42C idle at .93 vcore and 1.6 ghz -- i7 920 is a furnace!

[v8envy]

Not really asking for help, more griping about the thermal characteristics of the stock solution.

Running mprime (linux port of prime95, small fft) the CPU hits *80C* in approximately 9 or 10 seconds. My 3.2 ghz E2180 never went over 73 in the same case with the same peripherals running mprime for hours and idle temperatures were 10C lower under similar conditions. I have no idea what temperature the i7 tops out at, I kill the test as soon as my 80C motherboard thermal alarm trips and I see all those "80s" on my screen -- I'll try stress testing again after this beast is on water.

By now you might be thinking "you have the heatsink on wrong, fool!" Possible -- but unlikely. The temperature drops back to low 40s even more quickly than it rises (5-6 seconds). Oh, and I did verify all the heatsink posts were properly showing through the back of the MB before inserting into the case. The quick drop to idle temperatures tells me the heat is getting from the CPU to the sink just fine, the problem is shedding it quickly enough not to build up.

There are two 80mm extractor fans inches from the CPU heatsink, a 120mm side intake fan blowing almost directly into it and two distant 80mm front fans pulling air into the case. Plus the PSU fans. The case is a full tower. Removing the side panel did not appreciably change the idle CPU temperatures -- they seemed a degree or so higher at most. With the E2180 removing the side panel also increased idle temperature by about 1-2C and load by 5C.

Why am I bringing this up? I'm coming to the conclusion that the factory heatsink is completely and utterly inadequate for dissipating load heat generated by the i7 CPU at the rated frequency of 2.66 ghz, even in a reasonably well ventillated full tower case. It's complete and utter junk severely underengineered for the task of cooling the CPU it ships with at stock speeds.

 

[kazuyakun]

I can't agree more with you, and the same goes for AMD. The stock HSF doesn't even cover half the cpu's surface, this is probably to prevent folks from OCing it at the same time saving on material costs ~_~.

 

[PeteRoy]

my i7 idle at 60c

 

[v8envy]

How is that possible?

I mean, the i7 920 is not sucking more power out of the wall at idle than previous quads. In fact, it should be using quite a bit less. Unless you live in Death Valley with 50C ambients 60C idle seems completely unreasonable. Is your sink securely mounted? Is your PC locked in a closet? What are your prime95 small FFT load temperatures like?

 

[IntelUser2000]

You also gotta take into account the PCU. Now the CPU can measure temperature far more effectively than before. I remember reading something that said since i7 measures it closer to the CPU core compared to the previous generations, don't take temperature too seriously.

Really. If your CPU is truly running at 80C, you'll encounter really weird lag. I remember lagging in CS with my Pentium III system wondering why the internet lag is so bad. You gotta play around to really know how it feels like at that temperature. Surprisingly reinstalling the HSF with updated thermal paste fixed the problem. It was running at 70-75C at load and 50C at idle.

80C is rather unusually high for a motherboard though. Are you sure that's not the chipset at least?

 

[Schmide]

The I7 does some serious regulation of it's voltage for both power planes.

lostcircuits i7 power plays

By changing your voltage, even lower, you have told the system; you don't want it to dynamically alter it's power usage. I would try going back to stock settings and see what happens.

 

[v8envy]

Shmide, those ARE stock settings! With all power management options enabled in the BIOS the CPU is running at .93v and 1.6 ghz at idle, ramping to 1.2v and 2.66 ghz under load.

But even at that low power and frequency the factory cooler is starting to struggle with no load (with the fan at about 1300 rpm). At full load and the fan spinning well over 2000 rpm it's completely unable to even contemplate dissipating the "130 watts" of heat. My guestimate is this sink & fan is fine for a 65 watt CPU. For a 130 watt rated CPU it's not even close to adequate.

 

[OILFIELDTRASH]

Im totally with you on this. I went from a Q9400 to i7 920 and was shocked at how hot the 920 runs. I used the stock heatsink for about 3 days. Just long enough till my xigmatek hs came from newegg. I was careful how I used my cpu cause I was afraid of how hot it was getting with the stock hs. I was getting in the 80's running prime95 stock settings. I am now running my i7 @3.62ghz with stock voltage with a xigmatek dark knight hs with a scythe kraze fan. The only reason I didn't throw away the stock intel hs is so when I sell my cpu I can include the hs for a little extra cash.

 

[Thorny]

I wasn't having that big of an issue with my stock HSF, but I agree that it is barely acceptable IF you have a good case.

I had GREAT luck undervolting my i7 while still maintaining a stable overclock, and the difference in temperature was huge. Perhaps that may help you out. I'm running vcore @ 1.175 overclocked to ~3.3 and my temps only reach ~75 at full load. Ambient temp is ~34. I know you shouldn't have to undervolt your CPU to keep temps inline, but it remains a good option for you to do so.

 

[ChaosDivine]

1.052Vcore (LLC enabled but still droops a little) @ 4GHz, ~71C max.

 

[v8envy]

With a little bit of experimentation it looks like I'm stable at 2.8ghz at .93 volts. Load temps still hit 80 and beyond very very quickly though. I changed the BIOS voltage from auto to 1.1 as my first try, but that was worse. It never reduced the voltage from that level nor did it downclock to 1.6 ghz. Looks like I get to pick my poison -- power management with voltages ranging from .93 to 1.2375 and CPU frequencies from 1.6 to 2.67 ghz, or perpetual turbo mode and a static voltage. May need to read up on this BIOS and the linux power management daemons a bit more.

Idle temperatures are up to 44C, but the fan is staying at 1500 rpm. Something may be wonky with the fan control, that may be part of the problem.

In any event, I'm a bit irritated. It used to be you got more than you expect with Intel. The stock coolers were never great but they were always adequate, and sometimes enough for a bit of an overclock. With the current cooling solution the engineers were clearly told to design for a thermal envelope well below maximum -- just like they're advising OEMs to do for the rest of the system. It's like the Intel beancounters expect this CPU to only be used for light word processing. Yes, I'm planning on upgrading the cooling. But it's still irritating to need that rather than having it be an option.

 

[yh125d]

I'm sold on cool dual cores until they get the thermals of quads down. The only difference between the i7's and preshott is that i7 is keeping up with the competition


If I can run my e7200 oc'd to 3.5 cooler on my passive TRUE than an i7 at stock speeds with a low speed fan I'm damn sure not gonna pay $500 for the platform upgrade

 

[soccerballtux]

Originally posted by: yh125d
I'm sold on cool dual cores until they get the thermals of quads down. The only difference between the i7's and preshott is that i7 is keeping up with the competition


If I can run my e7200 oc'd to 3.5 cooler on my passive TRUE than an i7 at stock speeds with a low speed fan I'm damn sure not gonna pay $500 for the platform upgrade

Well that e7200 is using a lot fewer transistors and isn't hyper threaded...that right there means you'll be using more heat.

I'm no Intel fanboy, but Idle temps don't matter in the least if you're seated well.

 

[IntelUser2000]

Hey v8envy, what's your memory running at? Apparently memory overclocks raise the temperature of the CPUs by a lot(1600 to 1333 lowered it by 10C!). Stock configurations might even run it at 1066.

 

[thilanliyan]

You should try a Phenom I.

 

[PeteRoy]

realtemps and cpuz 1600mhz i7 running at 60c

 

[Thorny]

Originally posted by: PeteRoy
realtemps and cpuz 1600mhz i7 running at 60c

Looks like that screen shot was taken just seconds after load

What exactly are you trying to prove by posting a misleading pic like that? Even your GPU temp is high, is that the i7's fault too?

 

[PeteRoy]

Not trying to prove anything, I just have a similar problem like the OP with my i7, wasn't taken after load, summers are hot here and the temps in my room are 28C with high humidity

 

[Thorny]

There's only ~10 seconds difference between your max and current temps. Your temps are not anywhere close to normal even if your ambient temp is high, your temps are an indication of a problem elsewhere.

 

[PeteRoy]

It's because I took the screenshot 10 seconds after I ran realtemps, I have 60c at idle all the time,

I tried replacing the thermal paste using arctic silver 5 and did it like it said in their instructions and still have the same temps, in load I almost reach 90C

 

[v8envy]

To be fair, my problem is not quite as bad at idle. Ambients are pretty high here, about 25-27 as well. I'm coming to terms with idle temps -- after the HS compound had a bit of time to cure it's back to 42C idle even at 2.8 ghz& .93v. That's only 7C higher than my E2180, and I can live with that.

The problem is the stock HS can't keep up with load. The temps shoot up to 80+ fast, and show no signs of slowing down. I'd be hitting 100C tjmax in an estimated 30 seconds or less running prime95. That's the problem.

As far as RAM -- yes, I specifically got 1333 mhz ram, the 7-7-7-20 latency variety for this very reason. I figured for my workload lower latency combined with huge enough bandwidth would be my best bet. I'll see about raising latency of the RAM (which is running at 1.5v) a bit to see if that helps with load temps.

 

[Rubycon]

Why are people so critical about what a program tells them their cpu temp is? If it's stable all the way around why worry? Sure they're going to run warmer - you have FOUR cores in that little ah heck.

 

[Ben90]

Originally posted by: Rubycon
Why are people so critical about what a program tells them their cpu temp is? If it's stable all the way around why worry? Sure they're going to run warmer - you have FOUR cores in that little ah heck.

Because you can hit tjmax really really really really really easily under load

 

[Idontcare]

Originally posted by: Ben90

Originally posted by: Rubycon
Why are people so critical about what a program tells them their cpu temp is? If it's stable all the way around why worry? Sure they're going to run warmer - you have FOUR cores in that little ah heck.

Because you can hit tjmax really really really really really easily under load

Well in theory that is what TDP specs are for, but your point is well taken in the enthusiast world in the context that shmoo plots (Vcore vs. GHz) are temperature dependent, not power-consumption dependent, in the physics of the equations and as such our ability to extract stable clockspeeds at any given Vcc is temperature dependent and thus a high(er) temperature operating processor implicitly means lower maximum stable clocks at any given Vcore.

Threshold voltage temperature dependence

Note the basic form of the threshold equation is that Vtn = Vto + Y*(A+B) where B is the temperature dependent part.

Basically temperature changes the threshold voltage, the minimum voltage needed to form the channel underneath the gate oxide when attempting to switch on the xtor.

Higher operating temperature means higher threshold voltage, which means you have to set Vcc higher in order to still operate the IC in a stable fashion.

And since Vcore (and clockspeed) determines power-consumption, which feeds back into temperatures, you can easily get yourself into a corner where stable operation requires more Vcore which then raises your temperatures even further, requiring even more Vcore for stable operation, which raises your temperatures even further...so temperature matters to us OC'ers because the physics of transistor device operation is impacted by temperature.

I agree if you are not worried about maximizing the OC in your system or minimizing the power-consumption (even at stock clockspeeds) then there's really no cause for concern if your cpu operates close to, but below, the spec'ed TJmax.

 

[Rubycon]

Originally posted by: Ben90

Originally posted by: Rubycon
Why are people so critical about what a program tells them their cpu temp is? If it's stable all the way around why worry? Sure they're going to run warmer - you have FOUR cores in that little ah heck.

Because you can hit tjmax really really really really really easily under load

If you're overclocking, then yes. A stock cooler should never be used (for long anyways) for an overclock. If you properly mount a stock cooler using the pre-applied TIM and run at stock speed you should NEVER approach TjMax regardless of what you're running. Unless your ROOM is like a furnace.