Asus P5B Vanilla (non dlx!) vcore+dimm needed

[Thasp]

I'm embarrased to say, I don't know where the ground is.

I too have found asus to be accurate. A DFI board said the 12v rail was 11.68, the asus said 12.21. My multimeter said for the 12v rail on that PSU, 12.21!

I will definitely try this tonight. Thanks again for the assistance, I know it must be annoying answering this question so many times.

[ziddey]

There's grounds everywhere. Black cable would be ground. Metal lands where you screw the board to the case would be ground. Case itself would be ground.

I wonder about the dfi's readings though. I'm thinking that it might be reading it deeper in the circuit somewhere, and it's already taken a hit in terms of voltage drop due to resistance in the circuit leading up to where its measured. So, that said, I wonder if that's actually what many components are seeing, or if the poor conduct is due only for the measuring system.

[Thasp]

It was an $85 board that gave me a 1.05 GHz overclock, so I wasn't that critical of it at the time. I'm going to miss that DAGF, the last one I have is being sold this week.

Does it have to be a specific coil by the CPU or anything? This is a caseless system, so one to the coil and one to the casing of the PSU can work I'm guessing.

I'll try this later on the CPU I'm not using and report back later tonight. :]

[ziddey]

I believe the link I provided also has vcore read points for you. You can just put ground to the dots where you'd put a screw down. Those are ground. Or stick it in a black socket of a molex.

[Thasp]

This is weird, measuring from the black on the molex to that point, or any other ground I can think of, gives me 0.02v.. could it possibly be one of the other coils that I should be using? I tried any coil with the red wire and the PSU, ground on the board, or ground on the molex connector with the black and I get nothing. I tested the multimeter again on the PSU and it reports all the rails right.

[Thasp]

ok, I think I got it. I didn't measure vcore with the multi.. before I was getting 1.216 at 1.235v with 50K ohm resistance, with 3.7-5.6(I can't keep my hand still long enough to get a perfect reading), I get 1.224v at 1.235v and 1.392v at 1.4v on a dual, I will test with the quad soon as well. I just wanted to make sure it worked on the spare before the quad. I am going to see if the heat of the CPU agrees with the voltages I am setting before I put the quad in.

Thanks again for all of your help, now that I know how to do this I will be doing it much more often, and if I get more experienced maybe help some other people in the process.

I have 2.2-2.3v rated memory, and a board that does 2.1v max. I guess time to mess with the memory next.

I have a VR that goes from 0 to 10.1Kohms, is this good enough for a vdimm mod?

[Thasp]

I just spent about an hour and a half on the vdimm mod, and no dice. I have never had to solder something into that small a space before, and there is no way I can do that with my shaky hand. I guess I will have to buy a lower voltage set of memory.

[ziddey]

If you want to see if you can find a vdimm pencil mod, here's how:

Check for continuity between leg6 and the nearby resistors on each side. When you find one where there's continuity, check the other side of the resistor with ground. If it also has continuity, measure the resistance of the resistor. Then measure resistance between leg6 and ground. If they match up, you know to shade that with pencil.

[Thasp]

By continuity, do you mean one that measures with equal voltage to leg6?

I guess another one of the issues is I have doubt in my ability to hold a multimeter on leg 6 while the board is running without it hitting leg 8 or 4.. or the chip next to it. I have a very unsteady hand, any solder work I do on audio stuff takes me ages and it's much larger and easier work. If I short leg 6 to 8 or 4, am I looking at a harmless shutdown or a small fire/cpu burn?

[ziddey]

I mean 0 resistance. continuous

turn the board off. you can't measure resistance when there's current flowing anyway

[Thasp]

Ah now I get it. Continuity is the thing where I set it to beep when there's full current allowed through and it beeps. Thanks a lot - I will get around to this, but my new 2.1v ocz already shipped.. this memory will go into the ds3 dual box so not much need to test now except for curiousity.

[ziddey]

If you can indeed find and confirm a vdimm pencil mod for 1.03g p5b (which I'm asssuming you have), I'm sure the community would appreciate it. Besides, there's nothing really wrong with setting baseline to 2.0v anyway. That'd mean you'd only have to set 1.9v for 2.1v anyway. It'd be good practice to figure it out too

[Thasp]

If I do what you suggested above, is my risk 90% limited to killing just the memory or something else? Killing a cheap stick of memory is fine, but I'd rather keep the CPU that's in there now going. If there's risk to that I'll throw a P4 in there for testing.

Also, ridiculously stupid question, but when I touch the multimeter to ground then to any measure point when it's set to 20vdc(what I use to check faulty PSUs), I get 0.02v.. this is checking cpu vcore, memory vdimm.. I put one end to ground on molex connector and the other to the coil specified in that guide you posted, and I get nada. I've moved it around, and messed with different grounds, but I always get that as a result.

I am not sure what revision my board is, I've looked and I can't find it. It looks the exact same as the one on that website that's not all in english and looks different from the pictures you've posted. I'll take some pictures if I make any findings on this.

[ziddey]

just reset your bios and measure your vdimm initially. then, you can power on without even any ram after you d o the mod and verify again. that way you can see changed vdimm without having to even risk any ram.

[Thasp]

Yeah, that's the problem. Measuring vdimm gets me 0.02v, so I must be doing something wrong. I followed the instructions so I am not sure what, I can measure other voltages from the PSU or other components, but the coils on the board to the ground molex connector get me nothing.

Even weirder. I measured the cap to the pin for MCH as showed in the diagram for MCH, and I get 0.33v... this isn't anywhere near correct, is it? The options were always 1.55, 1.65, 1.75, in my p5w dh.

[ziddey]

the inductor has a covering on it. maybe you're not actually hitting the wire but the plastic instead. or at least not making good contact.

or, try the other side of the coil.

as for vmch, what if you use a black ground wire instead of the top of the cap

[Thasp]

You're a genius. :] now I get 1.35v.. WAY TOO LOW! Time to get to soldering I suppose.

btw, I found two things with continuity to pin 6.. I will take a picture later. But it's not like the vdroop mod. These are next to each other opposed to under each other. With the vdroop mod you were drawing straight across the resistor but here there is plain board inbetween the two things with continuity.. I couldn't find anything else with continuity. A picture will explain this better once I get a chance to take it.

http://www.wittekakker.be/projects/p5b/vmch1.jpg

In there, for the pin 9, is that the second to bottom pin on the right side that he has highlighted to solder to ground?

Also, would a 10K VR be that bad? All I have left is the thing that came with my tuniq tower, that goes from 4-10000 ohms.

Thanks again.

edit: nevermind, I looked at it again closely, there's another set of pins for another chip right next to it, and that pin has pins on both sides, there's no way in hell I could ever solder that without it shorting to everything.

[ziddey]

there may be a pencil mod if you probe with the same concept for vmch as well. there can be multiple resistors with continuity to leg6, but only one should also have continuity to ground on the other end. I believe the other one is used for feedback, based on the limited knowledge I have, and since shading that one will cause a drop in vdimm

[Thasp]

I got the MCH one to work

I am proud of this. I know it's simple and lame, but my soldering skills suck, this iron was free and it has a tip half the size of my thumb.

Unfortunately, even at 1.65v vs the old 1.35v I see little difference in my ability to overclock.

Since I have ram that works at 2.1, I did the vmch, and the vdroop(thanks to your help), all I have left to do now is find out why I can't set a multiplier and have it stick and I'm done.

[ziddey]

Whoa there.

Where are you at now and what are you trying for? 1.65v is a lot of voltage to be pushing, and certainly requires a lot better cooling of the northbridge. Most people suggest no more than 1.55v or so, and even then, getting rid of the heat is quite important.

What do you mean by getting the multiplier to stick? Is it falling to 6x at idle? Not really a concern. Just speedstep/c1e doing its job, but if you disable those, it should stay at what you set. Unless of course you're talking about something entirely different.

Maybe you need more vdimm to push higher?

[Thasp]

Whoa there.

Where are you at now and what are you trying for? 1.65v is a lot of voltage to be pushing, and certainly requires a lot better cooling of the northbridge. Most people suggest no more than 1.55v or so, and even then, getting rid of the heat is quite important.

What do you mean by getting the multiplier to stick? Is it falling to 6x at idle? Not really a concern. Just speedstep/c1e doing its job, but if you disable those, it should stay at what you set. Unless of course you're talking about something entirely different.

Maybe you need more vdimm to push higher?

I tried lower memory speeds, it's not much of a ram issue anymore. It's stock PC8500 5-5-5-15 2.1v and I had it with those timings and voltage under ddr900.

I've been stable for hours of stress testing at 9x385, 1.525v. I can do 1.6v and stay under 50c with my cooling but since 1.6v is actual 1.6v without the vdroop I am reluctant to do that to my CPU - I am not sure how kentsfield is with voltages yet.

I have a waterblock on the chipset, CPU, and GPU, a 3x120mm fan swiftech radiator that is mounted on an A/C unit so the water going to everything is pretty cold and a swiftech rebranded laing d5 pump so everything is nice and chilly.

If I set 6x, I boot into the bios, and I still see 3.6 GHz at 400 FSB. Obviously it's still at 9x. If I boot into windows at 8x450, it's ridiculously unstable, so I know it's doing it at 9x. I want to be able to set 8x in the BIOS and boot into windows at 8x450. My gigabyte had these stupid FSB holes and if this has them too I'd like to see if I can get around them.

[ziddey]

are you getting these values with cpuz? windows and a host of other programs will always use the default multiplier in speed calculations. For example, when I run my e6400 at 7x500, windows shows 4ghz, while it's actually 3.5ghz.

[Thasp]

I don't trust CPU-Z.

I checked the BIOS after I set it. It says the multplier is 8, but it says over 4 GHz, which is obviously wrong for 8x450. The lack of any stability after booting supports this theory that it's using 9x, because while the memory at that speed and the cpu at that speed have not been fully tested for stability, there's more than 1 second of stability for sure.

[ziddey]

ok...

what bios are you using again? I've never had issues with my p5b vanilla reporting incorrectly, but on my friend's dlx-wifi, we had great issues with random frequencies/multis until we flashed up to the new bios.

[Thasp]

Whatever the latest BIOS was as of Sunday when I flashed it.

The new BIOS is worse, maybe? Before I flashed it off the september BIOS, the temperature the asus tool reported was actually close to what I got in coretemp.. now it reports bull like 16c at full load. Do you think I should try reverting to an earlier BIOS?