ASUS P8P67 Pro/Evo/Deluxe OC discussion thread

[Gr1mR34p3r]

@post #429

What I was trying to say in my previous post. If load is stable but idle or off-idle is not increase the offset voltage. If the final load voltage is considerably more then necessary drop down to the next lower LLC level. Doing this may require bumping offset even more to bring the loaded voltage back to where stability is. This won't have a large effect on idle voltages because they increase considerably less then the loaded values per offset step.

Great post, all of this is true, but I have something to add.
If you want the lowest load voltage and still be stable in light load (LinX 1 thread), you are best with LLC Regular.
If you want the lowest light load/idle voltage, then you would have to go with LLC Extreme, but this results in very high load voltage. So LLC Ultra High/Extreme is not so useful in my opinion.
Even LLC High requires around 0.04 (tested on 4.6GHz) more load voltage than LLC regular. And in light load (only 1 thread used) Regular requires only 0.01V more than High.
So by my experience, LLC Regular is the best, everything higher has more disadvantages than advantages.

And now that we know that LLC Regular's weakness is load voltage, best way to test it is Prime95 Small FFTs.
On LLC High, LinX 1 thread proved to be most stressful. Same goes for Ultra High/Extreme.

[|ron]

Just closed a 32M@5,3ghz, this time with ram@1:8 insted of 1:7.
over 2seconds less, even with really bad latencies.
Anyway, at least, I'm under 6.29

[Sevenz]

@post #429

Great post, all of this is true, but I have something to add.
If you want the lowest load voltage and still be stable in light load (LinX 1 thread), you are best with LLC Regular.
If you want the lowest light load/idle voltage, then you would have to go with LLC Extreme, but this results in very high load voltage. So LLC Ultra High/Extreme is not so useful in my opinion.
Even LLC High requires around 0.04 (tested on 4.6GHz) more load voltage than LLC regular. And in light load (only 1 thread used) Regular requires only 0.01V more than High.
So by my experience, LLC Regular is the best, everything higher has more disadvantages than advantages.

And now that we know that LLC Regular's weakness is load voltage, best way to test it is Prime95 Small FFTs.
On LLC High, LinX 1 thread proved to be most stressful. Same goes for Ultra High/Extreme.

By regular you mean medium ?
otherwrise i agree with the whole post.

[Gr1mR34p3r]

No, Regular, as 0%.

[my87csx481]

Sure, re-reading 2. I can see how it would be confusing. Appeared to make sense at the time.

Well it seems currently there is a loophole with BUPDATER 1.22 in that if you replace the older GUID 77840A38-8AB1-4D7A-9181-3B2F5B34C0B module in the older flash firmware with the newer GUID 77840A38-8AB1-4D7A-9181-3B2F5B34C0B module found in the newer flash firmware then BUPDATER is happy to flash the older modified firmware believing it is the newer hardware version. Once that is done it seems EzFlash can be used to update from 0303 to 1253 or say you flashed a modified 1053 BIOS then it could be replaced by the original 1053. Luckily the module sizes are the same (at least for P8P67 EVO), about 2 blocks worth, so it makes the job fairly easy.

I wouldn't say it's recommended but since I also have an external flasher I can recover any fubars. Unfortunately though the external flasher runs of a parallel port and it takes ~2hours to flash the 32Mb Asus BIOS. IMHO it would be best if Asus provided/documented a force flash function so all of this wouldn't be necessary.

I find it strange Asus didn't update the hardware version for the Deluxe BIOS versions so rolling back for them is np. Maybe one of the Asus reps would be kind enough to explain why it's like that.

I know I'm getting in over my head but, what software do you use to mod the BIOS file? (I do write code for a living....for CNC machines.) I'm more of a mechanical person but, my employer also employs many electrical engineers. I'm sure we have something at work that could flash the chip if needed. What is the proper/technical name of the device used for flashing the chip? (so I can correctly word the question to the engineers )

This is all regarding the Pro version of the board. Is this the chip, just above the sata ports:


Pic clipped from the hi-res image available through here.

Thanks.

Sorry for cluttering up the thread. Maybe ASUS will take the hint and provide a fix/workaround for us.

[poke349]

@post #429

Great post, all of this is true, but I have something to add.
If you want the lowest load voltage and still be stable in light load (LinX 1 thread), you are best with LLC Regular.
If you want the lowest light load/idle voltage, then you would have to go with LLC Extreme, but this results in very high load voltage. So LLC Ultra High/Extreme is not so useful in my opinion.
Even LLC High requires around 0.04 (tested on 4.6GHz) more load voltage than LLC regular. And in light load (only 1 thread used) Regular requires only 0.01V more than High.
So by my experience, LLC Regular is the best, everything higher has more disadvantages than advantages.

And now that we know that LLC Regular's weakness is load voltage, best way to test it is Prime95 Small FFTs.
On LLC High, LinX 1 thread proved to be most stressful. Same goes for Ultra High/Extreme.

I've observed this effect enough to come back to this thread looking for answers. And it looks like I'm not alone.


So here's my conclusion on the problem. Feel free to correct me anywhere if I'm wrong.

The problem with the LLC is that the load voltages are different depending on your actual load on the processor.

With LLC on the load voltage is highest when all cores are under load.
Say that you're stable with 1.400v under 4 cores.
But when you reduce the load to say 1 core, LLC will pull back the voltage to like 1.200v. When that happens - instant BSOD. Reason: 1.200v is simply not enough for your current clock.

What you need is to hold 1.400v regardless of load. The only time it should be allowed to go down is when SpeedStep kicks in and actually lowers your frequency.

Over the past week or so, I was quite baffled at why my chip needed 1.400v to hold 4.6GHz stable over a 36 hour (not perfectly paralleled) task when it appeared to be prime and LinX stable at 4.7 GHz.
So I'll be playing around with lower LLC settings and higher voltage offsets.

[cstkl1]

try this
use Ultra High and set ure CPU pwm Frequency to 500

[poke349]

try this
use Ultra High and set ure CPU pwm Frequency to 500

I've tried that approach, but it doesn't seem to like anything over 380.
I haven't tried going as high as 500, but 400 is less stable than 350 and 380.

[Splave]

try manual phase control and put it on ultra , fixed alot of issues for me

[poke349]

try manual phase control and put it on ultra , fixed alot of issues for me

I've tried that, it fixes other unrelated issues. It doesn't fix the fact that loading 1 core results in a lower load voltage than 4 cores.

So if 4 cores is barely stable, 1 core is definitely not gonna be stable because LLC will lower the voltage.
My guess is that (and please correct me if anyone knows I'm wrong because I'm not sure myself), the phase and transient stuff seem to just affect how quickly the voltages respond to changes in load and current draw.
They don't have anything to do with the steady states - and that's what's failing. From what I can see, stock settings already have a bit of LLC.

What I haven't tried yet it to set a static voltage and use it with a bit of LLC. I've been using offset mode the whole time.

[McLaren__F1]

@poke349, can you post your OC template please? Thanks

[poke349]

@poke349, can you post your OC template please? Thanks

Here's a log I made over the past week or so...
This is all on Win7 SP1. So AVX was enabled.

These are all tweaks for my 24/7 OC settings. I intend to get 4.7 GHz prime + LinX stable and run the machine at 4.4 GHz 24/7.
I did some suicide runs at 5 GHz, but I didn't record the settings.

EDIT:
I can easily get 4.7 GHz stable with any settings by upping the vcore significantly, but that runs into the 1.40+ range... Right now I appear to have 4.7 GHz stable at 1.37 peak load.


4.7 GHz - Appears to be 24/7 stable @ 4.6 GHz
CPU Ratio = 47
BIOS Vcore = 1.225
BIOS Offset = 0.060
1-Load Vcore = 1.280
4-Load Vcore = 1.333

Load-Line Calibration = Medium
Current Capability = 100%
Frequency = 380KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

4.9 GHz - Bench Stable, not 24/7 stable
CPU Ratio = 49
Vcore = 1.230

Load-Line Calibration = Ultra High
Current Capability = 120%
Frequency = 350KHz
Phase Control = Extreme
Duty Control = Extreme

================================

4.7 GHz: 4 x 4GB LinX Stable (AVX + 4 threads)
CPU Ratio = 47
Vcore = 1.205 (offset +0.040)

Load-Line Calibration = High
Current Capability = 100%
Frequency = 350KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

4.7 GHz: >5 x 4GB LinX Stable (AVX + 4 threads)
CPU Ratio = 47
Vcore = 1.205 (offset +0.040)

Load-Line Calibration = High
Current Capability = 100%
Frequency = 380KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

4.7 GHz: More than 4 hours LinX + prime stable
4.6 GHz: Fails y-cruncher (100b) 8 hours in
Vcore = BIOS: 1.185, offset +0.020
Load vcore: as high as 1.408

Load-Line Calibration = Ultra High
Current Capability = 100%
Frequency = 350KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

4.7 GHz: More than 4 hours LinX + prime stable
4.6 GHz: Run 1 - Passes y-cruncher (100b) in 36 hours
4.6 GHz: Run 2 - Fails y-cruncher (100b) at 30 hours
4.4 GHz: Passes y-cruncher (250b) in 129 hours
Vcore = BIOS: 1.190, offset +0.025
Load vcore: as high as 1.408

Load-Line Calibration = Ultra High
Current Capability = 100%
Frequency = 350KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

Load-Line Calibration = Regular
Current Capability = 100%
Frequency = 380KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

Freq: BIOS vcore, Load vcore (stable?)

4.7 GHz: 1.240, 1.328 (prime (8) - 10 secs)
4.7 GHz: 1.250, 1.320 - 1.336 (prime (8) - 20 secs)
4.7 GHz: 1.260, 1.328 - 1.336 (prime (8) - > 1 min.)
4.7 GHz: 1.260, 1.344 - 1.376 (LinX (4) - > 3 min.)
4.7 GHz: 1.260, 1.296 - 1.312 (prime (1) - > 1 min.)
4.7 GHz: 1.260, 1.296 - 1.312 (prime (2) - > 1 min.)
4.8 GHz: 1.260, 1.304 - 1.320 (prime (2) - 30 secs.)
4.7 GHz: 1.260, 1.344 - 1.384 (LinX (8) - > 4 min.)
4.7 GHz: 1.260, 1.304 - 1.336 (LinX (1) - > 4 min.)

EDIT: For this last part, if I put a ">" in front of the time, it means I stopped the test manually and there were no errors. I doesn't mean that it failed after that much time.

[some_one]

Is this the chip, just above the sata ports:

That's the one. At least Asus make a bricked BIOS less of an inconvience by placing the chip in a socket. Quite a few boards have use smaller outline chips and solder them in.

I use my own SW for BIOS mods. If all your interested in is replacing the GUID module to flash back to an older version that can probably be done by just using a hex editor to replace the older BIOS module. I take it your using the Pro, I'll have a look at its BIOS later on and check for you.

The SPI chip is serially programmed and read. There seems to be quite a few cheap SPI flash programmers on the web if you google. Not sure how good they are but generally the cheaper USB programmers seem to go for ~$15. I think I even saw one as cheap as ~$6.

Here's a datasheet for a typical SPI chip.

[cstkl1]

lol poke

from ure voltage scaling
u were never stable in the first place

u want stable do this
pass all 3 in Windows Ultimate 64 Sp1

1. Linx with latest linpack lib
101 rounds with min 3500 mb in diagnostic mode
- testing vcore on on/off mode for high and low

2. Prime95 the official release one
min FFT 12, Max FFT 12, 3 hours min 3500mb in diagnostic mode
- testing vcore on stable run

3. Prime95 the latest beta
Use blend custom 3500mb 12 hours min.
- testing everything.

[Gr1mR34p3r]

If I'm already completely stable on 4.8GHz, offset +0.090V (around 1.35V), LLC Regular, everything else Auto/Default, what do I gain by playing with these settings:
-Phase Control
-Duty Control
-CPU Current Capability
Can these settings help me to be stable with lower CPU voltage?

[deaffob]

If I'm already completely stable on 4.8GHz, offset +0.090V (around 1.35V), LLC Regular, everything else Auto/Default, what do I gain by playing with these settings:
-Phase Control
-Duty Control
-CPU Current Capability
Can these settings help me to be stable with lower CPU voltage?

-Phase Control: Controls the response time
-Duty Control: Current feeding based on either temp/setting
-CPU Current Capability: Increases max possible voltage

First two can help you stabilize your CPU but won't help you lower CPU voltage.
CPU Current Capability will do absolutely nothing to help you stablize or lower CPU voltage.

[poke349]

lol poke

from ure voltage scaling
u were never stable in the first place

u want stable do this
pass all 3 in Windows Ultimate 64 Sp1

1. Linx with latest linpack lib
101 rounds with min 3500 mb in diagnostic mode
- testing vcore on on/off mode for high and low

2. Prime95 the official release one
min FFT 12, Max FFT 12, 3 hours min 3500mb in diagnostic mode
- testing vcore on stable run

3. Prime95 the latest beta
Use blend custom 3500mb 12 hours min.
- testing everything.

Those were my logs for borderline stability at 4.7 GHz. I knew from the start that most of those settings were not completely stable. Hence why I mentioned that I mentioned that I'd be running the rig at 4.4 GHz.

From my experience, XX hours of stability under XX test for XX hours at XX GHz doesn't really mean anything.
It's only when you get enough data points to analyze voltage scaling and then appropriately calculate a safety margin to back off do you get something that never fails.

The best example of this are these two that I posted in the log:

4.7 GHz: More than 4 hours LinX + prime stable
4.6 GHz: Fails y-cruncher (100b) 8 hours in
Vcore = BIOS: 1.185, offset +0.020
Load vcore: as high as 1.408

Load-Line Calibration = Ultra High
Current Capability = 100%
Frequency = 350KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

================================

4.7 GHz: More than 4 hours LinX + prime stable
4.6 GHz: Run 1 - Passes y-cruncher (100b) in 36 hours
4.6 GHz: Run 2 - Fails y-cruncher (100b) at 30 hours
4.4 GHz: Passes y-cruncher (250b) in 129 hours
Vcore = BIOS: 1.190, offset +0.025
Load vcore: as high as 1.408

Load-Line Calibration = Ultra High
Current Capability = 100%
Frequency = 350KHz
Phase Control = Manual - Fast
Duty Control = T.Probe

Look at that. 4 hours prime and LinX stable (with AVX)...
And yet it manages to fail a different stability test at a lower speed.

[Sevenz]

No, Regular, as 0%.

Well you're right,not sure why i didnt try it much further before.
my max load vcore on regular is a bit lower than medium and im still stable on linx 1 thread/8 threads.
So yeah llc regular is the best for me too

Thanks.

[Sevenz]

Look at that. 4 hours prime and LinX stable (with AVX)...
And yet it manages to fail a different stability test at a lower speed.

It should work better with lower llc(regular,medium) and higher offset.
Your "light" load vcore will be higher and and your max load vcore will be lower.

[poke349]

It should work better with lower llc(regular,medium) and higher offset.
Your "light" load vcore will be higher and and your max load vcore will be lower.

I mentioned that in one of my earlier posts. So I did a bunch of tests with regular and medium LLC.
Regular is turning out to be best as I can now run the same speed stable, but with peak vcore at 1.35 as opposed to 1.41.

[Gr1mR34p3r]

Testing low voltage, this is what I got with 48x100 DDR3-1600:
VCCSA Voltage: 0.80
VCCIO Voltage: 0.90
CPU PLL Voltage: 1.5125
PCH Voltage: 0.91

BTW, should I run 1600MHz 6-8-6-24 or 2133MHz 8-10-8-27? First requires 1.41V, second 1.61V.

[Sam_oslo]

Is there multiple BIOS on P8P67 Deluxe? In case where can you switch it?

I've been messing around in the case, as usual LOL, just moved around some fans, and then it booted with stock settings, and all my profiles are gone too. I'm wondering if I somehow switched the BIOS.

[my87csx481]

That's the one. At least Asus make a bricked BIOS less of an inconvience by placing the chip in a socket. Quite a few boards have use smaller outline chips and solder them in.

I use my own SW for BIOS mods. If all your interested in is replacing the GUID module to flash back to an older version that can probably be done by just using a hex editor to replace the older BIOS module. I take it your using the Pro, I'll have a look at its BIOS later on and check for you.

The SPI chip is serially programmed and read. There seems to be quite a few cheap SPI flash programmers on the web if you google. Not sure how good they are but generally the cheaper USB programmers seem to go for ~$15. I think I even saw one as cheap as ~$6.

Here's a datasheet for a typical SPI chip.

Thanks some_one. YGPM.

Hi Guys,

got my Asus p67 pro today. What´s the best bios currently so far ?

I'm getting decent 24/7 results with 1204. 1253 is slightly finicky at best.

However, I am not a skilled overclocker by any stretch and never tried 1003 or 1053.

Out of curiosity, where did you purchase your board? They've pretty much been pulled off most retail shelves.

[crossg]

You guy's think this could be Memory related? First try was with High LLC (Tried 4 times with exact same error after second pass) and second try was with Ultra High LLC and offset re-adjusted, had Vcore right up to 1.46 and it still errors.(Phase control Ultra fast) Cpu limit or memory, any ideas?

[orion23]

By regular you mean medium ?
otherwrise i agree with the whole post.

No, Regular, as 0%.

I haven't tried Regular LLC yet, so far I'm sticking to Medium LLC.

For my 24/7 overclock I have: (You guys should know what settings these are...)

4.8ghz
48x100
xtreme
xtreme
130%
Manual @ 350
Offset +.120

Linx loads me @ 1.4.08 V
Prime95 Loads @ 1.386 V

I've tested about 1 hour of each with no issues.
PC is on almost all day doing 3D stuff and games + everything else, really abusing it and no problems yet.

I stick to the 1253 bios because it is the only bios that always reboots for me on the 1st go. Every other bios gives me either a double boot, cold boot, or failed windows load from time to time.

About voltage / cpu scaling, I tried going for 5ghz but my CPU apparently needs 1.56 or 1.57 V to pass linx.

Very surprised that it can do 4.8ghz @ 1.4 and needs so much more for 5ghz.
I guess this could be just a CPU limitation and not just a voltage requirement.
5.1 and up could never load obviously...