P5K + Q6600 Vdroop issues [Archive]

View Full Version : P5K + Q6600 Vdroop issues

start

11-08-07, 06:48 AM

Hey guys,

Both me and my friend have the same PC, Q6600 G0, and Asus P5K (vanilla, I believe).

We're both getting some large (I think) vdroops going on, as well as the vcore set in BIOS being noticably lower than the reported vcore in both BIOS and Windows.

For example, on my PC, at FSB = 355mhz:

Bios is set to 1.45vcore
Bios reports lower than this
CPU-Z reports 1.376-1.384v IDLE.
Then, under load, this drops to 1.296v - 1.320v

On my friend's PC, at the same FSB

Not sure what BIOS is set to (but as with mine, it's set a fair bit higher than the reported voltages)
CPU-Z reports 1.344v IDLE.
Under load, this dropped to 1.256v
Thats 1.344v -> 1.256v, which seems a lot to me.

Can anyone tell me if this is normal for P5K, or what's going on, and what can be done to remedy this problem, if anything?

BTW, neither of us have a digital multimeter. :D

Thanks

bing

11-08-07, 07:11 AM

Those bios readings on the voltages or any other "software" based readings are never accurate because they're not calibrated at the factory at the Super I/O chip, the source of all voltage readings in the mobo.

DMM is the only way. ;)

start

11-08-07, 07:30 AM

Nooooooooooooo :D I have no multimeter and even if I did, I've got this phobia with regards to sticking long metal things onto my computer components.

Regarding voltage reading inaccuracies, I assume that the DIFFERENCES in voltages at idle and load are representative, right? Even if the actual reported voltages are incorrect?

I've been scouring these forums. I see that this is a somewhat well documented (if not very widely understood) phenomenon.

I've seen two suggestions so far, some kind of pencil mod (I'd rather not!), or enabling Voltage dampening. Anything else?

Thanks ;)

start

11-08-07, 07:35 AM

I just checked, and voltage dampening did nothing at all.

Also, I notice my Vcore is actually set to 1.45v (edited the original post), so the reported voltages are even less than they're "meant" to be (I know, DMM :P)

start

11-08-07, 09:05 AM

Triple post:

I increased the BIOS vcore setting to the setting after 1.45 (1.4625 or so?)

I also put the FSB up to 366mhz to hit 3.3ghz.

After about an hour or so of 4x Prime95, seemingly stable, CPU-Z reads 1.296V

That's down from about 1.39v idle, so about 0.1v vdroop. That seems quite bad to me, what do you guys think? (other than DMM)

I've another question:
Why does CoreTemp report 60 degrees avg. under load, and Sandra report 50 degrees?

Sandra also thinks my vcore is marginally higher than CPU-Z does. Why can't they all just agree!

bing

11-08-07, 09:11 AM

Nooooooooooooo :D I have no multimeter and even if I did, I've got this phobia with regards to sticking long metal things onto my computer components.

Yes, it is the fact that probing using DMM probe has it's risk, for example if the tip of the probe shorting two adjacents pins, it might kill the mobo instantly.

I remember there was a thread in Socket 775 section, its about an OCer who killed his mobo because of shorting two pins at the PWM area by the DMM probe. :(

So your phobia is perfectly normal ! :)

Regarding voltage reading inaccuracies, I assume that the DIFFERENCES in voltages at idle and load are representative, right? Even if the actual reported voltages are incorrect?

Yes, you could say that, the delta is "relatively" accurate, but from my own experience, even the cheap 5 bucks DMM is better on the delta. ;)

I've been scouring these forums. I see that this is a somewhat well documented (if not very widely understood) phenomenon.

In order to understand the IO chip, you need at least to understand few electronic's jargon in it's datasheet. ;)

If you want more detail on what your IO chip capable of, look for your Super IO chip datasheet, and once you understand that, you might be surprise on how many mobo makers are ignoring or disabling so many features provided by the IO chip just to save few pennies. :mad:

Btw do you know what is IO chip at your mobo ? CPUZ can show which chip is that at mobo tab under LPCIO field.

I've seen two suggestions so far, some kind of pencil mod (I'd rather not!), or enabling Voltage dampening. Anything else?

Since your mobo is P5K vanilla version, I guess their design or build at the PWM is weaker than more expensive version like P5K premium and such. This is not new, especially for Asus product range that has so many variants.

I just checked, and voltage dampening did nothing at all.

Not sure, but again as stated above, the quad needs high current, and even on "stock" speed, while loaded the PWM needs to supply 115 Amps to the CPU as Intel's spec, it will need much higher current when OC-ed even at the same VCore.

For good OC on quad, better PWM is needed and "most the time" its only available at higher model than the vanilla version.

start

11-08-07, 09:43 AM

Thanks a lot for the replies. :)

My IO chip is Winbond W83627DHG. Does this tell you anything?

When I bought my motherboard, I wasn't really aware of this issue - chalk it up to carelessness. On reflection, I still think I may have chosen this one, as the P5K Deluxe seems to cost about twice as much as my P5K, at least, in the UK. Shame.

bing

11-08-07, 10:03 AM

NP

No, it doesn't tell me anything, just for my curiousity, and my mobo also has the same IO chip.

Btw, if you decided to change mobo and aiming for P5K Deluxe, pick P5K Premium instead, it has better PWM than Deluxe and only cost "abit" more. :)

jason4207

11-08-07, 10:09 AM

Are you using CPU-Z 1.41? I find it's readings the most accurate. Earlier versions just don't read it right.

That being said your numbers don't look too off from other things I've read. You have a good board, but you need to do the vDroop mod if you want to alleviate this problem.

If you aren't willing to do that then you just need to compensate for the vDroop. This might mean setting the BIOS to a much higher voltage than you would think would be safe. What really matters is the voltage under load, so use that to compare your vCore to other users trying to keep it under 1.55v under load. Although, you might want to stay under 1.45v, so that at idle you aren't pushing more than 1.55v. Just watch your temps, and try to keep them under 75* if you can.

I'd trust the latest CoreTemp 0.95.4 temps.

bing

11-08-07, 10:17 AM

Jason, CPUZ does read the voltage at the same source as Asus Probe does which is the mobo Super IO Chip.

Asus Probe uses rounded number, CPUZ doesn't, hence it appears CPUZ "looks" more accurate. :)

start

11-08-07, 11:01 AM

bing

11-08-07, 11:07 AM

When it comes to volt mod if you finally decided to take a risk, I suggest don't use pencil tricks, use a trimpot version.

Its much more reliable and controllable, also most important stability on the mod.

Evilsizer

11-08-07, 12:05 PM

jason4207

11-08-07, 01:33 PM

Jason, CPUZ does read the voltage at the same source as Asus Probe does which is the mobo Super IO Chip.

Asus Probe uses rounded number, CPUZ doesn't, hence it appears CPUZ "looks" more accurate. :)

I don't use probe, so I can't comment on it, but I trust you.

I know CPU-Z 1.41 works pretty well, and that's why it is just easier to recommend that to everyone. While the voltage may not be calibrated perfectly I've yet to hear of anyone have a chip burn up b/c of a bad CPU-Z reading. I have heard of someone burning up their MoBo by using a DMM, though!

Even if you get perfect readings w/ a DMM there is still a bit of uncertainty as who knows how this particular CPU will handle high voltages. For most individuals I think a good recommendation is to use CPU-Z to read voltages, and try to keep it under 1.55v under load as long as temps stay in line. Sure you can get out the DMM (or go buy one), but how will this actually help? I doubt that the temps reported by CPU-Z will be off enough to cause damage to the CPU. One also has to use a little common sense. If your setting 1.8v in the BIOS, and only seeing 1.16v in CPU-Z then I wouldn't be trusting those readings very much. Again, CoreTemp is your friend.

If using a DMM I'd want to solder a lead to the proper point on the MoBo, and attach an alligator clip probe wire to it to avoid any incidents. I don't like sticking anything metal in there w/ the PC on. I like to learn from other's mistakes (GT Engineer!) when I can.

jason4207

11-08-07, 01:38 PM

its normal for the P5K, if you want to try to fix it. you will have to do the Vdroop mod. this though wont fix the issue of voltages not matching up from what you set in bios and what it shows. no board has ever been on the dot to what you set in bios, ie setting 1.45volts but the voltage monitor reports 1.4v.

Yep! 1.5125v set in BIOS, 1.488v at idle (and load thanks to voltage damper). I call this vSkew. It isn't a problem b/c all you need to do is skew your BIOS setting to account for it. Now the vDroop your seeing (0.1v from idle to load) is more of an issue. You can skew your BIOS setting to account for it as well, but then you'll be delivering extraneous voltage while idle.

Iirc the vanilla P5K has the voltage damper option in the BIOS, but it doesn't work. You have to do the mod, but people have very good results after doing so. I recommend using solder and a VR if you have decent soldering skills.

bing

11-08-07, 01:54 PM

If using a DMM I'd want to solder a lead to the proper point on the MoBo, and attach an alligator clip probe wire to it to avoid any incidents. I don't like sticking anything metal in there w/ the PC on. I like to learn from other's mistakes (GT Engineer!) when I can.

Agree, that is much more safer method to do it ! :thup:

I remember it now, it was GTEngineer, thx for reminder, going to search his thread and bookmark it. :D

start

11-08-07, 03:33 PM

Thanks for the advice, I appreciate it.

So I should keep my idle voltage to be no more than 1.55v (and I assume it'll be around 1.45v+ under load).

That should hopefully give me more room, hopefully to get a few hundred more mhz out of it without getting too hot.

If I do find myself getting limited, I can always bring down the multiplier can I not? Say I can't get more than 3.4ghz out of it, then bringing down the multi and bringing up the FSB should give a performance boost, right?

Cheers

Evilsizer

11-08-07, 03:41 PM

lets just say you ram runs stock at pc2-8500. yet your cpu at 9x cant go that high of fsb. you drop the multi to get that fsb up to run the ram at stock yet cpu speed is the same. your system will be faster with the faster fsb yet will still have the same cpu speed. your not just gonna be able to drop the multi for a higher fsb and get a higher cpu speed. its a balancing act between the compents and what they can do.

jason4207

11-08-07, 09:54 PM

Most likely max 24/7 OC will either be 400x9 or 450x8. 3.6GHz either way.

BobcatDan

11-08-07, 10:11 PM

FYI, http://forums.anandtech.com/messageview.aspx?catid=29&threadid=2055486&highlight_key=y

start

11-09-07, 07:42 AM

Vdroop has been causing issues getting to 3.4ghz - 1.408v idle, drops to 1.32v most of the time but some times during torture tests it fell as low as 1.288v, and subsequently tests failed (after about half an hour usually).

I'm now trying with idle vcore of 1.42v. Pretty high, for an fsb of 378mhz (x9 = 3400). At first the vcore was sitting around 1.34 (for me this is quite good, lol). However after a while, it likes to sit around 1.32v (and will probably droop even lower from time to time, but hopefully now it'll stay high enough as to not cause problems).

I noticed something really wierd - when the voltage was drooping heavily before (to say 1.288v) my core temperatures were up about 5 degrees, this happened a number of times, and then tests failed.

EDIT: low and behold, at the higher voltages, I get the same thing - when voltage drops to 1.304v now and again, (from the idle of 1.42v), cpu heats up by about 5 degrees. There is a definite correlation with large vdroop and higher reported temps. WTF?

E.g. test was running fine at 1.32v, then it dropped to 1.288v and core went from 60 to 65 degrees, and when the vcore went back up, the temps restored to their 'normal' 60 mark. This seems counter intuitive - less voltage should mean less heat right??

Thanks for the pencil mod link, I'm just not sure I'm prepared to do this yet. It seems like I'm risking damage not only to my motherboard, but also my CPU (which costs even more money) - as much as I would love to eliminate this very annoying vdroop.

Also, some very sensible comments in the vdroop pencil mod threads argue that this is a bad thing to do, such as:

Doing this seems like a really bad idea to me. Vdroop is used to cut costs - basically, they can use less capacitance on the output of the voltage regulator and still supply "good enough" power to the CPU by simply overvolting it slightly when it's not drawing much current, and undervolting slightly when it's drawing a lot of current.

The naive reaction would be, "well, I'll disable this and always supply the right voltage". A user does that, and verifies with some voltage measuring tools (a multimeter, software, whatever) that the voltages look better. But the user missed what really changed. With active voltage positioning enabled, the transient spikes in the power supply are reduced but the steady-state voltages look worse. With it disabled, the steady-state voltages look better, but transient spikes will be worse (and AFAIK there is no way for software to detect the transients). If the CPU suddenly transitions from idle to load, there will be a brief moment when it is significantly undervolted (and, like any undervolted CPU, may crash / corrupt your data / whatever).

Mark R found this very informative article that explains active voltage positioning / vdroop.

BobcatDan

11-09-07, 09:46 AM

Vdroop is designed into the board by ASUS. Why I have no idea, perhaps because they used it to cut cost. Or perhaps it's to help protect the CPU, you noticed more vdroop with higher temps, could it be that the vdroop was designed to help with CPU temps? The point is, I'm not sure ASUS would allow you to disable it in BIOS (on the deluxe boards) if it provided a serious and necessary voltage protection to the CPU (in terms of power spikes from the PSU). That, and I'm not sure I believe the guy about the power spikes being worse with no vdroop, I'd need a more technical explanation than that to believe it.
Dan

jason4207

11-09-07, 09:58 AM

All I know is many people have a lot of success w/ the mod.

If are unwilling to do it then you can sell/RMA your board and get the Premium, or you can just crank up the voltage to compensate.

BobcatDan

11-09-07, 10:23 AM

Ok, I misunderstood what he wrote to an extent, by power supply he meant (or at least the article he linked meant) the power circuitry on the mobo not the PSU. The basics is when you first load the CPU the input voltage drops momentarily (think of the lights in the kitchen when the fridge turns on), and when you unload the opposite happens. So lets assume some nominal value (like 1.4V vcore) and assume some transient change in voltage, due to the capacitors in the power circuitry, of 0.05V. Then when you load the CPU the vcore will drop to 1.35V, and when you unload it will spike to 1.45V, giving you a total movement of 0.1V. The idea behind Active Voltage Positioning is that if you start the voltage at 1.425V while unloaded, once it loads you'll go down to 1.375V, let it sit there while loaded and when you unload you'll pop back up to 1.425V, halving your total change in vcore.

Seems like a good idea, except the Vdroop on some of these boards is more like a 0.1V vdroop, which kills stability. Since stability is what we are after, and I, having done the pencil mod, can say that I have much better stability at lower temps (since I can set the vcore lower than with vdroop), I would say this is a good mod to do.
Dan

start

11-09-07, 02:08 PM

Thanks, valued input.

I've noticed 0.12v vdroop at times, I'm starting to think that this has to be worse than doing the mod, power spikes or no power spikes.

El<(')>Maxi

11-09-07, 04:08 PM

BobcatDan

11-09-07, 04:13 PM

This board has only 3-phase PWM, very weak for quad. Guys have blown caps on high clocks and as you already know the damper is not working on this board. I would never feel right running a quad on this board.

LOL, damn, well I bought this board for my E6420, so if it goes it goes, just have to get a better one...

start

11-09-07, 05:17 PM

This board has only 3-phase PWM, very weak for quad. Guys have blown caps on high clocks and as you already know the damper is not working on this board. I would never feel right running a quad on this board.

Heh heh, oh dear.

Can it break my cpu?

What does 3-phase mean exactly? How does it work on the premium boards?

BobcatDan

11-09-07, 08:43 PM

Heh heh, oh dear.

Can it break my cpu?

What does 3-phase mean exactly? How does it work on the premium boards?

Well, a few things. First, the posts I've seen (through googling) where caps have failed had cpu's with extreme cooling pushing 4.5GHz at 1.65V+. But then again, while trying to OC at 3.8GHz and 1.53V I heard a distinct humming coming from the case that would start when started the stress test and stop when I stopped it. I'm 99% sure that's a cap (as they are coils of material and when they begin to break down they will vibrate), turn down the OC and the volts and the noise goes away, so I get to stay at 3.6... From what I have read, the q6600 survived the blowout so I'm not too worried about it, just don't want to have to shell out for another good OC board.

3-Phase is a type of power conversion from AC to DC IIRC, I'm no EE so I'm not entirely sure about all the different types of power conversion.
Dan

jason4207

11-09-07, 11:26 PM

3-phase in this case doesn't apply to AC power. The better boards tend to have higher phase counts. You can count the phases on boards by counting the little black cubes (inductors) that have xR0 written on them or look like little copper coils in-cased in a black cube located around the CPU socket.

Most of these newer boards have 3, 4, 6, or 8 phase power. Some boards have 2 inductors per phase so it can be a little confusing. The P35-DQ6 from Gigabyte has 12 inductors but only 6 phases. The Premium has 8-phase power...I think.

The more phases usually means the power is more stable.

Worst case your board will go out. I think your CPU will be fine.

El<(')>Maxi

11-10-07, 01:04 AM

I don't think it's likely to kill the cpu, but anything is possible ;)

3-phase simply means the vcore circut uses only 3 'stages' to create the current used by the chip, less stages means the components are worked harder than on 4, 6 or 8 phase designs.

start

11-10-07, 06:30 AM

How many phases is the P5K Premium?

benscoobert

11-10-07, 08:23 AM

For reference i have a p5k-e and q6600 go, vcore 1.5 reads 1.46 under load in speedfan

Evilsizer

11-10-07, 12:53 PM

How many phases is the P5K Premium?

8-phases on the Prem/DLX boards.......

Dawgdoc

11-10-07, 02:54 PM

My CPU-Z version is 1.41, and my CoreTemp is 0.95.4

I read about the vDroop mod and honestly, I'm not sure it's worth the risk. It seems reckless to randomly add resistance to some component using a pencil! I'm sure it'd probably work fine, but even so...

Thanks for the replies guys.

Well it's stable at 3.3ghz after 3 hours of prime95 on all 4 cores
Voltages according to CPUz:1.4v (ish) idle, 1.30 ish load
Set to 1.46V in BIOS

FWIW.....I have a P5K vanilla on my main folding rig and have performed the Vdroop pencil mod. Been running stable with no problems for....2mo?

Its quick. Its INCREDIBLY easy, and many people have done it successfully without issues.

As Bing has mentioned there are other ways to perform it, but this is so easy so quick so reliable I would personally recommend it. If I can do it, ANYONE can do it.....

P5K Vdroop pencil mod (http://www.blazingpc.com/forum/showthread.php/asus_p5k_vanilla_vdroop_mod-7676/index.html)

Evilsizer

11-10-07, 03:07 PM

FWIW.....I have a P5K vanilla on my main folding rig and have performed the Vdroop pencil mod. Been running stable with no problems for....2mo?

Its quick. Its INCREDIBLY easy, and many people have done it successfully without issues.

As Bing has mentioned there are other ways to perform it, but this is so easy so quick so reliable I would personally recommend it. If I can do it, ANYONE can do it.....

P5K Vdroop pencil mod (http://www.blazingpc.com/forum/showthread.php/asus_p5k_vanilla_vdroop_mod-7676/index.html)
silver pen if you got it or some left over window defrogger kit from pad modding if you got that too.