Xtreme Addict’s Ultimate Tweaking Guide for B-Die Memory on ASUS Maximus VIII Boards

The HWBOT World No.2 Xtreme Addict has posted a very in depth and insightful guide centered on tweaking B-Die memory, specifically on ASUS ROG Maximus VII series motherboards. XA deals with all the issues involved with implementing the tightest timings possible, a massive edge in competitive benchmarking. Here’s a sample of what he has to say relating to the issue of motherboard binning, something that he regards as part of the ‘Ground Rules’:

“BIN MOTHERBOARD! I binned overall 10x M8E and with Uncle Mad around 15x M8I. Board matters, believe me even though IMC is inside CPU. IT WAS SAME WITH MANY VENDORS AND MANY GENERATIONS. JUST ESPECIALLY NOW IT'S MORE VISIBLE AND CLEAR THAT WE HAVE TO BIN BOARDS. It's not possible to get exactly same motherboard PCB, each PCB differs inside, copper layers which produce signals and so on. Those are very small and subtle differences, but influence OC potential a lot.”

“Note that mostly after LN2 benching, many motherboards degrade. Sadly but small sockets (BGA balls and traces under the socket) are very sensitive for full pot, especially fast temperature changes. That's why in Asus Shamino's/Elmor's Z170 guide we can find tips to slowly cool down CPU in order to preserve socket. Some motherboards are "RAMBO" edition, some sadly will degrade. For degraded boards you can try to clean socket + DIMMs with aceton and dry motherboard for a very long time.”

You can find the full guide from Xtreme Addict here on the HWBOT forum.


35

Poland Xtreme Addict says:

Time for a new guide. After some long tweaking and binning of B DIE memories I found out some tweaks in order to make B DIE fly on ROG boards :) It works for Maximus VIII Extreme, Maximus VIII Impact, Maximus VIII Gene (all tested by me), no idea if Hero/Ranger have same options in BIOS cause sadly I don't own those boards.

Influence on M8E/M8G:
- good boards will be able to boot and bench 2x8 GB 4000 CL12 1T TIGHT 2Ds like XTU and 3Ds, inside OS with BCLK raising we can even get more, though SPI32M CW will be very hard (I managed to do 4000 CL12 1T TIGHT only in single channel), and 3866 4x8 GB

Influence on M8I:
- good board will be able to do 4133+ SPI32M COPY WAZA

Ground rules:

BIN MOTHERBOARD! I binned overall 10x M8E and with Uncle Mad around 15x M8I. Board matters, believe me even though IMC is inside CPU. IT WAS SAME WITH MANY VENDORS AND MANY GENERATIONS. JUST ESPECIALLY NOW IT'S MORE VISIBLE AND CLEAR THAT WE HAVE TO BIN BOARDS. It's not possible to get exactly same motherboard PCB, each PCB differs inside, copper layers which produce signals and so on. Those are very small and subtle differences, but influence OC potential a lot. Note that mostly after LN2 benching, many motherboards degrade. Sadly but small sockets (BGA balls and traces under the socket) are very sensitive for full pot, especially fast temperature changes. That's why in Asus Shamino's/Elmor's Z170 guide we can find tips to slowly cool down CPU in order to preserve socket. Some motherboards are "RAMBO" edition, some sadly will degrade. For degraded boards you can try to clean socket + DIMMs with aceton and dry motherboard for a very long time.

How to determine if I have a good or a bad board?

1. Check both Channels (A & B). On M8I mostly Channel B (2nd slot) is weak/degrades, especially with 1T. On M8E/M8G vice versa, mostly Channel A degrades.

2. If both Channels boot same frequency (e.g. 4000 CL12) than try dual channel. If your board mostly produces 41 code on settings which should work easy, worked on other ROG board, you have a bad sample and you have to bin more boards :) If you are getting 41 codes too often, setting signals manually on a "weak" board might make it work like average/good board. If board stucks at 55 code post especially in Dual, it's not issue of motherboard, we just need to have a) better memories, b) tweak settings :) c) retry to boot, please remember RETRY button is the best friend of OCer^^ especially with very tight settings and high frequency sometimes it's worth to try retry button 3-5 times, we should try booting few times when we already had same booting in the past, finally it should work. Please note that on cold, when slots are getting frozen or socket area, our booting frequency will go down a lot when platform is tired or memories too cold. B DIE like to be warm and dry.

Retry button/safe mode @ M8I

For safe mode - just use Power on/off button (red one under code poster), press it and keep until boards shutdowns and power it on again to boot in safe mode. Retry button is not on the board, so the best option is just to turn off PSU and turn it on and power on the board again.

CPU IMC

To be honest most people say they have a limit of IMC, I think it's more board/memories or wrong settings. I tried 4133 12-11-11 CW with 5x CPUs (6500+ HWBOT PRIME from my own binning) and all could do that with 1.25-1.30v (depends from BIOS) VCCSA/VCCIO. I mean of course I am aware that potato CPUs might not do that.



BIOSES:

For M8I I don't see big difference between 0014-0019, though 0019 seems to be slightly more stable in booting, also we don't need LN2 MODE Enabled in order to set more than 2.0 VMEM. Retail newest BIOS also works really well.

M8E: for me 0015 was the best one, but newest retails are not bad at all either


MAGIC SETTINGS/BOOTING SETTINGS



RTLs/IOLs




[img]https://abload.de/img/bios2jnpa0.jpg[/img]





ROG boards don't have issues with tightening settings by hand, we don't have to perform special rituals or dances. Especially on M8I we can put everything by hand and it will work (sometimes we have to retry 1-2 times). M8E/M8G can't run that tight RTLs as M8I and it's harder to lock settings by hand, but it also works, but in fact for 2D/3D which are not XTU or SPI32M, I would just leave AUTO settings. It won't effect our efficiency in other programs, even if board boots 50/52 instead of 50/51.

SIGNALS - "SLOPES"







The most important are those 5:

Clk Rising Slope
Clk Rising Slope Offset
Data Falling Slope
Data Falling Slope Offset
Cmd Falling Slope

You can try different values of course. Each board is different and needs special attention. Personally I always change all "Slopes" at once, the most universal settings for M8I is "15-1" combination as you can see in the screenshots. For M8E/M8G mostly the best for me is "7-1". But as I said - try by yourself. I am just giving the most universal settings which worked best for me and the biggest amount of boards.

ODT settings




This worked the best for me on M8I board (1 sample only).
You can try and also finetune by yourself. On most ROG boards I leave that setting on AUTO, but some boards are much more stable with manual touch. On ASUS playing with ODT values doesn't influence efficiency.

EXTRA STABILITY SETTINGS

DRAM REF VOLTAGE CONTROL




On some B DIE this section can help booting in Dual, also improve stability and remove "79 SLOW BOOT BUG" with B DIE, when memories are on the edge (especially with 12-11-11 setting). Try 0.50000 - 0.52000 with 35-40 settings. Note that you have to change all lower settings same down to the end (really a lot of positions) cause with different values or half by hand half by AUTO board will not boot.

MRC FAST BOOT DISABLED

[img]https://abload.de/img/bios9rsp1i.jpg[/img]

For me Disabled is mostly better. Also keep in mind when setting TWR timing to set manually TWRPRE in order to avoid 41 code.

B DIE VOLTAGE-VTT SETTINGS

a) most B DIE (especially in Dual Channel) prefer lower VTT, 0.8-0.83v in order to avoid 55 code post
b) 12-11-11 mostly needs higher VTT than 12-12-12
c) with proper VTT setting we can avoid 79 code post "SLOW BOOT BUG" - stability issue


RESULTS


My good M8I (sadly the best one died, my new one is really strong too, but I can't do 4133 49/50 RTLs as on my old board. I present bunch of results with my 3x strongest B DIE memory kits (Galax ES, Retails & G.skill Retails).

1. OLD M8I 49/50 RTLs 4133 12-11-11 on Galax ES memories

[img]https://abload.de/img/spi32mb2au90.jpg[/img]


2. Galax HOF OC LAB Special Edition 001/100 4133 12-11-11

[img]https://abload.de/img/snaphsot0001kbu00.jpg[/img]

3. Trying to run 4220 12-11-11 CW on my Galax retails :)

[img]https://abload.de/img/snaphsot0003omuqj.jpg[/img]

4. SPI1M 4220 12-11-11

[img]https://abload.de/img/snaphsot0013fruv7.png[/img]

5. SPI 32M 4220 12-12-12 CW easy as piece of cake

[img]https://abload.de/img/snaphsot00177mucn.jpg[/img]

6. SPI 1M 4310 12-12-12 booting is easy and running SPI1M

[img]https://abload.de/img/snaphsot0019f7up8.png[/img]

7. G.skill retails 3600C16 4133 12-11-11

[img]https://abload.de/img/snaphsot0005uwuai.jpg[/img]

United States coolhandluke41 says:

memory settings extraordinary,this board is a rocket with the right memory ..well done XA

websmile says:

Awesome test work, good to see a helpful hand for a lot of people who wrestle these ics :)

Germany Dancop says:

That's a great info man... Like usual... Great guide! Gesendet von meinem HUAWEI NXT-L29 mit Tapatalk

Indonesia Hazzan says:

Nice Guide XA...I am agree with " Ground rules " you have

United States sabishiihito says:

Great stuff. Unfortunately with M8I, most of us have no access to a Retry button :( Sent from my iPad using Tapatalk

Poland Xtreme Addict says:

sabishiihito said: Great stuff. Unfortunately with M8I, most of us have no access to a Retry button :(


Sent from my iPad using Tapatalk


Turn on/off PSU. For safe mode - press power on/off button for 4 seconds.

Indonesia speed.fastest says:

Great Guide! On Gene try BIOS 0018, it worked better with B-Die when do 3800+, RTL are tighter than other BIOS and easier to boot at high frequency, even if you lock all settings. And retry button is your best friend :D

United States sabishiihito says:

speed.fastest said: Great Guide! On Gene try BIOS 0018, it worked better with B-Die when do 3800+, RTL are tighter than other BIOS and easier to boot at high frequency, even if you lock all settings. And retry button is your best friend :D
Will try my Gene again since I can connect OC Panel. I was convinced 4-DIMM boards couldn't do 4kC12 1T but I'm glad I was wrong.

Canada marc0053 says:

XA you are awesome man!! I've been struggling with mem OC on the m8e for so long and can't wait to give these settings a go. Not sure if it's the board or not but often I need to keep frequency at 3733mhz or lower if not it was constant 55 or 41. on the regular OC formula big board I could run the same mem at 3866mhz+ and tighter timings. Thanks for making a clear and concise guide! Marc

Indonesia speed.fastest says:

I can do 4K C12 2T Tight, 1T still not worked, will try XA Guide ;)

Sweden elmor says:

sabishiihito said: Great stuff. Unfortunately with M8I, most of us have no access to a Retry button :(


Sent from my iPad using Tapatalk


You can try this if you're not afraid of a bit of soldering.

United States GtiJason says:

Thanks for taking the time to write this, Asus mem settings in bios can be confusing, I was reminded last night testing a M7I cold for first time. Now I'm wishing I hadn't sold the M8I so I can test your guide out.

Asus Maximus Impact - 3 generations - Album on Imgur


Argentina nacho_arroyo says:

sabishiihito said: Great stuff. Unfortunately with M8I, most of us have no access to a Retry button :(


Sent from my iPad using Tapatalk


+1 And great job XA.

says:

Hazzan said: Nice Guide XA...I am agree with " Ground rules " you have
I agree too.i dont know why but board matter on imc overclock Sent from my UMI_SUPER using Tapatalk

Indonesia bboyjezz says:

Bunch of test...thanks a lot XA, will try ur settings tonight:) Sent from my SM-G930S using Tapatalk

Italy RULE says:

Nice job. Thanks for sharing. Inviato dal mio SM-A300FU utilizzando Tapatalk

Netherlands 5erveD says:

This is beyond awesome! Still dont understand half of it ;-) Verstuurd vanaf mijn MotoG3 met Tapatalk

Italy superpatodonaldo says:

Awesome!

Sweden Rauf says:

Very good guide!

Noxinite says:

Interesting, thanks!

Italy rsannino says:

Thank you for sharing your experiences!

Czech Republic Elkim says:

Thank you XA, I see alot of data to learn new things! :)

United Kingdom gavbon says:

Will be giving this a go with my B-die mems on my M8I! Thanks XA!

Serbia d0minat0r says:

Well done XA! Ultimate work and experience!

United States sabishiihito says:

Plugging away, can't get it stable though.[ATTACH]4713[/ATTACH]

Indonesia speed.fastest says:

sabishiihito said: Plugging away, can't get it stable though.[ATTACH]4713[/ATTACH]


Under cold maybe it will stable if your memory can handle that settings :D
Nice one!

United States sabishiihito says:

Only have AIO+air here, unfortunately. Sent from my SM-G935U using Tapatalk

websmile says:

Better options to test than me, no Asus board anymore - one killed, three sold^^

Sweden Rauf says:

board binning is very true indeed. First board managed 4000 12-12-12 with wazza. second board exactly the same. So I started to wonder if my memory was bad (Galax HOF) and that maybe my good g.skill had gotten degraded. But then I tried a third board and it just works. HOF will probably do 4133 12-12-12 or 4080 12-11-11 after a little tweaking and some cold on CPU, they passed a couple of loops just from dialing in timings. Great memory!

Edit: this is not only true for Asus impact. I also tried a MOCF and had the same bad results...

United States l0ud_sil3nc3 says:

Rauf said: board binning is very true indeed. First board managed 4000 12-12-12 with wazza. second board exactly the same. So I started to wonder if my memory was bad (Galax HOF) and that maybe my good g.skill had gotten degraded. But then I tried a third board and it just works. HOF will probably do 4133 12-12-12 or 4080 12-11-11 after a little tweaking and some cold on CPU, they passed a couple of loops just from dialing in timings. Great memory!

Edit: this is not only true for Asus impact. I also tried a MOCF and had the same bad results...



Every MOCF here with the same cpu (4133 c12 IMC) and same kit of ram performs the same. I have five MOCF some early ES but most retail. I only have one M8I so can't test on Asus but at least in my case binning boards has not been required for maximizing memory clocks.

Australia Bullant says:

^^Same,I've tested 3 ocfm and have a new forth here not tested and all work as good as each other on memory clocking For people testing imc,when on water check rtl strength. 50-51-7-7 or 50-50-7-6 or 49-50-6-6 @ say 4000 cl12,this is on ocfm

Australia Bullant says:

Nice guide XA :)

websmile says:

Boards can be quite different, no idea about mocf, only had one sample, I had two M8Hero, one did 3866 divider up to nearly 4k c12 stable 32m 1t, second had big problems even doing 3733 c12 stable, same mems and cpus. On air, rtls on MOCF are quite hard for me (4kc12 the mentioned 51/50 7/6 or 7/7 at best), apart from this the board worked like a charme up to 4133 c12 32m with all other timings set tight (of course with capable imc), additional sa and io voltage should help but cooling limits this for me, 1,45v+ might be no good idea xd

Germany Digg_de says:

Why the Boards are degrading? Is it the high SA/IO Voltage or maybe Vdimm?
I've bought me few Days ago a used Asus Impact Board. I hope it is not already degraded. ^^

I can start 2000 12-11 49/50/6/6 easy and Geekbench/Pi32m with Win8.1 is stable. Good sign?
[img]https://abload.de/thumb/snaphsot0001p3zan.png[/img] [img]https://abload.de/thumb/snaphsot000273ro1.png[/img]

Need to install XP32 for Pi32m. :)

edit:
XP32 installed, was a bit more complicated than a XP-Install on OCF.. ;)

[img]https://abload.de/thumb/snaphsot0006bhuzv.png[/img]

Most times i have the "79 Slowmode Bug"... i've tried the "DRAM REF VOLTAGE CONTROL".. mostly no success.. or first Boot after Setting is ok, after next restart 79 again.
Need to spend more time for tweaking Memory.

Please log in or register to comment.