Jump to content
HWBOT Community Forums

Recommended Posts

Posted (edited)
Lazy :) Auto volts on everything but Vcore, Vdimm, SA and IO.

AUTO with Ln2 mode ON then.

PLL OC stock is 1v if I remind correctly. Anyway this voltage only helps CB no need to push it otherwise, it won't do anything else for OC. :-D

Edited by pepinorang
Posted
AUTO with Ln2 mode ON then.

PLL OC stock is 1v if I remind correctly. Anyway this voltage only helps CB no need to push it otherwise, it won't do anything else for OC. :-D

Actual volts were 1.2. Misread because of T.0P version of Command center on non-T.0P bios.

 

Actually made 3600C11 TCWL9 now, 1.95V though:

guestaccess.aspx?guestaccesstoken=zi34PxLC2HyMtaY2Y9Gtn5cHGwAlWQdM%2f9bhCuEg3%2bU%3d&docid=0a3a393e6bfcd4b6bbbf4c5c0a935b83b

Posted

Don't want to turn this into a mem thread but just a FYI with regards to RJ's 3466 vs 3600 for those wondering...

 

Tested 8 sticks of mem today (3466 quad kit, 3466 dual, 3600 dual), two of which used previously in the thread. They dropped from 1.87v to 1.86v. All other sticks were quite poor, one 1.87, one 1.92, the rest i gave up at 1.95. Decided to change to 2 sticks per run, stuck the 3600's in with only one stick capable of 1.87, the other was way above 1.9v. Run after run of 32M i could trim the vDIMM. Ended up with the below. Appears that these need a little break in period before settling. Sound plausible? I'll see how the other sticks do tonight.

 

SA/IO volts not touched. (1.2v/1.15v)

 

1.9v on board = 1.906v on MM.

 

Anyway, seems whichever kit is the cheapest is the sensible buy. :) The 3466's i've posted previously i've now got running at the below timings with 1.86v.

 

p19V1VY.png

Posted

Yes, my team 3866 have also improoved since first tests on high volts. Bedore they could not do even spi 1m, now they do 32m. Same volts, even tighter subtimings.

Posted (edited)
Pepi, could you explain where ST volts come into the equation? Totally unfamiliar with the term ST.

 

On most ES CPUs that I tested, ST didn't made any difference from 1.25v to 1.8v but below gave higher CBB, that's why I recommended to use at least 1.25v in my guide.

 

However, I recently saw that some CPUs have a "wall" which CPU ST can help to lift once you push it.

For instance, Vivi's retail 6770K can do 6.3G Cinebench R15 but it needs +1.4v ST to clock higher than 5.9G and we tried all voltages one by one; nothing else helped.

Same thing with the now dead ES chip we used yesterday, it couldn't even idle at 6.4G with 1.25v and would crash straight when applying ratio in OS, but with 1.65v ST it ran 6.42G Cinebench R15 (we didn't try values in between, lower would probably have been OK as well).

 

That being said, you really have to test all voltages one by one for every chip to see if it helps, as it seems all Skylake CPUs need totally different voltages from one to another.

 

As stated in the voltage table of the guide, ST V6 needs to be set from 300mv less to same than ST as these voltages are linked together.

Edited by pepinorang
Posted
Pepi, could you explain where ST volts come into the equation? Totally unfamiliar with the term ST.

 

Vst = CPU Standby Voltage.

 

I'm not sure what its function is, but it appear to be related to the CPU PLL (frequency generator) voltage. It would be nice to know exactly what it does though.

 

On some motherboards, the PLL termination voltage and standby voltage are separated. That is for example the case on the Maximus VIII boards (not the Hero). I'm not sure if it has any major practical effect to dial the voltages in separately as even in the ROG guide it's written: "Now try to keep a less than 450mv delta between PLL Termination voltage and CPU Standby Voltage. When using 1.50v PLL Termination you will probably want at least around 1.20v on CPU Standby Voltage. Easiest would be to synchronize these 2 rails."

Posted

Ha, i'm going through exactly the same thing.....! Won't do 1t till down at 2800. Couldn't even tune quad till down at 3200 but for some bizarre reason it's now playing the game and now have 3600c11 quad passing max mem. Only thing i can think of is i'm now running a completely default bios and only changing SA/IO/vDIMM.

Posted

Like I said to Doug, I will check this when I get back to Taipei (15 Oct) with our memory guru Toppc and ask for some Quad channel settings. It really shouldn't be a problem as quad channel memory QVL is 3733 with this board and I even saw memtest pass at 3866 quad on G.Skill TridentZ.

Posted

I set bios to default, sa/io 1.2/1.15, vdimm to 1.65, set 3600 and primaries loose to 18/20/20/40/500, trefi circa 32k, all other timings auto then started moving primaries down. Now down to 11-18-18-28-300 @ 1.9v.

 

Haven't had a chance to run 32m with this, only max mem but passed the above on 32m with cl13 @ 1.65v. (ball park vdimm, same with 1.9v for cl11 as i had dual at 1.86/87v)

 

With xmp enabled for 3466c16 i pretty much couldn't change any timings till i dropped freq to 3200.

 

Not tested bumping core/cache with the above yet.

 

Will play about more later. Absolutely no chance of 1t right now but as pepi mentioned maybe an answer coming soon enough.

  • Crew
Posted

my experience with this motherboard!

 

1 - works easy in the bios.

2 - memory clocking was simple, xmp loaded on 10+ boards we tested. gskill though (2x4gb and 2x8gb kits)

3 - ln2 was easy going, but there is one mod needed to remove some 00 issue.

 

voltage breakdown:

 

Aav4cOb.png

Posted

Used this board for almost 2 days of 8 hours straight with LN2. Only thing that went wrong is that now one of the memory slots doesn't do XMP or any high speeds anymore.

Posted
Used this board for almost 2 days of 8 hours straight with LN2. Only thing that went wrong is that now one of the memory slots doesn't do XMP or any high speeds anymore.

Look at your socket, it is probably bent by tightening pot too much or going too fast on cold down/heat up... :(

Posted (edited)
my experience with this motherboard!

 

1 - works easy in the bios.

2 - memory clocking was simple, xmp loaded on 10+ boards we tested. gskill though (2x4gb and 2x8gb kits)

3 - ln2 was easy going, but there is one mod needed to remove some 00 issue.

 

voltage breakdown:

 

Aav4cOb.png

The voltage difference is as explained in my guide: -0.25v on CPU ST/ST V6 with Ln2 Mode ON in CCL, but fine in BIOS.

We are working to fix this in next CCL version already.

However IO reading on retail boards is perfectly fine, there's only a difference on some old ES boards like yours. ;)

 

The 00 issue happen when you push CPU ST above 1.33v and shutdown board.

I'll share the mod soon (really easy) but in the meantime you can bypass this by Power OFF PSU>Wait MB LEDs turn OFF > Power ON everything again and it will be gone.

If you still have 00 it is either CBB or CPU ST too high for temp as explained in guide.

Edited by pepinorang
Posted

Used this MB for a 6 hour session at -170 on my 6700K yesterday. It works really well! A couple of things I have noticed:

 

How the volts change the behavior of your CPU is really CPU dependent. It's not as easy as haswell, you will have to try all settings to see how your CPU reacts. ST-volts can make a huge difference on CB on some CPUs. Raising PLL volts might now work for some, but ST-volts will do the trick. For some ST-volts of 1.3V is enough, for one of mine it needed 1.6V to get below even -116.

 

Memory OC is really great on this board. Might contest the Asrock in this generation!

 

LN2 mode enabled in bios shuts of AES-instructions. You need these, at least for Geekbench or your score will hurt a lot.

 

My board is early retail, and does not have physical ln2 jumper. You don't actually need it though. It runs fine without it on LN2. Or you could solder, just connect two points.

 

For Samsung-E memory, you need at least 1.13 bios. The older T.0P won't run these mems well at all. Newer bios make them fly as you have seen in this thread!

 

One annoying thing however is that often after a bios flash the board enters a sort of mode where trying to reboot won't work 9/10 times. The errors are persistent and occur both at stock settings and under LN2. Both reboots after bios changes and reboot from windows gets stuck. You have to power off, reboot into bios because board reports failed boot, reboot again and it works most of the time. Sometimes it will take multiple tries to get successful boot. Sometimes a reflash of the bios will help cure this problem, but I'm hoping for a fix in upcoming bioses. I have seen this reported by another user so I'm not alone with this problem.

Posted (edited)
Used this MB for a 6 hour session at -170 on my 6700K yesterday. It works really well! A couple of things I have noticed:

 

How the volts change the behavior of your CPU is really CPU dependent. It's not as easy as haswell, you will have to try all settings to see how your CPU reacts. ST-volts can make a huge difference on CB on some CPUs. Raising PLL volts might now work for some, but ST-volts will do the trick. For some ST-volts of 1.3V is enough, for one of mine it needed 1.6V to get below even -116.

 

Memory OC is really great on this board. Might contest the Asrock in this generation!

 

LN2 mode enabled in bios shuts of AES-instructions. You need these, at least for Geekbench or your score will hurt a lot.

 

My board is early retail, and does not have physical ln2 jumper. You don't actually need it though. It runs fine without it on LN2. Or you could solder, just connect two points.

 

For Samsung-E memory, you need at least 1.13 bios. The older T.0P won't run these mems well at all. Newer bios make them fly as you have seen in this thread!

 

One annoying thing however is that often after a bios flash the board enters a sort of mode where trying to reboot won't work 9/10 times. The errors are persistent and occur both at stock settings and under LN2. Both reboots after bios changes and reboot from windows gets stuck. You have to power off, reboot into bios because board reports failed boot, reboot again and it works most of the time. Sometimes it will take multiple tries to get successful boot. Sometimes a reflash of the bios will help cure this problem, but I'm hoping for a fix in upcoming bioses. I have seen this reported by another user so I'm not alone with this problem.

EDIT: You DO need Ln2 jumper ON on many CPUs, it has been extensively tested and can greatly improve CBB and cold stability at boot.

However don't short points, use a switch or a jumper header if you can as I recently heard about retail CPUs that couldn't boot with Ln2 jumper ON under aircooling but are fine on cold, I will update guide soon.

Latest retail batchs have jumper header.

 

T.21 (based on 121) with all unlocked volts and tFAW working below 16 is coming (quickly succesfuly tried up to 12 with ViVi), just finishing some more test on it when I get back to Taipei and will post it here.

 

Not sure about the reboot issue, never had this @.@ but thanks for your feedback, will check!

Edited by pepinorang
Posted

The board that keeps on giving...

 

3600@1.91v, 11-18-18-28-2t, tcwl 10. QUAD! Passed 32m, still tweaking. Will post a screen when the rest of the house is awake. :D

 

Pepi, is DRAM or CPU SA Phase control going to have any impact on ram in quad? Both on auto now instead of disabled and seems to have lifted my quad curse.

Posted
The board that keeps on giving...

 

3600@1.91v, 11-18-18-28-2t, tcwl 10. QUAD! Passed 32m, still tweaking. Will post a screen when the rest of the house is awake. :D

 

Pepi, is DRAM or CPU SA Phase control going to have any impact on ram in quad? Both on auto now instead of disabled and seems to have lifted my quad curse.

Probably, it usually helps with dual/single at high frequency (+4000), so it makes sense it could also helps with quad.

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...