FEATURED G.Skill RipjawsX 2x4GB DDR3-2133 F3-2133C10D-8GXM

[Woomack]

Today's test will be about latest G.Skill RipjawsX series dedicated for Haswell platform so motherboards based on Z87 chipset.
As we see each newer series have more relaxed timings. Also here we see that older RipjawsX 2133 ( blue heatspreaders ) had 9-11-10-28 main timings and 1.65V voltage while new series are 2133 10-12-12-31 on slightly lower 1.60V.

Since I had this memory only for one day, there won't be many detailed tests. It's more an info what can you expect if you decide to buy this memory kit.


SPD/XMP - AIDA64 3.0



Memory is based on Hynix H5TQ2G83CFR H9C ( double sided modules ). You can also see it was manufactured in June 2013 so it's pretty fresh product.

 

[Woomack]

Tests 2133/2400/2600

Here are couple of tests. All were performed on i7 4770K@4GHz and Gigabyte Z87X-OC motherboard.
As you see new AIDA64 has already results based on multithreaded tests while older version the same as many other benchmarks on the market are showing only single thread transfers.

2133 10-12-12-31 1N 1.60V - settings really close to XMP #1/2 ( both profiles are about the same with some little changes not affecting overall performance )
I couldn't tighten main timings up to 1.685V so settings are not much better than what we see in XMP profiles.




2400 11-13-13-31 1N 1.65V - also here raising voltage didn't help much. Clock is higher but with more relaxed timings.




2600 12-14-14-34 1N 1.65V - My kit is scalling up to ~2600. It's not bad result but timings could be tighter. I wasn't able to make it pass HyperPi @2666 at higher voltage or @2600 using tighter timings and higher voltage.
However it can be possible, I just had not enough time to play some more with sub timings.



As always feel free to comment or ask questions but as I already said I had to return this memory so I won't be able to perform more tests.

 

[Sabishiihito]

Man that's some low binned CFR

 

[Woomack]

On Haswell tighter timings are not really so important as for older platforms.
I added tests in new AIDA64 also to show difference between each setting. Single threaded benchmarks are showing similar results at lower CPU clock ( here it was 4GHz ) no matter if it's 2133 CL10 or 2600 CL12 but multithreaded ( like new AIDA64 ) are showing +/- 4GB/s difference.
Actually well programmed sub timings can give more so it's not like first 5 timings are saying all about memory.
It doesn't change fact that I was counting on a bit better overclocking. More like 1 memory ratio higher for every timing set so up to 2800 12-14-14.