X3350 / Q9450 Overclocking Thread

[aigomorla]

Originally posted by: Foxery
Ok, Q9300 (Xeon equiv) came in today, and so far my thoughts are... WTF?

Temperatures make no sense

Perfectly stable at 3.2ghz (428 FSB, 1.25volts) but as you can see, it should have burst into flames if RealTemp is to be believed. CoreTemp, of course, reads even higher. And yet I'm sitting here writing this while Prime95 cranks away, and listening to MP3s to boot...

YOUR TEMPS ARE INSANE.

I dont know how your cpu is not shutting off when your EXCEEDING the TjMax.

You have 2 choices as of this point.

1. Get a new cooler
2. Inspect to see if its the thermal probe inside thats malfunctioning.

But as of right now YOUR CHIP IS NOT SAFE

 

[Foxery]

The motherboard is supposed to auto-shutdown at 85C, and the thing doesn't even feel hot. I was feeling suicidal, so I turned it up farther. You'll love this...

Nonsense, take 2

Larger pic of my whole screen. A few interesting points:
1. FSB 450x7.5 = 3375 MHz. Speedstep/C1E are turned OFF in the BIOS, but look at CPUz.
2. Now look at Prime95. Cores 0+2 are actually finishing work slower than cores 1+3!
3. And again with temperatures that can't possibly be correct, because I'm still typing and it hasn't exploded.

aigomorla: I'm not normally suicidal, but something is clearly not correct here. For reference, this IS using the stock cooler. It's half the size of what came with my E6750 and twice as loud. I had hoped not to spend more money on a real HSF, but perhaps I will.

First, though, I'd like to know why I've apparently acheived the impossible...

 

[runawayprisoner]

1) Faulty BIOS... so reported FSB is being messed up, obviously. Happens to me from time to time with the X3350.
2) Don't think you're safe. The reported temperatures from CoreTemp are from the sensors that are... on the CPU itself. This should mean there is a problem with the IHS, and that means...? Yep... you've exceeded TJMax.
3) Core 0 + 2 are not keeping up... after 10 minutes. I smell thermal throttling. If the system doesn't turn itself off automatically in 30 minutes then chances are... Intel Thermal Monitor is also off. That means sooner or later, your CPU would be... toasted.

And when the problem lies with the IHS, you won't feel the burn when you touch the die... at all!

 

[Foxery]

Ok, I should really start at the beginning

Idle, stock speed, stock voltage, SpeedStep ON

Prime95 at stock speed, stock voltage

Intel gives 71C as the max safe operating temp, and mine claims 94C at all stock settings. My thermal readouts are just plain borked.

 

[runawayprisoner]

That's why there's a problem with your processor. Sincerely, those sensors are inside the CPU. About right under the IHS, to be more precise. They are monitoring the temperature inside.
And by the way, I had my X3350 at 60C with all stocks... idle (with the stock fan. With ACF7 Pro, they went down to 45C). Which was why I lapped the hell out of that IHS, got myself a pretty insane HSF, and... basically did all I could because that 60C scared me to death. I swore to myself to never touch Intel HSF ever again. Sincerely!

P.S.: And my X3350 IHS was... tilting completely to one side instead of being concave or convex. Took me about 2 hours to actually flatten the die with a 180 grit sand paper.

 

[tenax]

bah...my penryn quad temps are the best!
http://i200.photobucket.com/al...enax66/realtemp224.jpg

 

[runawayprisoner]

Phase change? LN2? Chilled LC? Or do you just live in the NP?

 

[tenax]

Originally posted by: Foxery

Originally posted by: bryanW1995
foxery, did you ever install a q9300 in your ip35e?

It's due in Monday. Stupid FedEx.

Originally posted by: bryanW1995
Is anyone out there familiar with gtl settings on the ip 35 pro? I've heard here and there that they can help with stability, but right now I'm just guessing on whether to move them up or down, etc. The only gtl adjustments available are for core 0/2 and core 1/3, there aren't any for fsb,nb,etc.

Raising these very slightly may help stability at high FSB settings. By slightly, I mean like 3-5%. There was a technical writeup linked last week about exactly what this does... The key points I got from it, before it went over my head, was not to push this setting very far, and I got the impression that the writer's mobo had more options than our Abits do.

i believe that's the one i linked and it made my head overheat..fortunately, my brain's thermal shutdown was turned "on" and kept it from exploding i believe the board was a dfi from what i recall. and yes, i've read there and elsewhere it's very much a fine tune type item. i would go 1% up..test overclocks..another percent..test overclocks..or down for that matter from stock. 64% has been said to be the dualie penryn's default for this. 67% for the quads. (that's what mine defaulted too as well)

 

[tenax]

Originally posted by: runawayprisoner
Phase change? LN2? Chilled LC? Or do you just live in the NP?

nah..it's called locked sensors on the only B1 stepping 9450 i've seen anywhere

if you like that one..you'll love this one

http://i200.photobucket.com/al...123/tenax66/uguru3.jpg

 

[runawayprisoner]

One round of applause. That's a really rare chip there.

Uhm... how far can you push the chip... again?

 

[tenax]

not as far as i'd hoped..best stable has been 3600 and i had to pour 1.4 volts into it..not sure what other voltages i loaded on but i'm sure i upped the nbridge and everything else

 

[runawayprisoner]

Aww... that's too bad. Well, at least kind of better than some other here.
Anyway, I'm getting a real Q9450 this week. Nothing can stop that next-day shipping option! Once the processor arrives, it's finally time to find out if it's the CPU that's limiting the OC, or it's the board that's limiting the OC.

 

[bryanW1995]

Originally posted by: aigomorla

Originally posted by: bryanW1995
wow, that's great info! at least we now know WHY we can't find tjunction max anywhere with certainty... and it's good to note that realtemp does appear to be more accurate with their 95c tjunction max, too.

You cant ask what is more reliable because there both the same.

The results are identical in the aspect that one is 10C lower. Coretemp on every aspect is the same as realtemp, but coretemp is only 10C higher on everything nimus the VID.

I read a large portion of that link, and when the poster used his thermal sensor it appeared to show that real temp was more accurate in their guess at tjuction. Unfortunately, everyone is guessing because intel doesn't tell anybody what the actual tjunction is. The correct way to use realtemp, coretemp, etc is to have the program display the delta to tjunction, and they are both the same in that regard.

I do have another question, however. My motherboard shows my chip's default vcore as 1.20v, but coretemp shows it as 1.0375. I've heard others mention that coretemp shows an incorrect default vcore, but this seems like a pretty large difference. Is there a program out there that correctly identifies default vcore on these 45nm quads?

 

[bryanW1995]

Originally posted by: Foxery
The motherboard is supposed to auto-shutdown at 85C, and the thing doesn't even feel hot. I was feeling suicidal, so I turned it up farther. You'll love this...

Nonsense, take 2

Larger pic of my whole screen. A few interesting points:
1. FSB 450x7.5 = 3375 MHz. Speedstep/C1E are turned OFF in the BIOS, but look at CPUz.
2. Now look at Prime95. Cores 0+2 are actually finishing work slower than cores 1+3!
3. And again with temperatures that can't possibly be correct, because I'm still typing and it hasn't exploded.

aigomorla: I'm not normally suicidal, but something is clearly not correct here. For reference, this IS using the stock cooler. It's half the size of what came with my E6750 and twice as loud. I had hoped not to spend more money on a real HSF, but perhaps I will.

First, though, I'd like to know why I've apparently acheived the impossible...

I believe that this is a known issue with the ip35 and ip35e. go to the abit forums and/or mobo forum with this.

 

[tenax]

Originally posted by: runawayprisoner
Aww... that's too bad. Well, at least kind of better than some other here.
Anyway, I'm getting a real Q9450 this week. Nothing can stop that next-day shipping option! Once the processor arrives, it's finally time to find out if it's the CPU that's limiting the OC, or it's the board that's limiting the OC.

yes, i love seeing how these fare for others..meanwhile, i have fun with this as i can't read temps otherwise:

http://i200.photobucket.com/al.../tenax66/digthermo.gif

 

[bryanW1995]

Originally posted by: tenax

Originally posted by: runawayprisoner
Phase change? LN2? Chilled LC? Or do you just live in the NP?

nah..it's called locked sensors on the only B1 stepping 9450 i've seen anywhere

if you like that one..you'll love this one

http://i200.photobucket.com/al...123/tenax66/uguru3.jpg

yeah, uguru is obviously using a different sensor than the one that coretemp and realtemp are using.

 

[runawayprisoner]

Originally posted by: tenax

Originally posted by: runawayprisoner
Aww... that's too bad. Well, at least kind of better than some other here.
Anyway, I'm getting a real Q9450 this week. Nothing can stop that next-day shipping option! Once the processor arrives, it's finally time to find out if it's the CPU that's limiting the OC, or it's the board that's limiting the OC.

yes, i love seeing how these fare for others..meanwhile, i have fun with this as i can't read temps otherwise:

http://i200.photobucket.com/al.../tenax66/digthermo.gif

Nice. I bet we are going to see past-zero results?

 

[tenax]

Originally posted by: runawayprisoner

Originally posted by: tenax

Originally posted by: runawayprisoner
Aww... that's too bad. Well, at least kind of better than some other here.
Anyway, I'm getting a real Q9450 this week. Nothing can stop that next-day shipping option! Once the processor arrives, it's finally time to find out if it's the CPU that's limiting the OC, or it's the board that's limiting the OC.

yes, i love seeing how these fare for others..meanwhile, i have fun with this as i can't read temps otherwise:

http://i200.photobucket.com/al.../tenax66/digthermo.gif

Nice. I bet we are going to see past-zero results?

LOL...I can tell you that at the edge of the core cover, then temp on my quad is 31C. the northbridge at the edge of the heatsink is about 38-39C. for whatever that's worth

 

[bryanW1995]

Originally posted by: tenax
not as far as i'd hoped..best stable has been 3600 and i had to pour 1.4 volts into it..not sure what other voltages i loaded on but i'm sure i upped the nbridge and everything else

Maybe I was just a victim of decreased expectations, but I was very happy to get 3.6 stable at 1.38vcore in bios. How much vdroop are you guys seeing on your quads? I had .12 when I ran it at 1.41 and 1.40, now I'm seeing .11 vdroop down from 1.38 in bios to 1.27 in cpu-z. Since reading that AT article about vdroop I'm less concerned about it, but .11 vdroop does seem like a lot.

 

[runawayprisoner]

On my board, it's a .188v (almost .2v) drop. That also means... without vdroop and vdrop modding the board, I can never go beyond 1.50v max. Not that it's necessary... since 3.80GHz required only 1.304v for me. I hope my Q9450 will be the same.

 

[bryanW1995]

so you had it at 1.49v in bios and 1.304v after vdroop? I think I was at 1.29 after vdroop in cpu-z when I ran prime 95 at 3.64. What cooling are you using again???

 

[runawayprisoner]

Actually, 1.3500v in BIOS and 1.304v real. It's "up to" .188v, not "exactly" .188v all the time.

 

[bryanW1995]

wow, that's very impressive. You definitely have the best chip that I've seen or heard of thus far.

 

[Foxery]

Originally posted by: bryanW1995
I believe that this is a known issue with the ip35 and ip35e. go to the abit forums and/or mobo forum with this.

Yeah, found a few threads there and on some other forums. Seems this is not yet resolved in the final BIOS 16, nor in the first beta for 17. A few people see temps that are 30-40C higher than they should be, which matches what I'm seeing. I'll cautiously overclock with reasonable voltage and wait for a fix for the sensors.

I need to get back to Folding, dontchaknow.

 

[runawayprisoner]

Originally posted by: bryanW1995
wow, that's very impressive. You definitely have the best chip that I've seen or heard of thus far.

And... I'm getting another one this week. If the results are the same, I'd go out and scream that the ASUS P5E-VM HDMI is the best motherboard to get for Q9450.