The difference I found between the stress test and seti, is that the stress test keeps the CPU load @100% stable for the duration of the test. While running SETI, the load is fluctuating a lot between 15% and 100%, maybe that's what's generating the extra heat comparing to the stress test.

That doesn't seem to make any sense. If you use it less, on average, the heat generated is less. Ramping from 15% to 100% isn't like accelerating a car, where you pump in more gas, and stress the engine more.

Now SETI *may* stress some parts of the CPU more than the Intel test, hence generate more heat. My brother-in-law used to (among other things) design CPU stress tests for DEC and HP workstation CPUs, and by knowing the precise physical structure of the CPU and exactly how the instructions worked, he could cause the CPU to go nova if he wanted to. But SETI doesn't do that, it is just a program.

---

But still, 100% is a 100% either using the GPU or not.

What I meant is that your GPU is 250W and the CPU is mere 84W. If you have issues with pumping the heat out of the box these 250W might mean that your CPU is throttling because of heat produced by the GPU. In any case this should be easy enough to test. Just disable GPU computing for a while and SETI should behave more or less like the stress test. Am I making any sense?

---

Now SETI *may* stress some parts of the CPU more than the Intel test, hence generate more heat. My brother-in-law used to (among other things) design CPU stress tests for DEC and HP workstation CPUs, and by knowing the precise physical structure of the CPU and exactly how the instructions worked, he could cause the CPU to go nova if he wanted to. But SETI doesn't do that, it is just a program.

I understand. Different instructions use different sections of circuitry on the chip, and hence use a different amount of power. So the CPU may be busy 100% of the time, but there could be a 50% difference in power use depending on what instructions it's actually doing.

So a CPU test may cycle though a variety of instructions and external memory operations, to test all parts of the CPU and system. While the SETI code really gives the power hungry Maths co-pro section a thrashing.

This is where the TDP (Total Design Power) comes in. It's a worst case scenario. As in the old DEC case, someone could write code that was only high power instructions. If the chip / heatsink / case cooling wasn't up to handling that heat, the machine would overheat, crash, and possibly even die totally. It doesn't mean the CPU will be dissipating 85 watts, even if it's at 100%. But it MIGHT.

For the OP, any Electronics shops local? Plain Silicone thermal grease will work just fine on the heatsink.It doesn't HAVE to be the exotic silver stuff. A little tube of that is enough to do about 50 heatsinks.

Also, if you think it's a case cooling issue (heat from the GPU?) then take the side cover off and run it for a while. If it's cooler like that, then you need to look at the fans and airflow.

Ian

I believe the high temperatures are also a function of your Haswell generation chip, which is notorious for its poor heat dissipation from the CPU die to the internal heat spreader. The thermal compound between these internal components is poor, causing the cores to heat up VERY quickly under load. Idle is generally not an issue. This affected the previous generation Ivy Bridge chips as well.

Many overclockers have gone the extra mile to "delid" the internal heat spreader, remove the factory applied thermal compound, and reapply a better one (generally Coollaboratory Liquid Pro). This is a somewhat tricky process and voids Intel warranty; though the results are astounding, dropping temperatures remarkably by ~20C.

Unfortunately, I am also a Haswell chip owner (i7-4771) and saw those frighteningly high temperatures with my stock heatsink/cooler. I have since ordered an aftermarket all-in-one cpu cooler which should help out at least a little. In the meantime, I'm throttling the allowed CPU usage for SETI projects.

Finally got the new 4770k to not overheat when running Seti. I initially wanted to use a Corsair H105 LCS, but it wasn't immediately available, so I settled for a H75. It seems to be a good choice for the immediate future. Running with Seti with my normal 80% (5 CPU WUs & 2 GPU WUs), with .5 CPU for each GPU WU, I was hitting the 80s range. The ASRock Extreme4 board that I'm using automatically shuts down to avoid damage, so I couldn't get more than 20 minutes of work done before it shut down. Reseated the LCS several times until I got it just right. The other problem was with CPU allocation for GPU WUs. On the i7/950 & i7/930, I didn't have a heat problem allocating .5 CPU per AP GPU WU. This is not the case now as I have had to change the setting to 1 CPU per WU. Temps dropped at least 10~15c when BOINC started again. I will be keeping a close watch when it starts running 2 V7 GPU WUs at a time, but I don't think that will be a problem. Right now, I'm running with either 2 x AP or 1 AP & 1 V7 depending on the scheduling. With summer rapidly approaching, I will get the H105 and switch the H75 to the other machine when I finally get it back up and running. BTW, the temps are holding between 55~58c with the max at around 64c.

---

I don't buy computers, I build them!!

This is not the case now as I have had to change the setting to 1 CPU per WU. Temps dropped at least 10~15c when BOINC started again..

This is because each AP WU on the GPU actually needs a full CPU with the current app. If you allocate less (you had 0.5) you will find the CPU over-committed, which causes excessive task switching, which was the cause of your extra heat.

The MB WUs only need a small fraction of a CPU; the usual default of 0.04 works OK.

---

I guess that the stock cooler isn't suitable for this and I am not a big fan of other coolers due to their size factor. I will order the H100i and hope for the best. I will post the results once I get it (probably going to take a couple of weeks).

Meanwhile, I've changed the multiplier from 37x to 35x when the 4 cores are active: peak is now 80c and averaging around 75c, running on 100%. I think those temps are acceptable for a Haswell, room temp is 25c nowadays. I've installed a case fan in the bottom this morning as well, not sure if it helped. I will add a side fan and an extra top fan next week to see if that helps.

---

Here's the update.

I have finally got an H100i to replace the stock fan. I removed the fan, cleaned the CPU with ArctiClean and installed the H100i (it comes with stock thermal compound).

Temps with stock fan (peak values):
- seti running @50%, core multiplier x35: 79 degC
- seti running @80%, core multiplier x35: 82 degC
- seti running @100%, core multiplier x35: 85 degC
- seti running @100%, core multiplier x37: 100 degC

x37 is the default configuration when the 4 cores are running.

Temps with H100i (peak values):
- seti running @100%, core multiplier x35: 60 degC
- seti running @100%, core multiplier x37: 66 degC

The new setup has been running for a couple of hours. I am not sure if things will get better, since I read somewhere that it takes time for the thermal compound to saturate. In any case, I am happy with the huge improvement as it is now.

I got around 3 degC improvement in idle time. The only downside of the H100i is the noise when the fans are at full speed. I've got already 2 SP120 quiet edition, but I am not in a hurry to replace the stock fans.

---

Are you running with HT on?

---

Are you running with HT on?

Yes, it is.

---

You could try using the Corsair Link software, and set the fans to "quiet", it certainly reduced the noise on my 4670k/H80i combo, and I run Seti on all cores at around 85%

P.

---

A weird thing:
Enable the iGpu and you will lower the the core temps by 5-10C maybe more. It sounds weird but that how it works.

With my 4770k no HT and an H100i my core temps are around 50-60C at 4.4Ghz.

Took a while to dial in the different voltages in BIOS.

//LD

I'll check the Corsair link and try it. I don't mind the noise much as I am usually using my noise cancelling headphones while gaming.

Can you elaborate on iGPU? What's that and how do you enable it?

---

Can you elaborate on iGPU? What's that and how do you enable it?

iGPU = Intel GPU are an part of the CPU (ok a slow GPU).

In BIOS there are settings to enable the GPU part of the CPU.
Maybe you also must connect an monitor to the connector , Or an "dummy plugg"

This is depending on the motherboard. My PC at work has this feature is disabled. IE no connector on the MB.

You also have to install "gpudrivers" from Intel

http://www.cpu-world.com/CPUs/Core_i7/Intel-Core%20i7-4770K.html
https://downloadcenter.intel.com/default.aspx

and some more reading, READ ALL before you do antything
http://www.overclockers.com/3step-guide-to-overclock-intel-haswell

My 4770K is sitting on top of an ASRock Extreme 4 board, and I am using the Intel HD4600 IGPU as the monitor, which is set via the UEFI BIOS. The monitor is plugged into the boards DVI connector. Make sure that you install the latest upgrade from Intel (10.18.10.3496). There is a link to the Driver Update Utility http://www.intel.com/p/en_US/support/detect that I've found very useful, just ensure that what Intel has does not conflict with what your board manufacture has. It will tell you if the software on your machine is compatible with Intel's.

I'm cooling it with a Corsair H75 LCS and I've never seen the CPU temps get above 68c when crunching at 80% HT using a EVGA GTX660SC @ 2GB, which is free standing and not using a dummy plug. Prior to installing the LCS I couldn't crunch due to overheating (see above post). The LCS and case fans are controlled via ASRock utilities. We've had a couple of days in the hi 80s ~ lo 90s so far and the temps are righteous. I get more noise from the GPU fan, which is automatically controlled by EVGA PrecisionX. If the GPU is idle for some reason, then the machine is silent.

@Michel Makhlouta - your temps seem to be in line with what I have with a smaller LCS. Even though Corsair advises to have the radiator fan(s) pulling air into the case, I've found the opposite to be true in my case as the exhaust from the GTX660 will be sucked into it. Pulling air from inside the case in a push/pull configuration is a leading contributor in keeping the CPU cooled. Of course, this reasoning might have to change once I upgrade to the GTX750TI FTW. As far as the fans on the rad are concerned are you using them in a push/pull config? I've found that has been the best configuration and I've been using Corsair LCS for many years and every evolution have been a push/pull. I'm not sure which version of the GTX780 you have, but if it has rear exhaust, you might want to at least try how I do it and then check your temps.

---

I don't buy computers, I build them!!

I checked the Corsair Link software and set the fans on quite, the fans' noise is much much lower, I can't even hear them. Peak temperature is now 69 degC, which is a 3 degrees higher than before. Thanks for the tip. I am not sure about overclocking... I will give it a try at some point but for now I am happy with the setup.

I will look into enabling the Intel GPU and use it for SETI as well. I thought keeping that disabled would lower the CPU temperatures. Worth a try I guess. I've got an MSI GTX 780 Twin Frozer and using MSI tool's overclock mode. I am running 3 WU on the GPU, maximum temp is 68... I don't go over 61 while gaming. The 2 fans on the GPU are quiet.

I have a thermaltake chaser mk1 case with two exhaust fans on top, 1 exhaust in the back, and 3 intake (front, side and bottom). I've installed the radiator under the fans on top, and installed the radiator fans under it as exhaust. So I guess I have some type of push/pull configuration. The fan in the bottom is useless I think, the PSU cables are blocking it.

The motherboard is also an MSI (mpower max ac), which supports 2 SLI or 3 Crossfire (weird...). Assuming the electricity bill doesn't go over my budget, I will be adding another 780 soon and maybe in the future a 3rd card just for crunching, since 3 way sli isn't supported... number crunching is becoming an addiction.

---

I have a thermaltake chaser mk1 case with two exhaust fans on top, 1 exhaust in the back, and 3 intake (front, side and bottom). I've installed the radiator under the fans on top, and installed the radiator fans under it as exhaust. So I guess I have some type of push/pull configuration. The fan in the bottom is useless I think, the PSU cables are blocking it.

Just ensure the fans are pointing in the same direction in the push/pull config, had a problem when I first started using this config many moons ago, but things worked out once they were pointing in the same direction. I do find that this config does increase the cooling effect by a few degrees. I have the same problem with my bottom fan, but I keep it there as it does supply some additional air flow to the GPU, not sure how much, but some.

---

I don't buy computers, I build them!!

IMHO the biggest mistake people make is to buy thermal compound at all. I use plain old Vaseline (petroleum jelly to non-USA folks) and have for all my years at SETI. No problems. Just a tiny bit works wonders.

It performs the major job of TC - it fills the microscopic air gaps between the CPU and the HS; it never dries out or runs out (high surface tension). And it is non-conductive of electricity and easy to clean up - no alcohol needed. I've had machines run 24/7 for literally years without notable change in CPU temps.

(In case you are wondering, I got the idea from reading an article from Australia some years ago - actually from 2002:

www.dansdata.com/goop.htm‎

YMMV!)

Greetings JRavine,

Yesterday I transplanted my computer into a new case. I have a Corsair liquid cooling system for my i7 860 CPU. I cleaned off the old thermal compound and used a dab of Vaseline instead.

I'm not sure if you were joking when you made the statement above, but I damn near fried my 860. When it booted up I saw a 98+ C temp. TThrottle completely stopped BOINC from crunching.

I searched out my thermal compound, cleaned the CPU and cooler, applied a dab of paste and everything is running as it should be.

If your statement above was made in jest, you should have indicated that. If in fact you do successfully use Vaseline in place of thermal paste, then more power to you for such an inexpensive solution. I will be using thermal paste as before and from here on out.

Keep on BOINCing...! :)

---

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath

I wasn't joking. Both my crunchers are running on Vaseline right now (i7-4770K and I7-3820). Sorry you had a problem with it.

---

Dan at Dan's Data did a comparison of different types of thermal paste several years ago. He also used toothpaste & Vegemite.
And the best thermal compound- Toothpaste. Even the Vegemite did better than several of the well known thermal compounds.

The advantage of good thermal paste over toothpaste & Vegemite is that it won't dry out (as fast). The better thermal pastes don't dry out as fast as the not so good ones; the actual performance differences between the different thermal compound are very small, and the way they are applied will often result in bigger differences in heat transfer than the type of compound used.

Even without thermal compound, there is a reasonable amount of heat transfer, it's just not as good as using thermal compound (in the appropriate very small doses). The fact that your CPU got so hot when you first booted up indicates the heatsink wasn't attached properly.

---

Grant
Darwin NT