Intel Extreme 3960X Processor Temps, TDP, & Wattage Re Folding 24/7

WolfpupNX

Limp Gawd
Joined
Mar 13, 2007
Messages
167
Hi all,

I'm currently using a quite powerful gaming rig using an Intel 3960X (Extreme) processor (NON-overclocked, aside from the stock Turbo mode built in by Intel) and have just begun folding with World Community Grid, a project that works on various disease research and humanitarian projects. I would like to allow the computer to work as much as possible 24/7 to benefit this project (aside from when I'm gaming, etc.) but am concerned about a few things, so I wanted some input.

I've had 0 problems during real world use or gaming; this thing is solid as a rock, has plenty of fans as well as liquid cooling to the processor, but the tower itself isn't huge even though it is efficient (midtower). When folding at 100% capability (on the stock liquid cooling provided by the manufacturer), I've noticed my individual core temperatures approach the TJMax of 91C (most cores in the high 80-90 range, once in a while one of them going over 90 but never above 91 - it throttles if it gets to that point). I have noticed only very minor throttling (below the turbo speed of 3900 MHz...I've seen it briefly hit 3800 but it never stays throttled for long), and the processor maintains these temperatures without any crashes. I realize higher temperatures may impact processor life to a small degree, but are these temperatures and this workload (100%) safe for 24/7 (or close to it) operation? If not, is there a recommended temperature I should NOT go over for a 24/7 load?

I've also played with the default (60%) and even a bit lower (50%) which makes temperatures more variable and lower overall, but it still gets fairly warm (never really throttles at these settings though).

Additionally, the utility I'm using to monitor (Core Temp) reports TDP as 130 Watts, but I notice Power is reporting ~165W. Is this a concern? Again, the processor is NOT overclocked; I can't explain this unless it is a function of the Turbo mode.

I realize heavy use can affect processor life, but that's only a concern if it's going to greatly reduce expected useful lifetime. For example, if folding would reduce lifespan from, say, 7 years to 6 and a half, that's fine, as I won't be using the architecture for 7 years. But if it would reduce it to 2, obviously that would be silly to do.

Thanks in advance for any help/advice! I've heard everything from these are made to hit max temps 24/7 and you'll be fine to never go above 60% and everything in between, and I just want to know the real story.
 
I'm running my 3930k on a Sabertooth mobo at 4.7GHz with no voltage changes, and it runs about 50C on a Thermalright true black 120 cooler.

I've had no issues whatsoever, and it loves gaming.

I don't see that much power draw even using euler3d, which is the best loading program I've found. It hits all cores to max, and will find weak links easily. :) (It can be configured to do fewer cores.)

If your chip is really drawing 165W, it's drawing a lot more than my chip is, with my chip running at a higher clock.
 
DrBorg, thanks much for the reply! Can I ask what program you're using to monitor power draw? Is it Core Temp or something else?
 
If your chip is really hitting 91C, while NOT being overclocked, there's defiantly something amiss with your cooling setup. I would venture to say your block isn't seated well on the CPU and/or no TIM. My 2600K running at 4.8Ghz only hits about 72-75C, folding 24/7 and its on a Corsair H-60 with push/pull fans.

Realtemp reports my chip using between 103-106W at 100% usage across all the cores, but I have no idea how accurate that is. With your chip being 130W to start with, at 100% usage across the cores in Turbo, 165W isn't out of the question.

With your chip running at 85-90C, that will defiantly reduce the lifespan of it. How much, I have no idea and I would venture to say no one else would either. The best we could do is guess. I would defiantly look into your cooling setup though. 91C is NOT normal, especially for a closed circuit water cooler. Folding does stress your CPU MUCH more then nearly every other program around.

When I'm overclocking a new system, I run Prime 95 and IBT for 24 hours to start with, but more often then not, folding stresses the CPU much more then either of those programs. If it can fold for 24 hours without crashing, its stable.
 
I have a 3930k (very similar to your chip) overclocked to 4.2Ghz. My chip never goes above 70C on the hottest core (using corsair h100i liquid cooler). Even overclocked, your chip should NOT be hitting anything near 91C and it looks like you're running it at stock. I wouldn't expect your chip to go much above 60C at those settings.

I would make sure the fans on your radiator are all spinning and that you have the cooler mounted correctly (make sure there isn't a gap between the CPU and the cooler). Also make sure your pump is functioning correctly. I wouldn't stress your processor until you solve this heat problem. That kind of heat will cause a chip to degrade much quicker. For 24/7 use I would try to keep temps below 70C.

Also, you might want to consider manually setting the voltage. If you are folding all the time there is really no need for offset. You shouldn't need much more than 1.200v @ stock and this can help with temps as well.
 
Thanks so much for the help!

So my cooling solution at present (what shipped with the machine) is the Asetek version (550LC) of the Corsair H50. Figuring that out, I looked at the manual and realized they recommend using the rear 120 mm fan as an intake, not an exhaust, and on my rig (possibly due to my error when working on the memory previously), it was configured as an exhaust. I switched it, and I'm now getting significantly better temps, though still getting into the 80's at sustained max load. I'm happier it's cooler, but maybe I just need a bigger cooler if I want better temps? Appreciate all the input so far. I've never installed one of these, so am a bit nervous to upgrade.
 
Thanks so much for the help!

So my cooling solution at present (what shipped with the machine) is the Asetek version (550LC) of the Corsair H50. Figuring that out, I looked at the manual and realized they recommend using the rear 120 mm fan as an intake, not an exhaust, and on my rig (possibly due to my error when working on the memory previously), it was configured as an exhaust. I switched it, and I'm now getting significantly better temps, though still getting into the 80's at sustained max load. I'm happier it's cooler, but maybe I just need a bigger cooler if I want better temps? Appreciate all the input so far. I've never installed one of these, so am a bit nervous to upgrade.

You should really consider upgrading your cooler to a dual 120mm radiator. The best solution would be to build your own custom loop. You can also go with something like this:

http://www.newegg.com/Product/Product.aspx?Item=N82E16835181032

If you don't have room for a dual 120mm radiator you might want to consider upgrading your case as well.
 
Make sure your motherboards voltage is on "Normal" not "Auto" It should like the board has the voltages cranked for some reason.
 
I don't have room for a dual radiator unfortunately, but I could certainly do a much beefier 120mm than I currently have.

Will check the voltages and post an update.
 
Just checked the BIOS and you can manually set voltages, but there is no "Normal" setting per se (ie. you can set it to completely Manual, or set it to Auto; there's nothing even approaching a Normal setting). Auto for most things is also the recommended settings with the configuration sheet that came with the system - the only exception is DRAM, which is set to a constant Manual 1.5. Unless I'm missing something.

I should also mention this thing is rock solid stable. It may run a tad warm, but its also never crashed due to hardware problems. Also, I would think if it WAS over voltage, Turbo on the processor would turn off, which isn't happening. According to the specs for Intel's Turbo mode, it takes heat and power into consideration and will only go the faster speed if it's within certain criteria. And at this point it's pretty hard to get it to throttle at all (it can be done with extended 100% usage, but barely, and I'm not sure I've seen it at all since switching from exhaust to intake for the radiator).
 
So I installed a new cooler tonight (Corsair H80i), which is the biggest my case will fit, and my temps are much improved. Folding at 100% with fans set to maximum gets me down into the low 60's. With fans on Quiet, I do hit 75C on a few cores, but it seems to max out there, which is MUCH cooler than the 91C throttling point I was hitting before.

I wish I could run Max all the time but this thing is loud on Max - that's fine when I'm not home, but it's too much for normal usage.

So, considering I'm at stock voltage and TJMax on a core is 91C, should I be reasonably safe pushing this thing at 75C 24/7?
 
Core voltage is reading at 1.312 at full load according to CPU-Z. Thanks! And I will try to run that benchmark in a bit.
 
Intel's ARK lists the voltage range for this part at 0.6V-1.35V, so I appear to be within that range. Or are you saying at stock it shouldn't be using that much and can use less?

Keep in mind this is part of their Extreme line, and when I say full load, that means at constant Turbo mode (self-overclocking to 3.9); not sure if that might make a difference.

It was factory configured like this near as I can tell (I checked the settings sheet that came with it), and I see nothing in the BIOS that looks improper or overclocked; the only overclocking that should be in place is the Intel-designed/sanctioned version.

Thanks much for the input!
 
Also, Raghar, let me know what version of that benchie I should run for ya and if you want me to turn off HyperThreading (via the BIOS?) first.
 
Eh, I guess that about normal now that I do some research. Most of them seem to stop around 1.3 when under full load. You're seems a touch high, but that shouldn't change the temps that much. I think you might have bad core to IHS contact.
 
Thanks! If I'm reading you right, the IHS (Intel Heat Spreader?) is the metal cover over the chip, which I'm definitely not comfy trying to remove, so there's nothing I can really do, correct? It just falls into some chips have higher/lower properties (heat, voltage, etc.) than others category?

That said, I'm much happier with these temps than the previous ones, and so far no stability issues.
 
Also, Raghar, let me know what version of that benchie I should run for ya and if you want me to turn off HyperThreading (via the BIOS?) first.

I removed the old link and left only latest version. Do the test in any order you want. You have 6-core CPU thus you need to run pack6C without HT, and pack12C with HT. (In fact, you should also run pack.cmd without HT, just for verification.)

And for your problem. Well occasionally MBs use too high voltage on auto, either there is wrong VID in CPU, or they are paranoid. The other problem might happen because too aggressive LLC would increase voltage above BIOS settings. So carefully play with both.

For example I don't use LLC at all, and I have 1.200 V at full power in orthos. (down from slightly above 1.25) You have 22 nm CPU and I find very strange your voltage is larger than what I'd need for an extreme overclocking of 45 nm CPU. You should be at about 1.175 V for general usage.
 
Thanks! If I'm reading you right, the IHS (Intel Heat Spreader?) is the metal cover over the chip, which I'm definitely not comfy trying to remove, so there's nothing I can really do, correct? It just falls into some chips have higher/lower properties (heat, voltage, etc.) than others category?

That said, I'm much happier with these temps than the previous ones, and so far no stability issues.


Please don't try to remove the IHS, it is soldered to the chip and you will destroy it.

That being said, your temps are still a bit high for stock speeds. More than likely this is something your motherboard is doing on auto settings. Usually dropping the PLL voltage a little will help with temps, but if it is stable then I wouldn't worry about 75*C. I would worry about spending $1000 on an X-series unlocked CPU and not overclocking it though! :p
 
WolfpupNX,
I would set the CPU Vcore to 1.2V, boot into Windows, and run Prime95..If a worker thread errors out quickly, reboot and increase the voltage to the next lowest level (ie 1.211 etc) until you can run Prime95 stable on all 12 threads for at least 8-12 hrs...

The 3960X are top binned chips, and I really do not see them needing that much Vcore for the stock Turbo speed of 3.9Ghz..It's a very easy test to do, won't hurt anything, and will hopefully net you at least 3-5C better temps..

There is no way I would be happy with a $1000 cpu throttling, or even getting up in the upper 70's @ the stock voltage with a decent AIO WC solution..They shouldn't really be up that high with the stock Intel cooling solution and decent airflow in the case..

If reducing the voltage doesn't help, I would strongly consider RMA'ing the chip..I know it sucks, but Intel makes it really quick and painless...
 
Thanks again for the great feedback. =)

I realize the value of the Extreme series is questionable in the opinion of many for a non-overclocker, but I really just wanted stability, high speed, and reasonable temps, so it seemed like a good fit for me. And it has lived up to this, though for folding I did want lower temperatures if I could achieve them.

No worries I wasn't about to try removing the IHS ^.^ I know I would have killed the chip.

Also, Raghar, benchies run and results in thread.

So I'm thinking about trying to reduce VCORE, and I've read a bit about it. A few questions tho' before I try it:

1.) One thing I've seen is that your memory configuration can affect needed VCORE. I am running all 8 slots, 32 GB total, of DDR-1866; could this explain why my VCORE is so high, and do I risk any damage to anything by manually setting it at, say, 1.2?

2.) I've seen that it's best to disable LLC when setting manual voltage to keep things simple and assure the maximum voltage you set is the actual maximum voltage; is this correct?

3.) Do I need to touch anything else, like VCCSA, VTT, etc.?

Thanks! A bit nervous...I'm used to working on computers, but I've always been shy to screw with voltages.
 
1: no, you can't damage a chip or hardware by running at lowered voltages: you can however damage software: so make sure you prime95 test the CPU for several hours to ensure that the chip is stable: if it can run without showing errors: you have a successful undervolt.

2: I actually enable LLC when overclocking because this keeps the voltage stable. This is why some novice overclockers think their system is stable when it is not: at lower loads the system is fine, but as soon as a huge load is applied to the machine, the voltage droops down and the chip can no longer sustain it's frequency. LLC is A-OK.

3: you shouldn't have to.
 
I already said, your problems might be because of a too aggressive LLC, definitely disable it when you need low voltage.

The only problem that can be caused by too low VCORE is your CPU will not boot. (as long as you will not go under 0.8 V it can't cause worse problems) But some MBs are designed for overclocking, and return settings to safe values automatically, otherwise you need to do a BIOS reset.
 
Last edited:
Thanks for all the info! I don't have time to play with it right now, but I may give this a shot the next few days, and I'll let you know how I make out if I do. =)
 
Back
Top