How to tighten timings on LPX cas 18 3600

T

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
9,743
So im almost done building my 8700k and I just got a set of LPX 3600 at 18ns from Corsair at 16gb for 209 dollars.

Will faster CAS make a difference with kaby/coffee or will raw throughput be the biggest difference?

Is it possible to tighten timing, leave clockspeed at 3600mhz and hust add voltage? Or would there be any benefot tondropping to the ram to say 3466 and run cas 15 or 16 at higher volts....?

I am not a good ram OCr. Thanks for advice.
 
fire up aida64 and whatever gives you the lowest final latency will perform fairly well keeping in mind that increased frequency improves latency and bandwidth while tighter timings mostly just reduces latency
here is some aida64 results of mine for a rough comparison
http://www.overclock.net/t/1611359/...3-1600c11-2133c9-ddr4-2133c15-3000c12-4000c17

there is some great info here on overclocking memory that helped me a lot
http://www.overclock.net/t/1569364/official-intel-ddr4-24-7-memory-stability-thread
i found that the linux gsat test was best for quickly finding memory instability and the hci test was also good at finding memory controller instability
 
tangoseal said:
Will faster CAS make a difference with kaby/coffee or will raw throughput be the biggest difference?
Click to expand...
A rough guide for ram performance is frequency/CAS. The higher the number the better. If two sets are close, go with the higher frequency.

3600/18=200
3466/17=204
3466/16=217

If you can get it stable, the 3466/16 should be marginally better.

tangoseal said:
Is it possible to tighten timing, leave clockspeed at 3600mhz and hust add voltage? Or would there be any benefot tondropping to the ram to say 3466 and run cas 15 or 16 at higher volts....?
Click to expand...
As for the voltage, do your research and figure out what is considered safe. Set it to what you are happy to run at, then play with the frequency/timings.

On my LPX 3200 I managed to tighten timings from 16-18-18-36 2T to 15-16-16-28 2T or 15-16-16-35 1T with only adjusting the System Agent and VCCIO voltages. 1T required a much higher System Agent voltage.
 
Last edited:
7x1x said:
A rough guide for ram performance is frequency/CAS. The higher the number the better. If two sets are close, go with the higher frequency.

3600/18=200
3466/17=204
3466/16=217

If you can get it stable, the 3466/16 should be marginally better.


As for the voltage, do your research and figure out what is considered safe. Set it to what you are happy to run at, then play with the frequency/timings.

On my LPX 3200 I managed to tighten timings from 16-18-18-36 2T to 15-16-16-28 2T or 15-16-16-35 1T with only adjusting the System Agent and VCCIO voltages. 1T required a much higher System Agent voltage.
Click to expand...
That calculation is pretty helpful I have never seen that before.
 
3200\14=228 gives you a rough idea of the final latency which is the most important thing but bandwidth still matters and higher frequency will provide more bandwidth

any 3200c14 ram uses samsung b-die which overclocks really well making it some of the best value ram available
the flare-x has its sub timings tuned to suit am4 but it probably works fine on anything

my 3200c14 kit runs 3866c16 1.435v
this image was benchmark stable but i had to back it off to T2 TRFC 310 to get it completely stable
610de705_6700k204.7203866c17t1.jpeg

3000\12=250
4866\16=241
4000\17=235
as you can see that isn't perfect for calculating performance
3000c12 has a similar latency to 3866c16 but a lot less bandwidth and slightly lower performance
but it is simple and great for a rough idea
dd4a6065_6700k204.7203000c12.jpeg

71c7c2b3_Arma20III20cpu20vs20ram.png

http://www.overclock.net/t/1611359/...3-1600c11-2133c9-ddr4-2133c15-3000c12-4000c17
 
Last edited:
Just thought I would update with some findings.
15-16-16-35 1T wasn't stable. It was failing Prime95 blend after about 9.5 hours.
15-17-17-35 1T looks to be all good, so sticking to the stock timing differences might be the way to go.
 
7x1x said:
Just thought I would update with some findings.
15-16-16-35 1T wasn't stable. It was failing Prime95 blend after about 9.5 hours.
15-17-17-35 1T looks to be all good, so sticking to the stock timing differences might be the way to go.
Click to expand...
Hey what voltage are you at for mem and system agent? I have a similar Corsair LPX kit C15 3000mhz that Im going to work on, can't seem to get 1t yet
 
Last edited:
Speedeu4ia said:
Hey what voltage are you at for mem and system agent? I have a similar Corsair LPX kit C15 3000mhz that Im going to work on, can't seem to get 1t yet
Click to expand...

Ram: 1.35V
VCCIO: 1.065V
System Agent: 1.200V

On a really strange side note I couldn't get the ram stable using adaptive voltage (CPU) on my motherboard. Offset voltage is fine though.
 
7x1x said:
Ram: 1.35V
VCCIO: 1.065V
System Agent: 1.200V

On a really strange side note I couldn't get the ram stable using adaptive voltage (CPU) on my motherboard. Offset voltage is fine though.
Click to expand...
Thanks what version of p95 do you use?

doing some testing 3200 c15-17-17-35-1t currently using same voltages
 
Last edited:
7x1x said:
I'm currently using 29.1. I've heard LPX DDR4 can be a fickle beast for overclocking. Some people get good results, others can only run stock.
Click to expand...
I had to back my OC down to 4.7ghz and I think that was my issue, ran 3000mhz 1t for longer that ever p95. Guess this is kind of crappy chip, oh well thinking about working some magic and trading even up for a new 8600k and board in a few months
 
Last edited:
Higher MHz is a marginally better improvement for most things than lower latency settings.
(note that higher speed also decreases latency without changing CAS)
It may need a voltage boost but LPX can take pretty high voltage.
I wouldnt exceed 1.5V though as the CPUs memory controller could be harmed.

My LPX 3000 C15 clocks to 3733MHz C17 with 1.44V.
 
You must log in or register to reply here.
Back
Top