So I was thinking of overclocking when...

[ZeroRift]

Hi, I've been lurking around the forums for quite some time now, but I was hoping I could get some personal advice...

I recently dove in to the i7 party after desperately needing an upgrade from my athlon 64 3800+ system and upgraded to an i7 860 system. (I unknowingly shot myself in the foot a bit on that one, but that's another story...)

I had been toying with the idea of performing some minor OC (just increasing bclk to 140) with stock cooling / voltage after hearing about how people have had a lot of success with getting very high clocks with the i7 series processors. So before I did anything, I loaded up OCCT just to see where my temps were at before I tried anything and to my surprise my proc hit 80C within the first minute of testing!

Needless to say, I'm a bit concerned. I do quite a bit of gaming so I'm worried that I'm going to exceed that when I actually get around to loading up some games on it. When I first built the system, I actually hit a thermal shutdown when I discovered the CPU fan WASN'T TURNING. After some quick examination of the CPU heatsink, I found that the wires that run around most of the circumference of the fan had been pulled too tight and were actually blocking the fan's motion. Lame!

Fortunately, the thermal shutdown seems to have saved the cpu. I reinstalled that same fan assembly figuring it would be fine as long as it turned.... guess I was wrong....

I'm looking at replacing the stock cooler entirely, but I wasn't sure which route to go. I've heard good things about megahalems, but that seems like an overkill solution for me. I'm not looking at doing any more OC than pushing the bclk to 140 or so, and I also value having a quite system.

Any recommendations?

 

[happy medium]

Coolmaster hyper 212 is regarded as a fine cooler and at a nice price.

http://www.amazon.com/gp/product/B00...namitedata.com

 

[LoneNinja]

Coolmaster hyper 212+ is a great cooler for the price, quiet and performs well for all but extreme overclocks. I think microcenter has them for $19.99 still if that's an option for you than it's an even better price.

 

[Tsavo]

212p will be fine so long as you don't get too wild.

 

[n7]

Advice above is indeed excellent.

BTW, 80c isn't going to kill anything...shut down temp is well above that.

 

[Reincus]

Intel chips measure temp inside the chip, not at the surface or on the motherboard. 80C isn't going to kill anything.

AMD chips, on the other hand, should not be operated at above 55C steady state, but that has to do with where the temp is measured. Internally, the temps are similiar.

 

[nonameo]

Intel chips measure temp inside the chip, not at the surface or on the motherboard. 80C isn't going to kill anything.

AMD chips, on the other hand, should not be operated at above 55C steady state, but that has to do with where the temp is measured. Internally, the temps are similiar.

eh? I was pretty sure that starting with athlon XP, AMD started measuring from on-die as well.

 

[Reincus]

They are on die. I did a little research, the difference in the 2 brands of temps has to do with how the digital temp is calculated, not the location of the sensor as I had stated.

[FONT=Arial, sans-serif]This is how the program works: (Reffering to Core Temp)[/FONT]
[FONT=Arial, sans-serif]Intel defines a certain Tjunction temperature for the processor. This value is usually in the range between 85°C and 105°C. In the later generation of processors, starting with Nehalem, the exact Tjunction Max value is available for software to read in an MSR (short for Model Specific Register).
A different MSR contains the temperature data. The data is represented as a Delta in °C between current temperature and Tjunction.[/FONT]
[FONT=Arial, sans-serif]So the actual temperature is calculated like this 'Core Temp = Tjunction - Delta'[/FONT]
[FONT=Arial, sans-serif]The size of the data field is 7 bits. This means a Delta of 0 - 127°C can be reported in theory. In fact the reported temperature can rarily go below 0°C and in some cases (Core 2 - 45nm series) temperatures below 30° or even 40°C are not reported.[/FONT]
[FONT=Arial, sans-serif]AMD processors report the temperature via a special register in the CPU's northbridge. Core Temp reads the value from the register and uses a formula provided by AMD to calculate the current temperature.
The formula for the Athlon 64 series, early Opterons and Semprons (K8 architecture) is: 'Core Temp = Value - 49'.
For the newer generation of AMD processors like Phenom, Phenom II, newew Athlons, Semprons and Opterons (K10 architecture), and their derivatives, there is a differnt formula: 'CPU Temp* = Value / 8'.[/FONT]
[FONT=Arial, sans-serif]The sensor in AMD CPUs can report temperatures between -49C and 206C.[/FONT]
[FONT=Arial, sans-serif]*CPU Temp is because the Phenom\Opteron (K10) have only one sensor per package, meaning there is only one reading per processor. [/FONT]

 

[ZeroRift]

Microcenter is conveniently close, I'll pick up that HS and let you guys know how it goes.

 

[veri745]

AMD chips, on the other hand, should not be operated at above 55C steady state, but that has to do with where the temp is measured. Internally, the temps are similiar.

This is bogus. TCaseMax for AMD chips is between 68C and 95C depending on the model. I think Quad Phenom II's are right around 73 - 80C

 

[ZeroRift]

Well, I installed the heat sink last night and ran OCCT against it. Cooling is vastly improved. My proc now stays right below 60C even when being maxed out.

I think I may have goofed the thermal paste though... seems like other people get temps almost 10C cooler than I do at the same load. Of course my AC is also out so that may have something to do with it

Anyways, I've also notice that under load there is up to a 5C difference in temp from one core to the next when all four cores are under a full load. That variance was also there before I swapped coolers. Is this something normal to see? I figured they'd be within a degree of each other....

 

[n7]

Core temps definitely will vary.
One CPU will also get different temps than the next, so sometimes you get a hotter running one, or cooler, which can be that 10c difference alone.

Of course, it never hurts to reseat the heatsink if you think it's not quite right.

 

[Lark888]

Hyper 212+ is pretty different when it comes to applying thermal paste because of the direct.contact heat pipes. I had to reseat a second time after finding a good article on the net. Regardless, you got a good reduction in temps for a great price. This is my current favorite air cooler.

 

[JasonCoder]

This thread delivers on the cooler advice alone.

 

[Reincus]

This is bogus. TCaseMax for AMD chips is between 68C and 95C depending on the model. I think Quad Phenom II's are right around 73 - 80C

http://support.amd.com/us/Processor_TechDocs/43375.pdf

I don't know of anyone who runs their Phenom II at 70C, even under load.

 

[Carmen813]

I have the Hyper 212+, and it's great (i5 OCed to 3.6 Ghz, could probably go higher). It's a bit of a headache to install though, so watch this video:
http://www.youtube.com/watch?v=hSq_xbxsm7Q

Oh and make sure you grab a second 120mm fan for it so you can setup Push-Pull, drops temps by 3-5c.

 

[ZeroRift]

Already installed it but thanks!

That would have been useful, it really was a pain to install. I had to reseat it 3 times to actually get it mounted properly. It's my own fault though; I should have thought about it a bit more before I first set it on the proc.

I don't think I'll be messing with it again anytime soon. My temps are very good, and the improvement of a few C just isn't worth the trouble to me since I'm not exactly pushing any limits here. I also read up on applying thermal paste to Direct Contact Heat-Pipes and found out that, aside from having to reseat it a few times to get the mounting right, I had done the recommended procedure by "filling the gaps" between the pipes, so I'm not too worried about it.

Overall, I'm pretty pleased with the 212+, and at > 1C/$, it's tough to beat price wise.

Thanks for the advice guys!

It's interesting to me that different cores on the same die under the same load will operate 5C apart - I didn't think there was that much divergence between core thermal performance on the same package. Do you know of any cool resources that talk more about that n7, or is that just accepted fact?

Edit:

Ah, looks like 386DX just answered that yesterday:

"According to Intel tech documents the internal sensors of the Core/i CPU's are only very accurate at high temperatures (ie. near its thermal limit). At around 50C the sensors can be a 5-10% off, the accuracy increases to within 2% at the thermal limit (TJmax). So your temp of 60C could really be anything from 54-66C which is within the temps of the other cores."

I'll take a look there.

 

[slugman]

I just joined the i7 party and had exactly the same questions: temps (90+C with stock cooler @ stock clocks in prime95), temp differences (core 2 seems to run 3-5C cooler than the others under load). I have some threads really high up here asking the same questions.

Bottom line: temp sensors have +/-10% accuracy far from tjmax, 5C variation is totally normal. And the stock cooler sucks.

212+ made temps really great. Nice to see you've got it too.

Been playing around with overclocking and 160 bclk seems very stable to far with almost no vcore increase. If you wanna try:

Bclk 160
Vcore: Normal (1.1xxV)
Dynamic vcore: Normal (1.12V under full load, .864V idle @ x9 multiplier)
Line load calibration: ENABLED (stability problems without it @ these settings)

only other thing I changed was the ram voltage to 1.66V because mine calls for 1.65-1.84V at 1600Mhz and 9-9-9-24 (adata gaming ram).

went from 36-38Gflops in LinX to 42-43 reliably.

73C max load after hours (less last night, like 68C. I don't have AC and my temps are higher than most. The room I'm in might well be 35C right now.)

Post back success if you try the same? Or if you go higher.

 

[Tsavo]

^ Yep, my Core i5/750 will get a 4 degree spread using LinX.

I replaced the stock TIM with thermalcote and dropped 6 degrees C.

 

[slugman]

Just an update on my overclock advice: it's not all gravy:

LinX fails after 10 or so passes. Restarting LinX yeilds 15% performance decrease (37 Gflops @ 3.36 Ghz - was getting 43 @ 3.36 before). HW monitor reporting CPU using 207W instead of the previous 233W at that clock. Something weird. Anyway, back at stock everything seems fine, 38 Gflops, 34 passes cleared and counting. CPU using 200W as expected... Probably the voltage was too low? Hopefully no permanent damage done.

 

[n7]

Yeah likely you needed a bit more vcore.

LinX tends to do funny things & get poor scores when it's just a tiny bit off.

 

[slugman]

That's too bad... because in my bios there's seemingly no way to let it have, say 1.2V and still allow for .864V at idle. If I set Vcore to normal and dynamic vcore to +.10625 then I get 1.2xx at load but at idle it only drops to .99x V.

Anyone know if there's something I might be missing? It's a gigabyte board. (P55A-UD4P, bios F10)

 

[Tsavo]

Just an update on my overclock advice: it's not all gravy:

LinX fails after 10 or so passes. Restarting LinX yeilds 15% performance decrease (37 Gflops @ 3.36 Ghz - was getting 43 @ 3.36 before). HW monitor reporting CPU using 207W instead of the previous 233W at that clock. Something weird. Anyway, back at stock everything seems fine, 38 Gflops, 34 passes cleared and counting. CPU using 200W as expected... Probably the voltage was too low? Hopefully no permanent damage done.

Strange

My Core i5/750 at normal volts (1.152) and RAM at 1.56v @ 160 baseclock pulls 48.6 Gflops at 60c all day long, but then again my room temp is 22C.

edit: HWMon reports CPU @169 watts max.

 

[slugman]

Is it possible that hyperthreading is *decreasing* my performance? Wonder why my performance is lower? My RAM is at 1.66V.... could that be causing any problems? Its what Adata says to have it at.

Also noticing that your voltage is .03V higher than mine. Wonder if that makes the difference?

 

[Tsavo]

Is it possible that hyperthreading is *decreasing* my performance? Wonder why my performance is lower? My RAM is at 1.66V.... could that be causing any problems? Its what Adata says to have it at.

Also noticing that your voltage is .03V higher than mine. Wonder if that makes the difference?

Oh, ya. Disable HT and give it a go.