Do you get better OC with higher 5V/12V rails ?

[GreenBeret]

My rig is in my sig and I have an Antec TruePower 430W. It's giving me almost exact 5V and 12V under load. I wonder if I had higher 5V and 12V rails, would I be able to OC my CPU or RAM a bit more ? I've read some reviews and threads on the TTGI (Atlas in Australia) PSU and it seems that that PSU provides much nicer rails than the Antec. If you've upgraded to that PSU, have you got better OC (either with Intel or AMD) than your previous PSU ?

:)

[TheDude]

My response is "yes'. However note that I had low +5 and +12 numbers before I increased them and I only increased them to slightly higher than where you already are. I use the TTGI 520w PS with AMD systems. I got higher clock and more stability. This was quite a while ago and I don't have specific numbers for you. I guess a reply from someone who is running higher +5 and +12 rails than I am is needed to fully answer your question.:D

[Krizalid]

I doubt it, as long as the voltages are within spec. With some of the other PSU's you can tweak the rails, but I bet they don't have the consistency of the True Power ones (with PC Power & Cooling as the exception). In other words, I'm sure the voltages drop significantly when under heavy load. I have a True430 PSU and the only way I can make any of the rails budge is when I overclock my Ti4200 to 327/567 and run Prime95, my 3.3v rail will go from 3.29 to 3.28. :D

[Wolf]

You must have gotten a golden sample. On my true power 430w and 8k3a+ combo my cpu voltages jump from 2.7-2.1v, 3.3v goes from 3.23-3.28 , 5v line jumps from 4.89-4.95, 12v line is from 11.73-12.04. I think you will get a better overclock from a better psu but don't expect a miracle. As long as my 5v line is close to spec and my 12v line stays pretty steady I don't worry about it too much.

[Krizalid]

What are you measuring the voltages with? I used to use Motherboard Monitor for that, but it reported the voltages wrong on my 7VRXP. It reported the temps correct though. If that's not it, maybe you got a bad batch. I heard the earlier shipments of the TruePowers had some problems.

[JBELL]

voltages:
12v @ 12.416
5v @ 5.16
3.3v @ 3.376
vcore @ 1.648

PSU:
Enermax EG651P-VE FMA (STOCK!! no POT modifications)


I have never had a more stable rig....

[]JR[]

You want to get your 5 and 12v lines (theyre tied together which is unfortunate) as close to 5 and 12 as possible too high out of spec 12.2/5.2 is as bad as 11.8/4.8 in reality. Its the way the mobo has to regulate and deal with the voltage post psu, and you want it to have to do as little work as possible.

In my ideal world id throw the atx psu in the bin and run 12v and 5v from high quality ripple free independant psus... although there always seems to be more chips to buy... :p

EDIT : as already stated, you cant use the bios or mbm to read the voltage either its about as reliable as using grandma as a sherpa when climbing everest...

]JR[

[RacerX27]

First off there are not many power supplies that are better then the True Power Series. FACT!, esp if you understand switching power supplies. The True Power Series units do not have feed back pots because that are design to hold 5v and 12 dead on idle or under full load.

The single most important characteristic of a power supply is stable, clean DC voltages. This is what the True Power Series provides.

Second, every winbond chip monitoring IC's are pure junk. From wild temp variations on the same model board to lousy reporting of supply voltages. You MUST use a digital volt meter to read the power supply voltages. A digitial doc is far more accurate then the onboard monitoring functions. In fact the digital doc I had was with .01vdc of a certifed calibrated multimeter.
Radio Shack has many nice multimeters and there are some fine ones from Fluke and many others. A meter with a peak hold or min max funtion will tell you how stable the DC is over a period of time.
On my TP550 the 12v line was 11.99V and NEVER moved underload. The 5V line was 5.04V and never moved underload.

ANYONE that says you MUST have more then 12v and 5v to overclock needs to go to school. In some cases more 5v is needed because the majority of motherboards use the 5v line to create the CPU regulated voltages. If the CPU voltage regulator is poor underload more voltage will help. It helps because as more the CPU voltage increase the pulse width of the mosfets increases to create a higher voltage on the filter capacitors. This is where a 3 phase is better then a 2 phase the 3 phase does not "work as hard." It would be more accurate to say 2 channel and 3 channel voltage regulator.

I am an eleven year accomplished electronics tech. I trouble shoot and repair component level electronics in aircraft.

Motherboard monitor and the like are good for reference but SHOULD NOT be the final measurement device to fail a power supply. You must use a digital voltmeter for accurate measurement.

Most of you will notice if you track powersupply measurements in MBM or other, if you read 5.08v and 5.15v. You will never see 5.12v in a measurement. The winbond chip has a poor resolution of more then .08 VDC. A good multimeter, heck even a cheap POS multimeter will have .01 VDC resolution. The winbond jumps around alot, and is not accurate. I saw the winbond reading move when measuring with a min/max recording function. The true reading never changed and the BIOS readings were all over the place.

This is not my humble opinion this is fact!

[RacerX27]

]JR[, the True series has indepentent 12v and 5v supplies. :)

[RacerX27]

http://www.antec-inc.com/pdf/truepower_inclusion.pdf

[]JR[]

Nice post :) I knew tied 5/12v lines adjustable by a factory set glyptol'd pot were poo, my sparkle doesnt do too bad tho...

Might invest in an antec and retire the sparkle, its had a hard life :p

]JR[

[RacerX27]

]JR[, there is nothing wrong with the 'old' style power supplies. There are millions of them in service. The True Power Series is a new mouse trap, stronger and more stable.

Before the Antecs all I bought was Sparkles. Good power supplies are not cheap.



In order of importance for quality parts.
1. Memory, never skimp on memory.
2. Power supply, this is the heart of a system, dont be cheap.
3. the rest of it just about anything goes really. quality is better then price IMO. I dont look for the cheapest items, I look for the best I can get for the money I have to spend. :)

[charlie]

RacerX,
Excellent info! I've noticed my temp measurements on MBM work well, but the voltage jumps and hovers ! Now the BIOS voltage measurements, better? More accurate? I used the BIOS info WHILE I setup MBM5, for example under idle the MBM said my cpu was 69C...BIOS said 27C, changed sensor choice...27C
Now voltage, my Vcore in BIOS runs 1.50-1.51 ish, MBM hovers 1.46-1.52! BTW to measure PSU rails, you simply insert probes into 3 and 4 pin connectors, right?
C
BTW, my TTGI runs a bit weak at 12v of 11.76-11.80 under load. Rock solid 5.03-5.08 under load 5V though. Will probably give a small twist of the ole' pot (after I get a multimeter) thanks Racerx

[RacerX27]

Any 4 pin connector will do. Be very carefull not to short out the leads when measuring.
Yellow, black, black, red; 12v, ground, ground, 5v. Respectively. A digital doc or similar is the easiest way to measure the lines. If you gunna spend a few $'s on a meter, docs are around $40.

http://www.newegg.com/app/viewproduct.asp?DEPA=&submit=Go&description=digital+doc5

[sjohnson]

RecerX, agree with you in principle - but the two True Power PSU's I tried to work over for friends would pull down to 4.9 on the 5 volt rail under load. Measured with an older Fluke DMM, but a good one that calibrates nicely against reference voltage supplied by a HAM radio bud of mine.

I was asked to see what could be done about instability at high FSB/multiplier ratios on these (ABIT) boards, a KR7A and a KX7.

Putting in a Sparkle 350 and tweaking the 5 V to pull down to 5.03 under load lead to increased stability at the previous FSB/mults and lead to a higher overclock for the boards.

What I guess I'm trying to say is, the Antec True Power PSU's look like great design but for the two I dealt with the implementation was lacking.

I know this was an extremely limited exposure to the Antec PSU's, but I now don't trust what is promised versus what is shipped.

[STEvil]

I think I need to get a new PSU..

This Enermax 431 just about isnt cutting it.. at 1.8v and 1700mhz.. but could be board too.. 8-)

Cant imagine what my 5v will look like with 2.15v through the CPU.. lol! 8-)

[majormav]

would have thought so however ive yet to find a unmodded psu that can hold ALL the rails up on big overclocks on load even the big branded reccomended ones here even fall short in my own experiences

[GreenBeret]

Thanks for the info, RacerX :) I have a question regarding measuring the voltages with a multimeter:

Originally posted by RacerX
Any 4 pin connector will do

I think when measuring the volts from the 4 pin connector, it will only tell you what the PSU is providing to the multimeter itself, not what the motherboard is getting. Am I right ? I"ve heard that you should measure directly from the reading points on the mobo. Just that I don't know which points :p

[Svenn]

Using an Enermax 550W PSU and Digital Doc 5 is giving me 5.0-5.25 and 12.4-12.5... Is that bad? And is it a consensus that the TruePower Series is one of the better PSUs?

[RacerX27]

The motherboard connector as far as 12v and 5v line will be the same as the motherboard atx connector.

http://www.hioctaneracing.com/images/mainatxconnector.jpg

The orange wires will be the 3.3v line. There are many points that can be measured on the motherboard it self, its different for every board. There is no margin for error when using probes, it is very easy to slip and short 2 pins together on the small surface mount ICs. And it can be fatal.

[RacerX27]

Originally posted by Svenn
Using an Enermax 550W PSU and Digital Doc 5 is giving me 5.0-5.25 and 12.4-12.5... Is that bad? And is it a consensus that the TruePower Series is one of the better PSUs?

No there is nothing wrong with your power supply. The older style power supplies like the Enermax changes voltage underload, and that is normal. And it is not bad.

The Digital Doc is very accurate, and I am curious what are MBM or BIOS, or Sandra measurements compared to the Doc.

[Svenn]

4.89 and 12.16 from MBM while Digital Doc says 5.02 and 12.46...

[RacerX27]

The Doc is correct MBM is reporting what it reads but is wrong.

Now we need to ask the author of MBM to add compensation to the voltage readings like the temp reading has.

:)

The Doc is correct, looking good.

[Svenn]

12.46 is okay? Not too high?

[RacerX27]

Originally posted by Svenn
12.46 is okay? Not too high?

12vdc@5% (I think the ATX standard is 5% for 12v)
11.4-12.6vdc is ok, 12.46 is fine.

If its 3% then 12.36v would be max and .10v over is still fine. It runs Hdds CDroms and fans mostly.

[sjohnson]

From http://www.formfactors.org/developer/specs/atx/atx2_1.pdf

[STEvil]

Racer, what if the traces/resistors/capacitors/etc are creating resistance and the 5.05v that the digidoc/dmm is reading at the atx connector/4 pin molex connector is not the same as what the CPU/etc is being fed?

This is what the whole 5v/12v (asus) mosfet mod was about, was getting the 5v past the restrictions in the traces to the mosfets... Im thinking of doing this to my new mobo when it arrives, but with some caps in the way to smooth the voltage... 8-)

Im not saying your wrong, but maybe they arent as far off as it seems. Some CPU's also seem to draw more current for some reason. I'll use my 1600+ and 1800+ as examples...

AGKGA Y XP1600+ @ 1595 1.88v - 5v steady at 4.97v
AGKGA 4 XP1800+ @ 1666 1.81v - 5v unstable at 4.86-4.93v

The same enermax 430 was used in each case, and all the components were the same... and this really makes me mad cause I can get into windows at 1750mhz with this 1800+, but its not stable.. 5v drops to 4.7 8-(

[RacerX27]

The Atx connector and the 4 pin connector voltages are the same point electrically. What the board see at the atx connector is the same at the 4 pin.

Traces, resisters and such will affect the voltages as they pass thru the board. As current (I) increases, voltage will drop for any given resistance. Basic OHMs law V=I*R Volts, I(current), Resistance.

Part of the reason some boards can only (an example)'supports XP2000+' and down is because the board cannot handle the current demand of a faster CPU. The board is designed to the Xp2000+ current(I) demands and cannot supply any more with causing problems. There are other factors that I am not aware of and this is one possible example.

Yes some ppl have jumped around the board traces to create a second path for voltage, this takes the load off the board and increase the stability of the supply line. But there is a limit, based on OHMs law.

I would only be guessing here but good overclocking CPUs draw less current at a given speed allowing them scale higher. I say guessing because, measuring current draw is not available on most boards. I have not had a board that could but its coming someday to everyday boards. New steppings are almost always cooler overall. Cooler means it draws less current, and has a lower internal resistance overall. OHMs law for power.

P=V*I (uh oh its has a square or something). My memory is fading. I dont use power conversions much at all. I dont use the formula enough anymore and school was 12 years ago. LOL. Anyways power is drived from Voltage and current, which is all greatly affected by resistance.


The main problem with increasing the 12v and 5v at the 4 pin connectors, is that the HDDs and other components, (fans hardly matter in this case) have an increased out of spec supply voltage. Harddrives get hot enough, and this may be one contributing facter to HDD deaths. Running out of spec has consiquences(sp).

CPUs are like lightbulbs, complicated and expensive lightbulbs. Some use more current, or high wattage then others. Each CPU is unigue. Thats why out of a batch of the same speed and stepping some do better then others.

When the load increases on a supply line, the load causes in some cases the DC supply to ripple, and ripple is bad for a CPU. Ripple is bad on any supply line when a stable DC is needed. Your enermax 430 is underpowered in the case for overclocking. You prove that in your testing by seeing the supply line drop. A higher quality powersupply would hold up better.

end verbose mode

[felix88]

P = V * I just like you said earlier. i believe the current in newer cpus is about the same as in older ones. you have to remember that voltage has gone down quite a bit as well.

[STEvil]

Volts x Amps = Watts... I think thats the one your looking for..

The CPU needs a certain ammount of power, and when one drops the other must compensate..

I guess this is why my board is only rated to an XP2000+.. hehe 8-)

Gonna have to swap it for the Epox 8KTA3 in my room now.. T-Bred A's work in KT133A boards, sweet.. 8-)

Also, I think its more the mb than my PSU in this case.. the restriction at 1666mhz+ is getting a bit much for the power... But I still want a True Power PSU.. 8-)