G.Skill RipjawsX F3-17000CL9D-4GBXMD on AMD/990FX

[Woomack]

I think that some users noticed that I'm making memory tests from time to time. Today's short test is about G.Skill F3-17000CL9D-4GBXMD, memory designed for Intel Sandy Bridge about a year ago.
4GB kits aren't so popular lately because of really low memory price and 8-16GB in home pc isn't strange anymore. 4GB however is still good enough for almost everything, especially benching .

Link to producer's site:
http://gskill.com/products.php?index=367

and photo also from the same site ( maybe will add some my own photos but I think that all know how RipjawsX look like )

G.Skill guarantee 2133 9-10-9-28 2N 1.50V on Intel P67/Z68 boards using 2 sticks.
What is interesting, auto settings or basic SPD profile isn't like in most memory 1333 9-9-9 but 1600 9-9-9 ( at least my boards and software like cpu-z see it like that ).
For all who are interested what is inside then surprise, PSC IC like in most 2133 2x2GB kits designed for SB

Testing specification is available for most users on air/water cooling:
AMD FX-8120@4GHz, CPU-NB 2400 ( so can compare most clocks )
ASUS Crosshair V Formula / UEFI 1002
2x Crucial M4 64GB/RAID0
Gigabyte GTX470/stock clocks
ASUS Xonar DX
Corsair HX850/850W PSU

I will test most popular settings using 1.50, 1.60 and 1.65V and maybe later will add some more settings. Not many benchmarks but something for fast bandwidth compare and something for stability ( if anyone wish other tests then please pm me ).
Stability testing:
1600 6-8-6-20 1T 1.50V
1866 8-9-8-24 1T 1.50V
2133 9-10-9-24 1T 1.50V
1600 6-8-6-20 1T 1.60V <- passing this one as results are exactly the same as 1.50V and I wasn't able to drop timings and keep stability at the same time
1866 7-9-7-24 1T 1.60V
2133 7-10-7-24 1T 1.60V
1600 6-8-6-20 1T 1.65V <- exactly the same as 1.50V or 1.60V, I wasnt able to make stable 6-7-5 or 6-7-6 while both settings are working for benchmarks or games
1866 6-9-6-24 1T 1.65V
2133 7-10-7-24 1T 1.65 <- no difference comparing to 1.60V

Other tests:
2300 7-10-7-24 1N 1.65V
2400 7-11-7-28 1N 1.725V
2516 9-11-9-28 1N 1.60V
2817 10-13-12-28 1N 1.75V
3046 11-16-14-32 2N 1.85V

 

[Woomack]

1.50V testing

1600 6-8-6-20 1N ( tRFC 90ns ) 1.50V




1866 8-9-8-24 1N ( tRFC 90ns ) 1.50V




2133 9-10-9-24 1N (tRFC 90ns ) 1.50V

 

[Woomack]

1.60V testing

1600 6-8-6-20 1N ( tRFC 90ns ) 1.60V the same results as 1.50V so no point to upload, if I will be able to make something better then I will update it in the future


1866 7-9-7-24 1N ( tRFC 90ns ) 1.60V




2133 7-10-7-24 1N ( tRFC 90ns ) 1.60V

 

[Woomack]

1.65V testing

1600 6-8-6-20 1N (tRFC 90ns ) 1.65V

1866 6-9-6-24 1N ( tRFC 90ns ) 1.65V




2133 7-10-7-24 1N ( tRFC 90ns ) 1.65V

 

[Woomack]

other tests / max clock

Here are results that are not stable, playable, benchable or just screenable for max clocks.
Some results seem stable but couldn't pass IBT or Prime95.

2300 7-10-7-24 1N 1.65V




2400 7-11-7-28 1N 1.725V




2516 9-11-9-28 1N 1.60V , no crash or errors while playing games or benching but can't pass Prime95 or IBT




2817 10-13-12-28 1N 1.75V, cpu on SS , memory on air, benchable but games will crash




3046 11-16-14-32 2N 1.85V, max clock , memory still on air, barely screenable

http://valid.canardpc.com/show_oc.php?id=2227530
http://hwbot.org/submission/2248826_woomack_memory_clock_ddr3_sdram_1523.1_mhz?recalculate=true

 

[Woomack]

Some updates:
1866 6-9-6-24 1.65V
2300 7-10-7-24 1.65V
2400 7-11-7-28 1.72V

I noticed that anything above ~2300-2400 is crashing in IBT or Prime95. The same with low timings+higher clock combo ( like 2250+ CL7 ). Even that memory is stable, IMC just can't make more without single errors. It's possible to make it stable but not many users will run their PCs on sub 0 cooling 24/7
I was playing games for some hours on 2400 9-11-9-24 1N 1.58V without problems and all last night was running F@H on the same settings so I can call it just right for daily usage

 

[hokiealumnus]

Very nice IMC and sticks! Good work Woomack. Not too many people will test their memory so extensively. Two thumbs up!

 

[ivanlabrie]

Really interesting read!
Nice post, you got some great results so far; I'm sharing this with my friend Zes, who's getting a similar rig but with cl7 1600mhz gskills...

 

[Woomack]

Very nice IMC and sticks! Good work Woomack. Not too many people will test their memory so extensively. Two thumbs up!

Thanks hokie
I think you were right with that IMC limits. Even that I can make ~3000MHz for short tests it doesn't mean that I can make 100% stable ~2400MHz ( like longer Prime95 etc ). Probably new intel IMC will make more.
I got this G.Skill kit mainly for benching and I think it's not that bad

Really interesting read!
Nice post, you got some great results so far; I'm sharing this with my friend Zes, who's getting a similar rig but with cl7 1600mhz gskills...

Thanks , I'm trying to make some more tests each time when I get new memory kit ( other tests in sig, with my luck most are on PSC ).
BD cpus are maybe not fastest but are great for memory tweaking. At least better than SB where max is usually near ~2250MHz so you never know how high your kit can go ( at least for 2000+ kits ).

 

[Janus67]

Nice read, pretty interesting results there. Only issue I found (for comparison's sake) was that the CPU speed on the last set [high mem frequency] was quite a bit different from the others, but assuming it is like intel you had to increase bclk to get it at those speeds thus increasing your CPU speed at the same time (unless you dropped multi)

 

[Woomack]

All results except these in "other tests / max clock" post are on ~4013MHz cpu and ~2400MHz CPU-NB and are stable enough to pass IBT on maximum load / all memory settings ( I noticed that's about the same as Prime95 ).
Everything that wasn't able to pass IBT is in "other tests / max clock" post. In this post are also results on different cooling than air or water ( so like 2800+ with IMC on SS ) and higher or lower cpu/cpu-nb clock so hard to compare them.
Thanks for reading

 

[BugFreak]

Awesome tests.. and on an AMD. From an AMD person, Thank you! How were you testing for stability, tests, length of time, etc?

 

[Woomack]

Screenshots include ~35-40 minutes of IBT on full load + spi32m ~20 minutes. Except that I was running some other tests that aren't on screenshots like 8x memtest ( that is almost the same for memory as 10x IBT on max settings ).
So in total at least ~1.5-2h for each setting ( results with IBT on screenshot or marked as the same result as some other settings ) - 100% cpu + ~95% memory load. In daily usage it's almost impossible to achieve so high load 24/7.

 

[caddi daddi]

woomack, in post #5, what was the cpu/nb voltage to attain a 3052 cpu/nb?

 

[Woomack]

I don't remember. It was year ago and I don't have any settings from that time except some screenshots where is usually no info about voltages.