A [thermal] compound(ed) problem

[Josh Weston]

Idle temps do seem fine, but the higher temps are more relevant here. We know Ivy's have a somewhat heat transfer "issue", but I don't think you've hit that limits with those volts. i'm running mine at a crazy 1.56V on a Phenom stock cooler and stress tested my CPU would reach 60C max in a warm summer day, won't even hit 50C in BF3 which is a CPU demanding game.

You're talking about your Athlon, right? Completely different ball game.

To the OP, I just asked a friend of mine what his idle temps were like on his stock 3570K - high twenties/low thirties. That's with his CPU underclocking itself. The beefy coolers only begin to show their colours when the heat increases, so don't expect a H100i to have dramatically lower idle temps than the Hyper 612.
It doesn't surprise me that a chip could get into the 90s at 1.28v - some chips are better/worse than others, of course. I do think you should swap it out for the 612 again, though.

 

[dragoonk]

I take part of my last statement back. I use CorsairLINK to monitor my pump speed only, which is running at 2100 RPM if I remember correctly. And I'm not using the fans that came with it; Noctua NF-F12s in pull configuration @ 1200 RPM. Still I should get better performance than my 612.

My advice would be to go back to the stock config ( stock fans etc. ) and see if that changes anything. Troubleshooting sucks, I know.

 

[dragoonk]

You're talking about your Athlon, right? Completely different ball game.

To the OP, I just asked a friend of mine what his idle temps were like on his stock 3570K - high twenties/low thirties. That's with his CPU underclocking itself. The beefy coolers only begin to show their colours when the heat increases, so don't expect a H100i to have dramatically lower idle temps than the Hyper 612.
It doesn't surprise me that a chip could get into the 90s at 1.28v - some chips are better/worse than others, of course. I do think you should swap it out for the 612 again, though.

I agree, totally different. Wasn't comparing them directly though, point was the idea under thermal power and cooling. The H100i is still a pretty good cooler.

 

[Josh Weston]

The H100i's a great cooler, but Ivys get hot. They're cooler than Sandy up until the wall I mentioned in a previous post. When you hit that, every extra bit of voltage creates more heat than it logically should, when compared to chips of different architectures.
You take a Sandy Bridge chip from running at 1.35v to 1.4v and you might see a 5-6*C temperature increase.
You take an Ivy Bridge chip from 1.22v to 1.27v and you could easily see more than a 10*C increase.

 

[dragoonk]

I know the difference between the Sandy Bridge and the Ivy Bridge, I know about the heat thing Ivy's have, I guess you can consider yourself lucky to be able to overclock an Ivy somewhere near 5GHz with low voltage and decent temps.

I found a guy that was running a 3570k at 4.2GHz at 1,28V at under 55C . how's that possible. does it seem possible to you?

 

[seumasbeathan]

Well something clearly isn't right here, after decreasing the volts what are your new load temps? If they are ok then I really wouldn't worry about the idle temps, my cpu is at about 40-45c idle most of the time but at full load it only goes as high as about 76c and in games doesn't usually go above 70c, thats a 3570k @ 4.3ghz on the crappy stock cooler, so for you to be getting those kind of temps on a H100 sounds very odd.

Edit : Oh and btw, in future don't spread the TIM, just put a smallish blob on and the heatsink will do the rest

 

[XANADV]

I've been running Prime95 small FFTs for the past 20 minutes at 4.2 GHz and 1.17V with temps in the mid 80s. Time to shut down and swap in the old 612 and see what happens. If the 612 performs worse as expected then it might very well be the CPU limitation. Let's see what happens...

 

[Josh Weston]

I know the difference between the Sandy Bridge and the Ivy Bridge, I know about the heat thing Ivy's have, I guess you can consider yourself lucky to be able to overclock an Ivy somewhere near 5GHz with low voltage and decent temps.

I found a guy that was running a 3570k at 4.2GHz at 1,28V at under 55C . how's that possible. does it seem possible to you?

Not as lucky as I'd like - my CPU-Z banner was my first and only (succesful) attempt at a 5GHz run. I run my chip at 4.4GHz.

The Ivy chips seem to have the most outliers in terms of chip quality. We had a user who could get 4.8GHz at under 1.2v. It just seems to be as though his chip doesn't deal with voltages anymore.

With any luck, i'm completely wrong and it's nothing to do with his chip, but a malfunction with the H100, etc.

We'll see when he puts the old cooler back on

I've been running Prime95 small FFTs for the past 20 minutes at 4.2 GHz and 1.17V with temps in the mid 80s. Time to shut down and swap in the old 612 and see what happens. If the 612 performs worse as expected then it might very well be the CPU limitation. Let's see what happens...

Now, those temps are a higher than i'd expect. I think I was wrong about it being your volts, and it is, indeed, the cooling solution.
P.S. be sure to go with what Seumas said. No spreading allowed! It can cause a layer of air to insulate the cooling block from the CPU.

 

[dragoonk]

Not as lucky as I'd like - my CPU-Z banner was my first and only (succesful) attempt at a 5GHz run. I run my chip at 4.4GHz.

The Ivy chips seem to have the most outliers in terms of chip quality. We had a user who could get 4.8GHz at under 1.2v. It just seems to be as though his chip doesn't deal with voltages anymore.

With any luck, i'm completely wrong and it's nothing to do with his chip, but a malfunction with the H100, etc.

We'll see when he puts the old cooler back on

^_^ Yea, I'm pretty curious about what the real problem is.
I have mine maxed out, 3.93GHz is the highest stable clock speed. ( I know it's an Athlon and it's crap... ). I've known people that got one of these up to 4.6GHz at 1.54V. so... that speaks for itself.

 

[kully1972]

Have u considered your pump is not working at its highest rate. I had this prob with my H60 when connected to a 3 pin fan.

 

[dragoonk]

Well I think he said that the software is reporting his pump running at 2100 rpm and if I'm not mistaken that's the max rpm for that pump? Unless the readings are wrong...

 

[XANADV]

OK I'm back guys. Sorry for the long wait. Only took me half an hour to put the 612 back on. Took me much longer to fiddle around with Prime95. 4.5 GHz @ 1.28V with temps in the mid 80s at their highest with the fan running at max (2000 RPM). Kudos to Josh. Seems that my H100i really is somehow defective. I don't know how heavy they are supposed to be but mine kinda felt light. Possible there is not enough coolant in it maybe? Either that or the pump isn't working like it should. Either way I am going to have to RMA it. Thanks to all of you for your help with this problem!

Oh btw guys I just completed another build log update for those of you that are interested. [Just pretend there is still a Hyper 612 in the pictures. :lol:]

 

[Josh Weston]

Eeh, I was more confident that it was your chip's fault, so the kudos goes to DragoonK and Co.
I would put the H100 on one more time, just to ensure that it wasn't a mounting/thermal paste issue. And is that the lowest your volts will go for that clock?

 

[dragoonk]

Naah, no kudos, we weren't competing.

I agree with Josh, it could also be a mounting problem, but I don't know, I think if it were it probably wouldn't be hard to spot. Personally I would try everything I can think of before sending it back and .. potentially making a fool out of myself.

 

[XANADV]

Eeh, I was more confident that it was your chip's fault, so the kudos goes to DragoonK and Co.
I would put the H100 on one more time, just to ensure that it wasn't a mounting/thermal paste issue. And is that the lowest your volts will go for that clock?

I am pretty sure it is not a mounting issue. I must've repeated the whole process of removing and reinstalling the waterblock at least five times using a different thermal paste almost every time and, yes, I checked the mounting every time as well. That should preclude the possibility of me mounting it the wrong way also.

I've been running Prime95 for four hours straight now, changing stuff around every now and then (the wonders of TurboV EVO). Currently running 4.5 GHz @ 1.26V.