AIIx3 doesn't want to heat up when using prime95 (trying to remove heatsink)

[-Slacker-]

So I've been running prime95 for ~20 minutes now and the core temps aren't getting above ~46C. Idle is 30C.

I tried it before for about 10 minutes, then shut my machine off and tried to pull the heatsink out - it wouldn't budge (yes, I removed the heatsink's clippers from the cpu socket).

I'm monitoring the temps with 2 programs: Speed Fan, and CoreTemp - they both show 46C on the corewith a green check mark on the side, which is kind of annoying, but Speed Fan also shows another reading, the CPU, at 51C and a little red fire icon on the side, which is even more annoying, because now I don't know what to think of the temps - are they just fine and dandy at 46C core, or is my cpu about to burst in flames at 51C "CPU"?

Should I lower the fan speed?

 

[-Slacker-]

So it's been ~40 minutes of prime95 now, and I think I'm going to have to try removing the heatsink - if the thermal paste isn't loose enough by now, then it's probably not going to get any looser in another 40 minutes.

(the temps are the same - I even tried messing around with 7-zip at the same time, to try and get the cpu hotter bu, so far, nothing.)

I'd still like some opinions before I do this though, because if I fail, I'm going to end up with a heatsink that's no longer properly stuck to the cpu.

 

[Morg.]

Athlon 2 X3 . that makes more sense to me now ...

What is your problem ?
Why do you WANT to see those temps go up ?

 

[dawp]

yes, you can lower the fan speed and it will cause the cpu to get hotter. you can also unplug the fan for a short period of time without damaging the chip. that should cause it to get hot fast.

 

[dawp]

Athlon 2 X3 . that makes more sense to me now ...

What is your problem ?
Why do you WANT to see those temps go up ?

Makes it easier to remove the heatsink with warm paste.

also try to twist the sink before removing it.

 

[dma0991]

Now who says AMD processors gets so hot and bursts into flames again? Heard that a lot from my friends as well and they got a lecture from me. :awe:

There are better ways to do this than trying to run stress test software to get the temps up. You could point a hairdryer towards the heatsink but not too close to the processor as some claims there can be static damage but I am skeptical about it.

Another way is to take out your motherboard and put some solvents preferably isopropyl alcohol in between the processor and the heatsink if possible. It should soften the thermal paste a little.

 

[AdamK47]

Removing the heatsink would indeed help you in raising the CPU temps. I applaud you in your attempt at trying something different.

 

[Morg.]

I still have trouble understanding the point here... IF your CPU can't get hot, why change the rad or w/e ?.

 

[dawp]

I think he's trying to replace either the heatsink or the cpu and is having trouble removing the old one.

 

[BD231]

I still have trouble understanding the point here... IF your CPU can't get hot, why change the rad or w/e ?.

In his defense I've seen some paste turn to hardened looking Spackle and loose ALL it's thermal transfer properties in the process so it may be that he needs new paste.

But yeah, OP you'd have to pump up the voltage considerably to get anywhere near the temps you need to melt that stuff. Your idea is not going to work.

 

[-Slacker-]

What's all the confusion? I was trying to get my cpu hot enough so that the thermal paste between it and the heatsink would get looser, so that I could remove the stock heatsink and replace it with a cooler master hyper 212 plus.

Anyway, I'm finally done with everything. Cpu is running at <30C and the "core" (whatever that is, SpeedFan is dysplaying it too) at <20C. Idle, obviously.

I still have trouble understanding the point here... IF your CPU can't get hot, why change the rad or w/e ?.

I'm planning on overclocking my cpu, and I hear stock athlon cooling is garbage. Weather that's true or not, it didn't mater, because I already bought an aftermarket cooler, so I have to use it.

Your idea is not going to work.

 

[betasub]

I'm monitoring the temps with 2 programs: Speed Fan, and CoreTemp - they both show 46C on the corewith a green check mark on the side, which is kind of annoying, but Speed Fan also shows another reading, the CPU, at 51C and a little red fire icon on the side, which is even more annoying, because now I don't know what to think of the temps - are they just fine and dandy at 46C core, or is my cpu about to burst in flames at 51C "CPU"?

LOL - o noes my CPU is about to burst in flames.

SpeedFan has 50C as a default value for the "warning" icon. It is user-modifyable through Configure>Temperatures>CPU tab. The default generic setting does not necessarily apply to any particular CPU/cooling environment/OC circumstances. Adjust to taste.

 

[-Slacker-]

Ah, so the value that I should look after is the core temp.

That's great, because now my prime95 temps are 28c.

Going to overclock later today, will post results.

 

[dawp]

depending on your motherboard and cpu, you may be able to unlock that 4th core.

 

[-Slacker-]

^ASrock m3a770de --- yeah, I think it's possible but, for now, I'm going to contend with my measly oc at stock volts

Seems stable, but I haven't run a stress test yet.

edit:


Wasn't stable, stress-testing it at 240mhz buss speed now

edit2:

By the way, was I right to lower my DRAM speed from 665mhz to 533mhz, or is there some sneaky way to circumvent that? (the x3 435 doesn't have an unlocked multiplier..)




15 mins of prime95 so far, 30C:

 

[dawp]

that's how I do it, lower the mem multiplier and raise the fsb. I like to keep my mem around stock speed tho most do have some headroom.

 

[-Slacker-]

Glad to hear I didn't screw up, but now I'm sorry I didn't get the athlon2 x3 440 for the unlocked multiplier instead of the 435, just to save some pocket change.

I even got a 320gb hdd instead of a 500gb to save money, I feel like the biggest cheapskate

 

[Farmer]

The stock paste is fairly brittle when dry. Just pull it straight off. Don't twist, as that might actually twist the pins. There is absolutely no need to heat it to loosen it.

I just did this a week ago on a stock HSF that hasn't been touched in 2 years.

Side note, I tried to reach 3.7-3.8 on the 720BE (2.8 stock), but couldn't get north of 3.5, regardless of how I cranked voltage. Disappointing, since I was already at 3.3 on the stock cooler. At 3.6-3.7, Windows loads fine, but will blue screen at any significant CPU utilization. My BIOS doesnt have line load calibration, so vcore drops at load were in the neighborhood of 0.03-0.04 volts, which I believe to be the culprit. To add insult to injury, I can't unlock my 4th core. My RAM is so unstable that I cannot even increase FSB by 10MHz, even at 1.6V Vdimm, so I am at stock FSB. I will try altering the ratio.

Regardless, at 3.5, my CPU never goes north of 33 celsius, even after 12 hours of prime on all cores, so at least we know the cooler works.

You have a 1GHz overclock at essentially stock vcore, mostly because you could crank your FSB so much. I'd say that is a good overclock. 40-50MHz FSB bumps have been reported for my motherboard, so I must have the wrong kind of RAM.

 

[-Slacker-]

^I actually did twist it a bit before it broke off but, so far, the cpu seems fine, although it's been running for only a few hours now.

Seems reasonable to assume that the pins might get damaged if you twist the heat sink but if you think about it, that's a lot of pins we're talking about, and the "force of the twisting" gets distributed to all of them equally, so the pressure that each individual pin will receive isn't necessarily strong enough to break it.

That's my bro science reasoning, anyway.

 

[Morg.]

Well you can twist the pins it's no biggie .. as long as you don't break em it's still possible to manually readjust them (did it for an old athlon at some point)

For your RAM, may as well attempt some OC on there and give it some volts (although I know there are some limitations with the IMC, right ?)

 

[-Slacker-]

^I'm actually not sure if I want to downclock my ram any further... Assuming I can reach 3.7~3.8, what's better, having an extra 300mhz on the cpu, or having an extra 266mhz on the memory (from 800 to 1066, the later being my current memory clock)?


I'm also curious about something else:

Right now, my vcore is at 1.328 with some brief, ninja oscillations to 1.336

The last cpu-z screenie that I posted represents the last stable configuration that I'm still using right now and it reads 1.352v - I was running prime95 though.

2 questions:

Is the vcore supposed to fluctuate like that?...

...And

Does it fluctuate depending on the load of the cpu?

 

[eternalone]

^I'm actually not sure if I want to downclock my ram any further... Assuming I can reach 3.7~3.8, what's better, having an extra 300mhz on the cpu, or having an extra 266mhz on the memory (from 800 to 1066, the later being my current memory clock)?

Having the extra 266mhz memory IMO, is always better that the higher clock, in overall peformance and feel of the machine, but its up to you.

 

[Farmer]

2 questions:

Is the vcore supposed to fluctuate like that?...

...And

Does it fluctuate depending on the load of the cpu?

All your voltages fluctuate. I think your voltages are regulated via some feedback control (some kind of voltage regulator circuit) which measures voltage at some point and tries to correct until the measurement matches your input (i.e., what you set it at). Feedback control is not perfect, as there are many disturbances that could push your voltage one way or another, but your voltage should hover around a certain number. For some reason that is beyond my current understanding (but related to the fact that the circuit is nonlinear), it tracks a different voltage depending on how much load is on your CPU. For mine, for instance, load Vcore is consistently 0.025 volts lower than noload Vcore. In controls terminology, this error in tracking could be called "nonzero steady state error," but I don't know if it's actually an error in the feedback, or if the input voltage varies with respect to the CPU load.

"Load line calibration," which is a feature on some boards, seems to be a much stiffer control on the voltage, offering pretty much 0 steady state error. That being said, it is a much "stronger" feedback (i.e., the feedback gains are larger), so there is a risk of voltage spikes. If you have it you could try activating it. But you are at 1.3-ish volts, which is nothing to worry about at all. If your voltage dips become a stability problem, just increase your voltage.

I'm at a point where I can no longer increase voltage (in fact, I'm already above 1.55) but I suspect that fluctuations/particularly steep drops can cause BSODs. I really don't know how to diagnose the problem. Mobo/RAM/CPU/PSU? Ocing is an art because it is difficult to understand completely, as you are operating essentially within the manufacturer's margin of error.

 

[flexcore]

Try 1.4 and then 1.45 volts. and maybe a bump on dram and cpu/nb also.