bringing an I7 online
13 Apr 2011 22:24:52 UTC
Topic 195750
(moderation:
I was playing with overclock settings and believe I literally blew up my older IP35 Pro motherboard (Q6600 quad core). I heard a pop and everything went black. I managed to get back into the BIOS, but no drives were seen. OK, so I figure I did something bad and now it was time to repair or replace. I chose to replace.
I installed a new mobo with an I7 960. I've only just now gotten to the point where the system is stable and I can get online. Somehow the BIOS must have gotten corrupted, because it wouldn't recognize the onboard ethernet port. I cleared the BIOS and voila', everything is now recognized. BOINC has just been reinstalled.
I'm using SIV to monitor the cpu's/gpu's and I'm getting scared. SIV says that my cpu's are running anywhere from 90C to 99C. That seems awfully hot to me since the old Q6600 used to run in the mid-50's to low 60's C.
The boxed cooler is what is on the cpu now, but I have the CoolerMaster fan from the old rig I could possibly use. Short of taking everything apart again, is there a way I can reduce the heat generated? I'm not overclocking. Not just yet. This rig is running at it's base 3.2ghz.
bringing an I7 online
)
I would source a better TIM and re-fit the cpu cooler, assuming you are confident that it is installed correctly.
Mine is running at 65 degrees under a 90% cpu load.
HTH
dunx
P.S. What is SIV, I use coretemp myself.
!
)
!
SIV is a very cool program
)
SIV is a very cool program called System Information Viewer. There's more information available in there than I've ever seen from a single program. Some time ago there was a link in either these forums, or the Collatz forums that pointed to a place to download it from.
I wish I could be more specific on where I got it, but it's been almost a year since I downloaded it.
Now, I run an i3-540, with
)
Now, I run an i3-540, with the stock cooler but self-added arctic silver heat paste, and mine runs under full load at 43C-47C, depending on what runs. Looks like Einstein in combination with Milkyway generates more heat than EAH + Primegrid, or EAH + Rosetta.
Anyway, 90-99C is very close to its maximum of 105C, best stop putting a load on that CPU immediately, until you've taken it apart and checked that there's thermal paste between the CPU and the heatsink/fan, and that the fan is actually spinning... If you put thermal paste in there your self, make sure it wasn't too little, or too much. The best amount ever is about the size of a grain of rice, in the middle of the CPU and just pressing the heatsink into that... no need to spread it first.
I clean my system with compressed air as soon as I start hitting 60C (just cleaned yesterday, lots of cat hair in the HSF ;-)).
Really, unless you like the smell of melting plastic and burning copper, stop putting that system under load. Don't run with temporary workarounds either (software throttling). It's just not worth it.
RE: I was playing with
)
You don't say how much checking you did before deciding to replace, so please excuse my intervention if you've already checked your PSU. When you actually hear a pop, the most likely thing that has failed is the PSU. The PSU could still be giving some output but not good enough for satisfactory operation of the computer. It's well worth trying a replacement PSU if you can - even to the point of temporarily pinching the one in your new machine just for the test.
If you can actually get the machine to POST but the BIOS looks scrambled, or even if it won't POST at all, it's well worth removing the CMOS battery for 24 hours. Then put the battery back in, apply power from a known good PSU and try again. You'd be amazed how many times an apparently dead mobo will come back to life when given that treatment. Don't give up too soon :-).
It can't really be that hot. Your CPU's thermal protection would have kicked in before that and it would have shut itself down. Maybe you have an old version of SIV that doesn't properly understand the latest hardware in your new motherboard. You should make sure you get the latest version and you should try an alternative program like coretemp to see what differences there might be.
If your CPU is really too hot, check for the proper mounting of the heatsink. I've not seen an i7 mounting arrangement but if it's anything like that for the LGA775 socket, you really need to be careful with ensuring that all 4 mounts are properly locked in place. Even though I take a lot of care with these, I occasionally find a case where a machine starts running really hot when first started up. Sure enough, it will usually be a loose mount point.
If you're planning to run all cores at 100% load and eventually to experiment with overclocking, you really need a decent cooling solution. Take a look at this page on the hardware secrets web site for a large number of reviews of cooling solutions you might like to consider. If you pick the (currently) second review listed there and read the whole review, you'll get a good idea of what they do to test a cooler. On page 6 of the review you get comparison tables that will allow you to see how different brands and models compare against each other. This is invaluable information for picking a good performing unit that is within your budget. It's also very good for showing the poor performance of the stock coolers and some of the 3rd party offerings.
Cheers,
Gary.
RE: ... checked that
)
Actually, the thermal grease is of most importance when the surfaces of the CPU heat spreader and the base of the heatsink are either convex or concave (as they always are to some extent) since that can eliminate a significant air pocket. If you have perfectly flat surfaces then you don't seem to need any TIM at all. In the past, I did a number of experiments with lapping both surfaces and was quite dismayed to see just how many of the ones I experimented with were significantly convex or concave (or all shapes in between :-) ). Once I had both surfaces as flat as possible (and that doesn't mean scratch free or highly polished) I found that I could run for extended periods with no TIM at all, and that subsequently applying arctic silver made things slightly worse, possibly, but certainly not better.
I'm not at all suggesting that the average participant should be lapping the two surfaces, quite the reverse. What I'm really suggesting is that these surfaces do vary in their degree of flatness. Sometimes the irregularities are enough to make the gap between the surfaces relatively large both in distance and areal extent. The TIM may very well completely fill this gap and so be thick enough to act as a significant retardant to efficient heat transfer. The only way to rule this out is to clean off all TIM from both surfaces and use a straight edge to test both surfaces for abnormal levels of convexity or concavity. If it's visible to the naked eye it's probably too large to be acceptable.
I tried this with a number of samples a few years ago and when I thought I could see a problem I'd test it out by starting to lap the surface. The problem would be confirmed by the visible high spots and the extent of the low areas not touched by the lapping grit paper. The time it took to grind down until no low spots were left was quite indicative of the problem.
So, it's quite possible that running hot could be just due to a badly non-flat surface and not necessarily due to lack of TIM.
or not spinning too slowly.
I've had a case or two where the (quite new) fan is spinning about half speed and stops almost immediately when power is removed. Intel fans these days are of much poorer quality than they used to be, or so it seems.
Cheers,
Gary.
Thanks for the reply Gary.
)
Thanks for the reply Gary. The PSU is a-ok. It's running my nVidia card like it always did and a VOM shows proper voltages all the way around. What I can't measure is amperage under load, but then, only the gpu needs that high amperage and it seems to be working fine.
The fan had a little heat compound on it which I assume spread evenly. It has a mounting pattern very similar to the LGA775, only slightly larger, which is what the old PC was, however, I'm now not convinced that it is securely mounted. Right now, the boxed fan is using the posts that have the center tab to push down and lock. The old fan actually had nuts on it's mounting posts providing a firmer mount.
I was just hoping against all that I didn't have to take everything apart again to reseat the fan. If I turn off BOINC and all related processing, temp drops to a respectable 50 degrees or so. But the minute I turn BOINC back on, it jumps straight up to 90+, and does it very quickly too!
BTW - my SIV version is dated back in Dec, however I just went to their website and there was a new version released today. I will d/l and install.
RE: The fan had a little
)
Hold on, what do you mean with 'mounting pattern'? Was it where it showed the thermal compound from the last CPU you used the HSF on? Did you clean the underside of the HSF and put new thermal compound on, or left everything in place and reuse the old one?
Any air-pocket between the top-side of the CPU and the underside of the HSF is insulation, where the heat will stay in the CPU. So if the HSF doesn't mount firmly against the CPU...
50C for idle is HOT. I checked around through Google and the i7-960 has an idle temperature of between 20 and 30 degrees Celsius, depending on what environment it is in. That's without overclock, with all kinds of fans.
50 Celsius is not acceptable. Not for idle. It shows that you have a problem with your HSF, be it the air-pocket, a loose fan or both.
For comparison, if your car would stop every 5 miles because of an overheating engine, would you ask the garage to mount a bigger fan or would you want them to fix the damn thing? ;-)
Wow! Big Difference. I took
)
Wow! Big Difference. I took the Coolermaster fan off of the old fried rig and mounted on the new mobo with the I7. The boxed fan had a circular footprint, the Coolermaster has a square footprint. No wonder it wasn't keeping cool. Due to the circular nature of the boxed fan, I suspect 10-15% of the cpu was uncovered.
All processors are running full bore at 3.33ghz. Right now the system has been up for about 10 mins and the temps are at least 20 degrees cooler!
Core 1 = 79C
Core 2 = 74C
Core 3 = 70C
Core 4 = 74C
Fan speed is about the same as the boxed fan. Currently at 1900 rpm. I have it set in the bios to run wide open, but to warn me if it stops.
RE: Wow! Big Difference. I
)
The shape of the contact area is unlikely to be the source of the difference. Firstly because Intel designed the cooler and frying chips isn't in their best interest. Secondly because the heatspreader is significantly larger than the CPU die underneath it. A number of aftermarket coolers have intentionally curved bottoms so that all the contact force is concentrated in the center of the die while the thermal paste is all pushed out to the edges.
Those numbers are on the high side for what you should be getting with the stock cooler at default clock speeds. If your cooltermaster is a big model then there's something definitely not right. IF you're using the evil pushpins odds are one isn't seated all the way (they're hard to figure out and the pamphlet intel includes is IMO confusing enough to be worse than useless.) IF your aftermarket cooler is a baseplate and screws type, my thoughts would be to wonder what your ambient temperature is. The other possibility is that at least one mobo vendor screwed their software up. I don't recall which, but my friend bought one; he had temps in BIOS/their software 20-30C higher than what coretemp was reading, and apparently it's a commonly reported problem on the companies support forums.