should I be happy with this ram oc or keep pushing either frequency or timings here? (3000mhz cl15 kit at 3600mhz cl16)

Joined
Sep 14, 2016
Messages
33
hello all, so I have a crucial ballistix micron E-die 2x8gb kit rated for 3000mhz cl15-16-16-35 1.35v,

after some testing back and forth and a day or so of testing different combinations of vccsa/vccio/vdimm and different timings and frequencies,
i'm now running the kit at 3600mhz 16-17-17-38 with 1.15v on both vccsa and vccio and 1.4v on the vdimm, ran memtest up to 400% coverage and called it a day for stability testing,


I haven't touched tertiary timings at all and just slightly tightened the secondaries thus far after looking at some guides, anyway at lower frequencies than 3600 with tighter timings,
or higher frequencies with looser timings, it seems I always sooner or later eventually run into errors in memtest or BSODs, depending on the oc of course, but havent really gone that deep into timings yet I think,


just wondering if this oc is decent-good-average for this kit on a z390 cheap motherboard in general, should I continue or be happy with this? :p

got some decent (?) gains vs the xmp profile in aida64 anyway, but IDK how good these numbers are compared to what I could get with better timings on that frequency,
or tighter timings at slightly lower frequency perhaps:


3000mhz xmp (15-16-16-35 ):

read: 43200 mb/s
write: 42800 mb/s
copy: 38300 mb/s
latency: 50.5 ns


3600mhz oc (16-17-17-38):

read: 50851 mb/s
write: 51708 mb/s
copy: 46402 mb/s
latency: 45.5 ns

current subtimings:

DA2MxvN.png

mobo: msi z390-a pro
cpu: 9700k @ 5ghz 1.3v, 4.5ghz uncore/ring


if anyone sees any obvious subtimings I could tighten up easily or something let me know, and yeah, should I just be happy here and call it a day? :D

thanks for any feedback, cheers
 
Last edited:

Archaea

[H]F Junkie
Joined
Oct 19, 2004
Messages
11,551
What do you use the system for? If for gaming you’ll be hard pressed to ever tell the difference in real world use.
 

dasa

Gawd
Joined
Feb 19, 2012
Messages
572
Looks like you have them tweaked down fairly tight for e die

Have you tried 1T?

I wouldn't worry about tertiarys unless you are getting instability issues from the MB setting them wrong if so just lock them to current settings.

Try
tRRDS=5, tFAW=20 or tRRDS=4, tFAW=16
If your lucky you may be able to shave ~100 off tRFC

You may be able to set ~30000+ tREFI but if this is unstable it wont show up in a stress test but with crashes during idle usage.

Or if you want to try for higher frequency loosen these significantly
  • tRCD
  • tRFC
  • With tRRDS=8, tFAW=32
 
Joined
Sep 14, 2016
Messages
33
What do you use the system for? If for gaming you’ll be hard pressed to ever tell the difference in real world use.

well for gaming mostly yes! but I was under the impression that in my choice of gaming, which is competitive gaming with lowest quality settings to reach 240+ fps,
higher memory speeds / tighter timings could potentially boost perf from decent to a lot, but seems to depend on game/engine quite a bit too..

from what I saw in some ddr4 scaling benchmarks at least things like minimum and 0.1% fps and even average/max fps numbers could benefit from slightly to pretty substantial when tweaking memory?


I haven't benchmarked my games though with the OC since for example in fortnite, my main game of choice, there is no ingame benchmark with consistent scenario for testing,
but in the end I just wanted to achieve a decent to good OC on my ram to sleep better at night Lol! if that makes sense :p


so like a week ago I was running stock cpu/ram/gpu settings, and did so for the past months since I got my current system, but now after OCing cpu to 5ghz from 4.6 stock, ram to 3600 cl16 from 3000 cl15,
and +1000 mhz on the gpu memory (msi gaming z 2060) the system does feel slightly more responsive and snappy with less scenarios of stutters and hitching, just kind of feels a bit more fluid, nothing extreme in terms of
percieved difference but slightly better in my subjective experience ! maybe placebo maybe not :D



Looks like you have them tweaked down fairly tight for e die

Have you tried 1T?

I wouldn't worry about tertiarys unless you are getting instability issues from the MB setting them wrong if so just lock them to current settings.

Try
tRRDS=5, tFAW=20 or tRRDS=4, tFAW=16
If your lucky you may be able to shave ~100 off tRFC

You may be able to set ~30000+ tREFI but if this is unstable it wont show up in a stress test but with crashes during idle usage.

Or if you want to try for higher frequency loosen these significantly
  • tRCD
  • tRFC
  • With tRRDS=8, tFAW=32



hello, I did try 1T but that was immediately unstable even at lower frequencies, even at XMP profile if I recall correctly :(

on a sidenote I managed to run up to 100% coverage in hci memtest on 4000mhz cl18 on this kit, but somewhere around 150% I got an error and went back down again,
the timings were 18-19-19-40 in that case with vccio/sa at 1.2-1.25v and vdimm at 1.45-1.5v (tried few diff combos but couldnt get 100% stability),

and latency was very close in aida64 to my current oc , so I just went with this current OC instead of further tweaking, felt like a somewhat "sweet spot" in terms of performance and voltages for daily use, IDK.

the read/write/copy tests were a bit better though than now with 4000mhz cl18:

4000mhz c18-19-19-40:

read: 53711 mb/s
write: 55605 mb/s
copy: 48270 mb/s

and again current oc numbers:

read: 50851 mb/s
write: 51708 mb/s
copy: 46402 mb/s


but latency was same-ish around 45ns so I stepped down again to this stable oc with lower vccsa/vccio/vdimm..


think at some point I will just buy some high rated kit instead, was looking at G skills 2x8 gb 4400 mhz trident Z royal CL18-19-19-39 1.40V kit today and looked pretty nice to me, any thoughts on that? :p
think it would be possible to get that kit and tweak it down to like cl 16-17 ? and would there be any point for me? (mostly use pc for gaming, but only cpu-bound gaming at 240+ fps with low quality settings,
where tweaking memory *should* be beneficial? as far as I know ?)


cheers and thanks for the tips on the timings, if I decide to continue tweaking this crucial memory kit I will use your post as guideline :)
 
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
572
You may be able to run 4000 17-20-17 more stable and faster than 18-19-19 with e die.
Try with 750 tRFC just to see if it helps stability at 4000.
 
Last edited:
Joined
Sep 14, 2016
Messages
33
You may be able to run 4000 17-20-17 more stable and faster than 18-19-19 with e die.
Try with 750 tRFC just to see if it helps stability at 4000.

hm, I was certain that on Intel z390 platform tRCD and tRP have to be the same? so would have to be 17-20-20 or 17-17-17 or 17-18-18 and so on ?

at least that is what I saw in some buildzoid video on his actually hardcore overclocking youtube channel :S

but perhaps I could try 17-20-20-40 or something like that on the e die at 4000mhz ... in the coming days i'll give 4000 mhz another go with your tips in mind on the timings and see if I can get it stable anyway!
 

Wooshoo

Weaksauce
Joined
May 27, 2019
Messages
110
That seems really good to me but I’m on a Ryzen platform with hynix DJR chips that I had to add voltage to just to reach the rated 3600mhz
 
Top