Jump to content
HWBOT Community Forums

Recommended Posts

Guest Bullant
Posted

Few cold PSC stability test on Gigabyte Z97 SOC Force

5xo4rt.png

68gftu.jpg

2dv74fn.jpg

9kb12s.jpg

  • 3 weeks later...
Guest Bullant
Posted

Have you guys tried starting from 2400mhz first? So train memory 2400-2600-2666,quite often if you try boot 2600+ with this board with out training from 2400mhz first it could fail.

 

The other thing as well if you have a weak imc try a little cold on the CPU it could help imc to scale,ill post some of my bios settings later if you like?

 

In some recent testing I was able to boot 2800mhz 8-12-8-28 tight will CPU on water and memory -25,it was achieved by training from 2400-2600-2666-2800 and locking in rtls along the way,ill post it up a little later

Guest Bullant
Posted (edited)
When training from 2400 first, should you set your timings as if you were going straight to 2666?

 

Yes set them as if your going 2666 8-12-8-28,just start with the subs like mine 4-5-5-20-5-4-4-5-5-13-13-13 on the right of screen,ill pass some bios settings on later that should help,try low cpu VRIN

volts 1.73-1.8v,there another setting in advance tab I set to extreme I can't remember the name ill look when home and pass on imc,ioa,iod volts

 

 

In advance power settings I set PWM phase control to extreme performance,you can try this,also I set memory boot mode to disabled,also in the advance power settings leave CPU VRIN load line calibrations on auto to start with

 

Edit,if my sub timings are too tight to boot 2600mhz try one of gigabyte memory profiles,I haven't checked there profile in bios but I'm sure it looser ,doesn't have to be high binned memory to do 2600mhz and the sub timings I said

Edited by Bullant
Guest Bullant
Posted
to Bullant: provided you can train 2666 8-12-8 straight, is it more/less stable compared to 2666 8-12-8 trained via 2400-2600-2666 ?

 

Hey Sam,I haven't done a lot of testing on the board but when I was able to boot straight 2600+ the rtls and iols would go off line and even if it was stable my times would suffer.

 

The training of the memory did help and once trained stability did seem good,I've only tested two 4770ks on this board and one was good imc and the other not so good.Will test some more soon and try give some more feed back

 

Just to note,good imc and good memory will also help with memory training of rtls and iols,(keeping them closer in line)lower binned memory will need more care in training,I know you know all this but if anyone wanted to know.Will try the new Pentium CPU and 4790k soon and leave some more feedback

Guest Bullant
Posted (edited)

You may need to be on X04 bios mate,x04 is the better one for training rtls

 

Do the same again on X04,once you set 2600 after training 2400 save exit if it doesn't go to windows then you will have to set rtls before saving and exit on the 2600.

 

Also to note if you fail on the 2600,you may have to train 2400 again before setting 2600,first use X04 bios and try again

 

Edit ,after setting 2600 once trained it should go to windows if you don't like the rtls and iol this is normally were you go back to bios and set rtls and iol manually,so you should be able just to set 2400 save and exit,then set 2600 and save and go straight to windows with X04,so what I'm saying is before setting rtls if you have to you will have to train 2600 first because you will need the rtls to be in the 40s before locking them in,hope this make sense lol

Edited by Bullant
Guest Bullant
Posted (edited)
Can somebody explain in a video what does it mean "trained" in this context :)). At least some screenshots to kniw what you ard talking about...

 

It really is quite easy mate,the memory multis go in order of 2400-2600-2666-2800 and so on,you need to train the memory in its order from 24-26-2666 to achieve higher memory clocks,

 

The rtls and iol seem to train better when training from 24x onwards at least in my testing it does

Edited by Bullant
Guest Bullant
Posted (edited)
I can't even boot at 2400 using X04 :-O

 

Just make sure you haven't missed something if you haven't just go back to other bios,try memory boot mode on auto as well as disable see if any help,I normally only use disable mode as this helps my training,will do some more testing my self soon

Edited by Bullant
Posted (edited)

I think something is wrong with the board itself, takes a long time to boot, simple settings fail (even cmos reset).

 

EDIT stupid H100i usb header cable was the problem.

Edited by sabishiihito
Posted
It really is quite easy mate,the memory multis go in order of 2400-2600-2666-2800 and so on,you need to train the memory in its order from 24-26-2666 to achieve higher memory clocks,

 

The rtls and iol seem to train better when training from 24x onwards at least in my testing it does

 

So basically it just helps you out to reach a predefined(higher) frequency (like 2666) ?

Or it has other effects on spi32m run as well ?

Or both of the above combined ?

 

If the goal is just to achieve higher memory clocks,will it matter if i can put it straight to 2666 or do i need to "train" them to that frequency for other purposes also ,like efficiency?

 

Thanks for your help.

Guest Bullant
Posted (edited)
So basically it just helps you out to reach a predefined(higher) frequency (like 2666) ?

Or it has other effects on spi32m run as well ?

Or both of the above combined ?

 

If the goal is just to achieve higher memory clocks,will it matter if i can put it straight to 2666 or do i need to "train" them to that frequency for other purposes also ,like efficiency?

 

Thanks for your help.

Currently with early bios running PSC you quite often have to train the memory from 24x to get to 2600-2666,quite often it wont boot 2666Mhz straight on a i7 4770k.What happens is when you try boot straight 2666Mhz rtls quite often go a little to tight and can cause the 69 code,to get passed this we boot 24x 26x 2666x in this order

 

If you can boot straight 2666 it may be the same rlts as if you were training from 24x so efficiency would be the same,trained or untrained.Once in windows @ 2666x untrained or trained you could have wondering rtls and iols,this is when you would return to bios and lock in rtls and iols

 

On the rtls we would like to see 40-40 -41-41 @ 8-12-8 2666 with PSC,quite often when we achieve 2666 trained or untrained rtls they can wonder of line a little like 41-41 43-43 or even worse,this is were we lock in say 40-40 -41-41 for the two channels,only after 26x or 2666x is achieved

 

So another thing to note is rtls are efficiency

 

So the main reason for training from 24x onwards is to get to 2666 because you cant always boot 2666mhz straight,if rtls are the same when you get to windows with 2666 trained or straight boot then efficiency should be the same either way.If rtls are loose then you return to bios and lock them in tighter efficiency should then be better

 

Training from 24x =rtls 38-38 -39-39 26x = 40- 40 -41-41 i think it was and 2666 the same,so when I say train memory from 24x 26x what we a really doing is training the memory's rtls in a sense because quite often if you just try boot 2666 rtls can go 37-37 37 -37 and wont boot,its a good thing to always see what rtls are doing in bios, hope this makes sense

I quite often have to lock in rtls when pushing big Mhz on a straight boot when cpu is on say water,memory 2740Mhz air on other boards as well,imc can only handle so much when also pushing bclk

Great work Bullant

 

Thank you XA

 

 

This was some testing i did a little while ago,CPU on water and memory only -25 on my little single stage.

The EK adapter plate is great for pre testing memory,these 28x was ahceived the same way 24x 26x 266x 28x

 

dinx2.png

DSCN9536.jpg

DSCN9538.jpg

Edited by Bullant
Posted (edited)

Hey Guys.. Here some memory result.. all air, quick run, no wazza..

 

Samsung

snaphsot0110.png

 

PSC

snaphsot0105_hwbot.jpg

 

BBSE

snaphsot0097_hwbot.jpg

 

Hynix MFR

MFR.jpg

validasi_MFR.jpg

 

 

@Bullant, any tips for samsung and psc above? should it be more quicker? thanks man :)

Edited by ivaneza
Guest Bullant
Posted

I haven't tried the 125 strap yet but Dino did mention it to me a few days ago,just havnt had a chance yet to try it,looks good

 

Very nice Ivan on all your ics

  • Administrators
Posted

Nice run, Alby :D - and if you use 2200 divider, you might get nice surprise because you end up at 2750 even if you do not raise BCLK above 125 like you did :P - should be around 2800 then, I have doubts even your monster flares will do this on air :D

  • Administrators
Posted

Try to raise SA to +0,15 and the ios to +0,2, lower vccin drastically, you are at 4ghz^^ - 1,9v might be even more than you need - and try again, imc quality on 1150 cpus varies and voltages on this too low might be well the source of your problem

  • Like 1

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...