Routing CPU power.....help with mapping the right cables!

Hello everyone. I have some spare time and equipment and wanted to try something new: I wanted to isolate all cpu power from the rest of my system and then feed that power using a powerful dedicated powersupply. I want this PSU to be able to handle the CPU with lo load coming from other components. So I ask you: Is there a way I can isolate CPU power cables from the rest?

I figured the 8 pin aux supply cable is specifically dedicated to the CPu but what about the ATX board itself? Does the ATX connector carry any power supplying the CPU?

Is there any way I can go about doing this?

thanks!

no idea? I just need to know how much if any (estimate) goes to the cpu from the atx connector. IS he majority of the supplied power fed thorugh the auxiliary cable, the one close to the cpu socket?

Try measuring resistance between the 8pin EPS12V plug and the 12V pin in the ATX 24pin plug?

thanks, that is a good ide. I don't know why I had not thought about it before.

Do you think this could work? Most psus have a dedicatd 12v line to the cpu and a second for the vga subsystem. I have an estimated power consumption of 280w at 4.00ghz 1.7v with my q6600. My 12v1 line offers 22amps which provides 264w max consumption. Myabe this is limiting me? I get random reboots at those settings.

If I were to unite all 12v lines into one point and then have all necessary cables "spawn" from that point I should be able to have the cpu tap into the second 12v line. Currently my vga is powered with a separate dedicated 500w psu.

are you sure that your CPU is stable at 1.7V? that is suicide voltage for any Q6600 (AFAIK)

I am sure....I can boot easily, get into windows but then I get random reboots and when stressed one of the cores immediately fails. I fear it is a psu issue.

I have already supplied the vga with a dedicated psu but it won't do the trick. I will have to join the separate 12v lines to have margin in terms of power consumption. Anything approoaching 250w (not a big overclock) will stress the current dedicated 22a 12v line.

My guess...that it is not stable...try backing down the voltage and clock slightly?

1.7V on that CPU will fry it soon mate.

Nothing to do with heat even the voltage will surpass what the die can withstand just die.

Back it down is what I recommend

I don't plan on keeping 1.7v 24/7 but is shows that I may have power issues. I will merge all 12v lines (44A @12v). If this does not solve anything then I just hit my wall.

I'm afraid to tell you---you probably hit your wall before you needed 1.7v to hit 4ghz, you were just attempting absolutely insane voltages to even get it to boot
what kind of regimen do you have for testing stability?

When did this thread become about my personal overclock? I was just investigating th possibility of providing cpus with dedicated and stable (i.e fine tuned for stable 12v under load) power supply.

I brought up my personal situation because I thought it may be useful to have a benchmark in case of success. THis may provide some interesting ispiration for anyone out there with 6ghz+ chips but do not wish to buy a 2000w PSU to power the system.

Also by 1.7v I actually mean 1.68v and many actually keep their systems at 1.65 so I really don't see what the fuss is about.

Let me help you real quick, that chip you use? Thats a 65nm part, if you read more around these boards, 99+% of people that run those voltages are using phase/ln2/dice. The simple fact is, those voltages will burn out your processor very quickly.

On to your question. You are simply looking for a method to combine rails on your psu. If you are running it dedicated to your processor, then simply purchase a psu with a large dedicated 12v rail. Why do you think a lot of top benchers utilize PC&P PSU's? They have huge single 12v rails. Either way, I think you can make your life easier by abandoning the isolation of power to the cpu only, and just modifiying the psu itself so that it no longer has split rails.

Now, I am not saying I know how that can be done, but I was just providing my input.

Ok. Apparently I did not express myself correctly. English is infact my second language so I suppose there is a strong possibility of this happening.

a) I have never stated anything about 24/7 usage at 1.7v. I have only said that 1.7v and the impossibility of achieving a stable overclock in addition to the random reboots that I may have incurred in a power limit rather than a cpu + motherboard limit.

b) even ln2 or dice or phase do not affect the way that voltage interacts with hardware. It only helps with cooling which in turn increases stability but as far as I know ln2 or dice have never been labeled as effective ways to overcome voltages related issues. Also those runs are short and intense and as such can see high voltages applied.

c) I thank you for your other half of the reply because it is what I was looking for. I have already done some interesting experiments with combining parallel psus and have achieved almost 100A on the 12v rail. Joining rails is safe as far as I know if minor techinical issues are solved. I use high power common connectors and awg wires rated for 150A to carry all the current "around" in order to increase efficiency and be on the safe side should I venture in multiple tec setups. Plus it looks cool.

I am looking for info on where cpu power is taken from. I assume most is taken from the 12v supply line so any confirmation that 12v rails may add to stability is exactly what I was looking for.

Thanks.

I did some searching for you, hopefully this helps:

http://www.computerpoweruser.com/edi...06%2F36c06.asp

You could have always borrowed a big PSU from a mate... Tested it, if the computer still does its reboot trick,
then I dare say, like everyone else, your CPU is simply not stable at 4ghz, thus your wasting your time with
all this "dedicated CPU power line" bollocks.

Also is that 1.68 BIOS volts? if so, your probably only actually at around 1.63v or less even under stress.
Reality is, not all G0's do 4ghz, Especially as going by your sig, you already need 1.475v to do 3.6ghz.
So even trying this experiment, you will prove nowt.

1) I fixed the vdrop so 1.68 means actually 1.675 under load

2) I dont' actually have a friend with a better PSU

3) MAybe this can help those who don't have $150 to spend on a new PSU when maybe trying something like this might help. Also I am learning about CPU and power supplies. If this annoys you...

Hi Alex! If you have a single rail just try to remove some components and see if your OC improves. Otherwise this may help:
http://www.amazon.com/P3-Internation...3492098&sr=8-1

Quote:

---

Originally Posted by Alexontherocks

b) even ln2 or dice or phase do not affect the way that voltage interacts with hardware. It only helps with cooling which in turn increases stability but as far as I know ln2 or dice have never been labeled as effective ways to overcome voltages related issues. Also those runs are short and intense and as such can see high voltages applied.

c) I thank you for your other half of the reply because it is what I was looking for. I have already done some interesting experiments with combining parallel psus and have achieved almost 100A on the 12v rail. Joining rails is safe as far as I know if minor techinical issues are solved. I use high power common connectors and awg wires rated for 150A to carry all the current "around" in order to increase efficiency and be on the safe side should I venture in multiple tec setups. Plus it looks cool.

I am looking for info on where cpu power is taken from. I assume most is taken from the 12v supply line so any confirmation that 12v rails may add to stability is exactly what I was looking for.

Thanks.

---

A few things. Firstly, your comment about cooling setup (and thus temperature) having no effect on the lifespan of a processor and just affecting stability - this is not correct. One of the dominant causes of shortened lifespan in modern processors (*especially* those with small 45nm processes) is electromigration, whose effects increase as current/voltage increase (higher clockspeed and higher cpu voltage), as well as increasing as temperature increases. Even if a part does not immediately fail from the high voltage at "normal" cooling temperatures, electromigration will kill the chip a whole lot faster than if it's spending its life down in the negative degrees.

Wikipedia's handy factsheet on the problem

And, for the truly curious, Black's Equation has all the gory details about exactly what factors affect lifespan due to electromigration and by how much they affect it.

About your comment in *combining* 12V rails from multiple power supplies - this is actually a bigger can of worms than it may seem. If the circuits are not designed to have their outputs paralleled (and they definitely aren't across multiple PSUs) then slight differences in the output voltages can, especially at light load levels, cause a lot of power to be fed back into the outputs of the PSUs themselves - and that is generally a Very Bad Thing. Avoid this unless you have both an extremely good reason for doing so, and you are absolutely sure of the electrical/electronic factors involved. If you don't fully understand the behaviour of a switched-mode power supply and buck/boost converter, you might want to reconsider recklessly attaching PSU outputs together.

About where the power is derived for a cpu - you are correct in that nearly all the power used in a modern cpu comes from the 12V rail. Additionally, a large chunk of it will definitely come from the 12V power header attached to the board separately from the ATX connector, but whether that power is fully independent of the main ATX 12v lines or not I am not sure, and would have to do some multimeter probing on a board to get an idea of whether they're connected together or not. Considering the bootup behaviour of most boards that have their 12v connectors disconnected - I would think it likely that there is at least some separation between the two 12v sections. No guarantees though (Fancy Stuff may combine them just after boot or somesuch) - so proceed at your own risk.