Gigabyte Z170X-SOC Force - DDR4

[xxbassplayerxx]

Board wasn't playing nice last night. Would boot in at 59/59/49/50 on auto but the second I tried to set those manually it wouldn't stuck at Post Code 35.

 

Just as a bit of info, during the week I get about 20 minutes a night to mess with this. That's not enough to make much progress. Hopefully this weekend has a spare moment for me.

[Leeghoofd]

This should get some dialed it in a bit quicker XMP off, 1.24V SA & IO, 1.74V vDIMM, 3733 12-12-12 320 1T

 

Edited March 6, 2016 by Leeghoofd

[Rampagexx79]

Nice

[xxbassplayerxx]

Big update incoming. I managed to get 738 on XTU with my 6320.

 

 

[hwbot=3151555]submission[/hwbot]

 

The BIOS Settings that made it happen are below. Note that you have to manually set the RTL's for Channel B in Channel B Memory Sub Timings. Everything else should be grayed out. As you can see, there are still tons of settings that are on Auto. I'm going to begin to explore these myself but if anyone has any knowledge that you're willing to share, please let me know!

 

By using the settings below, realize that you are voiding the warranty on many of your components and failure is a real possibility. Only use the settings below as a reference if you are aware of this.

 

Advanced Frequency Settings

CPU Base Clock				102.74
Host Clock Value			102.74

CPU Clock Ratio				39
CPU Frequency				4.01GHz
FCLK Frequency for Early Power On	[Normal (800MHz)]
Advanced CPU Core Settings

Extreme Memory Profile(X.M.P.)		[Disabled]
System Memory Multiplier		[37.33]
Memory Frequency(MHz)			3835MHz

 

 

Advanced Memory Settings

 

Extreme Memory Profile(X.M.P.)		[Disabled]
System Memory Multiplier		37.33
Memory Boot Mode			[Auto]
Memory Frequency			3835MHz

Memory Enhancement Settings		[Normal]
Memory Timing Mode			[Manual]
Profile DDR Voltage			1.20V
Memory Multiplier Tweaker		Auto
Channel Interleaving			[Auto]
Rank Interleaving			[Auto]

 

Channel A Memory Sub Timings

 

Memory Timing Mode			[Manual]
Memory Boot Mode			[Auto]

Channel A Standard Timing Control
CAS Latency				12
tRCD					12
tRP					12
tRAS					28

Channel A Advanced Timings Control
tRC					36
tWR					10
tCWL					9
tRRD_S					5
tRRD_L					5
tWTR_S					2
tWTR_L					10
tCCD_S					Auto
tCCD_L					Auto
tRFC					220
tRTP					6
tFAW					16
Commmand Rate(tCMD)			1
N/1 Ratio Mode				Auto

Channel A Misc Timing Control
tREFI					30000
tREFIx9					Auto
tXP					Auto
tXPDLL					Auto
tPRPDEN					Auto
tRDPDEN					Auto
tWRPDEN					Auto
tCPDED					Auto
tAONPD					Auto
tCKE					4
RFR Delay				Auto

tXSDLL					Auto
tXS Offset				Auto
tZQOper					Auto
tMOD					Auto

tRDRD_sg				6
tRDRD_dg				4
tRDRD_dr				5
tRDRD_dd				5
tRDWR_sg				12
tRDWR_dg				12
tRDWR_dr				12
tRDWR_dd				12
tWRWR_sg				6
tWRWR_dg				4
tWRWR_dr				4
tWRWR_dd				4
tWRRD_sg				25
tWRRD_dg				17
tWRRD_dr				5
tWRRD_dd				5

Round Trip Latency(DIMM0/Rank0)		59
Round Trip Latency(DIMM0/Rank1)		Auto
Round Trip Latency(DIMM1/Rank0)		49
Round Trip Latency(DIMM1/Rank1)		Auto
IoLatR0D0				4
IoLatR1D0				Auto
IoLatR0D1				6
IoLatR1D1				Auto
IoLatOffset				Auto
ODT Read Duration			Auto
ODT Read Delay				Auto
ODT Write Duration			Auto
ODT Write Delay				Auto
Read ODT(DIMM0/Rank0)			Auto
Read ODT(DIMM0/Rank1)			Auto
Read ODT(DIMM1/Rank0)			Auto
Read ODT(DIMM1/Rank1)			Auto
Write ODT(DIMM0/Rank0)			Auto
Write ODT(DIMM0/Rank1)			Auto
Write ODT(DIMM1/Rank0)			Auto
Write ODT(DIMM1/Rank1)			Auto
ODT Override(DIMM0/Rank0)		Auto
ODT Override(DIMM0/Rank1)		Auto
ODT Override(DIMM1/Rank0)		Auto
ODT Override(DIMM1/Rank1)		Auto
RttNom(Dimm0)				Auto
RttNom(Dimm1)				Auto
RttPark(Dimm0)				Auto
RttPark(Dimm1)				Auto
RttWr(Dimm0)				Auto
RttWr(Dimm1)				Auto
WR Additional(Dimm0/Rank0)		Auto
WR Additional(Dimm0/Rank1)		Auto
WR Additional(Dimm1/Rank0)		Auto
WR Additional(Dimm1/Rank1)		Auto
Decrement WR Delay			Auto
Additive Latency			Auto
Output Driver Impedance Control		Auto
CS to CMD/ADDR Latency			Auto
C/A Parity Latency			Auto

 

Channel B Memory Sub Timings

Memory Timing Mode			[Manual]
Memory Boot Mode			[Auto]

Channel B Standard Timing Control
CAS Latency				Auto
tRCD					Auto
tRP					Auto
tRAS					Auto

Channel B Advanced Timings Control
tRC					Auto
tWR					Auto
tCWL					Auto
tRRD_S					Auto
tRRD_L					Auto
tWTR_S					Auto
tWTR_L					Auto
tCCD_S					Auto
tCCD_L					Auto
tRFC					Auto
tRTP					Auto
tFAW					Auto
Commmand Rate(tCMD)			Auto
N/1 Ratio Mode				Auto

Channel B Misc Timing Control
tREFI					Auto
tREFIx9					Auto
tXP					Auto
tXPDLL					Auto
tPRPDEN					Auto
tRDPDEN					Auto
tWRPDEN					Auto
tCPDED					Auto
tAONPD					Auto
tCKE					Auto
RFR Delay				Auto

tXSDLL					Auto
tXS Offset				Auto
tZQOper					Auto
tMOD					Auto

tRDRD_sg				Auto
tRDRD_dg				Auto
tRDRD_dr				Auto
tRDRD_dd				Auto
tRDWR_sg				Auto
tRDWR_dg				Auto
tRDWR_dr				Auto
tRDWR_dd				Auto
tWRWR_sg				Auto
tWRWR_dg				Auto
tWRWR_dr				Auto
tWRWR_dd				Auto
tWRRD_sg				Auto
tWRRD_dg				Auto
tWRRD_dr				Auto
tWRRD_dd				Auto

Round Trip Latency(DIMM0/Rank0)		59
Round Trip Latency(DIMM0/Rank1)		Auto
Round Trip Latency(DIMM1/Rank0)		50
Round Trip Latency(DIMM1/Rank1)		Auto
IoLatR0D0				4
IoLatR1D0				Auto
IoLatR0D1				6
IoLatR1D1				Auto
IoLatOffset				Auto
ODT Read Duration			Auto
ODT Read Delay				Auto
ODT Write Duration			Auto
ODT Write Delay				Auto
Read ODT(DIMM0/Rank0)			Auto
Read ODT(DIMM0/Rank1)			Auto
Read ODT(DIMM1/Rank0)			Auto
Read ODT(DIMM1/Rank1)			Auto
Write ODT(DIMM0/Rank0)			Auto
Write ODT(DIMM0/Rank1)			Auto
Write ODT(DIMM1/Rank0)			Auto
Write ODT(DIMM1/Rank1)			Auto
ODT Override(DIMM0/Rank0)		Auto
ODT Override(DIMM0/Rank1)		Auto
ODT Override(DIMM1/Rank0)		Auto
ODT Override(DIMM1/Rank1)		Auto
RttNom(Dimm0)				Auto
RttNom(Dimm1)				Auto
RttPark(Dimm0)				Auto
RttPark(Dimm1)				Auto
RttWr(Dimm0)				Auto
RttWr(Dimm1)				Auto
WR Additional(Dimm0/Rank0)		Auto
WR Additional(Dimm0/Rank1)		Auto
WR Additional(Dimm1/Rank0)		Auto
WR Additional(Dimm1/Rank1)		Auto
Decrement WR Delay			Auto
Additive Latency			Auto
Output Driver Impedance Control		Auto
CS to CMD/ADDR Latency			Auto
C/A Parity Latency			Auto

 

CPU Core Voltage Control

CPU Vcore				1.200V
Dynamic Vcore(DVID)			Auto
GPU Graphics Voltage (VAXG)		Auto
Dynamic VAXG(DVID)			Auto
CPU VCCIO				1.300V
CPU System Agent Voltage		1.300V
VCC Substained				Auto
VCCPLL					Auto
VCCPLL OC				Auto
CPU Core PLL Overvoltage (+mV)		Auto

 

DRAM Voltage Control

DRAM Voltage       (CH A/B)		1.900V
DRAM Training Voltage     (CH A/B)	1.850V
DDRVPP Voltage     (CH A/B)		Auto
DRAM Termination   (CH A/B)		Auto

 

Chipset Voltage Control

PCH Core				1.200V

Edited March 8, 2016 by xxbassplayerxx

[GtiJason]

What's going on with RTL/IOL's 56 58 13 13. Shouldn't they be 49 50 6 6 or so? Everything else looks money though, not many guys pushing GIGA boards like this, good work boss !

[zeropluszero]

Never touched my RTLS, left them auto personally.

My boards dead so I cant test this. but set tFAW to 12 and tWRWR_dr and tWRWR_dd to 9.

Surprised you got tCWL to 9, I could only get mine to 10.

try looser timings from tRDWR_sg and lower.

[xxbassplayerxx]

What's going on with RTL/IOL's 56 58 13 13. Shouldn't they be 49 50 6 6 or so? Everything else looks money though, not many guys pushing GIGA boards like this, good work boss !

 

Must have been a bug when opening Timing Configurator... I had to retake the screen shot after closing everything because I accidentally covered The CPU tab in the first SS. I'll check the original SS when I get home to see what it shows. If it's not a bug, I wonder if changing tRC from auto to 36 did it?

[xxbassplayerxx]

What's going on with RTL/IOL's 56 58 13 13. Shouldn't they be 49 50 6 6 or so? Everything else looks money though, not many guys pushing GIGA boards like this, good work boss !

 

To follow up here, they seem to just get borked at some point, even when set manually. I have to load another profile and then re-load my main profile to get them to show right again. Once I get everything up and loaded, I'm going to turn on quick book to bypass any training on the RTL's.

[xxbassplayerxx]

Anyone have luck running four sticks on this board yet? Just threw in two more sticks of B-Die (four total)... trying to figure out what I need to change to get them to do the 3866 12-12-12 or so that I was doing with two sticks.

[Strong Island]

Anyone have luck running four sticks on this board yet? Just threw in two more sticks of B-Die (four total)... trying to figure out what I need to change to get them to do the 3866 12-12-12 or so that I was doing with two sticks.

 

for some other boards changing twrwr_dd & twrwr_dr to like 8 and above helps me a lot with 4 sticks.

[GtiJason]

Anyone have luck running four sticks on this board yet? Just threw in two more sticks of B-Die (four total)... trying to figure out what I need to change to get them to do the 3866 12-12-12 or so that I was doing with two sticks.

 

Good luck, I hope you can get close but I can hear the imc screaming already. Even if you have to lower frequency or raise timings 4 sticks of B Die will be very efficient

[xxbassplayerxx]

It's definitely some sort of setting. It wouldn't even boot 3200 with my settings for 3866, even with the tWRWR_dd and _dr set to 8.

 

EDIT: Tried RTL's 49/50 6/6... going to keep trying more in a minute.

Edited March 23, 2016 by xxbassplayerxx

[xxbassplayerxx]

Thanks to @george\.kokovinis, I now have this board doing 3866 12-12-12 with no issue at all! Ignore my setting above and try 1.18V for VCCIO instead. It's suddenly completely stable where it wasn't earlier!

Edited April 6, 2016 by xxbassplayerxx

[Guest george.kokovinis]

Thanks to @george.kokovinis, I now have this board doing 3866 12-12-12 with no issue at all! Ignore my setting above and try 1.18V for VCCIO instead. It's suddenly completely stable where it wasn't earlier!

 

Very glad that I helped.

You are very kind.

 

George:)

[xxbassplayerxx]

Just a note... B-Die is very temperamental when it comes to voltage. It's very possible that your sticks may fail at 1.90V and require less to do 3866 12-12-12... Just have to test! It's also possible that they would require more, though less seems to be more common at this point.

 

For reference, I have tested three sticks so far. The first wants ~1.88V, the second only passes sometimes but when it does, it's around 1.86V. The third just passed at 1.86V and I'm testing lower now.

Edited April 6, 2016 by xxbassplayerxx

[Guest george.kokovinis]

Although it might sound weird, IOL offset plays a significant role both in training and stability.

Default at auto is 21.

By setting manually both channels IOL offset to 23, you tighten up both RTL and IOL to 49/49 - 5/4.

 

B-die likes that.

With auto offset, there come times it might not train.

 

Give it a go

[Fact11]

Hey Bass, I am running the same board as you, with Trident Z 3600C16 b die and when you are running at 3866 12-12-12-28 is that at 1T or 2T ?

 

Because I'm having difficulty getting it stable at 1T. It will pass training and boot, but then it will either load windows and then freeze, or it will sometimes just get to the logo screen and then freeze.

 

I have tried it with Bios versions X08, F6i, F6k and now the latest one F6.

 

I set my Vccio to 1.185 and my SA to 1.2 and then set both my Dram and Dram Training Voltages to 1.88volts. It will pass but then its just not stable.

 

I can get it do 3733 12-12-12-28 IT, and its fine, but will not complete a Super PI 32M if I run waza first. If I just run 32m with no waza it will complete it.

 

Does yours complete 32M successfully ?

[Fact11]

Although it might sound weird, IOL offset plays a significant role both in training and stability.

Default at auto is 21.

By setting manually both channels IOL offset to 23, you tighten up both RTL and IOL to 49/49 - 5/4.

 

B-die likes that.

With auto offset, there come times it might not train.

 

Give it a go

 

Hey George, I'll give that a try and see what happens.

 

I also tried your recommendation of setting the PCH voltage to 1.2, and I know you did it on a different model board, but when I try it on the Z170 SOC Force, and save and exit the bios, it will boot up once successfully, but then if I go to reboot or I shutdown the pc and then turn it back, it just powers up for a few seconds and then turns off and then just repeats that cycle, and then the only way to stop it is turn off the power switch on the PSU and press the CLR Battery button to reset the bios and then its fine again.

[Guest george.kokovinis]

Hey George, I'll give that a try and see what happens.

 

I also tried your recommendation of setting the PCH voltage to 1.2, and I know you did it on a different model board, but when I try it on the Z170 SOC Force, and save and exit the bios, it will boot up once successfully, but then if I go to reboot or I shutdown the pc and then turn it back, it just powers up for a few seconds and then turns off and then just repeats that cycle, and then the only way to stop it is turn off the power switch on the PSU and press the CLR Battery button to reset the bios and then its fine again.

 

 

Each board has its likes and dislikes.

Each mem kit also has its likes and dislikes.

 

What works for me, may not work for others.

 

The OCF has a totally different behavior than Soc-force and Asus M8E.

Given that we are working with the same mem set and THE SAME CPU,

after setting the basics that are common on all boards, we have to start a painful procedure of trial and error, providing that we are aware what each input we provide to the board means.

 

VccSA of 1.2V that you mentioned is extremely low for anything beyond 3600, even at XMP.

 

Since I own and have worked a lot on all three boards I mentioned ( and more ), the starting point for VccSA at 3733 is 1.25.

Climbing the ladder to 3866 I use 1.30 and then at 4000 I use 1.35.

 

3866 at 1T is not easy.

First and foremost it depends on the mems themselves.

I have two sets cl17 and one set cl16.

Out of the three only one set does 3866 - 1T and 4000 - 1T on the OCF.

 

The final result depends on many parameters and every set up has its own peculiarities.

 

With the exception of SPI, where a few milliseconds will make the difference, I have not noticed serious variances in many other benchmarks,

that could very well be in the acceptable margin ( +/- 1% )

[Fact11]

Each board has its likes and dislikes.

Each mem kit also has its likes and dislikes.

 

What works for me, may not work for others.

 

The OCF has a totally different behavior than Soc-force and Asus M8E.

Given that we are working with the same mem set and THE SAME CPU,

after setting the basics that are common on all boards, we have to start a painful procedure of trial and error, providing that we are aware what each input we provide to the board means.

 

VccSA of 1.2V that you mentioned is extremely low for anything beyond 3600, even at XMP.

 

Since I own and have worked a lot on all three boards I mentioned ( and more ), the starting point for VccSA at 3733 is 1.25.

Climbing the ladder to 3866 I use 1.30 and then at 4000 I use 1.35.

 

3866 at 1T is not easy.

First and foremost it depends on the mems themselves.

I have two sets cl17 and one set cl16.

Out of the three only one set does 3866 - 1T and 4000 - 1T on the OCF.

 

The final result depends on many parameters and every set up has its own peculiarities.

 

With the exception of SPI, where a few milliseconds will make the difference, I have not noticed serious variances in many other benchmarks,

that could very well be in the acceptable margin ( +/- 1% )

 

Thanks for that, I appreciate your help. I'll try up my voltages a bit and see how it goes. B-Die is very voltage sensitive, compared to my Ripjaws V E-Die, they are lot more forgiving.

 

So should I just set my IOL offset straight to 23 every time, or try 22 first ?

[Guest george.kokovinis]

First MANUALLY set 21-21 offset.

If this stabilizes things move to 23-23.

 

I like 23-23 because the RTL and IOL values are more correctly aligned and symmetrical on both channels.

 

49/49 and 5/4.

 

The more symmetrical these values get, the easier for the memory controller.

In simple words, it is as the controller is looking at one stick and not two.

Way of putting it extremely simple. Lol

[Fact11]

First MANUALLY set 21-21 offset.

If this stabilizes things move to 23-23.

 

I like 23-23 because the RTL and IOL values are more correctly aligned and symmetrical on both channels.

 

49/49 and 5/4.

 

The more symmetrical these values get, the easier for the memory controller.

In simple words, it is as the controller is looking at one stick and not two.

Way of putting it extremely simple. Lol

 

That makes perfect sense. Thanks George, will let you know how I go.

[Tordp]

If I may chime in on VCCSA/VCCIO on Z170 SOC Force... I learned that you need a lot more than expected. I'm running Corsair Dominator 32GB 3000MHz 2x16GB and need at least 1.27v VCCSA and 1.23v VCCIO. This with standard 1.35v and no overclock.

[xxbassplayerxx]

@Fact11, I was completing 32M at 3866 12-12-12 1T. However, I was binning sticks so only running in single channel. I'm having difficulties posting with two sticks. It makes it to 62 and graphics come on, then it goes up to 69 for a brief second then restarts. It will eventually fail on 32, 55, 4A, or a number of other codes when it tries to train one more time. I didn't have much time to troubleshoot last night so I'll keep trying tonight and updating here.

[Fact11]

Hey Bass, So far i've finally to managed to get two sticks in dual channel on my Z170X SOC Force, to pass training and boot into windows and its stable at 3866mhz 12-12-12-280 1T. Apart from setting the primary timings, tRFC and CR, I also have tFAW at 16. Everything else is still at AUTO.

 

I have the Trident Z 3600mhz - 16-16-16-36 16GB Kit.

 

It passed with the following voltages :

 

Vccio - 1.30

Vccsa - 1.35

Dram and Dram Training - 1.85

 

Upto 3733 Mhz, I can have my Vccio at 1.185 and SA at 1.225 and its stable as, at 12-12-12-220 1T.

 

Once I go to 3866mhz I need to up the voltages to the above. I still have to see how much lower I can go on my tRFC before it will no longer post. Then I'll try for 4000mhz as well and see how far I can get. I'm on the latest F6 bios, which came out at the end of March. I think its simply the bios still needing improvement, that is limiting how far we can lower our timings.

 

Asrock Z170 OC Formula simply has a better bios. Plus I noticed in the memory compatibility chart for our board, hardly any of the Trident Z B Die modules are in there, whereas the Asrock memory chart, pretty much has all the Trident Z in there. Which is why I think we are having so much difficulty getting it to pass training.

 

Let me know if that works on your board.