Resolved How-To: FIX MAC PRO FAN SPEEDS when using PC (non-EFI) Graphics Cards

[SteveJobzniak]

The temperature sensors in the Mac Pro are not tuned for things like modern nVIDIA graphics cards and certain SSDs or HDs, and will ramp up the fans tremendously, even though the temperatures are low. The sensitivity range was tuned for the super-cold, old graphics cards the system shipped with.

Some people have discovered that running and then closing a 3D application after every cold reboot would get their PCI "non-EFI" graphics cards into a colder/power-saving mode, which makes the Mac Pro fans silent again. But screw doing that after every cold reboot! That would get tedious fast.

I have the solution. This free software is actively maintained and signed by a professional software company and is the most advanced of all fan control software for Macs (the only software that lets you change minimum fan speed dynamically). It's by the same people who made AnyToISO. It's crazy that it's free since it's easy to use and well designed and they even updated their Fan Control software to El Capitan very soon after its release, so they're trustworthy and reliable. They should be charging for this... but I am glad they don't.

http://www.crystalidea.com/macs-fan-control

It's unfortunately necessary software when installing 3rd party components in Mac Pros.

In my case, a new, passively-cooled GTX graphics card is causing the case fans to spin up to an extreme amount because it raises the PCI ambient temperature a bit more than the old cards the system came with.

Adjusting the fan behavior is the only proper forever-fix for this problem for all us upgraders. Oddly enough, the PCIE Ambient temperature isn’t even high for me (about 30*C), so I don’t know why the PCI fan sped up so much in the Mac Pro. Either way, the only fix is to correct the fan behavior by changing the operating temperature range of the fan speed.

Mac Pro 2009 Setup (5.1-flashed), with an nVidia GTX 960, El Capitan and Macs Fan Control:
* Click on Preferences, and enable “General: Autostart minimized with system.” Unfortunately there is no way to hide the menubar icon, so it sits there as a constant reminder that the fans need assistance. Oh well… it's nice to have the software quickly available if the machine ever makes noise for another reason.
* Click on PCI: Custom…, and select “Sensor-based value”. It’s the fan that blows over the PCI cards, so set the sensor to “PCIE Ambient”. The lowest temperature is the one at which it runs the fan at minimum speed (800 RPM), and the highest is the one at which it runs at maximum (4500 RPM). Set the range to 34 (min) to 60 (max) *C. This was tuned during the winter, when “PCIE Ambient” said ~30-31, and it caused the PCI Fan to drop from ~1500 RPM to instead run at around 800-900 RPM. The max range may seem high but that’s just the range at which the fans would blast at 4500 RPM, which will *never* be reached! The minimum temp is what matters; and giving a higher max range just causes smoother changes in fan speed. After lowering the fan speed, the PCIE Ambient rose to about 33-35*C after the change, which is practically no temperature increase at all and is ridiculously low anyway (electronics can easily deal with ~80 when no capacitors are involved). Note that this particular PCI fan is very silent until about 1100, but it sounds best at 800. The temperature range may need adjusting during the summer and for your own climate in your part of the world.
* The only other “fast” fans in my Mac Pro are the BoostA/BoostB, which are the fans inside the CPU heatsinks, and those should be kept as-is. They’re very silent.
* If you still have a little bit of noise, it's probably the Mac Pro's Power Supply (PS) fan, at least in my situation, but tweaking that one is risky and advanced so I left it alone.
* Also a final warning: Do not use "Constant value of X" anywhere. If things get very hot you can destroy your Mac. You should always use sensor-based automatic speed adjustments, and make sure your ranges are still low-enough that they'll actually react if the heat begins to increase someday.

I have attached a screenshot to help people. And as I said, I set the "Based on PCIE Ambient" range to 34 (min) to 60*C (max), or whatever suits your particular ambient temperature.

 

[h9826790]

You can do the same thing for your PSU fan.

Sensor base
PSU component 2
min 43C
max 65C

 

[SteveJobzniak]

You can do the same thing for your PSU fan

Your suggested values there are dangerous.

Time to admit something: I have also done it for my power supply fan but I set it back to "Auto" for the screenshot because I didn't want to be responsible for people's power supplies dying an early death. They're one of the hottest, most dangerous components in the whole computer and need adequate cooling. When they overheat, they die and shut down the whole computer until the PSU is replaced, and that sucks...

Power supply capacitors are rated for a certain number of hours at either 85*C or 105*C. The lower the temperature, the longer before the capacitor evaporates and dries out. Temperature should be kept as low as possible to extend life, and that unfortunately means that fan speed is important.

Normal operating temperature for the Mac Pro 2009 Power Supply (which I think is the same in the 2010 and 2012 models) is around 30-35*C, which is what the default automatic fan speed aims for. This means that if they used caps rated to last 2000 hours (for example) at 105*C, they'd last tens of thousands of hours since they're only being heated to 35*C.

If you raise the temperature of the power supply to 45*C (which is what your settings allow), that's a lot more heat and evaporation and you'll reduce the PSU life by something like 40-50%. One day it'll go "Bang!" and the computer will shut off. To prevent that, you should aim at keeping the power supply cooled down.

So I did reduce my Power Supply fan speed but I used values that ensure that the temperature did not go above 35*C. Here are my values:

PS: Based on PSMI Supply AC/DC Supply 2. Min: 33*C, Max: 60*C.
(Note: Old versions of Fan Control called it "PSU1 Secondary Component". This post has been updated in March 2016 to reflect the new "PSMI 2" name in the latest version of Fan Control. Many other posts and screenshots here still refer to its old name!)

This got the PS fan to spin silently at around 700, which keeps the temperature around 35*C.

Both my PSU temperature value and the PCI values may need slight tuning in the summer, when the heat rises, since it may cause all fans to be audible again. But it may also actually be necessary to have a slightly audible computer just to keep it running safe and cool in the summer. We'll see...

 

[h9826790]

I use this setting for a long time, and I am living in HKG, hot and humid in summer time. My Mac run for 24/7 (most of the time without aircon).

And I can tell you that the PSU fan speed is NOT base on the PSU temperature (not either component 1 or 2). I didn't figure out the actual logic, but I can tell it's more related to the PCIe fan than it's own temperature. (Under native setting)

There is no way that the stock fan aim at 30-35C. My PSU under native fan control offen run at 50+C (ambient 33 or above) and the fan still stay at 600RPM. But when the GPUs work hard. The PSU fan will spin up more than it needed.

I am not saying that my fan profile is more safe then the native profile, however, I can tell you that my fan profile works well in both OSX and Windows. And it's more PSU temperature related. It keep my PSU never reach 50C and more quiet when the GPUs are working.

Anyway, it's just my sharing. Of course you can feel free to comment on it. And I cannot guarantee this setting is safe. Same as your PCIe fan setting suggestion, none of us can guarantee it's 100% safe to do so.

Extreme temperature of course will kill the Mac . However, if you check the spec, Apple state that the operating temperature of the 4,1 is up to 35C. Which means there is no way that the fan can cool the PSU to below 35C if the outside air temperature is already at 35C. I don't know why you think 45C is danger (just 10C above OAT, for PSU, I think it's very normal)? And where says it will shorten the life for 40%?

 

[SteveJobzniak]

I researched the Mac Pro power supply temperature around the net and that's what I got the 30-35*C number from, because that's the temperature on other people's power supply, so I strongly suspected that the PSU fan aims at that number. I saw it from many people and it matches my own numbers. Maybe it's not a hard rule, though. I don't know.

I also doubt your theory that PSU fan speed is based on the ambient temperature sensor of the PCI slots. When we install a modern graphics card, it draws a lot more power (75W from the slot, 75W per PCIe power cable, that's 225 Watts). This makes the PSU hotter and speeds up the PSU fans. Maybe it's based on power draw in watts. Maybe it's based on heat. Maybe it's based on multiple sensors. We don't know. But it's definitely linked to installing a new, power-hungry graphics card.


> "I don't know why you think 45C is danger (just 10C above OAT, for PSU, I think it's very normal)? And where says it will shorten the life for 40%?"

I'm an electronic engineer and frequently do repair work. The #1 cause of failure in electronics is dried out electrolytic capacitors. I have attached a chart showing that even "just 10C hotter" difference dramatically shortens capacitor life. As a general rule, every additional 10*C halves the life.

Examples from this chart (the file called EEOL), showing typical electrolytic capacitor life vs ambient temperature: 160,000 hours at 55*C, 80,000 hours at 65*C, 40,000 hours at 75*C, 20,000 hours at 85*C, 10,000 hours at 95*C, and 5,000 hours at 105*C.

That's for a *top-end* capacitor rated for 105*C at 5,000 hours. Most capacitors in power supplies are rated for 2000 hours at 85*C (about 500 hours at 105*C), but maybe Apple have gone more expensive and aimed at 2000 hours at 105*C... I don't know, I haven't needed to open it up to check since I take care of the heat by keeping it <= 35*C.

Since your PSU reaches 50*C in the summers, compared to my 35*C, you can expect that your PSU's capacitors will dry out and explode 2-3x faster than mine. Luckily it's pretty easy for someone with electronics knowledge to buy new capacitors and replace them without buying a new power supply. But it's even easier to make sure it lives a long life instead. Of course if you live in a very hot place, it's impossible to get both silence + low temperature at the same time. :\ Hey, the good news is that you're in Hong Kong with access to new capacitors cheaply when you need them.

Some people may not take my word for it, so I found a page for you which says the same thing about roughly halving life for every 10*C of extra heat:
http://electronics.stackexchange.com/questions/16499/hot-capacitors-is-that-a-problem

Edit: I attached a few more example charts, such as image048 which shows the failure rate vs time and temperature. For example, the brown line at the right shows that the tested capacitor reached up to 100,000 hours at 25*C, reaching a 50% risk of failure after about 80,000 hours. The orange line to the left of it shows that at 40*C, the capacitor reached up to 40,000 hours, with a 50% risk of failure after about 20,000 hours. The purple line to the left of that shows that 50*C, the capacitor reached up to 15,000 hours, with a 50% risk of failure after only ~8000 hours.

The #1 enemy to electronics is heat, the #1 failure in electronics is capacitors, the #1 enemy to capacitors is heat drying out the electrolytic material. :\

 

[h9826790]

Thanks for the info, good to learn something new.

Assuming Apple use the top end capacitor, then 160000hr for 55C capacitor core temperature, assuming that fit my 45C PSU temperature, which is 18years continuous operating time. I personally wouldn't worry about it, and further lower the temperature to gain another 18years life time.

If they use the 2000hr rated at 85C capacitor, then my PSU should die 3.5 years ago. So, I assume this is not the case.

Anyway, I know that there is nothing "danger", but at worse, shorten the lifetime to 18years (may be you are right, the capacitor is rated 2000hr at 105C, so my 4,1 should die at around 2017).

hopefully Apple will make another good Mac Pro, so that I don't need to run my 4,1 for 18 years without any good replacement

 

[benjaprud]

I've also noticed that PCI and PSU fan speeds are related to PCIe power consumption (maybe CPU too), not temperature, although they might accelerate above certain temperatures but that never happened to me, and would probably indicate abnormal overheating.

Using a power hungry graphics card (Titan X - 250W) overclocked and at full load would make my MP very noisy (and my PSU very cool) with 1500/2250 RPM PSU/PCI. Modding the PSU to draw current directly from the PSU thus bypassing the PCIe power supply made things quieter again under the same load (about 900/1450 RPM with only one of two PCIe power supply used). I guess Apple didn't intend the Mac Pro to be modified with such power hungry hardware and just went for an agressive cooling above normal scenarios just in case. It just seems too agressive to me, the more power I draw, the cooler is my PSU.

I agree nobody should play with the PSU fan. It's silent enough at default speed and starts being noisy only above 950 RPM. Also the temperature readings are meaningless unless we know where are the sensors in the PSU. Some parts are likely to get hotter than the reported temperatures.

As an electronics service technician I've spent years replacing electrolytic capacitors which is the most frequent cause of failure in cheap consumer electronics (but rare in well built quality stuff), I hope Apple didn't aim to fit in that category with the Mac Pro.

 

[ZombiePhysicist]

So,

a), thank you. This is a cool app!

b) I'm not messing with anything until I see a consens come out of the big heads here because I have no clue what is a good idea or not, so otherwise I'd leave well enough alone.

c) Weirdness. So I upgraded my card to a a visiontek 7870 (http://www.amazon.com/gp/product/B0085O90SQ?psc=1&redirect=true&ref_=oh_aui_search_detailpage). Nice card. And I get the fan kick up which I can get to go away by running Elder Scrolls for a seck.

I can make the fans kick up on demand by either using omnidazzle to leave lots of screen pixie dust or do use the trackpad for the show desktop gesture and leave the windows hovering half off the screen. And that will normally kick up the fans if I haven't done the elder scrolls run.

But here is the weird part. None of the 6 fans shown in the Mac Fan Control app kick up at all. It's the fan on the graphics card itself! I've confirmed this by snooping with the case open. Yet, run elder scrolls, and the problem goes away until the next reboot. Soooo weird!

and d) So for the app to be useful for me, it would have to be able to control the fan speed on the graphics card. Is this possible?

 

[SteveJobzniak]

Thanks for the info, good to learn something new.

Assuming Apple use the top end capacitor, then 160000hr for 55C capacitor core temperature, assuming that fit my 45C PSU temperature, which is 18years continuous operating time. I personally wouldn't worry about it, and further lower the temperature to gain another 18years life time.

If they use the 2000hr rated at 85C capacitor, then my PSU should die 3.5 years ago. So, I assume this is not the case.

Anyway, I know that there is nothing "danger", but at worse, shorten the lifetime to 18years (may be you are right, the capacitor is rated 2000hr at 105C, so my 4,1 should die at around 2017).

hopefully Apple will make another good Mac Pro, so that I don't need to run my 4,1 for 18 years without any good replacement

No problem. That's a lot of assumptions though, haha. I'll clarify a few things.

#1
First of all, you are looking at the maximum amount of hours. Most capacitors fail way before that, usually in the 50% timeline range. Whenever an electronic device breaks, it's the capacitors in 99% of all cases. Capacitors suck. They're unpredictable even when they're from a good brand, and the rated amount of hours is never guaranteed due to differences in quality between manufacturers, batches, materials, risk of some having random severe defects, etc.

#2
Next, I decided to look up what parts the Apple PSU uses and was pleasantly surprised to find the parts list at http://www.badcaps.net/forum/showthread.php?t=6264. Someone had taken it apart. From his list and the capacitor series, I was able to look up the spec sheets and see that they are all 105*C caps. That's a *huge* surprise since most power supply manufacturers are cheap and use 85*C for most caps and only 105*C for the main/hottest ones. So yes, that choice will contribute a lot to the life of an Apple power supply. A+ there!

#3
The capacitor brands are above average but not top-end, so they're going to fail before their rated amount of hours more often than higher-quality brands. Worse brands have worse ripple ratings which leads to more heat and earlier failure, and have worse quality electrolyte formula, etc. As someone pointed out in the thread above, the Apple power supply unfortunately has some pretty bad Taicon ones for "less critical" caps, whereas they only use better brands for the most critical ones. So they get a B+ there. I wish Apple had spent a few extra dollars on the best brands across the board. But at least they aren't noname brands.

#4
I didn't go into detail about ambient vs core temperature earlier, but the 45*C you are seeing isn't the actual temperature in the capacitor. That's the ambient temperature somewhere in the PSU. You can be sure that your capacitors are at least 10*C hotter in their cores (that's what we could expect if the cap itself was 45*C on the outside sleeve). So yes, you are correct to look at the 55*C values in the chart.

#5
No, Apple aren't using 5000 hour 105*C capacitors, so you can't use the example hour numbers in the chart I posted. Those capacitors are very rare and were just used in that particular example chart. Designers almost always use 1000-2000 hour rated ones, and Apple (well, Delta who designed the PSU and OEM'd it to Apple) are no exception, because the 5000 hour @ 105C ones are top-end and ridiculously expensive. And since 2000 hours @ 105C is 2.5x smaller than 5000 hours @ 105C, you need to reduce all numbers above by 2.5x.

So let's do the math for your situation. If your ambient is 45 C, your capacitor cores are at least ~55 C. So you were correct to be looking at the 55C numbers, but incorrect in the amount of hours it would give you. At 55 C you get 180,000 hours for a 5000 hour-capacitor, but since Apple doesn't use those you'll have to do 180,000 / 2.5 = 72,000 hours (~8 years continuous operation) for a 2000 hour-capacitor. That's the *maximum* time until *guaranteed* failure. The time until 50% of all capacitors with that rating have *prematurely* died is about half of that (~4 years continuous operation) but can be anytime before or after that. It could be a day if a capacitor is damaged from the factory, or 7 years and 349 days, etc. But most will be dead quite a bit before their maximum rating. Especially if the brands aren't the top-quality ones.

So at 45 C you can roughly guarantee 4-8 years of 24/7 operation, and at 35 C you can roughly guarantee 8-16 years. There are no guarantees though. Capacitors are awful and very unpredictable. I hope we invent some new capacitor technology someday that never dries out. Whenever I repair electronics, it's the capacitors in 99% of all cases. All electronics would pretty much last a whole lifetime without repairs if capacitors were as reliable as all other components.

The good news for you is that you (if you're handy) or an electronics repair shop can easily replace the capacitors when they blow up. Just open the power supply, safely discharge the capacitors by placing a huge resistor over the leads (while insulating your hands to avoid electrically induced death ), look at the ratings of all of them, buy matching replacements that have the same specs (diameter, temperature, voltage, capacitance and low ESR/good ripple ratings). A replacement set could probably run around $30-40, which is a lot less than a new PSU (seems to go for $200+). Then just de-solder and replace them all and make sure to put the new ones in with the correct polarity to avoid explosions.

Hey, lastly... thanks for making me look up the capacitors used by Apple, it's actually quite relieving to know that it's all 105*C caps from pretty-good brands. That means it will live longer than typical PC power supplies. I've had way too many of those die within 2-4 years at northern temperatures.

 

[SteveJobzniak]

As an electronics service technician I've spent years replacing electrolytic capacitors which is the most frequent cause of failure in cheap consumer electronics (but rare in well built quality stuff), I hope Apple didn't aim to fit in that category with the Mac Pro.

Then you'll be as happy as I was to read the news above. I mean, the power supply itself is an OEM design from Delta, and these capacitors aren't exactly top brands either apart from their Nichicons (which are awesome; I'd happily use those in the same sentence as Rubycon, Panasonic and Chemicon). *But* at least they used 105*C rated caps throughout the *entire* design, and they're all from above-average-to-great brands so we can pretty much trust their temperature and lifetime ratings not to be fake. They do use a *few* suspect capacitors but they're in non-critical locations so shouldn't get very hot and are probably expected to die at about the same time as the high-end, hotter ones. This is still better than almost all PC power supplies out there. I bet Apple didn't want people's Mac Pros to just randomly die too early.

I was definitely relieved by seeing the capacitor parts list. I've actually been expecting my Apple PSU to break any day since I've used it for about 6 years of almost 24/7 operation, but now I'm relieved to know that it may have a couple more years, maybe even 10 more years if I am super lucky. I don't expect any capacitors to ever reach 100% of their rated lifetime, but at least I won't have to worry about "explosion tomorrow?" anymore. ;-)

This is the longest-living PSU I've ever had in a 24/7 computer... Apple did a good job.

I agree nobody should play with the PSU fan. It's silent enough at default speed and starts being noisy only above 950 RPM.

Hah! I wish I had your PSU. The fan on mine is a bit noisy at 850 RPM (sounds like a very distant jet engine), which is why I lowered it. It all depends on the bearings in the fan. Mine probably has some trapped dirt particles.

And yes, nobody should play with their PSU fan temperature without setting up safe temperature ranges (like mine). I agree that the temperature of "30-35*C" is "meaningless" in the sense that we don't know where the sensor is, but it's labeled "PS component temperature" so it's probably on the circuit board somewhere near the components.

However, the temperature *is* meaningful in one way since I *do* know that the temperature in people's screenshots around the web is usually 30-35*C for the PSU, and that's when people are using the default/automatic fan speed and default hardware. So it seems like a safe bet to aim at a fan speed that maintains that ambient temperature inside the PSU, if that's the temperature that Apple's own fan speed management ends up with.

 

[SteveJobzniak]

So,

a), thank you. This is a cool app!

You're welcome. And I agree. It's amazing that they give it out for free.

b) I'm not messing with anything until I see a consens come out of the big heads here because I have no clue what is a good idea or not, so otherwise I'd leave well enough alone.

The PCI fan speed adjustment is safe, since it's just the gray fan near the front which blows over the PCI cards. But don't touch the power supply fan adjustments I also mentioned, because those are very dangerous if you don't carefully aim at fan speeds that still give safe temperatures.

c) Weirdness. So I upgraded my card to a a visiontek 7870. [..] But here is the weird part. None of the 6 fans shown in the Mac Fan Control app kick up at all. It's the fan on the graphics card itself! I've confirmed this by snooping with the case open. Yet, run elder scrolls, and the problem goes away until the next reboot. Soooo weird!
and d) So for the app to be useful for me, it would have to be able to control the fan speed on the graphics card. Is this possible?

Are you sure? Most modern graphics cards have super silent fans. I see that yours only has a single fan which means it has to work at a higher speed than two separate fans, so the extra noise may be innate to your card, but it should still be pretty quiet. To be sure that it's the card, try gently placing your finger on the center of the graphics card fan (where the label is) to stop it and check if the sound goes away. Then spin the fan with your finger to get it to start again.

If the noise comes from the graphics card fan, then there's nothing you can do unfortunately. Many graphics cards for Windows have Windows software to control the fan speed on the card itself, but such software doesn't exist for Macs. The Mac itself cannot see/control those fans, they're controlled by the firmware inside the graphics card chips.

But you probably have a combination of PCI/PS fans in the Mac revving up, *and* your graphics card fans revving up. So you can still reduce noise by at least checking and lowering the PCI fan speed on your system. See my first post here for that part of the tutorial. Avoid all the power supply stuff in later posts since that's dangerous and usually not necessary. But changing the PCI fan itself is safe as long as you follow my instructions to get a safe operating temperature range.

And as for your other question: Yes, running a 3D application and closing it again has been a "solution" used by people with the excessive fan noise. The reason that it "works" is that modern graphics cards have multiple power modes; 3D, 2D, etc. It seems that they start up in 3D mode by default (when the computer starts up from a cold boot), which draws more power and thus causes more heat/more fan speed/etc. By starting and closing a 3D application it seems the cards go "oh... we're back to 2D!" and thus reduce their power and heat so that the fans don't need to kick in. That's the only way I can explain why 3D applications are usable as a temporary fix. Because the Mac Pro itself doesn't know about the 3D application running, so it's got to be something on the card that changes after the 3D app is closed.

I hope this helps...

 

[h9826790]

Another good lesson for me. I come from the aviation industry, the "limit" or "rating" in our industry is usually guarantee (pre-mature failure may lead to disasister). So, I automatically assume that 2000 hr rated at 105C is the min time, in fact, it's the maximum time

Anyway, do you have HDD at bay 3 and 4? Just want to know if that make my PSU temperature read 10C higher than your "normal" temperature.

 

[SteveJobzniak]

Well, yeah it's an awful situation, and it causes me no end of electronics repair work... The reason they can't guarantee a minimum operating time is that they'd never see the end of the lawsuits. Capacitors are a joke. It's just a roll of paper with some special water (electrolyte formula). See the attached picture. They slowly degrade over time as they dry up, which makes them slowly lose their capacitance (how much electricity they can store), or more commonly causes them to explode when the evaporation pressure builds up. Sigh. It's the year 2015 and we still have paper and "water" in our electronic devices... It's no wonder they fail so rapidly.

The life rating of a capacitor isn't really saying when it will fail. It's the expected maximum time that it would take for an average sample of that particular capacitor to dry up enough to go out of spec so badly that the circuit fails. But they can and do go out of spec much earlier. Some brand new ones from the factory can already be duds. And to make matters worse, capacitors usually explode way before they go out of spec. And there are also huge variances in quality between brands and series and batches. A "5000 hours @ 105C" from "random brand" is not going to last as long as "5000 hours @ 105C" from Panasonic or another one of the top brands, and lies in the spec sheets are very common in all but the top brands. And if there's a problem with the electrolyte formula on a particular day, even a top brand can have a bad batch.

It's possibly to make "perfect" capacitors, and some manufacturers make ultra-high end capacitors for military use. But the prices are so high that there’s no general purpose market for it. So we're stuck with this garbage situation.

Worst of all: When capacitors die, there’s no way to know what will happen with the rest of the circuit. The death of a capacitor can also be the death of other components due to sudden voltage spikes, dirty electricity no longer being smoothed out by the capacitor, etc.

All electronics would pretty much last a whole lifetime without repairs if capacitors were as reliable as all other components. If there's a revolutionary new capacitor technology someday, it would put most electronics service technicians out of business since our shiny things would hardly break anymore.

But unfortunately the only way to help current-day electronics have long lives is to ensure that their capacitors don't reach high temperatures.

Anyway, do you have HDD at bay 3 and 4? Just want to know if that make my PSU temperature read 10C higher than your "normal" temperature.

Yeah I have hard drives in all bays, but they're almost permanently asleep via power-saving which means they're not spinning and not generating any heat. I see that my hard drives closest to the PSU are 2*C warmer than the others. But that's probably just from the graphics card underneath them, blowing warmer air up at those.

Unless your hard drives are super hot (mine are at ~20*C during idle and ~30*C-38*C during operation), they wouldn't heat up your PSU. Is your ambient temperature in the room hot? Maybe your PSU temperature sensor is faulty? Maybe you've got some bad capacitor generating extra heat? Are you overdrawing the power supply by installing 2x heavy graphics cards? Many possibilities.

 

[ZombiePhysicist]

You're welcome. And I agree. It's amazing that they give it out for free.



The PCI fan speed adjustment is safe, since it's just the gray fan near the front which blows over the PCI cards. But don't touch the power supply fan adjustments I also mentioned, because those are very dangerous if you don't carefully aim at fan speeds that still give safe temperatures.



Are you sure? Most modern graphics cards have super silent fans. I see that yours only has a single fan which means it has to work at a higher speed than two separate fans, so the extra noise may be innate to your card, but it should still be pretty quiet. To be sure that it's the card, try gently placing your finger on the center of the graphics card fan (where the label is) to stop it and check if the sound goes away. Then spin the fan with your finger to get it to start again.

If the noise comes from the graphics card fan, then there's nothing you can do unfortunately. Many graphics cards for Windows have Windows software to control the fan speed on the card itself, but such software doesn't exist for Macs. The Mac itself cannot see/control those fans, they're controlled by the firmware inside the graphics card chips.

But you probably have a combination of PCI/PS fans in the Mac revving up, *and* your graphics card fans revving up. So you can still reduce noise by at least checking and lowering the PCI fan speed on your system. See my first post here for that part of the tutorial. Avoid all the power supply stuff in later posts since that's dangerous and usually not necessary. But changing the PCI fan itself is safe as long as you follow my instructions to get a safe operating temperature range.

And as for your other question: Yes, running a 3D application and closing it again has been a "solution" used by people with the excessive fan noise. The reason that it "works" is that modern graphics cards have multiple power modes; 3D, 2D, etc. It seems that they start up in 3D mode by default (when the computer starts up from a cold boot), which draws more power and thus causes more heat/more fan speed/etc. By starting and closing a 3D application it seems the cards go "oh... we're back to 2D!" and thus reduce their power and heat so that the fans don't need to kick in. That's the only way I can explain why 3D applications are usable as a temporary fix. Because the Mac Pro itself doesn't know about the 3D application running, so it's got to be something on the card that changes after the 3D app is closed.

I hope this helps...

Thanks, but sadly, I'm positive. Only the graphics card fan kicks up as I used your app to also see if the values increased at all when you can see and hear the video card fan kick up like crazy.

Perhaps an easier solution than messing with the fans is if someone just writes a small app that puts it in the 3D mode, maybe rotates a cube for a second, and then just closes. Then we can make it a startup app that automatically closes, and switches back to that 2D mode, and then you dont have to mess with the fans! Sounds a bit easier and less engineering involved, no?

 

[SteveJobzniak]

Thanks, but sadly, I'm positive. Only the graphics card fan kicks up as I used your app to also see if the values increased at all when you can see and hear the video card fan kick up like crazy.

Perhaps an easier solution than messing with the fans is if someone just writes a small app that puts it in the 3D mode, maybe rotates a cube for a second, and then just closes. Then we can make it a startup app that automatically closes, and switches back to that 2D mode, and then you dont have to mess with the fans! Sounds a bit easier and less engineering involved, no?

Aww I am very sorry to hear that. Some graphics cards unfortunately don't have good automatic fan control and are more noisy than they need to be. Seems you've got one of those.

I use a totally silent card which I can recommend to everyone. I'm not even a gamer, so I went for this GTX 960 which is powerful enough to run all modern games on Ultra at 1080p resolutions: http://www.gigabyte.com/products/product-page.aspx?pid=5575

The most important thing for me was silence, so I researched and found this card which has low power draw and extremely quiet operation. It's literally inaudible, even during heavy load. And it only needs the 2x 6-pin PCIe power cables from the Mac Pro and doesn't need any external power supply. As an added bonus, it's part of the new nVIDIA generation that supports Metal on El Capitan which makes the whole operating system and all GUI animations smoother.

If more GPU power is needed, it is possible to find certain GTX 970 models that only use 2x 6-pin connectors but it's kinda hard (most are 1x 8-pin). And if even more is needed, the GTX 980s can be used but need a separate power supply since they only have 2x 8-pin connectors which need to deliver four times as much electricity as the Mac Pro's 2x 6-pin cables are capable of.

If changing graphics card isn't possible, you'll need some way to do the "3D trick". As for your idea of making a 3D application, you'd need someone to make it and sign it with an Apple Developer certificate so that it's launchable. But yeah that's doable. Doubt you're going to find anyone though. But one suggestion for a "lightweight" application I've seen is the free "OpenGL Extensions Viewer" from the App Store: https://itunes.apple.com/us/app/opengl-extensions-viewer/id444052073?mt=12 - You could make it autostart that application at startup and manually run the "3D Cube Test" and close it again after every cold reboot. It's only necessary after cold reboots, so maybe that workaround is acceptable, and it's cheaper than getting a different graphics card.

 

[h9826790]

Perhaps an easier solution than messing with the fans is if someone just writes a small app that puts it in the 3D mode, maybe rotates a cube for a second, and then just closes. Then we can make it a startup app that automatically closes, and switches back to that 2D mode, and then you dont have to mess with the fans! Sounds a bit easier and less engineering involved, no?

May be Automator can do that (even though it's relatively stupid programme if compared to a simple few seconds spinning cube apps).

I have hard drives in all bays, but they're almost permanently asleep via power-saving which means they're not spinning and not generating any heat. I see that my hard drives closest to the PSU are 2*C warmer than the others. But that's probably just from the graphics card underneath them, blowing warmer air up at those.

Unless your hard drives are super hot (mine are at ~25*C-30*C during operation), they wouldn't heat up your PSU. Is your ambient temperature in the room hot? Maybe your PSU temperature sensor is faulty? Maybe you've got some bad capacitor generating extra heat. Many possibilities.

My HDD at bay 3 & 4 is running at about 37C, I don't like to wait for them to spin up when I use it. So they almost never stop spinning. And there is a 2nd 7950 at slot 3 which keep "warming" these 2 HDD, which may eventually warm up the PSU a bit.

And yes, my room temperature is relatively warm. Even in winter, it's usually around 22-24. In summer, easier go over 33 (when I am not there, so no aircon).

My dual 7950 is mainly for FCPX, they almost never draw more than 75W via either 6pin. For gaming in Windows, they can draw up to 95W from each 6pins. However, my understanding is that may overload the mobo trace, but not the PSU. My overall power consumption never reach 450W, not even 50% of the limit. It's hard to believe that this "light" loading can cause a 980W rated PSU to run "too hot". (Again, I am using my aviation knowledge on this computer limitation. May be I am totally wrong)

The system ambient reads around 30 in winter, and above 39 in summer. So, for me, the other components reads 40+ is quite normal. However, I am living in a humid area, may be that will reduce the capacitor's "dry up effect" (even though moisture is not good for electronic generally).

 

[m4v3r1ck]

Great thread guys, thanks!

I think it's time I ASAP NEED TO ADJUST my fan settings too, so I would like some confirmation from you highly skilled tech-guys! Her are my readings for the two PSU sensors.

Setup Mac Pro 5.1 (2012) Mac OS X 10.10.5:
- PCIe-#1 GTX-670-FTW-4GB (PC)
- PCIe-#2 Apricorn DUO x2 with 2 840 EVO SSD's
- PCIe-#3 ATI 2600 HD (Mac) for bootscreen and maintenance
- PCIe-#4 CallDigit FastPro 6 USB 3.0 & eSATA
- BAYS-#1-4 7200 RPM spinners.

Running iStatsMenus & Daemon v4.22 (463)

PSU_1

PSU_2

Specially the reading in the 80*C's of the PSU_2 sensor is concerning me, after reading this thread.

1. Please advise me how to make a reasonable tweak in MacsFanControl for this.
2. Can I use MacsFanControl v1.3.2 in conjunction with iStatsMenus & Daemon v4.22 (463)?

Please see screen of the temps overview, setting my fans control to "default":

Installed MacsFanControl v1.3.2 and here is its Technical Information:

Macs Fan Control 1.3.2
MacPro5,1
Intel(R) Xeon(R) CPU X5650 @ 2.67GHz[100]
Mac OS X 10.10

Fans:
PCI-min{800}curr{1402}max{4500}mode{0}
PS-min{400}curr{872}max{2800}mode{0}
EXHAUST-min{500}curr{674}max{2800}mode{0}
INTAKE-min{600}curr{674}max{2800}mode{0}
BOOSTA-min{600}curr{1253}max{5200}mode{0}
BOOSTB-min{600}curr{1252}max{5200}mode{0}


TempSensors:
TA0P (Ambient) - 30.75, 30.75
TCAC (CPU A core from PCECI) - 64.0625, 35.9375
TCAD (CPU A Diode) - 60.375, 60.375
TCAG () - 80, 80
TCAH (CPU A HeatSink) - 55.5, 55.5
TCAS () - 55.5, 55.5
TCBC (CPU B core from PCECI) - 50.0938, 49.9062
TCBD (CPU B Diode) - 46.375, 46.375
TCBG () - 80, 80
TCBH (CPU B HeatSink) - 41.75, 41.75
TCBS () - 41.75, 41.75
TH1F () - 31.2305, 31.2305
TH1P (Drive Bay 1) - 31.25, 31.25
TH1V () - 34.4609, 34.4609
TH2F () - 31.8828, 31.8828
TH2P (Drive Bay 2) - 31.875, 31.875
TH2V () - 31.4453, 31.4453
TH3F () - 32.1445, 32.1445
TH3P (Drive Bay 3) - 32.125, 32.125
TH3V () - 32.1211, 32.1211
TH4F () - 34.6406, 34.6406
TH4P () - 34.625, 34.625
TH4V () - 42.2773, 42.2773
THPS () - 42.2773, 42.2773
THTG () - 58, 58
TM1P () - 44.125, 44.125
TM2P () - 46.5, 46.5
TM3P () - 48.125, 48.125
TM4P () - 48.5, 48.5
TM5P () - 47.875, 47.875
TM6P () - 50.75, 50.75
TM7P () - 45, 45
TM8P () - 34.75, 34.75
TMA1 () - 43.25, 43.25
TMA2 () - 47, 47
TMA3 () - 48, 48
TMA4 () - -1, 129
TMB1 () - 38, 38
TMB2 () - 37.25, 37.25
TMB3 () - 36, 36
TMB4 () - -1, 129
TMHS () - 48, 48
TMLS () - 36, 36
TMPS () - 48, 48
TMPV () - 44.7383, 44.7383
TMTG () - 75, 75
TN0D (Northbridge Chip) - 74.75, 74.75
TN0H (IOH HeatSink) - 65, 65
TNTG () - 65, 65
Te1F () - -1, 129
Te1P (PCIE Ambient) - 36.625, 36.625
Te1S (PCIE slot 1) - -1, 129
Te2F () - -1, 129
Te2S (PCIE slot 2) - -1, 129
Te3F () - -1, 128
Te3S (PCIE slot 3) - -1, 128
Te4F () - -1, 129
Te4S (PCIE slot 4) - -1, 129
Te5F () - -1, 129
Te5S () - -1, 129
TeGG () - 85, 85
TeGP () - -1, 0
TeRG () - 60, 60
TeRP () - -1, 0
Tp0C (PSU1 Inlet Ambient) - 34.5781, 34.5781
Tp1C (PSU1 Secondary Component) - 47.0938, 47.0938
TpPS () - 47.0938, 47.0938
TpTG () - 90, 90


Disk drives:
Samsung SSD 840 EVO 1TB - 48
Samsung SSD 840 EVO 500GB - 39
WDC WD20EFRX-68AX9N0 - 34
WDC WD20EFRX-68AX9N0 - 36
SAMSUNG HD501LJ - 37
SAMSUNG HD501LJ - 31
Samsung SSD 840 EVO 500GB - 0
Samsung SSD 840 EVO 500GB - 0

Directions are greatly appreciated!

Cheers

 

[h9826790]

That's what I want to say, the native fan control may gives you high PSU temperature. And PSU fan obviously not directly PSU temperature related, nor intend to keep it below 35C.

I don't know what setting is good for you, but what I am doing now is allow the fan to run at idle when below 43C (PSU 2) and every 1C above that will increase the fan speed by 100RPM. This actually make my PSU run cooler most of the time. And the fan won't go crazy when the GPUs are under stress (the crazy fan will make the PSU much cooler than normal, so I personally assume it's not necessary. Of course I can be totally wrong).

 

[m4v3r1ck]

Thanks for your prompt reply. I just simply don't understand why the PSU temperature sensor would read ~80*C, without software tools like MFC. As I understand that would be factory defaults then???

Setting up MFC with the following tweaks for my first temperature testing with a few render jobs. System Ambient = 32*C atm (room with no airconditioning):
1. PCI - Sensor-based value -> PCI Ambient: 32*C <-> 60*C
2. PS - Sensor-based value -> PSU1 Secondary Component: 32*C <-> 60*C
3-6. Auto

Cheers

 

[SteveJobzniak]

@m4v3r1ck: Your chart had a single spike to 80*C but there are no scale-lines so I cannot see what your average line before that is at. It looks like 30-40 though.

The aim for less than 40*C is the single greatest thing a computer manufacturer could do to ensure the power supply lasts for a long time, and that's a normal operating temperature for good PC power supplies. But it's looking like Apple's own fan control only specifically reaches <= 35*C under stock hardware setups and in northern hemisphere environments.

To me it's starting to look like Apple have based the PSU fan speed on something other than the two PSU temperature sensors (or at least not "just" temperature). Maybe the sensors aren't reliable enough to be used alone?

It's also clear now that Apple never predicted we would be modding their machines like this, so the default-programmed fan behavior is not ready for anything but what the machine shipped with.

Anything else and we get either too much fan noise, or waaaaay too much PSU heat.

And since your fans didn't kick in to reduce heat when it was temporarily at 80*C, I suspect the temperature sensors are a small part (or none) in the fan speed of the PSU.

I think the default fan speed logic is on the logic board firmware so unfortunately we cannot look at it and say "wtf, why did Apple do THAT?!" (but I am sure we would say that if we could see it). I suspect more and more that it's based on power draw in watts instead of heat, which means the fan will not react to actual heat.

In fact, one possible (sinister) reason for Apple to have done that is that they wanted the Mac Pro to be silent even though heat massively reduces the life of electronics. The PSU is a noisy fan and anything that causes it to spin up would get bad reviews. "We can always sell a new PSU later, but we cannot sell machines if they are noisy in the first place."

Unfortunately there's no way to keep a computer cold in hot countries without increasing the fan speed to compensate. But even the fan speed can't cure everything, since a hot ambient temperature just means it blows hot air over the computer. That is why your country's ambient temperature is the best predictor of an early death for a new computer. It is also why webhosts invest so much money and electricity in cooling their computer rooms to the point where it sounds like being inside a jet engine when you are in the room...

My advice is that you monitor temperature and PSU fan speed via Macs Fan Control.

Set all sensors to auto again so we monitor your default behavior.

Then just write down the values a few times under different loads, like immediately after playing a game for 15 minutes etc. And also what it's at when idle for 30 minutes doing nothing. Remember that it takes ~15 minutes for temperatures to stabilize during each workload.

We need to know your operating ranges. In fact, press Command+Shift+4 and then Space and click the Macs Fan Control window to take screenshots of it. Come back in a day or so with the results and we can try to figure it out.

 

[h9826790]

That 80C is on the 24hr graph. It looks like the PSU 2 temperature was above 77C for at least 1 hour.

And from the 7 days graph. It seems that PSU work constantly above 50C, and the average is around 60C.

The last 1hr chart shows a very normal temperature to me. However, that's follow a cold boot. And it seems that he was not "working" on his Mac, the relatively light loading keep the PSU much cooler than the average operating temperature.

What looks strange to me is the big split between the PSU 1 and PSU 2 temperature. They are very close on my Mac (PSU 2 is hotter, but just few degree C).

The very first thing in my mind is the fan (or inside the PSU) may be dusty, which greatly reduce the cooling efficiency. Or the optical bay is fully loaded, which cause the air much harder to reach the PSU.

But what we do know now is the PSU 2 temperature sensor should be located on a heat generating component. Otherwise, it's hard to get that temperature (assuming the sensor is working properly).

 

[m4v3r1ck]

Thanks again both for your replies! Will testdrive MFC for a few days, with several different workloads and leaving it idle.

Keep your posted!

Cheers

 

[SteveJobzniak]

@h9826790: Yes, the two temperatures in his graphs are meant to be almost the same. But I can't read his graphs because they don't have any lines to indicate what each represents. Looking at it again I think it seems he's around 45-50*C constantly, maybe even 50-60*C.

That's definitely not normal. Nor is yours. I mean, if it's supposed to be "normal" for the Mac Pro power supply to have the kinds of temperatures you two are getting, then the Mac Pro *itself* is mis-designed. I speculated above that Apple have made the PSU fan speed ignore temperature so that the computer stays silent even in hot countries and even when the components are literally cooking inside. This is the most likely explanation for the extreme heat you two are getting. The power supply fan clearly isn't reacting properly to heat, and yes it'll absolutely suffer for it. Silence being prioritized instead of 5-10x longer computer life, seems to be Apple's motto here. Like I said, we can always replace the PSU, but we can't live with noisy computers, so I am not saying their choice is wrong. Just that the computer is definitely hurting from that choice.

The specifications for the Mac Pro actually state it’s not meant to be used in hot countries:
https://support.apple.com/kb/SP506?locale=en_US

“Operating temperature: 50° to 95° F (10° to 35° C)” (meaning 35*C is the maximum recommended ambient temperature of the room, beyond which you’ll hurt the computer too much to be recommendable).

It’s understandable that we want to believe that Apple always does everything right. But your temperatures in the power supply (anything over 40*C) are not normal for open-grille desktop power supplies with fans and vents, and they’re very, very dangerous and will guarantee an early death. Every additional 10*C in the power supply halves the lifetime of it.

Apple bought the power supply from a company called Delta and slapped their label on it. It’s not full of magic “Applejuice”. It’s a generic, slightly-better-than-average power supply.

Heat hurts the Mac Pro as much as any other power supply. Temperature is temperature and Mac Pros do *not* handle it better than other computers. The lower you can get it the better. If we could get it to 10*C it’d run the best, since that’s cozily close to freezing temperature.

The heat in PSUs only comes from inefficiencies in the power conversion of the power supply - what doesn’t end up as power in the computer ends up as heat in the power supply. Better power supplies have better efficiency which results in less wasted power and lower temperatures. Fans are meant to bring the actual heat down to safe levels. A little bit of heat is normal, but a *lot* (like anything above 50*C) is very, very dangerous and definitely *not* what the Mac Pro power supply is meant to be handling, and I’ll show you why…

Most power supplies are rated for full power output and longest lifetime at 25 *C. The hotter it gets, the less watts it puts out and the shorter the life of the capacitors. This is called the de-rating curve.

In fact, if you look at the Life span information for power supplies at Wikipedia, they repeat the same number; an operating temperature at 25*C under full load is a common specification:

https://en.wikipedia.org/wiki/Power_supply_unit_(computer)#Life_span

Now let’s look at this quote which explains that we can never expect 10-20 years out of a PSU (and why I have to repair electronics constantly):
“An estimated MTBF value of 100,000 hours (roughly, 140 months) at 25 °C and under full load is fairly common.[24] Such a rating expects that, under the described conditions, 77% of the PSUs will be operating failure-free over three years (36 months); equivalently, 23% of the units are expected to fail within three years of operation. For the same example, only 37% of the units (fewer than a half) are expected to last 100,000 hours without failing.”

Here is more information about what de-rating does to a power supply, and how the output watts fall after reaching 25*C (usually, but can be stable up to 50 for better ones):
http://forums.anandtech.com/showthread.php?t=157636

Quote:
“How does the temperature inside of my case affect the performance of my power supply?
Power supplies can perform differently depending on the temperature at which they are operating at. When a power supply is rated for its total output wattage, it is rated to do so at a particular temperature. Anything beyond this temperature may take away from the power supply's capability. A power supply that is rated to put out 550W at 25°C or 30°C (room temperature) may only be able to put out 75% of that at 40°C or 50°C (actual operating temperature). This difference is called the "de-rating curve". A normal operating temperature for a computer power supply is 40°C.”

More information can be found by searching for “power supply de-rating” online.

But most damning of all, let’s look at the Delta catalog for 2015, by the people who manufactured the Mac Pro power supply:
http://www.deltapsu.com/download/delta-standard-power-supplies-catalog

Do a word search for “de-rating” in that document.

You will find something (unsurprising): They’re like every other power supply manufacturer and want us to run their supplies at 25-40*C. All of their power supplies lose ~3% of their output watts *per extra degree* once they reach 50*C, and they lose ~5% of their watts *per extra degree* once they reach 70*C.

So if you’re at 80*C, you are 30*C above the maximum.

You should all be getting yours lower than 40*C for the longest power supply life (my Mac Pro 2009 power supply is at 33*C during casual use with the PS fan at the lowest 600 RPM).

But it’s not a huge deal if you run them hot, like I said you can always replace the PSU or the capacitors when they inevitably break. When they do, I suggest getting them to a repair shop for re-capping with new capacitors, since a brand new PSU itself costs way more ($200+) than a few cheap caps ($30-40) + an hour of work.

Delta is a good PSU brand and Apple has picked a good model with better-than-average-to-great capacitors, and my PSU has so far been living for 6 years of 24/7 use at ~30-35*C operating temperature. I hope to get another 2-10 years out of it, thanks to this good operating temperature, but I've still got some valid worries since it's 6 years old and could blow up at any moment (remember what I said about capacitors being finicky little rolls of paper and water; they're unreliable as hell). Thankfully I'll be ready with replacement capacitors when it happens. It'll be the 1,430,473th power supply I've had to re-cap in electronics. I can't wait until we get new capacitor technology that never breaks. That day I can retire from constantly having to repair electronics around the house. ;-)

 

[m4v3r1ck]

@SteveJobzniak Temperature spikes (60*C - 80*C) are getting ironed out!

So far so good:

System Ambient ~31*C

PSU_1

PSU_2

RPM's of PCI & PSU (noise) are at acceptable levels!

Cheers

 

[SteveJobzniak]

@m4v3r1ck: Nice temperatures. I see that your PCI ("Expansion slots") + PS ("Power supply") fans are spinning very fast. Is that really with the default Mac fan auto-behavior, or did you adjust it with Macs Fan Control? (gosh I hate the silly name of that app)

I guess your computer is annoyingly noisy right now? Or maybe you've gotten lucky with silent fan bearings. Mine would have been really loud with those speeds.

To help you out for comparison, here's a screenshot of my statistics right now while everything on the computer is being heavily loaded, but I'm *not* using any 3D applications (so the graphics card is cool).

* You can see that my hard drives are all awake, and getting 35*C in the first two bays, 39*C in the last bay (above the graphics card heat, so that's understandable).
* The ambient temperature of the PCIE slots is 33*C thanks to the graphics card. The PCI fan speed is at the lowest.
* My power supply is 32*C at the inlet and 34*C at the "components" (wherever that sensor is). The PSU fan speed is at the lowest.
* The behavior of PCIE and PSU fan speeds are using the automatic temperature ranges I've described in my first two posts in this thread, so they will speed up briefly if they ever heat up more.
* The reason for the low CPU temperatures is that I've replaced the heatsink material with Gelid's (the best in the business at transferring heat), as well as installed a bunch of silicon pads to transfer heat away from surrounding components. My temperatures are much lower than Apple's stock CPU cooling setup. But I hope nobody asks about that since I didn't document the modification process and it's been 2 years ago.
* Everything is looking great. And nobody needs to worry about the only high temperatures in my screenshot; the "worrying" 65*C at IOH HeatSink and 79*C at Northbridge Chip, because both of those are meant to reach those kinds of temperatures. They don't contain any capacitors; they're just a very hot integrated circuit that is hot in any computer on the planet and is safe at that temperature.