Meet the new Spartans..err Ballistix!

[Alex-Ro]

Hello ladies and gentleman,as the star of our show we will introduce micron's latest addition to the D9-DDR3 flock,D9PSG&D9PSH.Why are they together?Because there are mixed chips on a lot of ballistix line.Same modules use D9QMT and are in the same park,so basically u can't be certain of what you get only by removing heatspreader ...

Well here is the guilty product:

http://www.crucial.com/store/partspe...4G3D1608ET3LX0





General characteristics:

-they react to cold,20 to 0 degrees yielded about 100 mhz more at same latency
-they run very tight secondary and tertiary,basically as tight as your imc can handle,only exception is trfc which has to be in the 120-140 range for high frequency
-they react to voltage.Now comes the tricky part,as they don't react in a linear way.Basically 2400 C9 is stable for memtest with 1.64,but over 1.70 NO POST.2400 C8 will post and be stable at 1.77V.2400 C7 will post and somewhat stable for spi at 1.89
-they will not run any twcl lower than 7,but they don't have that twcl 8 bug,so twcl8 is lightly slower than twcl7
-they will run 32M about the same they can post,for example 2800C9 was 32M stable but 2860 did not post no matter what
-they are not binned,basically looks like crucial just threw random chips with no binning,one module did not boot over 2000 no matter what,other 3 did that but only 2 can run 2800 stable
-they do not have any trcd wall like good ole PSC or BBSE,2000 7-7-7,2200 8-8-8,2400 9-9-10,2600 8-9-11,2700 9-10-12,2800 9-10-13 are achievable for 32M


Results:

2800 9-9-12 S-pi 1M:





2800 9-10-12 S-pi 4M:



2800 9-10-13 S-pi 32M




Testing was done using a 3570K with a powerfull imc(3160+ valid on air),Asus Maximus V gene and 0-1 degrees celsius ambient.However the bad fact is they are somewhat slower at 32M that comparative samsung 2800 9-12-12 eventhough they blow them out of the water at AIDA tests,i suspect it's either likely mobo is not tuned specially for this kind of IC's.

Will post some more results at lower frequency coupled with some memtesting for daily users later.

So,what are you waiting for?Start binning and pour some ln2 on them

[Shiranui Gen-An]

Maybe if I get rid of all the crappy Hynix I have (and weak D9PFJ)...

[Ivanlabrie]

Very nice!

I want a few sets now...xD

[TaPaKaH]

I'd be more interested to see someone try the normal sized Crucial sticks with same ICs ... as those easier to mount a pot on

[Alex-Ro]

I had four kits of normal sized ballistix elite,3 were with D9PFT K revision and were and 1 with D9PFJ.Now from what i;ve seen in micron datasheets there is a new revision J following the K,with standard DDR3-1600 C9,that should be interesting to test as well.If some company as Gskill would binn these and release some low-latency series instead of ty 3000 mhz with huge latency and single-sided then we should have competitive 4GB benching modules alternative to samsungs....

[NapalmV5]

hmmm that beats paying 5xx$.. enticing! merci Alex!

discontinued http://www.newegg.com/Product/Produc...82E16820148656

i guess only thru crucial..

[XBrytanX]

Great review ,Thanks

[Reefa_Madness]

Great review ,Thanks

So??? Are you going to go out and buy you some more Ballistix now?

You swore off of them, but maybe you've had a change of heart?

[xdan]

The new Micron D9 chips have an amazing potential.
To bad that i don't know that the hell management they have.
Through Crucial they not bin the modules, they use different chips on same modules on even on same partnumber, the don't work with mb's manufacturers to improve support, they don't make publicity to this memory....

Man with some highly binned modules they could crash Hnyix and give hard time to Samsung.

These LP modules where at 60$ on newegg, they blow even the worst modules, all kit up to 120-150$.
Not to say if you find a kit like what Alex-Ro.



Micron and Crucial they are just stupid as marketing with this kits.

[websmile]

Micron lost tons of money with oc memory in the past, this is reason why they focused on server market and ssds, which brought them back into profit zone again- I am sure on ddr3 they won?t invest more money in binning and high speced ocs lines, the oc potential does simply not interest them, if other modul houses want to bin these and use these can fight with warranty claims of killed ln2 sticks - it has its reasons why they do not even offer 2133 sticks anymore through Crucial, and it is not because they are stupid...

[XBrytanX]

So??? Are you going to go out and buy you some more Ballistix now?

You swore off of them, but maybe you've had a change of heart?

I wouldn't go as far as saying this are better then BBSE,PSC,etc since i can't see the evidence but this are definitely worth playing with ,it's still a "crapshoot" with this RAM ,will probably end up getting set of this

[xdan]

I might not have D9QMT on second kit or D9QMT works so....Will remove the heatspreader tomorow... 1.775v....seems to be stable 32M

[hiwa]

someone is binning them

[gungstar]

2800 9-10-13 S-pi 32M

voltage?

[Alex-Ro]

1.87

[Leeghoofd]

Nice one Alex

[xdan]

So i have removed the heatspread of one module from second dual channel kit and it's still D9QMT...
But still works as D9PSG&D9PSH or viceversa. Alex-Ro i think he bought single modules, i bought dual channel kits and all 2 had/have D9QMT.


2400 CL 8-8-10-24-1T 1.81v(raised from windows) . 1.775-1.8v were stable only 11-12 loops.
1.8v> might or might not boot from bios at 2400CL8. 1.77-1.79 boot just fine.
tWCL 8,- 7 boots.


2200 CL 7-7-10-24 1.88v in working.... The tRP seems to be a big wall.


32M takes forever on AMD to test, i can wait for Haswell to test this memory.

This memory is a mistery, there are so many things to try.

[Alex-Ro]

your kit is weaker could not pass 2712 and still not stable,i bought first one kit,and second order another three kits,2 modules can't boot over 2000 lol,2 are little weaker than this.binning can be done at 2400 by trcd and tcl,best modules do 9-10-10 memtest at 1.64,8-10-10 1.76 they get very hot at higher volts,need serious cooling

[xdan]

8-10-10 1.76 they get very hot at higher volts,need serious cooling

Yes they are, at 1.6-1.65v Memtest they are already hot.
Not sure if it works better or not on AMD but my kit has validation at 2850.
When i did it i did not know about 2800 9-10-13 and high voltage so 13-13-13-36 1.75V:
http://valid.canardpc.com/show_oc.php?id=2746783
If you say you're modules don't boot above 2800>

D9PSG&D9PSH and D9QMT anyway seems to have pretty much same comportament, same performance with +/- with bining .

[Alex-Ro]

Jeez,your kit was in my hands,did you think i did not test it before i gave it back to you ?

AMD memory controller is looser on memory,not so aggresive as INTEL,you can simply check the aida number and see how much performance these 2 memory controllers extract from memory...

It is not about the duration of the test,it's how aggresive it is,you will feel much more exhausted if you **** a whore very hard for 30 minutes rather than make pasionate love with your boring wife for 1 h...(i'm a retard i know)....

Micron chips are very nice to play with,especially for spi 32M,if it passes 3 loops of 32M 90% of the time they will finish it,it;s not like other IC's where you get a random crash at 10,20 or 24 and you want to throw the hammer on them.

[xdan]

Jeez,your kit was in my hands,did you think i did not test it before i gave it back to you ?

AMD memory controller is looser on memory,not so aggresive as INTEL,you can simply check the aida number and see how much performance these 2 memory controllers extract from memory...

Yes it was but you didn't told me all you tested or succeded with it.
I know, i know...still Gene V, your motheboard it's build to work with memory at 2800-3000+, not like my Sabertooth rev.1 on AMD which does not work dual channel above 2500 and it has multipliers at max 2400.

Micron chips are very nice to play with,especially for spi 32M,if it passes 3 loops of 32M 90% of the time they will finish it,it;s not like other IC's where you get a random crash at 10,20 or 24 and you want to throw the hammer on them.

This not really happens on AMD, may be due to not so good MB on memory....old.
Usually 7-8 loops stable if it starts. Than with more voltage 11-12 and than with more voltage complete.
But still very easy memory to work, clear.

It is not about the duration of the test,it's how aggresive it is,you will feel much more exhausted if you **** a whore very hard for 30 minutes rather than make pasionate love with your boring wife for 1 h...(i'm a retard i know)....

Man now i have AMD, on Haswell i will put them in flames of the hell....as for you practical exemple...i will not chose a boring wife....still with a whore one you will think more than 30 minutes a day that she cheats you.

[Alex-Ro]

This memory crashes and burns everything

2600 9-10-12-27-1T Ultra tight tertiary and secondary,1.73V real voltage===>memtest stable...

[Ivanlabrie]

Dang that's good stuff!!!

What about 32m at those speeds with 5ghz cpu?

[Crabby]

Great clocks Alex and thanks for sharing, love your "practical example" This ram looks awesome on the OC Formula btw.

[l0ud_sil3nc3]

how are the 8GB dimms doing?

Also what about this kit?

http://www.amazon.com/Crucial-Ballis...cial+ballistix