I7 3770k with h100 just isn't performing.

[Blake7787]

I recently upgraded my mobo and CPU to an Asus p8z77 pro and a 3770k.
I am trying to get at least 4.5 on my CPU but the temps are just through tHe roof.

Current setup is:
Bclk 100
Multiplier 43
Voltage 1.175 non offset
LLC normal or 0
Power phase is extreme
HT is on

The temps on full load using Aida64 are averaged around 86-87 C with some spikes to 94 or 95. Idle is about 40 on core 1 and 2 and 30 on 3 and 4.

Cooler is a new H100 with 4 Corsair SP 120 high performance fans in a push pull exhaust at the top of my Nzxt Switch 810. Max RPM on all 4 and all 3 clicks on the H100.

I don't know if there are other settings I'm missing or what. I've re seated the h100 with AS5 and I've tried offset mode for the voltage of .055.

I can get 4.5 if I want with around 2.6 non offset voltage adding some LLC but the temps hit 100 c within 10 seconds on full load.

Any help or advice would be appreciated. Let me know if more info is needed. I know ivy bridge runs hot but seems like I should be able to get at least 4.5 before moving to an actual liquid system.

 

[r3pshow]

check the CPU block is attached correctly, tighten the screws...Or try to replace the thermal compound before you apply the new thermal make sure You clean your IHS and cpu block very well !

 

[Blake7787]

I did. Took off the back plate and all. Re did entire set up. Cleaned water block o h100 and top of CPU with rubbing alcohol applies AS5 and spread across whole CPU with a credit card. Thin layer yada yada haha. Any other ideas?

 

[Termie]

This seems like a question for the cases and cooling forum, because something isn't right with the cooler.

But for starters, what are the temps at stock?

 

[Schmide]

I did. Took off the back plate and all. Re did entire set up. Cleaned water block o h100 and top of CPU with rubbing alcohol applies AS5 and spread across whole CPU with a credit card. Thin layer yada yada haha. Any other ideas?

There's your reason. Too much TIM. Application methods are a grain of rice or BB in the center of the heat spreader then mount.

 

[Smoblikat]

There's your reason. Too much TIM. Application methods are a grain of rice or BB in the center of the heat spreader then mount.

That doesnt happen if you think about it, if he put too much TIM on it would just squish out the sides due to the pressure.

Welcome to IB, its almost identical to SB but much hotter.

 

[Schmide]

That doesnt happen if you think about it, if he put too much TIM on it would just squish out the sides due to the pressure.

It happens, especially with thick thermally activated TIMs like AS5.

If it's a small amount it has room to spread out. A larger amount has enough friction and viscosity to hold up against the pressure and becomes a cake.

 

[Termie]

This isn't an IB problem. He needs to work on identifying the real issue, which has to be the cooler. While he's at it, he should use something better than as5.

Again, as a baseline, we need to see stock temps.

 

[SheHateMe]

I wouldnt blame the TIM, a lot of people have been reporting faulty radiators on those Corsair closed loop coolers. Isnt there an application where you can see what the temperature of the liquid inside the cooler is? That is how most people found out it was the radiator.

 

[BonzaiDuck]

This isn't an IB problem. He needs to work on identifying the real issue, which has to be the cooler. While he's at it, he should use something better than as5.

Again, as a baseline, we need to see stock temps.

I'm guessing that the H100 has a built-in pump . . . No more comment from me on this, since I never had one. I'll be interested in the outcome, though . . .

Only after solving that problem would it be advisable to pick a better TIM and "do some additional stuff" as he's inclined . . .

Just dawned on me . . . Is the waterblock on the H100 bare copper? Or nickel-plated? Perhaps he should check to see that the surface of the waterblock is flat, using a metal straightedge and a strong light. If it needs "some work," he can conveniently get rid of the nickel plate as well . . .

 

[Bill Brasky]

I have an h100 on an oc'ed 3930k and it's working like a champ. It is pretty important that the fans are intake, not exhaust (I think this is your problem along with possible improper mounting). Also, you should thumb tighten the waterblock screws until they stop or feel firm. Overtightening can cause problems. Also, make sure the 4 pin molex connector that powers the waterblock is plugged directly into the power supply.

How many times have you tried mounting the block? It took me 3 times to get good TIM patterns.

edit: BonzaiDuck, The block on mine is all copper, fyi.

 

[SheHateMe]

I have an h100 on an oc'ed 3930k and it's working like a champ. It is pretty important that the fans are intake, not exhaust (I think this is your problem along with possible improper mounting). Also, you should thumb tighten the waterblock screws until they stop or feel firm. Overtightening can cause problems. Also, make sure the 4 pin molex connector that powers the waterblock is plugged directly into the power supply.

How many times have you tried mounting the block? It took me 3 times to get good TIM patterns.

edit: BonzaiDuck, The block on mine is all copper, fyi.

You have all 4 of your fans setup for intake?

 

[Blake7787]

This seems like a question for the cases and cooling forum, because something isn't right with the cooler.

But for starters, what are the temps at stock?

Ok sorry just helped my buddy rebuild his comp and just now getting home.

I have seated it twice. First time with the stock h100 thermal paste and the second time with the AS5. Im going to set the bios to stock real fast and take a look at the base clock temps.

 

[Blake7787]

Stock temps from AIDA64:
Core1--->Core 4
53, 50, 54, 45

Thats at 3.9 and about 1.175 average Volts. Which to me is way too much.

On Full load for 30 seconds they get up to and average of 90 degrees across all 4.

I was getting lower temps @ 4.3 with a lower voltage for obvious reasons.

 

[Blake7787]

I have an h100 on an oc'ed 3930k and it's working like a champ. It is pretty important that the fans are intake, not exhaust (I think this is your problem along with possible improper mounting). Also, you should thumb tighten the waterblock screws until they stop or feel firm. Overtightening can cause problems. Also, make sure the 4 pin molex connector that powers the waterblock is plugged directly into the power supply.

How many times have you tried mounting the block? It took me 3 times to get good TIM patterns.

edit: BonzaiDuck, The block on mine is all copper, fyi.

Mine is all copper as well. Do you think instead of switching from exhaust to intake, a set of 120 AF120 Corsair High Performance fans for the pull function would help? I like it to exhaust out the top bc there is no filter up there. Granted if it is proven that intake works better i can also rig something up.

 

[BonzaiDuck]

Mine is all copper as well. Do you think instead of switching from exhaust to intake, a set of 120 AF120 Corsair High Performance fans for the pull function would help? I like it to exhaust out the top bc there is no filter up there. Granted if it is proven that intake works better i can also rig something up.

I can't speak with any firsthand authority about this, since I've never used a Corsair H100, nor have I replaced my Sandy with an Ivy.

If the pump works, the air is going through the radiator fins, the cooler is filled with coolant, the waterblock is perfectly flat and the IHS of the Ivy Bridge is reasonably flat, then I can't imagine what this might be.

Except. Except the possibility that the F***y new approach Intel uses -- replacing some fluxless solder with a TIM between the bottom of the IHS and the processor -- has some variability in how effectively it was applied at the factory.

Someone else may know. Maybe someone might advise you to go RMA and replace the processor. Otherwise . . . .there's a remedy that's posted in a thread or two of this forum, and you may not be inclined to go that route.

If I tell you to look at Graysky's new thread, it's only to get some discussion here about whether I could be right. Because . . . as I said, I have no firsthand basis to know if I am.

 

[HURRIC4NE]

i have an h100 on a clarkdale cpu (i5 650)... from what ive heard all round, ivy bridge is HOT HOT HOT HOT HOT for everyone, dont remember why but isnt the die shrinkage supposed to make it cooler? wth intel...

ya'll better not mess up haswell or else i'm switching to my calculators to play skyrim.

 

[BonzaiDuck]

i have an h100 on a clarkdale cpu (i5 650)... from what ive heard all round, ivy bridge is HOT HOT HOT HOT HOT for everyone, dont remember why but isnt the die shrinkage supposed to make it cooler? wth intel...

ya'll better not mess up haswell or else i'm switching to my calculators to play skyrim.

That's it . . . really . . . I've seen thermal data on OC'd IB's running at 4.5 which tells me I could "live with it." But if the quality -- the consistency -- of the product has taken a big dump with these initial Ivy releases, given that issue with the IHS and TIM, I have to wonder . . . People don't just undertake "IHS removal" out of curiosity, and in this economy, nobody wants to damage $300 in hardware or void a warranty without good reason.

See -- the die shrink makes it faster, and the thermal design power -- the TDP -- is lower at 77W. But it's all concentrated over a smaller area. The thermal sensors are apparently located so they respond instantaneously. That was the problem with using a TEC plate for cooling: the ability to remove heat depends on the area of contact.

As someone said, it's important to understand the difference between heat and temperature.

Somebody else can "throw in" about this -- I don't want to be chastised for bringing it up. But like I said -- if the H100 is working, the block is mounted correctly, etc. . . . . something else may be going on. If Intel changed their approach to the IHS and processor after so many years of doing it the same way, it's possible that someone didn't pay enough attention to the "consistency" issue for what's rolling off the assembly-line. . . .

 

[Bill Brasky]

Mine is all copper as well. Do you think instead of switching from exhaust to intake, a set of 120 AF120 Corsair High Performance fans for the pull function would help? I like it to exhaust out the top bc there is no filter up there. Granted if it is proven that intake works better i can also rig something up.

Intake really is that much better unless you have retarded air flow or an extremely cool system. I have two 1200rpm pwm fans (scythe slipstreamed) in the pull config, which gives me about 10-15C cooler temps. Granted system differences can have a large impact. I have 6 hard drives and a graphics card that dumps all heat inside the case, but I still think that switching to intake will help a lot.

 

[Idontcare]

The temps on full load using Aida64 are averaged around 86-87 C with some spikes to 94 or 95. Idle is about 40 on core 1 and 2 and 30 on 3 and 4.

Cooler is a new H100 with 4 Corsair SP 120 high performance fans in a push pull exhaust at the top of my Nzxt Switch 810. Max RPM on all 4 and all 3 clicks on the H100.

I don't know if there are other settings I'm missing or what. I've re seated the h100 with AS5 and I've tried offset mode for the voltage of .055.

I can get 4.5 if I want with around 2.6 non offset voltage adding some LLC but the temps hit 100 c within 10 seconds on full load.

Hi Blake7787, I have an H100 and a 3770k. A couple things that can give you less-than-ideal temperatures are your TIM application method and your mounting orientation.

On the topic of mounting orientation, you want to mount your H100 water-block such that the text on the waterblock (the word "Corsair") is aligned with that of the text on the IHS (the words "Intel Core" etc).

On the topic of TIM application...you mention using the credit card trick, I am familiar with this and recommend you do not do it, at least not in this phase of the "problem solving" process.

Instead, for now, do the "grain of rice" dollop method. Once we get your temperatures figured out then you can go back to the credit card method if you are keen to continue using it.

I use Noctua's NT-H1 for TIM, its cheap and I found it works better with my H100 than AS5. (see my H100 vs NH-D14 thread in Cases & cooling)

Here is what a "dollop" of TIM should look like on your 3770k:

To confirm that this TIM was appropriate, after concluding my tests (data at the bottom of this post) I pulled the H100 and took a post-mount pic:

^ that might look like too little TIM but I assure you it is not, what you can't see (but I could with my eyes) was that both the upper and lower sides of the IHS (the parts not covered by the TIM) were completely metal-on-metal mated to the H100. There is no TIM there because, under pressure of the mount, none was needed there.

And the results?

I haven't lapped my 3770k (or delidded it as is the plan), so I consider these preliminary myself, but they show you what you should be looking towards.

Your hardware is definitely capable of delivering sub-80C for 4.5GHz provided the H100 itself isn't borked.

So double check that mount orientation, consider getting some NT-H1 (its cheap) or MX4 (performs the same as NT-H1), and ditch the credit-card pre-spread method in favor of using the simple rice-grain dollop method (at least until you test for yourself to confirm it is helping, not hurting, you).

 

[Bill Brasky]

IDC to the rescue! :awe: I also went from AS-5 to MX-4. Maybe my arctic silver was old, but I thought the MX-4 spread out better and resulted in a more even application with the "line" method.

Why does aligning the block with text on the IHS help? Do you have any theories, or is this just what you've found from experimentation? And how much does it actually help? I think I have mine rotated 90 degrees CW.

 

[BonzaiDuck]

IDC to the rescue! :awe: I also went from AS-5 to MX-4. Maybe my arctic silver was old, but I thought the MX-4 spread out better and resulted in a more even application with the "line" method.

Why does aligning the block with text on the IHS help? Do you have any theories, or is this just what you've found from experimentation? And how much does it actually help? I think I have mine rotated 90 degrees CW.

I'm curious why the orientation would matter. IDontCare's temperatures up to 4.5Ghz look perfectly acceptable.

 

[Yuriman]

My Ivy runs particularly hot too, it seems to vary from chip to chip but I'd bet yours and mine would benefit significantly from delidding. Some people get chips which I assume have better contact between the core and the IHS or something, and have much lower temps.

 

[BonzaiDuck]

My Ivy runs particularly hot too, it seems to vary from chip to chip but I'd bet yours and mine would benefit significantly from delidding. Some people get chips which I assume have better contact between the core and the IHS or something, and have much lower temps.

. . . . Which is what I also timidly suggested . . . so I'm not alone. . . .

Here . . . there are all sorts of forum exchanges on this topic, but this is a "review" article. Several here -- people like Graysky and IDontCare -- may have read it already.

http://www.overclockers.com/ivy-bridge-temperatures

If the use of silly-cone based TIM shows such a wide disparity in thermal conductivity compared to a silver-indium solder, what might that say for the variability of temperatures between units of the same product? I don't know -- I'm just posing the question.

As for TIM replacement one might use, I'm a big fan of nano-diamond paste, but Indigo Xtreme (with ingredients not so different from the solder of earlier CPUs), is better. But I'm not suggesting that anyone HAS to do it.

It just seems that we're taking one or two steps backward just to take a step forward. If Intel departs from common sense for cost reasons with the IB IHS and TIM, that's the steps backward. Also, apparently some have contacted Intel about this issue. The web is filled with comment and posts about it.

That's my two cents worth -- and some have already heard it, regardless who it's coming from . . .

 

[Ferzerp]

Nice graph. for reference, I've been able to keep my 3770k after delidding cooled in the 1.4+V range with an H100, so I can't wait to see your results.