3600mhzCL18 vs 3600mhzCL14 impact on Cinebench20 score

[Noctua Fan Fan]

I upgraded my RAM today from 3600mhzCL18 to 3600mhzCL14 (Both 2*8GB kits).

 

My Cinebench score improved, but not by as much as I thought, as below:

 

 

When I've OC'd RAM in the past, improving the frequency, I saw a huge uptick in Cinebench. This time I've stuck with the same frequency, but naturally 4 steps faster in timings. Is it likely that the timings improvement is going to have a bigger effect on the 1% lows than the ceiling?

 

Other useful info:

 

 

Full system spec in signature

 

[porina]

Cinebench, old and new versions, were never that sensitive to ram performance. That difference in the screenshot is less than 1% so unless you did multiple runs to optimise it, that in itself could be a bigger factor.

[mr cheese]

Kinda looks like a little bit of run-to-run variance but hey, that's something at least, now that you mention it I'm gonna try that with my own ram

[Noctua Fan Fan]

Just now, porina said:

Cinebench, old and new versions, were never that sensitive to ram performance. That difference in the screenshot is less than 1% so unless you did multiple runs to optimise it, that in itself could be a bigger factor.

The runs I've done have had an average score of:

 

CL18: 4450

CL14: 4550

 

So ~100pts+ with the tighter RAM timings. I expected a much bigger bump though.

[Noctua Fan Fan]

1 minute ago, DaJakerBoss said:

Kinda looks like a little bit of run-to-run variance but hey, that's something at least, now that you mention it I'm gonna try that with my own ram

I'd love to find out your results! Would be super interesting to me.

I won't be pushing an OC on this for a while because my bottom rad is in the way of me clearing CMOS. I'm going to add an old reset switch to have a 'clear cmos button' in the summer when I rebuild the loop.

[Noctua Fan Fan]

An entirely 'gut feeling' observation is that my system feels snappier, little more responsive.

 

I have a feeling my 1% lows will improve.

[porina]

6 hours ago, Noctua Fan Fan said:

The runs I've done have had an average score of:

 

CL18: 4450

CL14: 4550

 

So ~100pts+ with the tighter RAM timings. I expected a much bigger bump though.

Your screenshot shows 7526 and 7481, so what's the difference? If you have the individual scores, I'd compare maximum of each case. It is what competitive benchmarkers will do to show the "best" run. Other things on system can slow it down, but nothing (short of cheating) will make it run faster than it can.

[xg32]

6 hours ago, Noctua Fan Fan said:

The runs I've done have had an average score of:

 

CL18: 4450

CL14: 4550

 

So ~100pts+ with the tighter RAM timings. I expected a much bigger bump though.

you'll benefit more with 4 sticks if your imc can handle it, i'd suggest trying 4x8 cl18 to see, and 2% sounds about right. ram upgrades are mostly for capacity and minmaxing.

[Analog]

7 hours ago, Noctua Fan Fan said:

An entirely 'gut feeling' observation is that my system feels snappier, little more responsive.

 

I have a feeling my 1% lows will improve.

 

Run Geekbench 3 and see what scores you will be getting there. The difference should be significant. 

[Noctua Fan Fan]

11 hours ago, Analog said:

 

Run Geekbench 3 and see what scores you will be getting there. The difference should be significant. 

I also have a gut feeling a better mobo would yield higher clocks.

[Guest]

2 hours ago, Noctua Fan Fan said:

I also have a gut feeling a better mobo would yield higher clocks.

The board is fine..... lol.

 

So you changed memory kit. CL14 3600mhz. 1.45v.

All you netted is lower latency. There is a very small bandwidth improvement not even worth the measurement.

 

That memory would be fine easily 1.50v daily. You want that 1:1 mem to IF ratio at the highest frequency you can get. 

End Goal 3900mhz. 

This memory will likely easily do 16-16-16-16-36 at 3900mhz. Maybe even at the XMP 1.45v.

 

After you've actually OCed a bit, then compare your benchmark scores. So far you've only left that part out.