Super Pi32mb Shoot out 1600mhz standard Timings vs 2133,2600,2666 + 2800

8 Pack

8 Pack

8 Pack

OcUK Staff
Joined
20 Feb 2012
Posts
10,163
Location
John Smiths Stadium
I did some testing inspired mainly by Surveyor who has asked a couple of times if I could run some average 1600mhz Dimms and Timings vs some higher binned and tuned dimms in this benchmark. So here we see 1600mhz vs 2133 Tuned, 2600 Tuned 2666 Tuned and 2800 semi tuned I say this as my best timings at this speed are on an ES CPU and I could not use these with this retail sample so I had to slacken a few vital ones as the controller on this one simply cant match my ES in terms of tight timings and high memory frequency. CPU speed on all these is constant at 5.8ghz my full benching speed is 6ghz but that needs a lot of volts. I thought 5.8 was enough to show up any differences RAM speed could make.

1600mhz

Please resize image to 1024 pixels wide (max)

2133 mhz Tuned

Please resize image to 1024 pixels wide (max)

2600mhz Tuned

Please resize image to 1024 pixels wide (max)

2666mhz Tuned

Please resize image to 1024 pixels wide (max)

2800mhz Semi- Tuned

Please resize image to 1024 pixels wide (max)

Any comments / questions let me know.
 
Thanks for the tests.

So there's very little difference after 2133MHz and around 6% difference between the bog standard, non-optimised (not even 1T) 1600MHz settings and your best results.

With the type of overclock on the processor most users are going to achieve I imagine the difference will be even less.

In my opinion these tests demonstrate:

1) The vast majority of Ivy Bridge users will be perfectly fine with 1600MHz RAM.

2) If the vast majority of users fancy having a tinker with higher RAM speeds then something like the Samsung Green is all they need. Not too expensive and it will get them nearly all the extra performance you can get from faster RAM.

3) The much more expensive RAM is really only for extreme overclockers/benchmarkers.

Thanks again.
 
Rossi~ said:
Samsung Green can go 2800MHz, it's a prime candidate for extreme overclockers/benchmarkers as well.
Click to expand...

Overclocking results are not guaranteed.

It would appear that some batches overclock better than others.

The best bet for the extreme overclockers/benchmarkers is to buy the fastest RAM and then try and overclock it a bit more if they wish rather than buy 1600MHz RAM and try and overclock it 75%.

But as the benchmarks above show up to 2133 MHz gives a modest but reasonable performance increase (with the processor at 5.8GHz) and beyond that it's diminishing returns.
 
Last edited:
Surveyor said:
Thanks for the tests.

So there's very little difference after 2133MHz and around 6% difference between the bog standard, non-optimised (not even 1T) 1600MHz settings and your best results.

With the type of overclock on the processor most users are going to achieve I imagine the difference will be even less.

In my opinion these tests demonstrate:

1) The vast majority of Ivy Bridge users will be perfectly fine with 1600MHz RAM.

2) If the vast majority of users fancy having a tinker with higher RAM speeds then something like the Samsung Green is all they need. Not too expensive and it will get them nearly all the extra performance you can get from faster RAM.

3) The much more expensive RAM is really only for extreme overclockers/benchmarkers.

Thanks again.
Click to expand...

I think in this your correct but I personally care about scores and about over clocking lower spec RAM too so I have a few different kits. I am gonna throw some PSC Based Gskill at tight timings at 2400 into the mix when I finish tuning. Thats another option for all I believe 2400mhz Cas7.
 
You must log in or register to reply here.
Back
Top Bottom