Up for sale is 10 sticks of the infamous Avexir PSC that was made for Splave.  The IC's were hand binned and then re-balled on to the KO-8155 PCB, and the XMP is 2200 8-12-8 1.65v.  The memory is a little bit picky if you want to get 2666 8-12-8 32M stable, but all of them can easily pass 32m using the 2600 8-12-8 "Tight" profile in the Z97 OCF.

GTIJason Spent some time playing with this memory and getting 2666 to pass, here is an FB post he made which might help you: https://www.facebook.com/photo.php?fbid=10154622123748579&set=pcb.10154622126778579&type=3&theater

I tested each stick in the outside black slot (farthest possible from CPU) and each stick had a heatsink + fan on it.  I should also note that the ambient air was VERY warm, 82F (27.7c) which has a negative impact on passing volts and overclocking in general.  With a really good IMC and better ambient temps, I expect the memory will perform better for you.  Also, if you spend some time tweaking the profile you can maybe achieve better results.  I chose to use the basic OCF profile just to make my results as re-producible as possible.

The price is $25 USD per stick + shipping.  Each stick is numbered in the top left corner, please tell me which one(s) you want.  They are sold naked, with no heatsink or anything else.

SOLD:  #11, #1, #10, #7, #3, #8, #6, #5

 

Edited September 7, 20187 yr by mllrkllr88

What is the importance of that specific pcb? 

2 hours ago, AutisticChris said:

What is the importance of that specific pcb? 

Perhaps others can answer this more succinctly but from my understanding its an 8-layer PCB which provides better ground shielding for the sensitive signal bus layers.  Possibly there are more power/ground/signal layers than comparable PCB's, but I cannot attest to that.  In theory this leads to higher overclocks :)

@Splave is looking around his garage wondering where all his PSC sticks have gone ?

49 minutes ago, mllrkllr88 said:

Perhaps others can answer this more succinctly but from my understanding its an 8-layer PCB which provides better ground shielding for the sensitive signal bus layers.  Possibly there are more power/ground/signal layers than comparable PCB's, but I cannot attest to that.  In theory this leads to higher overclocks

I didn't know this specific PCB was good.

What can it do sub-ambient

I've owned a pair of these since 2015, was very excited when Allen sent me a kit. Took me a while to figure them out and they are still a bit of mystery to me because to my knowledge they were hand picked and binned IC's

http://www.xtremesystems.org/forums/showthread.php?286435-I-bin-PSC-per-IC-at-Home-!!

I believe all of the sticks should be able to do 2666c8-12-8 tRFC=88 tCWL=6 RTL's 40-41-4-4
, Also know that Miller's test were quick n dirty since he didn't contact me to get the goods on em. They should all be able to do  the 2600c8 he tested but with tCWL=6 at 1.850 tops down to 1.78v or so for the better ones. The timing I found to be the pickiest is tRDRD, the rest seem solid. 
I have taken them cold a couple times, first time I was still pretty "green" but managed 2750c7 at low volts (2.04v if I recall correctly). Second time I was running 2866c7 @ 2.12v. Both quick sessions and was running MaxxMem, GB3 and some stupid bench by Intel.
Own a piece of OC History that are worth using as well, they are in great condition and have the PCB  you want for ddr3

Edited September 4, 20187 yr by GtiJason

Guess I'll be the first to bite. 

mllr, you've got PM.

KO-8155 pcb is double copper and expensive #1 for psc and bbse

9 minutes ago, Splave said:

KO-8155 pcb is double copper and expensive #1 for psc and bbse

Hi,
What's the difference with KO-8117?
 

3 hours ago, bolc said:

Hi,
What's the difference with KO-8117?
 

Well since it's old news 8155 usually runs with slightly lower voltage than 8117. Both 8 layer but 8155 is of higher quality and has additional copper
8155 is usually better/more stable at the limits, high freq/high volts c6 and -165c++. Your chances of finding a crazy 2900c6-9-6 64 35-36-4-4 are greatest with 8155 but still Rare AF
As of right now, out the very best 4 kits I've found 2 are 8155 and the other 2 are 8117. Tho somehow I have another dozen sticks of good bin psc waiting for a rainy day

Edited September 5, 20187 yr by GtiJason

thanks for the info.

my best stick is of ST pcb and 2nd best is 8155, 2666 8-12-8-28-1t super tight, 1.87 and 1.88 V, spi32m, on air.
but on LN2, things may change indeed

Ok so these are supposed to be like insanely good or something

PCB KO-8155 seems to be the preferred on LN2, yep. then KO-8117. finally ST pcbs
on air, it does not matter I think. some chips on ko 8117 and 8155 sucks much more than chips on St pcbs..

That sounds correct. I got a lot of 6 ST pcb sticks off ebay once and none of them could do 2600 tightish (CWL 7 and tighter Tertiarys with random tight secondaries) yet my very best stick that does  2750 mhz at same timings is ST pcb.

Pm sent.

On 9/5/2018 at 3:41 AM, mllrkllr88 said:

I tested each stick in the outside black slot (farthest possible from CPU)

Why pick this slot? I'm just asking, I have no idea about memory binning lol

47 minutes ago, unityofsaints said:

Why pick this slot? I'm just asking, I have no idea about memory binning lol

I'm only speculating here, but I would guess you only have interference coming from one side (the other slot) unlike the other slots which also have the traces towards the further out slots on both sides.

1 hour ago, ground1556 said:

I'm only speculating here, but I would guess you only have interference coming from one side (the other slot) unlike the other slots which also have the traces towards the further out slots on both sides.

I thought the conventional wisdom was closest slot = shortest traces = highest MHz but your explanation also makes sense, thx.

Black slots are what you use on Z97 OCF, so 2nd and 4th. 4th is usually what is used for Max Freq
On my Z97 OCF I prefer to bin psc using #2 the closest black slot to cpu as I feel it gives tighter timings but #4 almost as good.
As you guys can tell Miller has been super busy testing mems, benching TC, living life so I wouldn't put a lot of stock in this test
Up until TC he has been focused on newest ddr4 stuff and it's not easy revisiting something after an extended amount of time
Thats why I have a crazy amount of notes and screenies from my testing but you won't find any $30 sticks either if you come knockin' on my door lol

 

EDIT, oh and btw there are other good 8 layer pcb's,  B83U874 2.00 / (BP ML 94V-0 E186014) and 8 layer ST's .  .  . ST-SG38U816 V 1.0
Adata has a really good one too but the IC's are not 

Edited September 6, 20187 yr by GtiJason

6 hours ago, GtiJason said:

EDIT, oh and btw there are other good 8 layer pcb's,  B83U874 2.00 / (BP ML 94V-0 E186014) and 8 layer ST's .  .  . ST-SG38U816 V 1.0
 

Same ref for ST pcbs here
You can see it has less copper (it is meshed while plain in the KO-8117 8155) but on air I got really good results with PI, XH and Trident with those ST pcbs

All G.Skill above 1600 speedbin and normally all PI use 8 layer, 6 layer st you only find at low bins and series normally. Brainpower pcb is a pain in the a.. imho, but seems to depend on how you look at it xD. This was far from good for me on 2gb sticks. A-Data usually used HJ PCB for their higher psc bins like 2400c9 and sometimes 8155 or 8117, but also other 8 layer revisions of same manufacturer, but they were often black.

I reckon people prefer KO-8155 on G.Skill because that's the only type of PCB they used on 2400C8 and 2400C9 PI. So every time they buy 2200C7 with KO-8155 they must think that these are just underbinned 2400C8

On practice, there is hardly any difference between the three (8155, 8117, ST). On kits like 2200C7 PI that came in all three variants, week is a much better indicator of IC quality than PCB. Also, my best PSC stick ever was a 8117 so ...

that could be but surely the best ics on the best pcb would be even better no?

 

 

This is giving me flashbacks but I'm not sure if they're good ones or bad ones...

Almost all your psc was great :D.