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

[Falkentyne]

Bullet,
Make sure you test at 1.25v before being sure about the choke legs. I saw an earlier thread where they were discussing this, and testing at the wrong choke tests the vcore *before* it goes through the VRM's that apply the vdroop. At 1.60v, with LLC disabled, the real vcore should be between 1.45 to 1.50v at full load (e.g. LinX or Prime95 small FFT). 1.590 with vdroop disabled at load definitely can't be correct. That looks like the vcore before any droop is being applied.

[Bullet92]

Okay, I'll check the measure points with few guys from gigabyte to be sure.. thx

[FL00D]

Falkentyne,

What about CPUZ 1.57? Still doesn't show the correct Vcore?

[Falkentyne]

Not on B3. At least not on GB boards.
Might work on other boards.

HWinfo32 works and shows a lot more information.

[Gaul]

what best value for 5 GH ?

Turbo Power Limit (watt)
Core Current Limit ( amps )

[Falkentyne]

If you don't have "per core" settings enabled (meaning you're using a fixed multiplier and have OS changes disabled), leave it at default, since it's grayed out; it only will use that value at POST time only.

If you're using the per core settings or have OS changes enabled, set it to 200/200.

[irev210]

Basically, the grass isn't greener on the other side.
There are just as many people switching to Gigabyte as switching to Asus. The freeze at idle was driving asus users crazy, and I only saw one GB user who was freezing at idle at stock, and we were trying to troubleshoot it.

Another asus B3 user just had the first PCIE slot completely die, after a few days of use. Asus DID fix the cold boot problems MUCH sooner than GB, though. Although I'm happy with the progress GB has made; I flashed f3c on my B3 UD5, used the clear cmos button after power off, and the board didn't even double post at all. I was quite impressed. It does shut off after certain bios option changes, but I think all the boards do that.

I will say that the boot looping on the GB boards was the most annoying issue ever. That probably drove more people away than the asus idle BSOD's.

If someone gets a defective board, jumping ship may or may not help.

The freeze at idle (on B2 Asus P8P67 Pro and Gigabyte GA-P67A-UD3) is driving me nuts.

Very basic system:

2500K
ga-p67-ud3
hd5770
1600MHz corsair vengeance @ 1333MHz (stock timings) @ 1.5V
Seasonic x-400 PSU

Windows 7 64

Stable under load, crash @ idle, passes all mem tests (tried different ram also)

completely random freezing, it's really frustrating...

[Falkentyne]

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

Profit.

[stasio]

please see my superpi 32m test

Polypenko,
Can you run your Corsair 2133 with 8-9-8-24(CR1)?
Post some pic and SPD.
Thanks

[Bullet92]

Can anyone get me correct measuring points for vCore? (other voltages are also welcome for sure)
stasio?

[C-N]

Can anyone get me correct measuring points for vCore? (other voltages are also welcome for sure)
stasio?

This is from the UD5 but the read point is the same for UD4, there's also another point on the back of the board but it reads just the same as the front.

http://www.xtremesystems.org/forums/...30&postcount=2

[mAlkAv!An]

Don't use inductors for voltage measurement... you can either use the back of the 6.3V 560µF caps or MLCC caps at the back of the CPU socket. The closer to the CPU the better is the measurement.

[Polypenko]

Polypenko,
Can you run your Corsair 2133 with 8-9-8-24(CR1)?
Post some pic and SPD.
Thanks

was not able to boot at 2133mhz with settings at 8-9-8-24

could boot with 8.9.8.24 at 1333 / 1600 and 1866.

[C-N]

Don't use inductors for voltage measurement... you can either use the back of the 6.3V 560µF caps or MLCC caps at the back of the CPU socket. The closer to the CPU the better is the measurement.

Same reading here front or back, what difference are you seeing?

[stasio]

was not able to boot at 2133mhz with settings at 8-9-8-24

could boot with 8.9.8.24 at 1333 / 1600 and 1866.

I am sure you can boot 8.9.8.24 at 2133,maybe with Vdimm 1.68.

Btw,this is SPD what I ask you:

[Bullet92]

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

[Polypenko]

I am sure you can boot 8.9.8.24 at 2133,maybe with Vdimm 1.68.

Btw,this is SPD what I ask you:

here the CPU-Z 157 readings.:

[stasio]

Thanks,
did you try CR1.
Btw,here is grapf what this sticks can do:
2133 6-10-6-24 !

[Polypenko]

Stasio, thanks for the info.

I have not much experience in Overclocking the Cpu and Memory Modules.

Just went into the Bios, but coulds not find at setting CR1 or 2.

could you kindly advise were I can select same in the Bios.

thanks.

[Falkentyne]

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

Ok, these are MUCH better now
In the first test you did, you had LLC disabled giving only 0.010v lower than the BIOS setting (like at 1.6v, result was 1.590v with LLC disabled at load). Now on your new test, it's showing 1.538 and 1.536v much better and more logical now.

So basically, the sensors are underreporting the vcore by as much as 0.035v, and LLC enabled is actually performing EXACTLY as it should. Voltages either right on, or about 0.09v higher than the bios setting.

If that's the case, then, then if the UD5 and UD7 boards "underreport" the voltages in the same way, then holy hell...that means LLC2 is pumping as much as 0.080v more vcore than the BIOS setting, and 0.030v more than CPU-Z/ET6 is showing. And people using B2 UD5's (which had broken LLC1 and useless LLC2) were using LLC1 that was actually using LLC2's voltage scale.

Sounds like no one should be using LLC2 on any board, then.

[stasio]

Stasio, thanks for the info.

I have not much experience in Overclocking the Cpu and Memory Modules.

Just went into the Bios, but coulds not find at setting CR1 or 2.

could you kindly advise were I can select same in the Bios.

thanks.

MB Intelligent Tweaker(M.I.T.)/Advanced Memory Settings/Channel A/B Timing Settings


P67A-UD4 Virtual BIOS

[Bullet92]

Ok, these are MUCH better now
In the first test you did, you had LLC disabled giving only 0.010v lower than the BIOS setting (like at 1.6v, result was 1.590v with LLC disabled at load). Now on your new test, it's showing 1.538 and 1.536v much better and more logical now.

So basically, the sensors are underreporting the vcore by as much as 0.035v, and LLC enabled is actually performing EXACTLY as it should. Voltages either right on, or about 0.09v higher than the bios setting.

If that's the case, then, then if the UD5 and UD7 boards "underreport" the voltages in the same way, then holy hell...that means LLC2 is pumping as much as 0.080v more vcore than the BIOS setting, and 0.030v more than CPU-Z/ET6 is showing. And people using B2 UD5's (which had broken LLC1 and useless LLC2) were using LLC1 that was actually using LLC2's voltage scale.

Sounds like no one should be using LLC2 on any board, then.

Yes, it more logical now. But the thing which is not logical is why I can't run benches on same voltages with LLC Enabled and Disabled when using the same final voltage (after vdrop). But okay..
Also, guys who have done some researches on UD7 have already told LLC2 is dangerous to use in combination with higher voltages because it's overvolting too much then.

[Falkentyne]

So you're saying that the voltages shown in ET6/software correlate more to what benches you can run successfully?

[Bullet92]

This is what I'm saying

[ante_ante]

I have a UD4-B3 on the way, should arrive on monday or tuesday. replacing my old ud3p