[MemoryReaction] G.Skill F2-6400CL4D-HK

[EnJoY]

Introduction:

I received a sample from G.Skill around two weeks ago and I've finally had the chance to finish up my review on this exciting new memory kit. When I first received this kit, not much was known about it, and frankly I assumed with it's timings that it had to be Micron. I was quite surprised after asking around that the kit contained new Promos IC's.

It's great that memory manufacturers have been able to find a performance substitute to the Micron based IC's. Micron IC's are extremely expensive, and while they perform well, they aren't the only game in town, and I suppose that's what G.Skill and other manufacturers are trying to prove with these alternative IC kits. Now, without further ado, let's get on to the review.

The Memory:



-------------------------------------------------------------


Test Setup:

Intel Core 2 Duo "Allendale" E6400 - Lapped
Asus P5B-Deluxe rev1.03G - Stock, Bios: 908beta
2x1GB G.Skill F2-6400CL4D-HK
BFG GeForce 7950GT 512MB
OCZ PowerStream 600w

Cooling:

CoolTechnica AQX MP-05 SP LE, Swiftech MCW60, Black Ice Pro III, Swiftech MCP655, 7/16" ID
120mm Yate Loon @ 12v over memory

-------------------------------------------------------------


Testing Methodology:

The testing process I used is quite simple, and I'm guessing quite close to the same procedures the bulk of you use to test your new hardware. I simply, hopped in bios, set the lowest reccomended voltage by G.Skill themselves, which was 2.0v, and set the stock rated timings of 4-4-3-5-3. I then proceeded to exit bios and enter Memtest v1.70. I tested using tests 1, 2, 4, and then 10 loops of test 5 before moving on to the next speed up. I worked in the 2:3 ratio and moved up 2MHz on the FSB at a time. I also used full vFSB and vMCH voltages the entire time to ensure the chipset wouldn't hold me back. I increased clock speed until I would error in memtest, then I'd back down 1MHz on the FSB at a time until the errors disapeared, then I'd log my results and push up the voltage another notch and do the same testing with that. Once this was done, I entered windows and began testing in Orthos using 2hr test periods to attempt to document some level of stability, atleast higher than SuperPi 32M, while still maintaining my social life and other responsibilities. I experienced fairly large discrepancies at some points throughout the Orthos testing between what Orthos was saying was stable and what Memtest had previously passed. This just goes to show all those of you who think Memtest is the end all be all test suite for memory, it isn't, it's a starting point.

The Review:

Each piece is separated by timings and voltages. I started with 4-4-3-5-3 and ended with 5-5-5-15-5 to achieve maximum MHz.

-------------------------------------------------------------


Timings: 3-3-3-5-3

2.35v


2.4v

No Gain, Actually Lowered Stability.

2.45v

No Gain, Actually Lowered Stability.


Timings: 4-4-3-5-3

2.0v


2.1v


2.25v


2.35v


2.4v-Higher

No substantial increase in stability or clocking was achieved with voltages above 2.35v. It's quite apparent that these sticks like to stay in the lower voltage zones, especially when compared to Micron D9.

Let's see how they do with 5-5-5-15-5 timings shall we?

[EnJoY]

The Review (Continued):

Timings: 5-4-3-10-3

2.0v


2.1v


2.25v


2.35v

Extra voltage with these timings netted me only an additional 5MHz of stability in windows, however they were Memtest stable at 1080MHz (50 passes). Not windows stable.


Timings: 5-5-5-15-5

2.0v


2.1v


2.25v

Not windows stable

2.35v

Not windows stable

-------------------------------------------------------------


Conclusion & Verdict:

We can see by loosening the timings, even slightly (CAS 4 -> CAS 5), we gain a substantial boost in the MHz/Voltage ratio. The sticks clock nearly to 1000MHz at their stock voltage with this somewhat minor timing change and surpass it with a small voltage boost. Knowing this, I can say that both CAS Latency and RAS to CAS Read/Write Delay are by far the most influencial timings for these memory chips. TRP hardly seems to effect them at all in regards to stability, and really only comes into play onces the sticks pass the 1100MHz point, which from what I've seen they are quite capable of doing.

Regarding my 5-5-5 high latency testing results, I was able to acheive nearly 1200MHz MemTest stable (50 passes at 1170MHz to be exact), with 2.4v. Unfortunately my chipset seems to be a brick wall in that no matter what I do, I cannot seem to stay windows stable anywhere above 1060MHz or so, regardless of CPU speed or FSB. I'm quite confident that if these sticks were on a volt modded 965 board, or especially a 680i based board, they would certainly soar through windows at 1200MHz or more with relative ease.

Given these results and my experiences with the sticks, I can happily recommend them to anyone looking for cheap and reliable performance memory modules. If you are using a p965 based board, or 975x, you'll see excellent results with these up to nearly 1100MHz, and if your chipset is volt modded, you should see even more! If you're on 680i, you're most certainly in for a real treat with some high clocks and possibly even tighter timings at those speeds than even your Micron can offer with the same voltages.

Concluding, I'd like to thank G.Skill for the review opportunity. This has been a tech reaction by yours truly, EnJoY.

You can find these sticks here at Newegg.

[Hutch]

Good stuff man!!! 520mhz cas 4!

Get back inside and do the cas 5!!!! :-P

[Hutch]

Bazi you're running looser timings... good results nonetheless.

[McPhisto]

Nice Promo chips here!
Why don't you run Blend test in Orthos?

[r00n]

Originally Posted by McPhisto

Nice Promo chips here!

It's ProMOS, but I would have to agree on the quality of their new ICs.
Who would have thought ProMOS to become the second best next to Micron...
Really an amazing achievement.

EnJoY, good job on the review.
Looking forward to the second part with some CAS5 clocks.

[dinos22]

looks like they are all pretty much doing 600MHz 5-5-5-x ~2.5v

[EnJoY]

Thanks guys, I promise the 5-5-5 timing results will be even more exciting.

And incase anybody is curious as to why I skipped 4-4-4 results... Basically, I saw zero increase in stability or overclock from RAS Precharge going to 4 from 3. I don't know why, but it just didn't help, so I decided to just shoot right up to 5-5-5, which is what I'm testing right now.

[G H Z]

Nice pics

[Hutch]

With Elpida and Promos usually you don't see any increase in stability between ras to precharge 3 4 or even 5... try 4-5-4 timings or 4-5-3 timings. Ras to Cas delay 5 usually increases stability a lot on both elpida and promos.

[ozzimark]

try 5-4-3 as well

[EnJoY]

Originally Posted by ozzimark

try 5-4-3 as well

Yea, I think I will. I'm also gonna try 3-3-3. I'm having a ton of trouble getting into windows with my system. I'm seeing DDR2-1140 with only 2.25v, and it's memtest stable, but I can't for the life of me get into windows anywhere near that. It's disturbing and I've played with tons of settings. I'll keep toying, but because of this, I won't be posting the 5-5-5 results tonight. I'll have 3-3-3 results tomorrow without fail to tide you guys over until I can get 5-5-5 worked out.

[-n7-]

Damn nice results...

Now the question is how can we tell the crap Promos from this good stuff?

[Hutch]

Bring them here... I've got both the P5W DH and the DS3 setup on the bench

[GrimReaper]

doing soo well for promos are these sub 300$?

[GSKILL TECH]

it is a nice review and great photo taken

I can't wait to see CL5-5-5-15, 5-4-3-x, & 3-3-3-x tests

the price for 2GBHK is 239 @ newegg

http://www.newegg.com/Product/Produc...82E16820231114

[EnJoY]

3-3-3-5-3 Results posted in the first post above the 4-4-3 results. I could get 3-3-3 working at 2.35v, but I couldn't get any higher than 400MHz fsb, and more volts actually made it worse.

Next is 5-4-3 or 4-5-3 results.

[nn_step]

is that 3-3-3-5 T1?

[EnJoY]

Originally Posted by nn_step

is that 3-3-3-5 T1?

This is 965p, so it's always 2T.

[nn_step]

Originally Posted by EnJoY

This is 965p, so it's always 2T.

run everest latency benchmark

[rvd]

nice pics

[Stapler]

Enjoy, how much time do you spend tweaking sub times when you review ram? Do you mostly leave them be and just adjust the main timings or do you play with them to get every last mhz out of the kit?

[EnJoY]

Originally Posted by Stapler

Enjoy, how much time do you spend tweaking sub times when you review ram? Do you mostly leave them be and just adjust the main timings or do you play with them to get every last mhz out of the kit?

I played with the sub timings a lot with these, and I actually ran them at different timings throughout the review depending on what the other timings were or what speed I was shooting for. For the 3-3-3 testing I tightened the sub timings up a bit, but I played with them a lot to see if they made any difference in the low MHz clocking. They really only come into play regarding stability when you get above 1000MHz, before that they're just there to tweak performance.

[kakaroto]

Very nice and clean review you made! Keep the good work going

kakaroto,

[GrimReaper]

just 1 more thing

how do you get selected for reviewing the memory? I too want to review some memory