Q9450 OC / Temps / Settings

[JohnnyV]

Not many posts here so I'm just putting down my next (and probably final) score:

[Slovnaft]

Ooh I'd forgotten about this thread.
I'd better post my 3.9ghz 1.29 vcore stable before I RMA this board.
I'll do that tonight, if anyone is still reading.

Oh and, Johnny, try burntest there. I think you'll see some...surprising...temps.
I'd bump down the vcore if i were you.

[JohnnyV]

Yeah Vcore is a bit too high for my liking but temps under stress test progs like prime and OCCT are acceptable.
Think I may need to increase the VTT to reduce the Vcore right?

[Slovnaft]

I haven't touched VTT on my P5Q3.
i do have 1.6 vNB though.
how much have you toyed with vcore? I bet you could go lower =]

[ReLiK]

Is the CPU V-Core that CPU-Z reports accurate?

I ask because on CPU-z my Q9550 @ 3.4Ghz (8.5*400) shows a voltage of only 1.152v. That sounds really low to me. Can the CPU-z voltage be trusted?

[SNiiPE_DoGG]

Is the CPU V-Core that CPU-Z reports accurate?

I ask because on CPU-z my Q9550 @ 3.4Ghz (8.5*400) shows a voltage of only 1.152v. That sounds really low to me. Can the CPU-z voltage be trusted?

as long as u are using the latest version, yes

there are some mobo's where the voltages are a little screwey though, not sure about the blitz...

[JohnnyV]

I think it's pretty accurate... do you have an E0 Q9550 Relik?

@slovnaft.... I had prime95 give an error twice after about 3 hours so I increased CPU PLL, VTT and Vcore a bit and now it was stable. Could it be that by increasing vNB I could do with less Vcore?

[ReLiK]

Actually not using the blitz anymore, gonna sell that. I am on a Gigabyte GA-EP45-DS3R right now. I think your right about the voltages. I have it set at 1.25 in BIOS but it is reported 1.152 in CPU-Z latest version.

Anything I can do to fix the discrepency?

Johnny - It is actually C1. Nice chip though.

[SNiiPE_DoGG]

using load line calibration? turned off C1E and throttling options? on my board I have VID and VID special add, I do not know about the gigabyte bios though...

[ReLiK]

Don't think the board has Load Line Calibration. C1E and all the throttling options are off. I'll have to check on the VID Special option. Pretty limited on this board I think though.

[Slovnaft]

that sounds about right.
CPUZ is generally accurate (with the exception of BlackOops..lol...) at volts
I've never used anything else.
And that kinda voltage sounds OK and pretty good for that CPU.

[Slovnaft]

@Johnny

Yes.

and maybe it's me, but I don't like playing with PLLs and VTT, just like i don't like playing with gtl refs. I don't entirely understand them and they seem to be the less flexible and more dangerous voltage settings.

[ReLiK]

Thanks for the help Slovnaft! Any idea how to get the board Voltage more accurate with the drop in windows?

[Slovnaft]

hard volt mods to the board.
what you're experiencing is vdrop. It meanns that the BIOS is setting one voltage, but because of resistance and inefficiencies in board design your CPU is really experiencing another.
When this differential is constant it is called vdrop, when it gets larger with cpu load it is called vdroop. many boards have Load Line Calibration which attempts to compensate for vdrop or vdroop by estimating it and delivering higher voltages than specified at load, attempting to maintain the set BIOS voltage.
In essence, CPUZ is registering correctly, your board just has vdrop. What you can do is set the voltage a touch higher than the voltage you want. As long as you're not shooting for huge volts, you should be fine.
If you've got a soldering iron handy and don't mind taking a risk, you could look around for a hard vdrop mod; usually soldering a variable resistor here or there, to remove the discrepancy.
But these are board-specific and depending on the amount of attention your board gets from the community there may not be any solutions for your board.
what I like to do is decide the max voltage you're willing to put through your cpu for the kind of clock you're trying for (24/7, benchable, suicide run), set that, and baseall your OCing off that voltage, reach your max clock, and undervolt later. basically accepting vdrop as a limitation of your board and not risking setting voltage higher than you may want, as it can spike up in transitions from idle to load or vv.

[ReLiK]

Great, thanks for that info!

I might as well leave it alone for now. It is completely stable at that voltage and clock.

I don't even think I can take it any higher. I have my Ram at 400 which is what it is rated and I'm not sure how well it can overclock. Im sure the chip has quite a bit more in it though.

[ReLiK]

http://www.newegg.com/Product/Produc...82E16820145194

That is the Ram I have, here is some more info on the chip they use TWIN2X4096/CM2X2048-6400C4DHX XMS640X v1.X PSC 128MX8

You think these have any room to overclock at all?

[cliffy]

I have a 10 Degree difference at Idle between cores1 and 2, should i be sending this back for a replacement?

[Slovnaft]

by 1 and 2 do you mean 0 and 1 as in the first two cores or the middle cores?
if you mean the first two cores then this is normal. If you mean the middle cores then, well, i mean it's not terrible either. you could send it back as an excuse for a batch update, get a new chip take a chance at being higher binned, but it doesn't really matter.
reseat your hsf if you're really concerned, and you could probably try lower voltage to get your temps down.

[cliffy]

yeah i mean the middle cores, tried reseating made no difference don't know that i can kower vcore much cause i am at stock. This is my last upgrade for a while so want everything just right Think its going back

[Slovnaft]

eh why not.
get a new chip, roll the dice again. i love doing that.
stock volts for 45nm is like, 1.1 i believe. at least, thats what my P5Q3 set it to.

[JohnnyV]

with what are you measuring? When I installed realtemp the first time I had a 10 degree difference as well but found out that the tjmax for 2 cores was set differently from the other two.

[cliffy]

coretemp, realtemp and everest

[Slovnaft]

and btw tjmax for Q9450 is actually 100 not 95.

[mikey976]

ok i just got a used q9450 and popped her in, now for some oddball reason its idling 33-34-22-31

ive reseated it twice benched it, and stressed it and all 4 cores are working but for the life of me i dont understand why the temps are funky. btw i just upgraded from an e6600

i am using a storm g4 block. i think that might be where my problem lies cause i am noticing cores 3-4 temps are solid at idle where as cores 1-2 are fluxuating up and down by 2-3c
should i invest in a dtek with quad core insert? or are there any better quad friendly blocks. i havent OC'ed yet because i want to get this temps thing cleared up before i cook my new chip

[Slovnaft]

People need to read up on the diodes on these chips
they are not meant to give accurate idle temps their primary purpose is to provide accurate load temps and trigger thermal throttling.

Your cores recieve unequal load and thus have varying temperatures.
As long as none of your temps are crazy under load, don't worry about discrepancies, especially in the cooler direction.