First test of GSkill 16GB kit (F3-17000CL9Q-16GBXLD)

[Massman]

Kit and content of box

 

 

First set of results

 

I received this kit a few weeks before I was allowed to bin Sandy Bridge CPUs at Tones. Now, I have a CPU that can do ~ 108MHz BCLK on air, so I will have to retest the kit to see if it goes any higher than on the previous chip (max 5.4GHz). Also will try a benchmark that makes full use of all 16GB.

 

Tests will be continued after MOA!

 

Test setup:

 

- Core i7 2600k

- Asus P8P67 Deluxe

- MSI P67A-GD65

- Antec Truepower Quattro 1KW

- WD Raptor

 

Results:

 

This kit doesn't really scale that well with voltage. More than 1.75V makes the scaling go down quite rapidly. Also, CL9-11 isn't easy to push much further than the stock rated speed; with CL10 everything goes a lot more smoother, but then again ... CL10 looks so dirty .

 

- DDR3-2133 CL9-10-9

- DDR3-2146 CL9-10-9

 

- DDR3-2133 CL9-11-9

- DDR3-2144 CL9-11-9

- DDR3-2199 CL9-11-9

 

- DDR3-2234 CL10-12-10

- DDR3-2252 CL10-12-10

 

Edited April 11, 2011 by Massman

[Masterchief79]

Nice testing, doesn't seem to be real OC Ram if I compare it to my ADatas (2000C9) which can do CL9-10-8-24 _1T_ 2150MHz on 1,68V primestable. 1866MHz run with CL8-9-8-21 1T and only 1,475V (check screens).

But therefore you got 16GB ^^ And the Ram really should be fast enough @stock, too.

 

I guess the FAN is completely unnecessary and not exactly too silent, right?

 

Edited April 11, 2011 by Masterchief79

[Massman]

It's not really OC ram; you would buy this kit pretty much because it's 16GB .

[Hyperhorn]

@Masterchief79: Yes, capacity is the important difference here. DDR3-2133 is pretty much ok for 4 x 4 GiByte and for daily use nobody should expect a giant leap latency-wise anyway.

@Massman: Given the weak scaling at which voltage did you test the sticks actually? What about Memtest 86+, HCI Memtest, Prime 95 Blend or Hyper Pi?

[Massman]

I tested with 2x 4GB and the scaling was pretty similar. I assume there would be a bit less IMC overhead, so a bit better results. But it was not significant enough to continue pushing with 8GB.

 

What's the simplest stability benchmark of the 4 you mentioned?

[Hyperhorn]

The simplest to use is probably Hyper Pi as it is multi-threaded Super Pi leading to more IMC stress and higher RAM usage. For RAM reviews PCGH uses Memtest 86+ (bootable e.g. from USB, focus on #5 and #8 looping) and Prime 95 Blend with >90 % coverage, which works pretty decent even with Sandy Bridge.

Sometimes HCI Memtest delievers error detection very quickly, but without the full version there's a limit of usable RAM, so you have to open it multiple times and confirm that you use the unregistered version. If you want to prevent to boot with to aggressive settings, which can corrupt important OS files, Memtest 86+ should be the first step before a Windows tool of your choice.

[Leeghoofd]

think memtest86 has proven too many times already it is not that good (recall Thuban PJ )... Also Hyperpi 32M can pass, but HCI memtest will still error out... and is alike eg LinX that it needs a voltage bump or looser timing to stabilise...

 

For maximum ram stress HCI memtest is the way to go... just my 5cents...

[rdizz]

nice post OP

[Massman]

Even though this is not remotely interesting for 24/7 users and makes the huge timings make this stuff look oh so dirty, I still kinda like this screenshot

 

 

Oh right. I do use a special force to mount my CPU cooler for this: gravity.