[GIGABYTE-Z77] DDR Issue report

[Hicookie]

I like to open this thread and welcome you guys report incompatible issue with your own memory with GIGABYTE Z77 motherboards. Here is the basic settings, I tested many 2666/2800 modules and stress more than 100 hours without error.

 

Let me explain somethings about ivy-bridge

 

1. CPU internal PLL:

 

Allow to open the internal PLL voltage to go higher, but how high? and what's the default PLL voltage, this is part of Intel NDA, can't leak.

 

What this function be open help?

 

• CPU total frequency be improved once you enabled this option
  
• by pass IMC 2800, help your chip doing DDR3-2800MHz if enabled this function.

 

2. IMC: 2800 for every chip?

 

Yes, almost every chip can do 2800, we know every CPU has IMC margin, raise Vsa(vIMC) and Vtt can help you to reach and stable with 2800 if your chip is not good to handle 2800 by default voltages (Vsa 0.925V. Vtt 1.05V). So its worth to try if does improve your CPU IMC by raising Vsa and Vtt. To get more information, with GIGABYTE's PWM design, Vsa can't be higher than Vtt, beside the UP7.

 

Take a example, if you set Vsa to 1.25V, then you also need to set Vtt as 1.25V or higher. Some example for you to set those voltages.

 

• DDR2666, Vsa 1.1V Vtt 1.1V
  
• DDR-2800, Vsa & Vtt 1.1~1.25V, some good IMC CPU just need keep them defaults.

 

So any incompatible issue, please reply on this thread, GIGABYTE will look into your issue.

 

Overall BIOS settings for your X.M.P memory (if you have XMP memory and don't know where to start just follow those settings)

 

• Vtt: 1.15V
  
• Vsa( vIMC) : 1.15V
  
• DDR voltage : 1.65V ( or apply by your X.M.P spec )

 

tRC------------------auto

tRRD------------------7

tWTR-----------------10

tWR-------------------16

tWTP------------------auto

tWL--------------------8

tRFC-------------------214

tRTP-------------------10

tFAW-------------------30

CMD--------------------2

tRW Stability level -----auto

tWR Stability level -----auto

tREFI-------------------10920

tREFIx9-----------------95

tRRDR-------------------3

tRRDD-------------------3

tWWDR------------------5

tWWDD------------------5

tRWDR(DD)--------------5

tWRDR(DD)--------------3

tRWSR-------------------5

tRRSR--------------------5

tWWSR-------------------6

 

For Corsair 2800C12, the TAONPD might need to open, but not on recent BIOS yet.

[zeneffect]

Will give this a shot. home sick today so ive got a little time to tinker around.

 

Thanks for taking the time to write this up and help debug the issues we are seeing. I think the UP7 has alot of potential, its just very hard to work with unless you really know the board and your memory.

 

*edit*

 

quick test

 

3770K - UP7 - Kingston HyperX Preadator 2666 kit

 

xmp profile 1 = 51

xmp profile + following your guide = 51

xmp profile + following your guide + tRW Stability level 32 + tWR Stability level 32 = full post, not 32m stable

 

will check to see if slew helps any for gaining 32m stability.

 

 

xmp sets ram voltage to 1.68? (opened gtl, 1.68 is automatically set)

Edited December 13, 2012 by zeneffect

[Hicookie]

It should be compatible with GIGABYTE Board.

 

http://www.overclock3d.net/reviews/memory/kingston_hyperx_predator_8gb_2666mhz_review/2

[Christian Ney]

Already sent an email to Dino

[Leeghoofd]

It should be compatible with GIGABYTE Board.

 

http://www.overclock3d.net/reviews/memory/kingston_hyperx_predator_8gb_2666mhz_review/2

 

That website also uses Wprime95 for RAM testing... tells the tale...

 

 

 

 

BACK TO UP7:

 

UP7 F5J doesn't seem to change RAM divider, XMP profile loaded for 2666C10 Platinums, boots at 667MHz, forced 2666 divider: result 667Mhz c'mon. Bad bios flash ? Board came back from Guru3D, no idea how they tested it.

 

Flashed back to F5f, RAM dividers sort of working, but now I lost the CPU multiplier OC button function...

 

 

 

XMP profiles tested:

 

CORSAIR 2666C10 PLATINUM Need to loosen tWWSR to 5 iso 4, to be 32M stable (1.68Vdimm is set by the board) tWWSR at 4 works fine on other brands with two sticks

 

CORSAIR 2800C11 PLATINUM 51 error

 

GEIL EVO VELOCE 16GB (2 x 8GB) 2133C10-11-11-30 Passed 32M

 

GEIL EVO Corsa 16GB (2 x 8GB) 1866C9-10-9-27 Passd 32M

 

GSKILL RIPJAWSZ 2200C7-10-10-27 51 error Boots fine with CAS 8 set iso CAS 7 from XMP (sticks do 2500 C7-11-7 on other boards) Only way to get CL7 booting is setting tWL to 8

 

GSKILL FLARES 1866CL7-8-7-24 CAS 7: 51 error, CAS 8 boots fine (sticks do 2400 C7-11-7 on other boards) Only way to get CL7 booting is setting tWL to 8

Edited December 13, 2012 by Leeghoofd

[zeneffect]

with f5j, if you lose ability to set ram multi, clear cmos, then set it. ive had this problem a few times as well, clear cmos always does the trick. you will be able to change multi's for a while then when you 51 too many times it kicks back to default and stays there. mabye i try flashing both bios and take it off of single bios mode.

 

f5f gives a weird value for me in dram voltage settings.

 

@ hicookie

 

testing in that review was done on UD5, not UP7.

Edited December 13, 2012 by zeneffect

[Leeghoofd]

@offtopic Also why doesn't the Gigabyte Tweaklauncher 1.0 work with Sandy CPUs on the UP7 ? ( I know it says IVB )

 

Thanks Zen, need to get the hang of this board asap, will be working on it for a demo show next week...

[zeneffect]

will you be using more than 1.85 vcore?

 

if so, best bet is to do to the f3 oc bios and use bbse. the board is pretty smooth with that combo and it gives proper oc volts. why are you using sandy on up7?

Edited December 13, 2012 by zeneffect

[Leeghoofd]

Nope normally not more then 1.85, it's just demo BBSE are still at Nick Shih's place sigh... you got a link for F3OC ?

 

Sandy for country cup

[zeneffect]

you might want to use a proper sandy board then. p67ud5 was pretty killer iirc

 

http://www.mediafire.com/?a9iilopes78c56z

 

f3oc. got the link from the tt giga bios thread.

Edited December 13, 2012 by zeneffect

[Hicookie]

For 51 issue is the bios problem, this is odd should be fix by latest biases.

Do you guys test them? The 51 is issued manual set bulk cause x.m.p fail.

[sin0822]

zeneffect- that is the predator 2666C11 right? I have that kit without issues with XMP.

[sin0822]

@offtopic Also why doesn't the Gigabyte Tweaklauncher 1.0 work with Sandy CPUs on the UP7 ? ( I know it says IVB )

 

Thanks Zen, need to get the hang of this board asap, will be working on it for a demo show next week...

 

Because of where the clock generator is, GTL uses ICCS driver Integrated Clock Controller Service, Ivy has clock generator built into the CPU.

 

 

For the UP7 is you want OC touch to work with Ivy, what i do is I go into GTL after making sure ME, OC Touch, and GTl are properly installed, and then i change the multiplier 1x up or down and then that allows me to use OC touch for multiplier with Ivy Bridge. You can even close GTL after you change the multi, you just need GTl to initialize ICCS driver I believe.

 

I made a video:

[Hicookie]

Because of where the clock generator is, GTL uses ICCS driver Integrated Clock Controller Service, Ivy has clock generator built into the CPU.

 

 

For the UP7 is you want OC touch to work with Ivy, what i do is I go into GTL after making sure ME, OC Touch, and GTl are properly installed, and then i change the multiplier 1x up or down and then that allows me to use OC touch for multiplier with Ivy Bridge. You can even close GTL after you change the multi, you just need GTl to initialize ICCS driver I believe.

 

I made a video:

 

add, if your CPU ratio not working, please check if EIST be disabled.

[Massman]

Cookie, remember I had the '51' error too after testing the bios you sent me? Needed the "cookie profile" for everything to start working.

 

But, eventually here's a workaround:

 

1) load default options

2) change memory divider to 2666 - do not touch anything else

3) save and reboot => should be no 51

4) change bclk to 101 or 102

5) save and reboot => shutdown and boot and no 51

 

This worked for me on the UP7. If I would change too much after a CMOS clear, I would still get the 51. Especially if I would change the BCLK together with the memory multiplier. Also, I needed to load the optimised defaults if 51 fail.

[Leeghoofd]

So what happens with a hard crash under LN2 ? 00 error alike Asus if the board resets 100 bclock ? Will put the board under it's first paces with LN2 next week as a pretest for the demo. To be honest atm not really impressed, far too many small issues to workaround...

Edited December 14, 2012 by Leeghoofd

[sin0822]

Ill post an OC guide for this board including memory in there too, you can use that Leeghoofd if you don't want to explore the board for yourself. CPU OC works very well, if you don't want to have EIST enabled for OC Touch, then you can use my methods above, i know it works for F3OC. I did Ln2 OC with it, it worked just fine, not many issues.

 

Just make sure to change from dual BIOS to single BIOS mode, oh and if your ratio is stuck at 16x, check the Ln2 switch.

 

I asked for an LN2 BIOS with the Slew rates btw(I am waiting for a version that can be put into public), but if you are OCing hynix or something then you can just use F3 BIOS.

 

If you really don't want to use it, use a UD4H instead.

[Leeghoofd]

UD4H is no option Gigabyte wants me to promote the UP7 thx for any info...

[zeneffect]

if you hard crash, just hit reset button. 50% of time it will recover and you can still boot with full pot.

 

its alot easier to set like 108 bclk with low multi, and ram set for the "target" freq + timings (i.e. you expect xxxx clock so you set timings accordingly even though you havent reached those clocks yet. basically giving yourself some headroom... why am i even telling you the things you already know... time to ramble and play with the special characters *((@&(&(*!(*)(#@!*$)(!)

then setup in gtl once in windows or use the OC buttons.

 

if you crash, hit reset and it will go to "safety" 108bclk clocks you set in bios and you will boot right up with full pot. ive been able to bench for hours @ full pot without ever having to warm up. this works 100% of the time for me. you can be even speedier and use the team.au utility and set a preset button so once you boot, you set the preset and fine tune from there. boot to benching if done right can take less than 40 seconds from low clock to full tilt. the key is setting 100% stable boots and fussing with the memory before going below cbb. with the right pre-setup you will get (almost) never get reverted back to 100bclk. thats the advantage of the oc buttons

 

i dont mean to be telling anybody "this is how they shoud run" this is just how i run and find the most trouble free benching.

 

 

on a side note... i have been manually setting vtt + imc with the xmp. xmp wants to put vtt at 1.3something. im not too comfortable with that and set 1.2vtt 1.15 imc. chip is capable of 2800 with this kit on "other" boards with these volts.

 

*edit*

 

Leghoofd, take a look at bios F5m (latest) vcore is now 2.1v max + slew rates. Lets play with some PSC

Edited December 15, 2012 by zeneffect

[stasio]

 

on a side note... i have been manually setting vtt + imc with the xmp. xmp wants to put vtt at 1.3something. im not too comfortable with that and set 1.2vtt 1.15 imc. chip is capable of 2800 with this kit on "other" boards with these volts.

 

 

In reality,

XMP set Vdimm to 1.680V (UP7,UP5,...) and some Z77 boards to 1.668V, but never 1.650V ,Vtt and IMC are always set to default.

 

Also my TeamX 2600C10 XMP set CMD to 2 instead 1.

[zeneffect]

In reality,

XMP set Vdimm to 1.680V (UP7,UP5,...) and some Z77 boards to 1.668V, but never 1.650V ,Vtt and IMC are always set to default.

 

Also my TeamX 2600C10 XMP set CMD to 2 instead 1.

 

just set xmp on f5m. 2666 boots 0 issue now. checked vtt with dmm and it does in fact set default. in bios it says profile vtt is 1.35, but actual is default as you said. thanks for the heads up.

 

*edit*

 

scratch that. im having 0 problms because its not setting 2666. stuck at 1333 currently, will try re-flash. (xmp + setting multi manually)

 

*edit edit*

 

cant change memory freq with F5m. dram voltage stays 1.5 regardless of bios setting. it seems there is alot wrong with this bios release. looks alot like f5f as the bug in dram ref voltage shows weird value instead of auto.

 

leggy, id skip this one. many things wrong (but please do test to see if you getting same results)

 

still wanting/waiting for a good oc bios with sku and enough things that work to bench with for some 32m fun. dino, hicookie, massman, anybody? id be willing to try "test" bios just for fun id send an in and out tshirt to match the lanyard *wink wink* for a good ln2 bios.

Edited December 15, 2012 by zeneffect

[Leeghoofd]

Great work sigh, no RAM multipliers indeed, no OC buttons, way to go... for me not a good idea to continue building on this bios... maybe time to explain to the bios team 1333MHz in bios is not 2666MHz

 

Tried 3 reflashing, CMOS resets,... will try the OC bios... this should not be... possibility to be a great board, crippled by mediocre biosses...

 

F3OC bios works great with Samsung, that's where I stop my testing with other IC types... far too annoying experience in my book...

Edited December 15, 2012 by Leeghoofd

[Blue Storm]

Next Beta BIOS and more errors than with previous one?:confused:Still waiting for good BIOS for LN2 and PSC:(F3OC is still best BIOS for LN2 OC but not for mem clocking...

[Bobnova]

Both the Z77-UD5H and the Z77-UP4 couldn't run XMP timings on my 7-10-7 ripjaws. Needed 8 CAS. Never did figure out what the problem was.

Other z77 boards (ASRock professional, TZ77XE4) could do it fine. One other couldn't (asrock extreme6) and needed the same +1 CAS.

 

Just a heads up really as both boards have been rage-sold at this point

[stasio]

Edit to my first post........

placing memory near to CPU socket (DDR3 3 & 4) give less code 51.

 

One more thing is interesting.......replacing the memory sticks also help.....?!

 

Just 2 days ago trying my TeamX 2600c10 @ 2700 in slots 1 & 2......(boot only twice with 10 try)....51

 

 

slots 3 & 4 no problem......

 

 

cant change memory freq with F5m. dram voltage stays 1.5 regardless of bios setting. it seems there is alot wrong with this bios release. looks alot like f5f as the bug in dram ref voltage shows weird value instead of auto.

 

leggy, id skip this one. many things wrong (but please do test to see if you getting same results)

 

still wanting/waiting for a good oc bios with sku and enough things that work to bench with for some 32m fun. dino, hicookie, massman, anybody? id be willing to try "test" bios just for fun id send an in and out tshirt to match the lanyard *wink wink* for a good ln2 bios.

 

You are right......I am with UP5 back on F12c BIOS......F12e give me only headache.

 

And another 4 months Z87 coming.........!?

Edited December 15, 2012 by stasio