit more stable (even though I’m still running Prime95 tests to be sure). Plus - I can keep it for $5 less than the G.SKILL :slight_smile:

My guess is if I loosed up the timings just a bit on the G.SKILL I can probably get more stability, though I’m kind of wary about being able to go for 12+ hours on Prime95 with no issues 2x in a row and then just have a failure @ 30 minutes. But, if I relax the timings I’m probably lining up closer with the PATRIOT speeds/latency anyway…so what’s the point.

Any good reason here I should try to work more with the G.SKILL?

G.SKILL is fairly characterized as a manufacturer that pre-selects (heavily) their DDR4-RAM chips. You will get DDR4-RAM kits from G.Skill that are guarantede to run at least the specified speed, e.g. my F4-3600C19-8GVRB runs fine w/my Ryzen 7 1700 and I can only get it to boot into 4000 MT/s but it will result in freezes, 3933 is somewhat unstable, esp. when heavy on RAM, 3818 (w/36.66 blck) is stable but not in long run RAM-stress tests (mprime gave a rounding error after days of working fine), only 3744 MT/s is running solid so far (all @ CL19).
This is my 24/7 Ryzen 7 1700 build:
@ 37 * 102=3774 MHz (CPU), 36.66 * 102=3739 MT/s (RAM) Cl19 XMP, 102 MHz (BCLK)
tweaked from optimal defaults @ max=71°C (ambient=22°C) status: pass

Hello, my pet. I see you’ve violated one of the many rules of the Level1Techs… This thread has been locked. Return to layer 7 and message the abused admin that has done the deed to reopen.

Now it’s time to pay the piper…