Overclocked RAM but Benchmarks are slower!!

Associate
Joined
31 Aug 2004
Posts
44
I`ve just upgraded my PC and have been busily overclocking since Thursday!
Haswell Core i5 4670K was increased from 3.4Ghz to 4.2Ghz and is stable at 1.22v with temps under Prime 95 large FFT test maxing out at approx 62/63 under a Corsair H80i :)

Managed to get a 10% core overclock and a 12% memory clock on my GTX 770. :)

I am puzzled by my RAM results though and would appreciate it if any of you sages might understand what is going on:
I have Kingston Hyper Beast 2400Mhz (2 x 4Gig) running at 2400, 11,13,13,31,1T and ALL benches (including everyones favourite - Windows Experience Index) show a drop in performance if I run the same tests at 2600,11,13,13,31,1T with a small voltage bump to 1.69v. I don`t think voltage is the explanation because I left it at the same value when I down-clocked back to 2400Mhz and got faster results. :confused:
Any explanations as to why my overclocked RAM actually slowed down because I `m somewhat confused. Thanks in advance!
 
Windows experience index is not a benchmark!! Its garbage!!! utter garbage!!!

Use 3D 11 Physics test for example to measure the improvement memory clocking can offer in games for example.

As wazza says overall memory bandwidth and perf can be measured using sofware showing read write and copy speeds.

Often what also happens is when you clock up RAM you slacken not only primary timiings but auto timings on sub timings also slacken in memory training so performance gets worse overall. When you overclocking RAM keep all the timings as close to default as possible and push frequency then if possible tweak timings further.

TWCL and TRFC for example effect performance a lot in the sub timings.

Oh and Prime is not good. Use your PC to test stability.
 
Thanks guys. I have an ASUS Z87-A board which has the UEFI bios and is quite overclock friendly. Any suugestions for RAM timings other than the 2600,11,13,13,31,1T that I have already shown you. CPU-Z confirms that the RAM IS running at 2600 by the way.
 
you can use aida64 to test the read/write/cop figures or maxmem

could try lowering the 31? maybe 28? I doubt youd be able to improve timings much at 2600mhz
 
Back
Top Bottom