Fluctuating voltage on A8N-SLI Premium

[Kaine]

Hey folks, I'm having some strange issues with my ASUS A8N-SLI Premium. I run an Opteron 170 in it, overclocked to ~2.4ghz. Now Opterons being Opterons, they generally overclock well. I've seen this one run @ 2.6ghz on air, however i struggle to get past 2.4ghz on watercooling, even running up to 1.5v (lots) Over 2.4ghz stability becomes extremely erratic. sometimes it won't boot, other times it will run for a few hours or a few minutes. I've removed RAM from the equation, i run top of the range OCZ Platinum 1gb sticks and i've relaxed the timings tinkered with the dividers to little success.

CPUZ's "Voltage" indicator on the CPU tab shows the cpu voltage fluctuating between 1.52v and 1.48v... it doesn't sound like much but i'm positive that can't be healthy. The bios shows similar fluctuations. I run it on an OCZ Modstream, which are generally rock solid, consistently producing good clean power. Compare that to this machine, which runs an A64 3700+ also at 2.4ghz and the CPU voltage is dead still at 1.35v. This is with an Antec 430W (which has been great). The opteron also runs a fair bit hotter than it really should - no doubt because of the 1.5v i'm pumping into it to keep it stable, although from what i've read here if there is much ripple in the supply it will cause the chip to run hotter than it would otherwise.

Does anyone think i might have some dying caps in there? could replacing them with something lower ESR and better quality stabilize the CPU voltage? I can't see any bulging or leaking caps, but we all know that doesn't mean they're not there. there seems to be a mix of brands, some Chemi-con KZG, others i can't identify without taking the board out (major pain in the ass with watercooling)

anyone had a similar experience or have any experience with these boards?

[kikkoman]

Re: Fluctuating voltage on A8N-SLI Premium

Be aware that the entire VRM circuitry normally gets at least a minimal airflow with conventional air cooling. You wouldn't be the first guy that fried his board this way.

Depending on how long you've been using the board, it's well possible that the caps have suffered overheating. As you say, they don't necessarily bulge or leak.
You might consider adding an extra fan (doesn't have to move much air) for the VRM, too.

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

the VRMs are actually heatsinked on this board; having said that i agree entirely that the VRMs might be a little cooked. The heatsink is really not adequate. The heatpipe setup on this board really makes me wonder whether it actually does any good at all. The heatpipe runs from the chipset just behind & below the vid card up to the VRMS which have a ~12mmx50mm aluminium heatsink on them. I have tried to get as much airflow as possible up there but it's hard in a restrictive case. When/if I do the recap i'm strongly considering setting up watercooling on the chipset and possibly the VRMs too. To top it off, some of the caps are almost hard up against that scorching alu heatsink.

Ooh i forgot to mention that the board in the 3700+ system is an ABit AX8.

[kikkoman]

Re: Fluctuating voltage on A8N-SLI Premium

yeah, just had a look at a few pictures and all I can say is 'WHAT THE HELL?'
If there's no airflow in the VRM area, one part heats up the other (whichever is hotter).
The engineers at ASUS seem to have been struck by the heatpipe plague too, i guess. I wonder when they will sell cables with heatpipes.

The (active) cooling of the non-premium version makes much more sense. A big passive heatsink would be possible, too, as the chipset sits perfectly between the two SLI slots.

My biggest concern isn't the MOSFETs, but the caps. Watercooling the VRMs wouldn't help as much as an extra fan I think since the caps do not only take some heat from the PCB but generate quite a lot themselves.

[starfury1]

Re: Fluctuating voltage on A8N-SLI Premium

Really an oc noob here, so take what I say with that in mind

As kikkoman has said the caps would I think most defiantly be getting a workout.
if you have been running it in an oc state for fair while they could well be suspects.

(general comment)
If it worked ok in that setup from new then that to me would be a further pointer.

As always heat is the enemy.

The heatsink thing I am not so sure about either, might be fine for bog standard operation but doubt very much its really up to the task in OC case
extra air flow would be needed...anyway you are well aware of that by the sounds of it.

Caps are to put it simply, a part that will ware out and degrade overtime so maybe treat them like a grease and oil change and change them every so often in the Case of OCing
but i'll leave that comment open to those that do OC

To me, in this case "premium" Low ESR caps would be what to look for, whats best suited to these Conditions I dont know.
Also possibly taking the "total capacitance"of the VRM output up is an idea really, and possibly a larger can size
Thats if you can get a larger uf value cap to fit in those positions.

I am not so sure I subscribe to the idea of "maxing out" the VRM capacitance wise as its has certain design criteria that must be met but some extra capacitance I think won't hurt especially in the case of OCing
(as its been stated here manufactures probably use the minimum they can get away with anyway...not sure with Asus thought since they direct their Mb at OCer's)

If I was into heavy OCing, looking at ways to battle harden the VRM would be my first step
I think a system stands or falls on how well the VRM can deliver the current demands required.
Its basic design principal is to deliver huge current demands (swings) in very short order of time while keeping the voltage regulation very tight
(a tall order for any psu...it is at the end of the day that)

Anyway thats my thoughts on it and thats if the caps are the issue.

There are a few that oc round here so they would be best to advise.

Cheers

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

ok so i dropped the overclock back to stock (2ghz) and the voltage instability has disappeared. now rock solid at 1.344v. Temps have dropped considerably (i also have a 70mm AMD fan pointed directly at the VRM heatsink. I think i'll be keeping it as-is for a while.

[acstech]

Re: Fluctuating voltage on A8N-SLI Premium

Sorry I'm late to the party!

Anyway, I've got an Asus A8N-E (similar board w/o SLI) and the voltage will vary between 1.408v and 1.424v in CPU-Z. IIRC, I have 1.3875v set in the BIOS. This is with the system in my sig.

Restarting computer to check voltage in BIOS... Edit: Yep, 1.3875v.

IIRC, my board uses a lot of Panasonic FL series caps. I think it's FL, might have been FJ. One of those. Anyway, I doubt the problem is in the caps.

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

well that's comforting in that it tells me that kind of fluctuation is symptomatic of the board design. Doesn't mean i'm happy to settle for it. The recap shall go ahead at some point in the future. I have some sound issues i hope to remedy anyway. I'll try to get my hands on a multi that can measure capacitance and i'll be sure to post a report on the whole affair. Worst case scenario (bar killing the board) is that i'll get some practice in. I have access to a replacement anyway. I'm not so attached to the board that killing it would cause me any great discomfort.

could i bother you with what bios settings you use to achieve your overclock? where and how do you adjust the RAM divider?

[starfury1]

Re: Fluctuating voltage on A8N-SLI Premium

actually meant to get back to this one....like I said no ocer here and mostly never bother to look at the voltage things....I did notice that this box Asus running a C2D 6750 the voltage goes up and down (cant recall what they call it in the bios) I thought it might be just that I know this cpu is suppose to change depending on demand..(well think thats it)...anyway it did make stop and think...I don't really know if its set at a fixed voltage

Anyway yeah the Vrm does rely on the CPU telling it what it needs

But regardless of that if you are overclocking then a recap with decent caps isn't going to hurt and adding some more capacitance want go astray I think.
but remember I have no real practical on over clocking

Be interested in what its says then......also when people mention voltage I tend to think multimeter...so I tend not to think monitoring SW or Bios monitoring.

BTW capacitance ranges on an MM are useful
(although a lot don't seem to make it into the 1000's of uf)

but in this case your real concern is ESR

which basically means how well (or badly) the cap will act as a series resistor.
The higher this value the more ripple and heat that will be generated.
A capacitance meter wont show you this.
Only an ESR Meter.

[kikkoman]

Re: Fluctuating voltage on A8N-SLI Premium

SW/bios voltage monitoring is shit anyway. I was a bit upset when my bios showed 2.75V on the 3.3V rail (and so did Speedfan/Sandra etc) although everything was running stable. A quick look at the DMM told me that everything was fine, though.
Same thing occured with the -12V rail, which showed -7.49V in the bios once.

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

SW/bios voltage monitoring is shit anyway. I was a bit upset when my bios showed 2.75V on the 3.3V rail (and so did Speedfan/Sandra etc) although everything was running stable. A quick look at the DMM told me that everything was fine, though.
Same thing occured with the -12V rail, which showed -7.49V in the bios once.

my -12v rail shows something like -7.49 constantly. it's normal.

Turns out the voltage isn't stable at 1.344, it's still fluctuating, just less noticably. I understand what you're saying about voltage monitors not being the most accurate things but even if the voltages are off, i can't really see dodgy monitoring practices somehow showing me a ripple when there isn't one.

In other news; my best mate's dad has an awesome prototyping board he picked up a while back for sale. it has a sine wave/square wave generator, AC & DC variable voltage supply and even a basic oscilliscope. He's willing to let it go for $150. I think i'm going to grab it- should be able to set up a basic ESR tester without too much trouble. it's already got the major parts i need. would be great for testing & prototyping audio circuits too. i'm stoked!

[Spacedye69]

Re: Fluctuating voltage on A8N-SLI Premium

I thought ASUS sold (or at least recommended) a small fan setup for the heat pipe fins if you choose to water cool. You MUST have air flow over the evaporator (finned section) or the heat transfer slows or comes to a complete stop. Heat pipes don't cool anything. They move the heat from one place to another. You still have to remove the heat.

[starfury1]

Re: Fluctuating voltage on A8N-SLI Premium

They do I believe have a heatsink fan as you said spacedye69, not sure which models in came supplied with and yeah think it is available as a separate part.

Didn't get one with mine, although its mentioned in the manual
(it doesn't come with the MB I got Asus P5K deluxe wifi.)
Couldn't find it in the site thought.

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

i didn't get one with mine either but yeah, it's mentioned in the manual. From memory they say it's only "required" when using watercooling (because there's no CPU fan to blow air over the heatsink). I haven't been able to find them for sale separately anywhere- besides that i'm not sure i could even fit it in there, i have a radiator in the way.

[Kaine]

Re: Fluctuating voltage on A8N-SLI Premium

Quick update.

I now run at 2.6ghz stable with a ram divider that puts my ram at about ddr440. Attached is a screenshot of Speedfan. I just happened to switch over and have a look at the Vcore graph when my cpu finished a Folding@Home work unit. So what you're seeing is the difference between the CPU under load (left of the graph) and the CPU at idle (right of the graph)

pretty dramatic eh? I actually had the system booting at 2.75ghz but it was unstable. 2.8ghz posted but wouldn't boot regardless of the voltage. My temps are great at 2.6ghz, 30C idle, 40C under load @ 22C ambient. Even when it hit 30C yesterday temps were barely tipping 50C.

My voltage in the bios is set to 1.4v, so i have a Vdroop of .01V which is stuff all really. I think i'll recap my power supply on my next major system overhall along with checking out what power management circuitry the A8N-SLI Premium has and investigating if there's room for improvement. I'll most definitely be changing the cooling setup, probably a waterblock on the NForce 4 Ultra chipset and some sort of active cooling on the VRMs. They would probably even benefit simply from the removing of the silly heatpipe system on this board that basically carries the heat from the chipset and dumps it on the CPU power circuitry.

[gonzo0815]

Re: Fluctuating voltage on A8N-SLI Premium

The biggest share of heat generated by the VRM MOSFET will be sinked by the motherboards PCB.

To put a HS on top of the well insulating plastic package is probably more an pr thing then anything else.
Sure, it will add some cooling, but i think this is just not that much on the small MOSFET case.

From my point of view it is inevitable to cool the PCB and the components (inductors, caps, mosfet and surrounding copper pads) with some airflow, not much, but it have to flow directly over the VRM area.

Especially those heat pipes or heat lanes are a very stupid design, without active air cooling of the fins.
Any Zalman NB32K would probably sink more heat in passive mode than those little fins.

If the caps are operating in high ambient temperature, the endurance will be not very high, considering the rated life of 2000h @105°c.
2000h is not much for an everyday running system.
Considering a 8h running time per day, a 365d /y would account for about 2920h.

Considering the design of the VRM, i think this board would be predestined for an recapping with solid polymer capacitors like Oscon SEPC or Samxon X con 1000uf there are even some unpopulated silkscreen markers, which could allow to do this without loosing any bulk capacitance (if they are really connected to the V core).