Hi everybody,

I have been lurking for more than a year now, thanks for all the good information on this site!

My build was done from scratch in 09-2012, in the meantime I changed a few things (will mention the original part in brackets):

• Antec Solo II
• Seasonic X-400 FL (replacing a bequiet! Pure Power L8-CM 430W, which replaced an L7, both started whining under load)
• Gigabyte GA-970A-UD3
• AMD FX-8120 125W, 8x 3.10GHz, boxed
• Noctua NH-U14S (replacing a be quiet! Dark Rock Advanced C1, which handled the 8120 nicely; the NH-U14S is a bit better under high CPU loads)
• Corsair Vengeance DIMM Kit 16GB (2x8GB), DDR3-1600
• Samsung SSD 840 EVO 250GB (replacing an Intel X25-M G2 Postville 80GB, which I took over from my very old build)
• Seagate Barracuda Green 5900.3 2TB, SATA 6Gb/s ST2000DL003
• Sony Optiarc AD-7280S DVD Drive
• ASUS GTX760-DC2OC-2GD5 DirectCU II OC (replacing a fanless ASUS ENGT520 SILENT/DI/2GD3)
• Noctua NF-P12 PWM as rear exhaust (moving the Antec TrueQuiet 120, 1000rpm to the front, which came with the case; it's nice in the "low" setting, but barely moves any air)
• 2x Noctua NF-S12A ULN 120mm, (right, the Antec fan is out of work for the moment)
• Scythe Kaze Master II (to dial down those front fans when idling)
• BenQ GW2460HM 1920x1080, 24" (replacing a Samsung SyncMaster 226BW, 22", now sometimes used as secondary screen; it's humming quite awfully)
• win7 pro 64

Use cases next to the usual office and online activities are Lightroom, Sigma Photo Pro (to handle RAW files from Sigma DP Merrill cameras, it's quite resource hungry), Handbrake, VirtualBox, and a little gaming once in a while (but then I like all the eye candy I can get). ATM 1440, EtherOne, Osmos.

The CPU and exhaust fan are controlled by SpeedFan, which does a much better job than the BIOS of this board.

The PC used to be placed on the floor right of my desk, which happens to be a room corner, so I always could hear the exhaust fan. At the moment it is sitting under the desk, which makes it inaudible when idling, but it does restrict my leg & feet space (the desk having drawers left and right). I live in a quiet neighborhood, at night it is dead silent. Ambient temp in this room is about 20-23°C throughout the year.

As you can maybe already see, I like tinkering, so here are a few observations:

• The Noctua CPU cooler sounds different, the bequiet Dark Rock Advanced had a very gently signature, even at max RPM, however, the Noctua can take higher loads without ramping up (and thus staying inaudible) and it also improved the "Temp2" reading (in SpeedFan).
• The Seasonic PSU is a blessing.
• Ever since I mounted the Barracuda HDD in the rubber suspensions, it is quite good, there is a little airborne noise coming from there. I tried the rubber grommets in all three HDD trays, they all sounded different (the lowest being the worst).
• The front fans hardly make any difference at all for temps (1-2°C when gaming), but still create a lot of whoosh when dialed up all the way.
• The NF-P12 in the back is very efficient but very well audible when ramping up. I got used to it's sound signature, but don't like it too much.
• I once turned the bequiet PSU so, that it sucked air out of the case. This improved temps a lot. However, the PSU itself would heat up after a while and start whining
• Handbrake ramps up the CPU fan a little, but it is not disturbing
• Temps when gaming: CPU ~58°C, GPU ~68°C
• Power consumption when gaming: ~320W (reading a Voltcraft Energy check 3000)

Now there are two considerations:
1. The loudest part is of course the GPU. It is absolutely inaudible when idling but under load it is plain loud. Reading the forums I can see, that some ppl replace the GPU fans & shroud with 120 or 140mm fans. The heatsink of this card seems to be not so bad. Is there anybody around here, who replaced the fans of a recent ASUS Direct CU card and got better results?
Another possibility would be to replace the whole cooler with something like an MK-26, but does it even fit this card, being a non standard design?

2. As much as I like the Solo 2, it has only one 120mm exhaust. When gaming there is a lot of heat, which needs to get out of this case, so the NF-P12 will ramp up to 100% eventually (not that it matters much at this point, the GPU fans already being hysteric). I am eyeing an R4, everybody seems to be so fond of, but wouldn't I just get a lot of openings , so the CPU fan and HDD will be audible even when idling?

Sorry for this long post, thanks for reading.

Have you tried Afterburner to cook a custom fan curve, in order to (try to) lower the graphics noise level at load?

Have you tried to undervolt the CPU, in order to lower the load temps (even if I suspect a Gigabyte is not uv-friendly)?

I wouldn't bother with the stock cooler on "custom" cards. You can try strapping those NF-S12A's to see how temps go, but MK-26 is the preferred way to go. It should fit since there is good clearance.

I would suggest open case (test bench), which is nice and compact while having completely unrestricted airflow which means fan can spin very low but provide much better cooling. There are some people who can't have a build like that though (pets, small children, clumsy owners, etc).

Mounting the Mk.26 is not a problem, as it will fit on almost anything. But squeezing it into a system is another matter. On my 6950, it stands a smidgeon taller than my Hyper 212, takes up all but the last slot on my mobo, and makes the GPU sag quite alarmingly when in a tower configuration.

It's fantastically effective, but I kinda regret buying it, because it is extreme overkill for my purposes. Right now, I've got a single TY-143 just leaning against it (it's a long story), and according to MSI Afterburner, during several hours of gaming, the fan never spun up past 719 RPM and my mildly overclocked 6950 never went past 51C.

Thanks for your input guys!

Here is what I did so far:


I never touched the voltage or clocking of any CPU yet, so I'm hesitant about it, don't want to mess anything up. There are some settings in the BIOS of this board and I will start experimenting eventually. The main culprit for now is the GPU.
So I downloaded and installed MSI afterburner . It's pretty similar to "ASUS GPU Tweak" but the fan curve interface is much nicer. It misses a temp target slider, the beta version comes with this, so I will try that later on, if I feel the need for it. Made a custom fan curve for the GPU, allowing for a little higher temps.
Additionally I revised my SpeedFan settings, adding a second curve for the rear fan, linking it to the GPU.
The result of this first step is a much better balance. At 60% GPU power (this is the max load of 1440 on this card) the GPU temp settles at 75°C and the GPU fans stay at around 60% (above that and they start howling). The CPU temp is now even better at 55°C and the rear fan settled at about 75% (it gets annoying at 80%).
So that's already a success! But now my curiosity was taking over...


Well, about the open case suggestion: I guess I am one of these people. The reason being a lack of space on the desk.
I left the NF-S12A's in place but strapped on a bequiet silent wings 120 and the Antec true quiet. No howling any more, but quite a bit of whoosh. Enough cooling capacity for 60% GPU power but at about 75% GPU power the temp hits 80°C and the thing starts throttling.


In order to unplug the GPU fans I had to take off and reseat the heat sink anyway and realized that it is not such a big deal. So I will get the MK-26. The Hyper 212 is 5mm shorter than the NH-U14S, I guess the MK-26 will fit in the case. The other slots are free and I will tie it up with some wire or string to keep it straight.


I like overkill! I am way beyond reason anyway. Looking forward to those amazing temps!

I will post an update, once the MK-26 is installed, thanks for now!

Case fans dont have as big impact as direct fans like GPU/CPU heatsinks fans on the temperature of each component, think of the case fans help inside fans get fresh air and get hot air out, and while higher rpms still net a gain on temperatures, the gain is not as high, for example in my setup having the 4x Noctua NF-A14 PWM at 1k rpms nets around 5C on all cores coparaed to having them at 650rpms, but the changes on noise a dramatic, so much that i gladly give up that 5C, now the CPU fan for me going from 1100rpms to 700rpms increases temps more than 10C on all cores, and this is a single fan, and given that its not in a on an exit or entrance its not as noticeble as the Noctua (case fans), but this is something that you have to play with and find what works for your case and hardware.

For me a very easy quiet build can be done with Fractal Design Define R4 + Antec True Quiet 140s + Asus FanXpert2 motherboard, the fans tonally are very good, and a FanXpert2 motherboard can drop them to 200rpm on idle (on cha_fan headers). But it can also be achieved with other setups, just is so simple with that combo, but you also have to take into account that those fans are so low rpms will cool very little, so really depends on your needs.

MK-26 is a very versatile cooler, mostly it needs that the holes for the mounting around the GPU are standard that the memory is not so close that touches the pipes, but i seen it install in a lot of gpus. While this is not the same as yours, its worth checking as its DCUII GTX670 on mk-26, Malik Asus GTX 670 DCU II with Prolimatech MK-26, weather it fits the asus gtx760 idk. One warning that i give you is that in my GTX780 i coudlnt control the fans with the mk-26 PWM adapter, but i blame nvidia design on that card, was not only me, but a lot of reference gtx780 owners had this issues, but i still like to warn people just in case it happens.

Btw the fans that i would use on the MK-26 are the Prolimatech Ultra Sleek Vortex 140mm x 15mm Slim Profile PWM Fan, decent PWM range as long as you can control it with the card, at 450rpms imo they are quiet and still move a lot of air, and you should have sub 50 temps with it at those rpms, im guessing maybe below 50s. Prolimatech also sells the clips to attach it to the MK-26, Prolimatech Ultra Sleek Vortex 14 Fan Wire Clip Set - MK-26 Series. I leave you a graph of the fan bieng tested on FanXpert2, for you to check the range of operation under PWM,



NF-S12A PWM would be my second choice, really nice PWM range, although not all cards will allow it to go that low, unless you use software or mod the vbios.

Currently using NF-S12A on my MK-26. Very happy with the noise and performance. But I'm using a Gigabyte GTX 670 that lets it drop to 400rpm.

update:

the heatsink arrived today, looks really nice!

It does not fit this card though, at least in my opinion, the reason being the "DIRECT POWER" thing on the backside of the card. I know it's not a bug, it's a feature, but just have a look at this:





I was hoping, that the rubber spacer would be thick enough, but alas:



I'm afraid, that even when cutting the rubber spacer, this would be just too dangerous. I guess, that the screws are grounding the cooler construction (a lot of moving air is bound to create electrostatics), so as soon as the backplate touches the direct power bridge, I would shortcircuit big time. Unless this label is not metal, but some kind of plastic (I will check and update)






Additional info:

The GPU chip only touches the inner two heatpipes of the ASUS heatsink.



I still proceeded and pseudo-mounted the MT-26 to this card, just to check, if it fits inside the case. It does:








What do you guys think, am I being overly cautious, or would you also rather not mount this cooler to this card?

update #2:

Reading some web review stating that the second layer of the Direct Power thingy is some kind of isolation made me a bit bolder, I also checked with a voltmeter, there is no current getting through to the PCB. Thus, after another good night's sleep I proceeded:

The rubber spacer needs a little cutting.



Just to be extra safe, I added a small piece of cardboard, cut to shape and glued on with a tiny blob of thermal goop.



The rest of the procedure is pretty straightforward. I like the fact, that this card is so small, leaving a big area of the heatsink unrestricted.



I did not trust completely in this construction. I still have a disused but functional old PC in the basement, which acted as a guinea pig. (Not all connections are done yet in this photo)



(I would not have been able to close that case...)



Nothing bad happened in a test run, so I thought, I would give it a try in my main rig. Mounting the fans first (NF-P12 PWM and NF-S12A ULN)



At long last, the card found it's way into the case.




It works and I'm happy, that I did not have to sell off the card, just to get another one, which fits the cooler.

PWM does not work with standard PWM fans. I put a PWM splitter on the sys fan header, which is now running the exhaust fan (bequiet silent wings) and the NF-P12 on the video card. The NF-S12A is connected to the fan control and dialled down to 600 RPM. I like the silent wings fan a lot as an exhaust fan. It's pushing enough air but does not have this droning hum, when ramping up.

So, how does it all work?
It's great.
In idle the loudest part of the rig is now the HDD. It spins down, when not accessed for a while, after that, it's inaudible, even at night with no noticeable ambient noise. Changing the exhaust fan was a good idea, the PC is now standing in it's original place with it's backside close to a solid wall and for the first time ever, I just can't hear it at all.
When stressing just the CPU (P95), there is a very gentle whoosh.
When stressing the GPU (furmark), the whoosh becomes substantial (coming from the CPU and exhaust fans, which cover the noises of the GPU fans), but well, the fans have to get rid of all this heat somehow. Compared to the jet-engine sound of the Asus fans, this is very, very good.

Some numbers of synthetic testing: (stating temp-PWM percentage-RPM; power draw in W; please note that GPU PWM and RPM also apply for the back fan)

idle: CPU 32°-40%-328RPM GPU 33°-37%-521RPM
GPU only, power limit 60%: CPU 50-60-813 GPU 60-56-792
GPU only, power limit 80%: CPU 51-63-860 GPU 67-60-843; 320W
GPU only, power limit 100%: CPU 49-58-803 GPU 73-68-945; 355W

GPU 100% CPU 3 cores p95: CPU 53-69-930 GPU 73-69-968; 428W

Adding more cores resulted only in more power draw, the temps stayed in that region. The PSU (Seasonic X-400) could stand 440W (furmark + 7 cores P95). When I added the last core to P95, things got sort of unstable, the system limiting the GPU power.

In real world scenarios everything is fine. I let Unigine valley run for a while: power draw is around 320W, the temps are in the low to mid 50s for both CPU and GPU with just a gentle whoosh coming from the fans.

This is, how I wanted this system to be all the time, I'm a happy camper finally! Thanks again everybody for your input!

edit: typo

Is 600rpm the least that the asus gpu allows you to take down the pwm fan?

The Asus GPU seems not to work with regular PWM fans. With it's own fans I could take them down to 29%, but when I connected the NF-P12 through the Prolimatech connector, it just ran at full speed.

I will take out the card and apply additional cooling to the RAM chips later on (there were none with the Asus cooler, so I did not bother for the moment). At this occasion I will check and compare the wiring. It's four pins, I don't think they cooked up some super special extra Voodoo just to drive a couple of fans.

Thanks for the feedback on the fan control with the prolimatch, sad that its the same situation as with my artic accelero III and my gthx780, nvidia must have done something different eith the pwm conector, maybe they wiered differenly, maybe they are no in the same order like a stadard pwm fan. Either way this even discorages me more to go aftermarket, maybe prolimatech and artic should investigate this, since their products are aimed for this, if its jus the wiring they could easily change it on their adapter.

So either drive the GPU fans through SpeedFan (using a dedicated mobo fan header), or go with a Radeon .

Less explored yet quite important point particularly for silent computing enthusiasts seems to be that a variety of specifications are used in PWM headers of graphics cards. As the two people in the below have found out, ASUS and some other vendors seem to have adopted non-standard PWM headers, incompatible with standard PWM fans. To do justice to graphics card vendors, they don't have any obligation to follow the standard PWM specification.

In constrast, my MSI GTX 660 turns out to have a standard PWM header, into which I have plugged two (daisy-chained) NF-S12A PWM fans. PWM percentage and tachometer readings are completely correct as reported by Afterburner and GPU-Z, and also in line with my human sensory experience. I suggest whoever interested in fine controlling aftermarket PWM fans on graphics cards should buy MSI graphics cards. Also, though I'm not completely sure, as far as I can recall, I have seen some posts saying that EVGA claims to produce aftermarket-friendly graphics cards.

Well personally i dont blame nvidia nor the video card manufacturers, we need the manufacturers of the aftermarket coolers to buy cards from each gen and test it, even if they are reference models, and once they do, to release it with the appropriate adapters, its not hard nor expensive, its just part of what they should do to ensure their customers will be happy buying their products. For me.... its hard to justify a $60-100 aftermarket cooler if im not going to be able to control the fans on it.

I been to the red side with a ATI Radeon 9800, and had the worst experience i can say with a gpu, i was raiding on a competitve guild back then, and every time i look at the back wall of chromagus on blackwing lair... i get dced... for months i had this issue, i almost got gkicked by how unreliable i was, out of this, there is so little chance AMD will ever get my money again, even if they release a much better product than nvidia, even if they are totally fanless, i just can't see my self dropping money on a company that grief me for such a long time.

Btw can speedfan control GPU fans based on GPU temperature on motherboard header?

Why do you ask me so? Maybe is there a misunderstanding?

However, as you know, it's trivial to drive those Noctuas Mk.26 fans using SpeedFan and any controllable mobo header, while, with reference to your setup, you have to crimp a regular PWM connector on the Accelero fan cable before, in order to do the same (you could also need a PWM extension/splitter, depending on Accelero's cable lenght and header's placement).

Maybe i didnt explain it well, usually motherboard software (fanXpert2, easytune, etc) cant read GPU temperature or any other sensor that its not in the motherboard, so even if the motherboard header is controllable (voltage or pwm), you cant bind it to the GPU temperature to have it idle at low temperature and ramp up as the temperature goes up. So i just wondering if Speedfan could do this?

Yes, at least in my case SpeedFan can read the GPU temp and I linked the Sys1 fan header to the GPU temp (as well as the CPU temp).

Yes, at least since 4.45 release, SpeedFan usually can read GPU temp sensors (with Radeon cards it may also control fans).

ok, as this PWM issue raised a little discussion, I went ahead and played with the fan pins. Information about standard pinout can be found here (page 17), how to depin a PWM fan plug is shown nicely here.

Things I found out:

• Asus actually does use the standard pin layout, BUT
• The percentage shown in MSI Afterburner is not related to a PWM signal, but to the max speed of the original fans.
  Example: You set the speed to 30%. This card comes with 80mm fans, which spin up to ~3300RPM, thus 30% of that is roughly 1000 RPM, which is the speed the card drives any connected fan. I tried both the Noctua and the bequiet fan, which have different max speeds, both were running at 1000RPM at 30%.
  In a way this is a clever design, as it does not just provide a PWM signal, which will be interpreted differently by different fans, but it makes sure, that the fans will run at the "right" speed, even if the bearings get old and show more resistance.
• When booting, the GPU fan spins very slowly, at the time the win7 login screen shows, the fan spins up. Maybe fan control is taken over by the GPU driver, as soon as it loads? (I will try to boot with a Linux Live CD and watch the fan)

Conclusion: I stand corrected, Asus does use the standard pin layout, but you have no real control over the PWM signal with the standard user interface. The signal from the card is pretty useless when using bigger fans, as a lot of silent potential gets lost. I can hear the NF-P12 at 1000RPM when running inside the case, so I will revert back to my previous setup and let SpeedFan control the GPU fan.

I think you are discovering invaluable information about ASUS graphics cards (GTX 760) if you are making correct observations, about which I'm not entirely convinced but it is highly likely to be the truth.

Provided that your graphics card sets the speed of the fan to "x %" of 3300 RPM regardless of the fans plugged into the PWM header, you have every reason to be sanguine about the possibility to finely control the fan. There is an aftermarket tool, "Kepler BIOS Tweaker", where you can adjust such numbers associated with the fan on the GPU PWM fan header.

I have been using this tool for some time with my MSI GTX 660 and I haven't encountered any issue. My two Noctua NF-S12A PWM rotate at around 360-390 RPM in idle state (GPU temperature = 30 Celsius, PWM Percentage = 25%) and 630-660 RPM in full load state (GPU temperature = 69 Celsius, PWM Percentage = 42%). It is absolutely silent. Admittedly, I can hear very gentle subtle airflow sound in the full load state only because my PC is composed of dead silent components, e.g., PSU, SSD, and etc. These numbers are rather close to the worst case scenario because I turned off the front fan of SilverStone TJ08-E, implying that there is only one rear case fan attached to my TJ08-E, i.e., NF-S12A PWM rotating at around 300-400 RPM.

Maybe I was a bit rash with drawing conclusions, but these are my observations:
When connecting a fan with an official max speed of 1300RPM (NF-P12 PWM) it spins at ~1000RPM (oscillating between 990-1060) at 29%, which is the minimum setting in MSI Afterburner. Using the slider I can control the speed between 1000-1290RPM. The max speed is reached at 38% in MSI Afterburner.
When connecting a fan with an official max speed of 1500RPM (bequiet silent wings 120) it spins at about the same speed at 29%. When using the slider I reach the fan's max speed at ~45%.


I downloaded & installed GPU-Z and Kepler Bios Tweaker, read out my card's BIOS and loaded it into KBT. It shows a stock fan range of 10-100%. There are no actual fan speeds (RPM) shown in the BIOS. I don't see much point of messing with the BIOS at the moment. The BIOS states a minimum of 10%, which the card happily ignores (29% being the minimum in MSI Afterburner or ASUS GPU Tweaker)


Apparently some things changed in the fan control setup when the 7-series cards were released. The web is full of forum posts in various places questioning about the possibility of changing the min speed of 760, 770 and 780 cards. People tried changing the numbers with Kepler Bios Tweaker without success. I found posts about Asus, EVGA, Inno3D and Palit cards, just to name a few.
Here are some examples:
http://forums.evga.com/tm.aspx?m=2014160#2017187
http://forums.evga.com/tm.aspx?m=2014160&mpage=3#2040399
http://www.computerbase.de/forum/showthread.php?t=1314565&p=15261437#post15261437

I could not find any statements of people who managed to tweak the card's minimum fan speeds with 7-series cards. It seems to be a software issue, though, as EVGA provided fixes for some of their cards through BIOS updates.

I'm a bit at a loss at the moment. I will stick with my SpeedFan / SysFan-header solution for the moment, as it works well enough. If there are new ways or possibilities, I'm still happy to try things, of course.

I can confirm that currently there is no known solution.

After tinkering more than a while, I was unable to reduce the minimum fan speed, editing the relevant Vbios setting, on several Kepler cards: ASUS GTX-650Ti DC-II, ASUS GTX-660 DC-II, MSI GTX-650Ti PE and MSI GTX 760 Gaming.

I'm pretty sure the same thing happens on EVGA cards, but IIRC our co-forumer lodestar told me he was able to get some satisfactory results tweaking a Gigabyte Kepler.

This makes me sad, i really treat my PC like a hobby more than a tool, losing the chance of controlling aftermarket fans / coolers is very unappealing to me, i guess ill have to stick with the best one that comes with the aftermarket cards, probably will be MSI if they continue their trend. That said Artic Acccelero III has been really good even at steady unchangeable 7V.

UPDATE:

CPU Temps got sort of high (60-62° is hot for an AMD chip) in the last 2-3 months during gaming. First I attributed it to the really hot summer (at least for my country it is), but still, I wanted to see, if there was a chance of improving things.
(System is still FX8120 & Noctua NH-U14S, GTX-760 & MK-26, Seasonic X-400FL, Antec Solo2)

"Thou shalt not breathe hot air", I told the CPU fan and what shall I say: This cardboard duct has been the cheapest addition to the system. Very efficient. Improved the temps by ~3°.




Furthermore I kicked out the 3,5" HDD (replaced it with a 500GB SD, all the rest sits on a NAS) and experimented with intake fans, the GPU fans and the exhaust fan. Here are my findings about fans:

GPU: I have 1 NF-P12 and an NF-S12A on the GPU. The NF-P12 is capable of higher pressure (1.68mm H₂O), the NF-S12A being a typical case fan (1.19mm H₂O). My assumption was, that I would get better temps with the NF-S12A in the left position (having less heatsink to work and provide more direct airflow to the PCB and the RAM chips) and the NF-P12 in the right position, cooling the bigger part of the heatsink. The opposite is true. I get better temps with the NF-P12 left and the NF-S12A right.

EXHAUST: This fan is crucial. I tried four different ones. The bequiet silent wings 120 and the original Antec TrueQuiet 120. Both sound good, but they just don't push enough air, once the system gets hot.
The NF-P12 pushes very well indeed, it just develops some sort of drone as soon as it ramps up a little. The NF-S12A is the best solution. It's capable of getting rid of the heat, all I would hear is some whoosh. It's now daisy-chained to the CPU fan.

INTAKE: I know, this is going to be unpopular, as I read a lot about positive pressure setups in these forums. I mounted 3 NF-S12A (1 bottom, 2 top in a row to provide some real airflow to the CPU fan). All I can say is, that temperature-wise it's negligible (~1° improvement), noise-wise it is absolutely not. At least with this system inside this case, it simply does not make much sense.

Finally, two days ago I did two things: Look into the case and wiggle the CPU heatsink a bit (sort of checking, if it still sits tight) and look into my system settings and change the energy settings from "power" to "balanced". Temps during gaming have improved by ~5°! I don't know exactly which of the two is responsible, but now the CPU temps are between 52-54° during gaming, no intake fans running. At these temps the CPU and back fans are barely audible at night, during the day, I can't hear them at all.

UPDATE 2:
Today all the nice temps were gone. Playing the same sequence drove the temps up to 59-61° again.
Naturally I was very confused. I had not touched the PC except for pressing the power button. What was different? - After some thinking I realized, that during the last days I had put the PC into sleep mode only, but yesterday I had powered it off completely.
Thus: Put it into sleep mode, resume and fire up the same sequence once more. Back are the magic temps of 52-54°. Repeat a number of times. Always the same.

I found some evidence that the FX-8120 has issues with the clock speeds after resuming from sleep mode. I'm not OCing here and watchings the clocks shows no difference, but "CPU power" shows slightly lower numbers: 60-90W after cold boot, 50-80W after resuming from sleep mode. Do those roughly 10W explain a temperature difference of ~7°?

Whatever bug it is, I will use it as a feature...

If you run some tool like y-cruncher or ibt either after powered it off completely or after resumed it from sleep, do their results change?

Short answer: No

Long Answer: See below...

I did five runs of y-cruncher in each mode and the results are practically identical. The first example is from a run in reboot mode, the second from a run in resume mode.








However, the temperatures behaved quite differently.

I'm running an FX-8120 on a Gigabyte GA-970A-UD3 mobo, vcore, clocks, etc. all on auto. No software OC. The BIOS claims to be able to control two fan sockets with PWM, but it regulates the speeds with voltage anyway. The BIOS fan control is not satisfying to me, so SpeedFan takes care of the fans. The CPU and Back exhaust fan are daisy chained to the CPU socket, the GPU fans are hooked up to the "System1" socket.
This is the fan curve in SpeedFan of the CPU socket.




The following is a collection of curves inside SpeedFan:

Left column: Two or more runs in reboot mode
Right column: Two or more runs in resume mode
Middle column: One run in reboot mode (second to last curve) and one run in resume mode (last curve)

Top row: Temp curves: grey=CPU; red="Sys2" (probably VRMs. It goes down when CPU fan is ramping up)
Middle row: Vcore: green=Vcore1; red=Vcore2
Bottom row: Fan Speeds: green=CPU Fan; red=Back Fan (but they are chained anyway)





You can see, that the CPU temp in reboot mode reaches about 58° (with a few extra spikes in the end), therefore the CPU fan spins up to ca. 900 rpm (cooling the "System2" place better, so the red line goes only up to ca. 52°). In resume mode the CPU only goes up to about 50-52°, therefore to CPU fan stays around 640 rpm. That's why the "System2" is way above CPU temp in this scenario. The middle Temp curve shows, how these two temp lines practically swap positions in the two different situations.

I also did some logging with HWInfo. Still have to get excel to make some useful curves out of the numbers. Will post this later on.

Probably the most useful number is the vcore values in both circumstances (as they should be somehow different).

OK, here is a comparison of the Vcore values. Looks pretty similar to me.



Here are links to the two original CSV files created by HWInfo. Bunches of numbers, but nothing I had not seen in the SpeedFan curves. But maybe I'm looking in the wrong places?
https://app.box.com/s/n62b8afxgb78o289j1od
https://app.box.com/s/znllcv43nhv4ajsppjoa