ASUS Maximus Formula SE

[ninogui]

You could be right, except none of the other boards I have here exhibit this issue. Once the cpu draws a wee bit over 165a, the "system" process will load up one core. As you breach 165a, the system process will use a full core, preventing Prime from running 100% on all four cores.

Now, as I said earlier, you must read before you post...I clearly stated that my G0 DOES NOT have the same issue at the same voltage. My B3, however, does show the behavior GAR descibed weeks ago...at a LOWER voltage than his cpu does. It's another user here that has the issue at the same voltage.

BTW, before you claim it's the board sample I got, I have 2 boards, and RMA'ed one already. Each of the three showed the same behavior w/ the B3...my G0 does not show this all the way up to 1.725v, real, but @ 1.7375 it does...



My dualcore doesn't cause this at all.

About your "user error" claim...if such was the case, changing cpu but not changing settings would exhibit the same behavior....as all that a user can do is change settings in bios.

However, against your opinion, this is not the case. @ 1.525, 400x9, all other settings on auto besides setting FSB strap to 333mhz, and ram to 2.1v...B3 shows it, B2 E6600 does not, G0 does not.

So if it my error, then what error is this that we(GAR and I) are both making? If you could explain how you came to this conclusion, maybe we could fix the error?


Other than simply blaming it on me, how about some contructive criticism so others here may benefit? Simply stating that you do not experience the problem does not mean there is no issue...nor can you deny the fact that each cpu is different, so how your cpu performs on this board is NOT how everyone's cpu will behave...

Please feel free to look back a few pages at my settings for 450x8, and tell me where I went wrong...maybe it would help others...

There´s an interesting link to see at

http://support.microsoft.com/kb/896256

I wonder how this will affect the throttle up and vdrops. My registry setting on this is set to 1 (on)

Windows messes around the processor a lot, sometimes we just don´t notice it at first sight.

[Jeremyz]

Hey guys and settings to share?

I have a Q6600 G0 with me.

[Leeghoofd]

I wanted to update the bios yesterday to the 0903 version for the Maximus Formula SE, though it always gives Asus signature not found in EZ flash and aborts the flash...

Any ideas ? I'm currently on 0701 bios, maybe a newer version updates the EZ flash too... tried to the 0903 bios from the ftp and official site , same warning pffffffff

[Aldy402]

told ya its not a voltage issue jonny, i knew it
testing with my e6300 and the qx9650 confirms what you and cadaveca said
really dissapointed..

[Zucker2k]

told ya its not a voltage issue jonny, i knew it
testing with my e6300 and the qx9650 confirms what you and cadaveca said
really dissapointed..

It is a voltage issue. Your Q6600 is sucking more power than any of your other chips. If you run the same tests under stock conditions, you wouldn't have that problem. I am currently trying to swap my PSU for the exact same reason.

By the way, P95 small ffts is much more demanding on CPU than any of those apps you listed, except maybe OCCT, with slider adjusted all the way to the left.

Pumping as much as 1.6v to a q6600 is probably bothering on 165watts! Not many 3.3 rails can sustain than in addition to rams, nb, sb, and so on and so forth. But if you believe your hardware is up to the task, then you may consider finding the right combination settings as that could also be the culprit. My vote is on voltage though.

[leoftw]

Has anyone noticed that CPUZ , AISuite , & Everest read different Cpu Voltages ? I don't know what to trust ?

I jacked it up to 1.5v and its says 1.42V . What is lying ?

[kup]

For me, CPU-Z and Everest are EXACTLY the same, both being correct. Latest versions of both.


Prime95 Bug
=========
That patch from M$ seems to help/fix the problem with the "System" resource stealing clock cycles! I've not had it for a while, but I can no longer reproduce it now I've installed that update.

[leoftw]

wtf !

I put in 1.5V in asus and I get low 1.4's in CPUZ and Aisuite

[Nuckin_Futs]

Of course you can't see the bug. It was only described at a voltage above 1.50-1.55v if I do remember. Running the QX9650 at stock vcore won't rule-out the bug. That bug is also seen only on the Maximus as I know, so not really related to the Q6600

I run my E6750 @ 1.57v(1.58v loaded in Prime95 25.5. What bug, I still haven't noticed anything. Can somebody with the so-callled bug explain what they are/were getting in Prime95, and what version and core?

[CERO]

Hey guys I got a Maximus on my way, I want to replace the stock heatpipe for a Thermalright HR05 IFX and for the SB a enzotech low profile copper block. But since I want to mount my TRUE Vertical, should I pick the regular IFX or the SLI? btw I am using a OCZ XTC cooler, but if thats in the way it could be replaced by a simple 80mm fan over it......Any one got this combination and can confirm it will fit?

[ElEctric_EyE]

wtf !

I put in 1.5V in asus and I get low 1.4's in CPUZ and Aisuite

You're not alone. cpu-z and everest seem to agree with each other on cpuV, but they don't agree with what i put in the BIOS. I think the BIOS is compensating for the "actual" value we put in. I would trust 2 apps that agree with each other. If my board was open to measurements, I would measure volts at the pins...

[kup]

You guys are factoring in vDroop, right?

[pphx459]

By the way, P95 small ffts is much more demanding on CPU than any of those apps you listed, except maybe OCCT, with slider adjusted all the way to the left.

You sure about this????

I can run my chip at 1.3v with Small FFT's but at that voltage, blend/custom tests will not complete successfully. I need at least 1.35v.

[leoftw]

You guys are factoring in vDroop, right?

Not me. How could I factor this in ?

[Nuckin_Futs]

Eveest and CPU-Z read bout the same, just that CPU-Z like CMOS reads in the .000 so Everest can only round up or down. CPU-Z has read for me the vdrooped working vcore in CMOS all the time and even responds much faster then most other apps in fluctuation, if any.

[Nuckin_Futs]

Nothing really to factor if you enable Loadline Calibration.

Simply select an estimated voltage in settings , reboot, reenter CMOS and read what effective, working voltages became under "Power" tab.

For those who leave a lot of the mobo voltages in "Auto", you will see how much this board overvolts on its own. mine used a PLL of 1.712 from the start with no major OC (barely calm 400x8). I get 500x8 w/ as little as 1.632v PLL

[kup]

You know you can adjust Everest to read the voltages to the same decimal place as CPU-Z...

[safan80]

BTW, before you claim it's the board sample I got, I have 2 boards, and RMA'ed one already. Each of the three showed the same behavior w/ the B3...my G0 does not show this all the way up to 1.725v, real, but @ 1.7375 it does...


About your "user error" claim...if such was the case, changing cpu but not changing settings would exhibit the same behavior....as all that a user can do is change settings in bios.

However, against your opinion, this is not the case. @ 1.525, 400x9, all other settings on auto besides setting FSB strap to 333mhz, and ram to 2.1v...B3 shows it, B2 E6600 does not, G0 does not.

So, please, don't post such affirmations as "should be user error". You can just propose us to increase priority, which we'll tell you we did make :-) (no bad feelings)

both of you should list your setting and I will try to create the problem.. chances are you guys have different options than me set in the bios.

[ElEctric_EyE]

You guys are factoring in vDroop, right?

vDroop is nullified when loadline calibration is enabled right? I've seen your pencil mod's on this board and I am puzzled why you did this.

[cadaveca]

both of you should list your setting and I will try to create the problem.. chances are you guys have different options than me set in the bios.

I'm sry for this attack, as that is all I can call this post of mine, safan80, however, once again, you fail to read my post completely...my settings have already been posted in this thread.

Again, in the last post, which you just quoted, has yet ANOTHER set of settings that show this behavior...I don't know how to epxlain that everything was set at default other than those things I mentioned any other way...


And really now, do you seriously think I would be so lax in my overclocking to not have already tried EVERY SETTING COMBINATION to eliminate the problem?

Obviously you do, and I must then be aware that you are too quick to judge how I overclock without even knowing what parts I have...you simply state"I don't have it, so it's a non-issue", without first facing the real variable between ou rsystems...the cpu.

NO 2 CPU's will draw the same power. A L737B with 1.25v VID should draw about 75A-80A @ default speeds. Notice I said should...ask Fugger if the two cpu's he got with sequential serial numbers clocked the same....I bet you can already guess his answer...NO!!!. They were close, sure...but not exactly the same.

SO, let's keep cpu in mind....

what's differnt between them? Say myself and GAR...him with G0, me with B3.

OK, so stepping is different.

But what else is different about those cpu's?


How they clock?





NO!!!!

How much power they draw?

YES!!!

So, having spent some time with GAR on this issue, and about 72 straight hours digging into it, I could find no answer to remedey the problem.


I changed each and every cpu setting, tried every bios ever released...could not get rid of the problem over 1.477v real(read with DMM).

So, I went and bought a new cpu...

Popped in new cpu, issue was gone.

cranked up the clocks, and volts...issue remained gone.

Tossed my pot on the cpu, cranked up the voltage to extremes...the problem came back.

New bioses have come out(8&9-series), and not one of them made any difference.


So, as much as you want to blame this on something other than the board...there is no such possibility. Now, what part of the board that causes the issue...EPU most likely...why...well...I dunno.

Other than that, unless you can actually duplicate the issue iwth the cpu's you have, I highly doubt you'll be of any help...it's directly related to power consumption, and some CPU's just simply quit scaling before they ever reach the amperage required to trigger the issue...

but if you'd like to try, here's how to attempt to replicate:

set cpu @ 400x9, 1.525v.

load up prime95 v25.5, run four threads of small FFT.

Open task manager, and verify on the process tab that prime95 is using 99% cpu time.

let prime run for 20 minutes...check task manager again.

If prime is still continuosly loading @ 99%, then up voltage by .025. Repeat prime test, and verify cpu time in task manager process tab. Continue to repeat until you see that prime95 does not get 99%...if just getting close, it will take 20 minutes or so(depending on your cooling, and how long before you reach stable coolant temps) before prime95 will loose just 5% cpu time.

As you get higher in cpu consumption, System process will "steal" more cpu time from prime95...until you get 75% cpu time w/ prime95, and 25% with system process.

I will re-interate...XBX2, P5WDH, P35 Neo2, P5K, 965Dark all do not exhibit this behavior...only Maximus Formula does. I cannot verify whether the non-SE parts do as well...I do not have one of these boards. But for 5 boards to show no problem, but one does, really points towards the board.

[ElEctric_EyE]

I'm sry for this attack, as that is all I can call this post of mine, safan80, however, once again, you fail to read my post completely...my settings have already been posted in this thread.

Again, in the last post, which you just quoted, has yet ANOTHER set of settings that show this behavior...I don't know how to epxlain that everything was set at default other than those things I mentioned any other way...


And really now, do you seriously think I would be so lax in my overclocking to not have already tried EVERY SETTING COMBINATION to eliminate the problem?

Obviously you do, and I must then be aware that you are too quick to judge how I overclock without even knowing what parts I have...you simply state"I don't have it, so it's a non-issue", without first facing the real variable between ou rsystems...the cpu.

NO 2 CPU's will draw the same power. A L737B with 1.25v VID should draw about 75A-80A @ default speeds. Notice I said should...ask Fugger if the two cpu's he got with sequential serial numbers clocked the same....I bet you can already guess his answer...NO!!!. They were close, sure...but not exactly the same.

SO, let's keep cpu in mind....

what's differnt between them? Say myself and GAR...him with G0, me with B3.

OK, so stepping is different.

But what else is different about those cpu's?


How they clock?





NO!!!!

How much power they draw?

YES!!!

So, having spent some time with GAR on this issue, and about 72 straight hours digging into it, I could find no answer to remedey the problem.


I changed each and every cpu setting, tried every bios ever released...could not get rid of the problem over 1.477v real(read with DMM).

So, I went and bought a new cpu...

Popped in new cpu, issue was gone.

cranked up the clocks, and volts...issue remained gone.

Tossed my pot on the cpu, cranked up the voltage to extremes...the problem came back.

New bioses have come out(8&9-series), and not one of them made any difference.


So, as much as you want to blame this on something other than the board...there is no such possibility. Now, what part of the board that causes the issue...EPU most likely...why...well...I dunno.

Other than that, unless you can actually duplicate the issue iwth the cpu's you have, I highly doubt you'll be of any help...it's directly related to power consumption, and some CPU's just simply quit scaling before they ever reach the amperage required to trigger the issue...

but if you'd like to try, here's how to attempt to replicate:

set cpu @ 400x9, 1.525v.

load up prime95 v25.5, run four threads of small FFT.

Open task manager, and verify on the process tab that prime95 is using 99% cpu time.

let prime run for 20 minutes...check task manager again.

If prime is still continuosly loading @ 99%, then up voltage by .025. Repeat prime test, and verify cpu time in task manager process tab. Continue to repeat until you see that prime95 does not get 99%...if just getting close, it will take 20 minutes or so(depending on your cooling, and how long before you reach stable coolant temps) before prime95 will loose just 5% cpu time.

As you get higher in cpu consumption, System process will "steal" more cpu time from prime95...until you get 75% cpu time w/ prime95, and 25% with system process.

I will re-interate...XBX2, P5WDH, P35 Neo2, P5K, 965Dark all do not exhibit this behavior...only Maximus Formula does. I cannot verify whether the non-SE parts do as well...I do not have one of these boards. But for 5 boards to show no problem, but one does, really points towards the board.

System process is controlled by software isn't it? So the real problem here is that the loading percentage is not being reported correctly? I've noticed in OCCT, during loading with no OC'ing it has 100% load. As I've increased all the settings to OC, the reported load decreased. I originally just thought it was some sort of auto-protection for the CPU. Maybe it's built into the BIOS. If I built this board and wanted a small amount of RMA's, I might put in some simple formula to auto monitor the CPU core temps.... Now how the software like OCCT and system process and lower the load had got to be some kind of shared temp register maybe between the BIOS and software?... Just a shot in the dark here...

[NLight95]

Here's a review comparing the most known NB coolers

http://mickaeln2006.free.fr/index.ph...1&limitstart=2

The ES II s**ks. In that test, the HR-05 and Noctua are passive, while the ES II is active. There are reviews showing also an improvement in the IFX model and furthermore if you add a fan for active cooling

The HR-05 IFX will be better than stock. The ES II, I can't say. But, on thermalright site, only the IFX version is shown to be compatible with X38 chipset, sadely

In that review the ES II is shows a 10C improvement over the stock heat sink and it's quite a bit smaller than the HR-05, so I wouldn't say it sucks. The ES II actually worked better than the HR-05 on my ICFX3200. If anything sucks, it's that I'd have to purchase the IFX because TR didn't design the HR-05's mounting mechanism right in the first place.

[cadaveca]

System process is controlled by software isn't it? So the real problem here is that the loading percentage is not being reported correctly? I've noticed in OCCT, during loading with no OC'ing it has 100% load. As I've increased all the settings to OC, the reported load decreased. I originally just thought it was some sort of auto-protection for the CPU. Maybe it's built into the BIOS. If I built this board and wanted a small amount of RMA's, I might put in some simple formula to auto monitor the CPU core temps.... Now how the software like OCCT and system process and lower the load had got to be some kind of shared temp register maybe between the BIOS and software?... Just a shot in the dark here...

well it seems as a certain amperage threshold is reached, second part of CPU PWM phases should kick in, It seems as though it is(vdroop becomes vboost), but due to Prime95 load style, the ACPI driver responsible for this action is either not properly exiting, or is failing to load properly, and in either instance, creates an endless loop. As the threads bounce around the cpu getting executed, the load will vary slightly, and so does the cputime that the system process steals, making me beleive that you may be very correct in your assumption.


I came to the same conclusion long ago, but wondered what would make this board so different in it's ACPI function that it showed this behavior while others did not...and I could only come up with the "EPU" as the culprit. I lack the knowledge to understand whether it is a failed command from EPU, a halt state triggered by EPU, an unknown register not programmed in bios...could really be many things.

I do understand, however, that there will not be a fix, other than a new board or cpu...cpu is a crap shoot...I got lucky...

This problem does also affect performance...3dmark cpu scores @ 3.6ghz are low on cpu that causes issue compared to same cpu with same settings on other boards...then verified by G0 cpu that does not show this behavior...you can relate teh loss to about 220 points on '06 cpu tests @ 3.75ghz.


I now have that B3 running on MSI Neo2 @ 3.75ghz, 1.55v, 24/7, without problems. NO prime load issue, nor anything else...but that B3 does not like Maximus Formula.

Neither did Gar's G0...and we were very close to the same clockspeeds that trigger the problem...i thought at first it was cpu speed at fault, however when i lowered speeds and upped voltage the behavior remained, again, another indicator pointing towards current drawn by cpu as teh true culprit...of course, combined with EPU implementation.

[ElEctric_EyE]

Just FYI, when I was on 0701 and tried to load EPU drivers, my keyboard would lock up. After a total reinstall of vista, several times, step by step and typing on the keyboard with notepad, I finally narrowed it down to EPU drivers... Since then I don't load them, who needs EPU anyway. It's some power saving BS. OC's don't need that sh*t. I say open all the lanes up and pour that current in!... B3=hot=current leakage=inferior process.

[cadaveca]

well, you must undersatnd that by default the board only uses half-phases, not full, so possibly without driver, you can never get proper load-balance. But not to fear...I have not isntalled drivers from ASUS on most recent board...just simply used Intel's .inf's for chipset and SB, as ASUS driver CD's have been lacking for eons... even just using Intel drivers, I still have the problem. It could simply be something caused by X38 driver...I did not see this problem when I played with X38 and DDR3 tho(was not an asus board tho, so no EPU..)


To be honest, dealing with the problem by buying new cpu, or getting different board, is the only solution. It's a low-level driver issue caused by hardware implementation...change the hardware, and the driver is not a problem.

PWM = Pulse width modulation....meaning the voltage supplied to cpu from PSU is not constant..it's pulsed. It could be possible that due to how current is "shared" bewteen phases, and how this is regulated that is the issue...EPU could NOT be the problem, but maybe problem lay in PWM, and reading given back to EPU by PWM created teh loops...i dunno. I lack the expertise in this area...too many SMDs in circuit that do not have proper documentation for the public.