Voltage Mod

efcoins · Post by **efcoins** » Fri Jun 06, 2003 6:14 am

I have an FSP PSU, but the 5volt rail is a bit low (4.8v at idle).

Is there an adjustment pot that can be tweeked or some other mod to raise the voltage a bit ? The PC works OK and all oter voltages are within 1% of nominal.

I have modded the fan so can not send it back.

dukla2000 · Post by **dukla2000** » Fri Jun 06, 2003 8:40 am

In my FSP300-60BT(12V) there is a pot close to where all the DC wires connect to the board (sort of in the triangle between the DC wires, the nearest heatsink and the AC socket). It effects all voltages simultaneously (or at least +5V & +12V which I was watching on a meter, and 3.3V which I was watching in MBM).

efcoins · Post by **efcoins** » Fri Jun 06, 2003 10:26 am

mmmmm
12v and 3.3v are both 1% high, 5v is 4% low
Suppose I could make a small adjustment

Does anyone know which voltage lines are the most important on an Athlon board ?

DeadBySundown · Post by **DeadBySundown** » Fri Jun 06, 2003 10:31 am

Are you measuring using a digital multimeter, or using something like MBM? MBM is a bit off on some rails on some boards.

dukla2000 · Post by **dukla2000** » Fri Jun 06, 2003 11:02 am

efcoins wrote:Does anyone know which voltage lines are the most important on an Athlon board ?

Well the VCore will be derived from the +5V (unless your mobo is one of the few recent ones that also uses the 4 pin 'P4' power connector as well). And a heavy load on the +5V is what tends to cause the +12V to rise. You may find you can tweak the 5V up a bit without the 12V rising too far. On a typical Athlon setup the main +12V users are all your (hdd & optical) drive motors which are usually +-10% spec (the ATX power spec is +-5%), so having a high +12V (say 13V) is preferable to a low +5V IMHO.

But as per DeadBySundown use a multimeter on a spare molex to check the voltages (before, during & after tweaking) - your mobo sensor & software reporting has unknown accuracy.

Yomat · Post by **Yomat** » Sat Jun 07, 2003 12:45 am

How to adjust one FSP. Might not be the one you have but it should not be that different:
http://www.thenakedreview.com/index.php ... icle&id=71

dukla2000 wrote:Well the VCore will be derived from the +5V (unless your mobo is one of the few recent ones that also uses the 4 pin 'P4' power connector as well). And a heavy load on the +5V is what tends to cause the +12V to rise. You may find you can tweak the 5V up a bit without the 12V rising too far.

This seems to be very true from the behaviour of many boards. However I see on several sites they say that all P4 and all Athlon use 12V. Is this a big missconception? Or does some Athlon boards use 12V? There is 12V on the ATX connector so why would they need that extra 4-pin as P4 use?

For some reason 3.3V and 5V are coupled when it comes to maximum effect on those lines simultaneously. This is all stupid. This means that one might have to use a much higher effect PSU on 5V-using-CPUs than on 12V-CPUs.

You also said once to answer a thread from me that 'cheap PSUs' have this 12V/5V behaviour. Does this mean that better PSUs will not have their 12V increase if 5V is loaded? Or not as much as on a cheap one perhaps.

dukla2000 · Post by **dukla2000** » Sat Jun 07, 2003 1:27 am

It is a misconception that 'all Athlon use 12V'. As you say there is a single +12V wire in the standard 20 pin ATX connector. I haven't looked up my gauges/power stuff (and am no engineer so dont remember

) but the maximum current a single 18 gauge wire can carry is not enough for one of todays 70/80W CPUs. Now if you add that additional 'P4' connector from your PSU (that has 2 more +12V wires) there is no problem. Again I am no Intel junkie, but seems Intel required ALL P4 boards to use 12V for VCore input as well as to have a connector for these extra two 12V wires.

But a quick look at Athlon boards shows very few have the extra 4 pin power block. For example, if you look at the 11 way KT400 review at LegionHardware only 2 (IIRC) of the boards (FIC & Soltek) have the 'P4' power connector. Certainly my next mobo is likely to be for AMD, and definately I will make sure it uses +12V for VCore generation with the extra P4 connector.

I think part of the problem is that psu manufacturers are trying to stick to a single range of psus that can work for either '5V powered CPUs' (most AMD and older Pentium) or '12V powered CPUs' (P4 and a few others) which means they cant be sure whether the main load will come on +5V or +12V and so have to compromise. (The single range makes sense to me as the average Joe buying a psu hasn't a clue what loading he needs on the psu rails anyway!!)

I think (again not an engineer) the root cause is the transformers that derive the DC rails - typically 1 is used for +3.3 and +5, a second winding is used for +12V. On a cheap psu these are probably on the same core (or some other way to reduce cost) whereas a more expensive psu has more independant components which raises the cost. A quick example:

On my FSP 300-60BT with XP2700 the swing between STOPGRNT mode and Prime95 was +5V went 5.08V to 5.03V, +12V went 12.46V to 12.63V. (This was after I had tweaked the 5V up.) The sag on the 5V and rise on the 12V is noticable but only just. Again evidence that FSP make better psus.

On my cheapo psu that I am running the XP2700 on now I haven't done the same experiment, but simply observe my +12V reading under Folding load is 13.1V !!

(So why, I hear you ask, do I use the cheapo? Well, I am more comfortable running it fanless so much quieter. And if t overheats and melts I wont be heartbroken. And I never could silence the PFC choke in the FSP, but it lives on in my children's machine.)

[edit] Just checked my ATX specs, they say maximum load is 6A per pin and recommended wire is 16AWG. But this means the 3 +5V wires in the 20 pin connector can supply 90W which is getting marginal for todays Athlons, and certainly causes those with dual CPUs problems. The single +12V has a spec limit of 72W, but with the 2 extra wires in the P4 connector the ceiling is much higher. [/edit]

efcoins · Post by **efcoins** » Sat Jun 07, 2003 3:41 am

Yes I was relying on MBM, using a multimeter the 5v is actually 4.95 not 4.8, the 12V rail was correct at 12.15v. AOpen SilentTek reports the same voltages as MBM, so the measurement error must be on the motherbord. The only voltage they disagree on is the 3.3 volt line where SilentTec gives 3.47v to MBMs 3.34v

My motherbord does not have the P4 connector.

Vcore is another story the CPU wants 1.65v. If I set the BIOS to 1.65V, MBM measures it as 1.675, so I changed the BIOS to 1.625v producing a measurement of 1.65v in MBM
Which do you think is more accurate, should I believe the BIOS or MBM, presumably I can not measure this with a multimeter.

My FSP350-60-BTP has active PFC, but I can not hear any buzzing from it, the PC as a whole is now very quiet, I have to close the windows to hear it at all even under load.

dukla2000 · Post by **dukla2000** » Sat Jun 07, 2003 4:02 am

efcoins wrote:Which do you think is more accurate, should I believe the BIOS or MBM, presumably I can not measure this with a multimeter.

Well you could measure VCore with a multimeter, except it relies on finding the output of the onboard regulation which is probably one of those power transistor pins. Except if you slip with the multimeter probe and short 2 pins you could have extremely undesirable effects!!

I believe MBM is accurate, or at least it is plausible for me that the actual VCore is higher than what you set with BIOS (or jumpers). On most ASUS boards the feeling is actual VCore is 0.05V higher than it is set. It makes sense for mobo makers to err on the upside - slightly higher VCore is likely to tend toward stability, whereas slightly undervolting could induce instability and get them a bad reputation. (Same argument for car makers to set speedometers that report faster than your actual velocity: it means they cannot land up as a party to the cause of your speeding offences!)

Yomat · Post by **Yomat** » Sat Jun 07, 2003 11:25 pm

efcoins: If you have some health tool for your graphics card you can check 3.3V value there. My mobo measures its own regulation, not what the PSU delivers. Perhaps you have sensors for both.

Thanks yet again dukla2k.

Help on the way to stable voltages. I wonder how to solve this properly. More juice in the PSU would reduce the fall/rise of 5V/12V, you think? Even though the Fortron is better quality, 12.6V seems a little high value for HDs (speced at +/- 5-10%). I preferably go under 12. But I suppose the difference in 5V is relatively much less. So if its possible to run on f.ex. 4.9V the 12V would normalize on a good PSU like your Fortron. That even with high load on 5V that is.

I did some research and it seems that newer PSUs have higher rating on
5V/3.3V effect these days. Mine is 180W and a friends was 170W. They are usually 200W now.

I also did PM MikeC about including documentation of voltage changes under different loads in his upcomming PSU tests. We will see if it sinks in.

efcoins · Post by **efcoins** » Sun Jun 08, 2003 3:09 am

Thanks, most useful

halcyon · Post by **halcyon** » Sun Jun 08, 2003 9:49 am

An interesting addition to the FSP potentiometer voltage adjustment.

I have FSP-30060PN(PF) 300W Fortron made AOpen PSU.

It has two potentiometers inside.

While having BIOS Hardware monitor (shows 3.3, 5 and 12V reading off the MB) and having my Fluke True RMS multimeter connected straight to the 4-pin molex connector of the PSU, I adjusted the potentiometers.

The 3.3V and 12V rails were adjusted by the same potentiometer.

The second one seemed to do nothing, even if I waited a long time for the effects.

So, I could not adjust the 5V rail (MB reading was 4.73 and multimeter reading was 4.98) regardless of what I did to the potentiometer adjustments.

Maybe I did something wrong, but I could not make the 5V rail reading change.

regards,
Halcyon

Yomat · Post by **Yomat** » Mon Jun 09, 2003 3:43 am

Halcyon, man. You're one brave guy. Or you just know what you are doing.

Strangely enough some articles I've read says that you adjust 12V/5V and 3.3V separately. Would be much better if one could adjust the _relationship_ between these ratings instead. One pot for 12V/5V relationship and one for level of all three. Then one could adjust for either a 12V board or a 5V board. Then some tricky vendors can adjust and sell as "AMD optimized PSU".

Btw Halcyon. How much difference do you get in 5V and 12V when PSU is unloaded and with load?

halcyon · Post by **halcyon** » Sun Jun 29, 2003 9:44 pm

Yomat,

sorry this has gone under my radar.

I haven't tested the difference between load and non-load conditions, sorry.

I did open my PSU again and tried to re-adjust the voltages again with the potentiometers (two of them) inside my FSP/Fortron (same as above).

I had again the BIOS readings and my Fluke True RMS digital meter connected (meter connected to PSU 4-pin power molex).

This time I could not make ANY of the readings (3.3, 5 or 12V) move consistently to any direction. I could turn the potentiometers all the way clockwise and wait more than a minute for no effects on the voltage lines. Turning anti-clockwise all the way and nothing.

Weird.

Yes, I'm sure I adjusted them

Yes, I'm sure I had my meter at proper readings and waited for a long time to see the effects.

I'm still running 4.70-4.75 on the +5 V line according to MBM, but the 4-pin Molex output on the +5V line reads almost consistently 4.99V (measured with a multimeter).

regards,
Halcyon

Yomat · Post by **Yomat** » Sun Jul 06, 2003 8:09 am

Sorry for the delay myself. I have been away for a bit.

Could it not be that the PSU tries to go to that voltage but dont have the power to do so? Did you try to adjust without connecting it to the machine?

halcyon · Post by **halcyon** » Mon Jul 07, 2003 12:22 am

No, I didn't try adjusting without connecting to the machine, I'm afraid. I had almost all of my devices (2x HD, 1 CDRW, 3 PCI cards, 1 AGP card, some USB devices) connected.

regards,
Halcyon