******Gigabyte GA-P67A-UD4 Discussion Thread******

[Falkentyne]

This is what I'm saying

Ok I understand now, and that is indeed very perplexing.
Now I'm going to "assume" that the UD5 overvolts the same as the UD4, just for sensor purposes, so we have to go by the software (as I have no DMM), but with LLC disabled, I get a different vdroop than you do, even at lower voltage:

1.25 bios, 1.224 idle, 1.164v load, difference = 0.086 vdrop, 0.06v vdroop, 0.026v bios to idle drop
Yours: 1.35v Bios, 1.320v idle, 1.284v load = 0.066 vdrop, 0.036v vdroop,0.03 bios to idle drop.

Now what's interesting is, your BIOS setting to idle drop performs like it should, but (just going off the sensors), and noting that the higher the vcore, the higher the vdroop on both boards (as expected), then on my board, the vdroop at 1.224v is significantly higher than your vdroop at 1.320v (and that is NOT expected!)

So your board is drooping less than mine, but you're BSOD'ing at voltages that have you stable with LLC on. That would make perfect sense if you were actually getting the same vdroop as me and the measurements were 0.024v too high (then the BSOD is obvious: too low vcore and the 1.284v is actually 1.272v), but since the measurements prove that wrong, then I have absolutely NO idea what's up.

Since Vdroop is supposed to be a certain slope under VRD 12, it seems odd that your board has less vdroop than a UD5. But the BSOD would only make sense if the vdroop were more than what it's showing. But you can't break the laws of physics...so I'm clueless. And I suck at electronics, anyway...

[irev210]

Extremely simple fix.
Disable EIST, C3, C6, C1 (forgot if that's there).
Use fixed Vcore.

Profit.

Disabling C3/C6 option does the trick on this board. Don't need to disable anything else.

Very annoying that ASUS/Gigabyte/Others? are having this problem.

[dinos22]

hey man thanks for testing that

looking good now

Okay, I've tested some voltages.

i5 2500k @ 4GHz
Gigabyte P67A UD4-B3 @ F3b

Idle means... idle?
Load means running wPrime 1024M.

I've tested all chokes and all caps near CPU socket. There are few chokes that "give" 0.01v more but the results are from other ones. The difference between cap and choke is small but i guess choke vs. MLCC difference should be larger (but have no time to test that).

Enjoy:

LLC Disabled




LLC Enabled

[Bullet92]

Ok I understand now, and that is indeed very perplexing.
Now I'm going to "assume" that the UD5 overvolts the same as the UD4, just for sensor purposes, so we have to go by the software (as I have no DMM), but with LLC disabled, I get a different vdroop than you do, even at lower voltage:

1.25 bios, 1.224 idle, 1.164v load, difference = 0.086 vdrop, 0.06v vdroop, 0.026v bios to idle drop
Yours: 1.35v Bios, 1.320v idle, 1.284v load = 0.066 vdrop, 0.036v vdroop,0.03 bios to idle drop.

Now what's interesting is, your BIOS setting to idle drop performs like it should, but (just going off the sensors), and noting that the higher the vcore, the higher the vdroop on both boards (as expected), then on my board, the vdroop at 1.224v is significantly higher than your vdroop at 1.320v (and that is NOT expected!)

So your board is drooping less than mine, but you're BSOD'ing at voltages that have you stable with LLC on. That would make perfect sense if you were actually getting the same vdroop as me and the measurements were 0.024v too high (then the BSOD is obvious: too low vcore and the 1.284v is actually 1.272v), but since the measurements prove that wrong, then I have absolutely NO idea what's up.

Since Vdroop is supposed to be a certain slope under VRD 12, it seems odd that your board has less vdroop than a UD5. But the BSOD would only make sense if the vdroop were more than what it's showing. But you can't break the laws of physics...so I'm clueless. And I suck at electronics, anyway...

Yeah.. That was weird for me too but those are correct readings from DMM so I've no idea what's up with UD5 if you measure the voltages with DMM. Maybe the SW readings are far from correct on it, don't know, you have to check it with DMM

hey man thanks for testing that

looking good now

No problem man. I'll keep testing it

[sin0822]

I found that with differente LLC settings CPU-Z reports wrong by a certain offset, and that offset is not the same depending on LLC setting. I just wouldn't trust software unless you know the exact offset.

It is fine to measure off teh choke unless you wanna solder a wire toe the back of the board. All the CPU chokes are all connected on the power plane side so they can't read any different. All teh MLCC caps on teh back of the socket, can give you the same reading as a choke, i have a wire soldered to choke and MLCC, i see no difference through all my LLC testing, but its very nice to see someone take teh time and do it too. Great job! I am very happy to see its working well ont eh UD4, now just for teh UD5 i heard its fixed, and the B3-UD7 i shoul have soon and ill test that.

[Falkentyne]

Yeah.. That was weird for me too but those are correct readings from DMM so I've no idea what's up with UD5 if you measure the voltages with DMM. Maybe the SW readings are far from correct on it, don't know, you have to check it with DMM




No problem man. I'll keep testing it

You will have to increase the voltage for both the LLC disabled and enabled test, to get a stability reading (without BSOD) with LLC disabled, but if you don't mind doing this (by calculating what vcore you will need for LLC disabled at load, and then matching the same vcore, but with LLC enabled, with no more than a 0.004 difference, the smoking gun question is:

Are the temps different?
From your results, you should be getting a noticablly higher temp with LLC enabled and 1.308v final, than with LLC disabled and 1.305v (perhaps both will have to be increased to around 1.320v-1.323v).

You should also use Prime small FFT for this test, and not LinX, because LinX runs at different alternating loads, and with LLC disabled, a low load= higher vcore, and we don't want that to mess things up

All we really want to see is if the max temps are going to be different at that vcore point that is arrived at with LLC on and off.
Will definitely be interesting to see...

[Mafio]

finally my UD4-B3 is here, too bad that my 2500k is crap...maxed out near 5300 MHZ



BTW, the biggest problem is that i can't run none of my Hyper kits (both MGH and MNH) with 2133 divider.
if i'm lucky windows boot up, but super-pi 32m instantly freeze.
tried with Kingston 2000 cas8 MGH, SuperTalent 2000 cas8 MGH and Adata 2133X MNH but nothing change.
i read somewhere that high vdimm helps to stabilize but nothing changed in my case, i've tried with every kit up to 2.1 volt.
also tried playing with System agent and VTT voltage but even this time nothing changed.

p.s. dinos, is it possible to have a bios that can handle more than 1.7 volt ?
my chip here scale pretty well even at air with 1.7 volt, would be great to be able to feed it with more and see what happens.
thanks in advance

[Bullet92]

Mafio you're not the only one with Hyper kit problems.
I installed them yesterday and tried to run 2133MHz 8-8-8-24 1T - can bench pifast, 1M, wprime but 32M freezes.
Tried to go lower 7-7-7-21 and 7-8-7-24 1T/2T - the same as cas8. Then I've tried with 6-6-6-23 and 1866MHz and I could bench 32M but with 1.82V what is way too much for that speed/latency.

I've tried with 1.7v to 1.82v for vdimm and 1.08v to 1.24v for VTT for those runs.
Hypers on UD4 = fail.

@Falkentyne: I'll retest it and make a graph with temperatures this time.

[Mafio]

hi bullet, what kind of RAM are you using?

i'm quite sure that dinos some time ago posted some screenshots using UD4 + Hyper and 2133 divider.

[Bullet92]

2*2GB Dominator GT 2000C8 with MGH-E chips

edit:

you mean this one? if yes, this was on UD7 with Cors 2000C7

[Andrea deluxe]

finally my UD4-B3 is here, too bad that my 2500k is crap...maxed out near 5300 MHZ



BTW, the biggest problem is that i can't run none of my Hyper kits (both MGH and MNH) with 2133 divider.
if i'm lucky windows boot up, but super-pi 32m instantly freeze.
tried with Kingston 2000 cas8 MGH, SuperTalent 2000 cas8 MGH and Adata 2133X MNH but nothing change.
i read somewhere that high vdimm helps to stabilize but nothing changed in my case, i've tried with every kit up to 2.1 volt.
also tried playing with System agent and VTT voltage but even this time nothing changed.

p.s. dinos, is it possible to have a bios that can handle more than 1.7 volt ?
my chip here scale pretty well even at air with 1.7 volt, would be great to be able to feed it with more and see what happens.
thanks in advance

hi mafio..

holy.... 1.7v... is too much....

is better to search another 2500k/2600k....

[Mafio]

hi mafio..

holy.... 1.7v... is too much....

is better to search another 2500k/2600k....

hi

it is a crappy cpu, i'm not afraid to kill it

[dinos22]

i think i see the possible reason for your conclusion about hypers Marko, keep testing you will figure it out

[Bullet92]

Just say is it subtimings or voltage related?

[Lsdmeasap]

Hypers are a huge fail at 2133 here as well, even with up to 1.86v!

Dino you have to tell us what the secret is!

[Bullet92]

Guys try to disable Rank Interleaving and see if you can run your Hypers 2133MHz C7. I've tried to get it stable whole day but didn't success.
But now I can run 2133MHz 8-8-8-24 with 1.72v which wasn't doable before Disabling Rank Interleaving. With 7-7-7-21 it finishes 8M and few loops of 32M. I've tried vmem up to 1.82v and vtt from 1.02v to 1.22v.

Say bye bye to my nerves

[Mafio]

this mobo act just like DFI UT X58 and as in that case i think that try to make it stable with Elpida Hyper at high frequency isn't a doable task...
to be honest i'm rather pissed off, i get my first UD4 2 months ago and it arrived DOA, then send it in RMA, waited for more than 1 month and now the new one have alot of problem and act weird with all the DDR3 i've here.
i'm a long time gigabyte fan but this one is by far the crappiest gigabyte i've ever used......

@bullet: already tried 2 or 3 days ago to play with Rank and Channel interleave but here nothing changed...

[Bullet92]

I've tried just one stick in DIMM 2 and it works. But dual is impossible for now.

1st stick


2nd stick

[FL00D]

I noticed a very annoying 'bug'. If I restart my PC and try to enter the BIOS, when I hit the del button my PC would just restart. This is a very weird phenomenon: I hit the del button the PC restarts but powers off during the process and stays powered off for like 5 seconds then it powers on again and everything is fine. This only happens if I attempt to enter the BIOS after restarting the PC. Is this a known issue? Could someone with a P67-UD4-B3 check this for me please?

[Bullet92]

I noticed a very annoying 'bug'. If I restart my PC and try to enter the BIOS, when I hit the del button my PC would just restart. This is a very weird phenomenon: I hit the del button the PC restarts but powers off during the process and stays powered off for like 5 seconds then it powers on again and everything is fine. This only happens if I attempt to enter the BIOS after restarting the PC. Is this a known issue? Could someone with a P67-UD4-B3 check this for me please?

That's correct. It happens if you are hitting delete button before the PC restart. Just wait few seconds after seeing black screen on your monitor and then start hitting delete or better, when you see logo screen, then hit it.

[FL00D]

That's correct. It happens if you are hitting delete button before the PC restart. Just wait few seconds after seeing black screen on your monitor and then start hitting delete or better, when you see logo screen, then hit it.

It doesn't matter whether I press Del before or after the logo screen. I usually press it when the logo screen appears.

[Bullet92]

Which BIOS are you using?

[FL00D]

Which BIOS are you using?

F2. I reckon this issue is BIOS bug. What do you thing? If you have the same BIOS, could you please check this?

[Bullet92]

I have the same mobo as you could see last 2 pages. I did some researches over LLC and now I'm trying to stabilize Hypers on it.
I use beta bios and don't have that kind of bug. In my case, when I'm press "delete" button before the PC actually restarts, it'll shut down. If I press it on the logo screen, it'll enter BIOS normally.

[FL00D]

I have the same mobo as you could see last 2 pages. I did some researches over LLC and now I'm trying to stabilize Hypers on it.
I use beta bios and don't have that kind of bug. In my case, when I'm press "delete" button before the PC actually restarts, it'll shut down. If I press it on the logo screen, it'll enter BIOS normally.

I see. Well, then the new BIOS will solve this issue. I hope it will be released soon... To be honest, I don't really want to mess about with beta BIOSes.