Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Their temperature targets are much better rooted in practical experience than the silliness you often see in overclocker's forums and guide.

I take it from your lack of any comment though that you feel this article refutes what I said, which it doesn't.
I was just replying to the last bit about lasting the longest. I imagine other builders would agree with Puget that temps above 95C would shorten the system's life considerably.

Yeah, it's fine. Unfortunately there's a lot of unsubstantiated "wisdom" in the enthusiast community that a CPU has to be at 60C or 50C or 80C or whatever magic temperature in order to run best and/or last the longest.

80-85C is where this company felt comfortable speccing out their systems in 2014. An article (well reasoned as it is) from a custom shop does not refute what we've been empirically observing of Intel CPUs since they started pushing their 14nm tech to its limits. They run very hot when fully loaded.
Intuitively, smaller processes would be more sensitive to factors like heat and electromigration. Intel may just be relaxing its standards with regards to the longevity of their products.

If we start seeing higher than normal defect rates due to CPU or nearby component failures then it's worth revisiting this discussion. We almost certainly won't.
Studios and developers would be running their computers at much higher loads and for much longer durations than casual users. Do we have stats on whether Macs that run at just under 100 last comparably long to the custom PCs that run at 80-85C in professional environments?
 
  • Like
Reactions: dabotsonline
Here's my fortnite temps.
 

Attachments

  • Screen Shot 2018-11-09 at 10.24.20 PM.png
    Screen Shot 2018-11-09 at 10.24.20 PM.png
    61.6 KB · Views: 633
I was just replying to the last bit about lasting the longest. I imagine other builders would agree with Puget that temps above 95C would shorten the system's life considerably.

Yes they would. And "considerably" would be the correct word. However, considerable reduction from what? A CPU can easily live for 20 years before failing. Cutting the life of a CPU in half from 20 years to 10 by running it 24/7 at 95C is a consideration Apple's engineers had to make, balanced with the criteria of noise, cost and maximum performance for the vast majority of use cases.

Intuitively, smaller processes would be more sensitive to factors like heat and electromigration.

More sensitive to electromigration, yes, definitely. Heat is universally entropic no matter how small you are :p

Intel may just be relaxing its standards with regards to the longevity of their products.

Yes. For which they would be justified because for at least the last decade (more really), CPU failure rates have been extremely low, even in 24/7 environments.

Studios and developers would be running their computers at much higher loads and for much longer durations than casual users.

Yes, again though, context is important. "Much longer durations" does not usually mean 100% load 24/7 for 10 straight years. It can certainly mean potentially weeks or months per year where the CPU is nearly fully loaded, but almost never is it 24/7 for years on end.

Do we have stats on whether Macs that run at just under 100 last comparably long to the custom PCs that run at 80-85C in professional environments?

Unfortunately no (at least none that I know of that are in scientific journals).

--

To round this out, I can provide an anecdote if you're interested. One of my businesses puts machines down in the Bahamas. We frequently use consumer grade hardware (to wit, not Xeons and ECC RAM) to save costs, but they are i7s. One of the problems with the Bahamas is that the scattered islands make service times very long (the company we deal with can take weeks to get to a machine sometimes - both by negligence and the physical difficulty of getting to places after heavy storms).

Over the last five years that we've had machines down there, several times now we've gotten calls about "the system running slowly." A quick scan of the logs shows the CPU fan died MONTHS before they started to notice "something." Note that every one of these machines runs 24/7 with system load averages around 100% (load average =/= CPU load, but the point is they aren't idle) in pretty hot ambient conditions. The system logs are megabytes worth of line-after-line kernel temperature throttling messages.

I'm talking CPUs that have been bouncing off TJmax for months like a tachometer's needle bounces around when you hit the rev limiter.

After they eventually get around to replacing the CPU fan, the system immediately returns to normal and continues to run just fine for many years after. This has happened several times now.

Just an anecdote, I know. Not empirical to a standard I would remotely consider scientific. But it is interesting none-the-less how much abuse these consumer grade CPUs can actually take.
 
Yes they would. And "considerably" would be the correct word. However, considerable reduction from what? A CPU can easily live for 20 years before failing. Cutting the life of a CPU in half from 20 years to 10 by running it 24/7 at 95C is a consideration Apple's engineers had to make, balanced with the criteria of noise, cost and maximum performance for the vast majority of use cases.



More sensitive to electromigration, yes, definitely. Heat is universally entropic no matter how small you are :p



Yes. For which they would be justified because for at least the last decade (more really), CPU failure rates have been extremely low, even in 24/7 environments.



Yes, again though, context is important. "Much longer durations" does not usually mean 100% load 24/7 for 10 straight years. It can certainly mean potentially weeks or months per year where the CPU is nearly fully loaded, but almost never is it 24/7 for years on end.



Unfortunately no (at least none that I know of that are in scientific journals).

--

To round this out, I can provide an anecdote if you're interested. One of my businesses puts machines down in the Bahamas. We frequently use consumer grade hardware (to wit, not Xeons and ECC RAM) to save costs, but they are i7s. One of the problems with the Bahamas is that the scattered islands make service times very long (the company we deal with can take weeks to get to a machine sometimes - both by negligence and the physical difficulty of getting to places after heavy storms).

Over the last five years that we've had machines down there, several times now we've gotten calls about "the system running slowly." A quick scan of the logs shows the CPU fan died MONTHS before they started to notice "something." Note that every one of these machines runs 24/7 with system load averages around 100% (load average =/= CPU load, but the point is they aren't idle) in pretty hot ambient conditions. The system logs are megabytes worth of line-after-line kernel temperature throttling messages.

I'm talking CPUs that have been bouncing off TJmax for months like a tachometer's needle bounces around when you hit the rev limiter.

After they eventually get around to replacing the CPU fan, the system immediately returns to normal and continues to run just fine for many years after. This has happened several times now.

Just an anecdote, I know. Not empirical to a standard I would remotely consider scientific. But it is interesting none-the-less how much abuse these consumer grade CPUs can actually take.
I'm getting my new mac mini and heading for the Bahamas!
 
Yes they would. And "considerably" would be the correct word. However, considerable reduction from what? A CPU can easily live for 20 years before failing. Cutting the life of a CPU in half from 20 years to 10 by running it 24/7 at 95C is a consideration Apple's engineers had to make, balanced with the criteria of noise, cost and maximum performance for the vast majority of use cases.



More sensitive to electromigration, yes, definitely. Heat is universally entropic no matter how small you are :p



Yes. For which they would be justified because for at least the last decade (more really), CPU failure rates have been extremely low, even in 24/7 environments.



Yes, again though, context is important. "Much longer durations" does not usually mean 100% load 24/7 for 10 straight years. It can certainly mean potentially weeks or months per year where the CPU is nearly fully loaded, but almost never is it 24/7 for years on end.



Unfortunately no (at least none that I know of that are in scientific journals).

--

To round this out, I can provide an anecdote if you're interested. One of my businesses puts machines down in the Bahamas. We frequently use consumer grade hardware (to wit, not Xeons and ECC RAM) to save costs, but they are i7s. One of the problems with the Bahamas is that the scattered islands make service times very long (the company we deal with can take weeks to get to a machine sometimes - both by negligence and the physical difficulty of getting to places after heavy storms).

Over the last five years that we've had machines down there, several times now we've gotten calls about "the system running slowly." A quick scan of the logs shows the CPU fan died MONTHS before they started to notice "something." Note that every one of these machines runs 24/7 with system load averages around 100% (load average =/= CPU load, but the point is they aren't idle) in pretty hot ambient conditions. The system logs are megabytes worth of line-after-line kernel temperature throttling messages.
I am not going to continue the discussion with you over CPU temps, I disagree with you and I am not going to spend more time on it.
I do have a question about the average 100% statement you made. You say that load average is not the same as CPU load and you say that the average load was 100%. How can you have an average of 100% when you can't go higher than 100% load? Also, how is load not the same as CPU load when we are talking about load of a CPU? I am confused.
 
I do have a question about the average 100% statement you made. You say that load average is not the same as CPU load and you say that the average load was 100%. How can you have an average of 100% when you can't go higher than 100% load? Also, how is load not the same as CPU load when we are talking about load of a CPU? I am confused.

Load average, at least as far as it pertains to Linux, is a measure of the average number of threads awaiting execution divided by the number of cores. You could have, for instance, 20 threads writing to disk on a 4-core system and if the disk can’t keep up your load average can easily go into the hundreds of percent.

In a typical desktop scenario load average is a combination of raw CPU utilization (compute tasks that don’t complete before the next thread is ready) and waiting for disk I/O.

Load average is generally understood to be a more useful measure of a server’s true loading than just the CPU load.
 
Last edited:
  • Like
Reactions: LeeW and auduchinok
Yes they would. And "considerably" would be the correct word. However, considerable reduction from what? A CPU can easily live for 20 years before failing. Cutting the life of a CPU in half from 20 years to 10 by running it 24/7 at 95C is a consideration Apple's engineers had to make, balanced with the criteria of noise, cost and maximum performance for the vast majority of use cases.



More sensitive to electromigration, yes, definitely. Heat is universally entropic no matter how small you are :p



Yes. For which they would be justified because for at least the last decade (more really), CPU failure rates have been extremely low, even in 24/7 environments.



Yes, again though, context is important. "Much longer durations" does not usually mean 100% load 24/7 for 10 straight years. It can certainly mean potentially weeks or months per year where the CPU is nearly fully loaded, but almost never is it 24/7 for years on end.



Unfortunately no (at least none that I know of that are in scientific journals).

--

To round this out, I can provide an anecdote if you're interested. One of my businesses puts machines down in the Bahamas. We frequently use consumer grade hardware (to wit, not Xeons and ECC RAM) to save costs, but they are i7s. One of the problems with the Bahamas is that the scattered islands make service times very long (the company we deal with can take weeks to get to a machine sometimes - both by negligence and the physical difficulty of getting to places after heavy storms).

Over the last five years that we've had machines down there, several times now we've gotten calls about "the system running slowly." A quick scan of the logs shows the CPU fan died MONTHS before they started to notice "something." Note that every one of these machines runs 24/7 with system load averages around 100% (load average =/= CPU load, but the point is they aren't idle) in pretty hot ambient conditions. The system logs are megabytes worth of line-after-line kernel temperature throttling messages.

I'm talking CPUs that have been bouncing off TJmax for months like a tachometer's needle bounces around when you hit the rev limiter.

After they eventually get around to replacing the CPU fan, the system immediately returns to normal and continues to run just fine for many years after. This has happened several times now.

Just an anecdote, I know. Not empirical to a standard I would remotely consider scientific. But it is interesting none-the-less how much abuse these consumer grade CPUs can actually take.

To be honest I could not care if my CPU lasts 10 years instead of 20, it’s throttling that is the issue. I don’t want a 6 core to be slower than a lower spec machine cause it throttle , like when they launched the MacBook Pro’s.

Where Apple engineers are going wrong these days , and with today’s turbo boost , is thermals , they are not making good choices in compromises , they are gimping the chosen CPUs with inappropriate cooling , which is something they could get away with using older CPUs . Today you need to keep a CPU from throttling

And don’t forget while CPUs are engineered to deal with heat, heat kills other components in a computer ;) hence why cooling is very important. Even more worrying as today’s macs cannot be user repaired , even for a simple fan etc
 
Last edited:
After my mini experiences with the i7 I’d avoid them like the plague until Apple really improve cooling in the thing . How about some side vents or even top vents or would that destroy the aesthetic / cost more ?

Not impressed with this upgrade especially after 4 years...
 
To be honest I could not care if my CPU lasts 10 years instead of 20, it’s throttling that is the issue. I don’t want a 6 core to be slower than a lower spec machine cause it throttle , like when they launched the MacBook Pro’s.

Where Apple engineers are going wrong these days , and with today’s turbo boost , is thermals , they are not making good choices in compromises , they are gimping the chosen CPUs with inappropriate cooling , which is something they could get away with using older CPUs . Today you need to keep a CPU from throttling

And don’t forget while CPUs are engineered to deal with heat, heat kills other components in a computer ;) hence why cooling is very important. Even more worrying as today’s macs cannot be user repaired , even for a simple fan etc
But it's only throttling when it hits the limit intel set. This would be the same if you had it in a desktop you built. Intel set the limit. Intel's stock coolers are only designed for the 8700 to 65w TDP at 3.2ghz. Not for turbo boost frequencies. Anandtech has a nice write up about this.
This thing is going all the way to the thermal limit without any additional throttling from apple.
So let's say you could put in a better cooler and drop those temps 10c to say, 87c. It's just going to ramp further and temps go right back in to the 90's. That's what it's designed to do. And you can keep putting on better coolers and it's going to keep ramping up until it hits 100c and throttles or hits the turbo boost upper limit. And then you'd need an even better cooler to back it off from that.
My 5820 6 core on my desktop is in the 80's while encoding and I have to use water cooling just to achieve that! Short of putting a water cooler in the mini I don't know what else apple could have done.
Apples design is very good for getting heat from the CPU out of the case. My mini's case stays cool under load. Only the air coming out is warm. So it's not holding a lot of hot air in the case. I have opened it up to upgrade the ram. I'm not an engineer(I only work for them) but it looks like a good design overall. The heatsink and fan take the heat directly from the CPU straight out of the back of the case.
The only concern is the heat on the board around the CPU itself. Is there much heat there? Did apple design the board so more heat sensitive components are further away from that area? It would be easy enough to test some of this.
 
Let me preface by saying I am not well-versed enough to offer anything to the discussion about what is and is not safe for CPUs and the machine as a whole. Just want to offer a couple observations.

I remember this exact discussion about 12 years ago when the Core Duo MacBook Pro was released. Same exact things being said, people reapplying thermal paste, laptop coolers, etc. I find it interesting how some things never change. Also, I find it interesting that the people tasked with designing a machine are never trusted as competent by enthusiasts. We see this in the car tuner world as well. I am not saying the enthusiast is wrong, but isn't it prudent to trust the people that are tasked with engineering something that has to be built in the hundreds-of-thousands while having a low fail rate and are familiar with the entirety of the project over the person behind a keyboard that is hyper-focusing on a single component and mostly drawing from "family remedies?"

I know we have instances where the enthusiast is correct (in fact I think the Core Duo MBP was verified as having a manufacturing issue with thermal paste), but far more often it turns out the people making and designing these did just fine.
 
Every time there’s a high-temp discussion, people talk about lifespan of the CPU. I’ve never read of a CPU burning out on a Mac. Seems like it’s all anecdotal worry (if that) if you ask me, - all fear-mongering.

AMD graphics giving up (burning up?) is a different matter than intel cpus (e.g. radeongate on mbp 2011). I have personal bad experience with this problem.

Having said that my 2011 mbp 15” still runs reliably after hacking it to run only on discrete graphics, though.

I actually see the lack of discrete AMD card as a good thing in the 2018 Mac mini.
 
AMD graphics giving up (burning up?) is a different matter than intel cpus (e.g. radeongate on mbp 2011). I have personal bad experience with this problem.
This was an issue on the 2011 imacs as well. AMD Video card burned up. Apple even extended the warranty on them as it's a known issue with that particular model of GPU. And replaced them with the same chip. However no matter how much I pushed the CPU, no matter how hot it got(to the point the case was almost too hot to touch), it never died. For the record the CPU and GPU were on opposite sides of the 27" imac and each had their own cooling system and fan exhausting to opposite sides of the case. The video card was not part of the main logic board. It was a daughter card that could be removed.

Aside from that video card, which was a known issue, i've never seen heat damage in any computer or component in my life. I have several computers here and containers of motherboards and cpus. All still functioning(last I used them). Some nearly two decades old.
 
  • Like
Reactions: dabotsonline
Every time there’s a high-temp discussion, people talk about lifespan of the CPU. I’ve never read of a CPU burning out on a Mac. Seems like it’s all anecdotal worry (if that) if you ask me, - all fear-mongering.
Yes where the heck is this coming from! CPUs were not invented yesterday!
[doublepost=1541867842][/doublepost]
Yes they would. And "considerably" would be the correct word. However, considerable reduction from what? A CPU can easily live for 20 years before failing. Cutting the life of a CPU in half from 20 years to 10 by running it 24/7 at 95C is a consideration Apple's engineers had to make, balanced with the criteria of noise, cost and maximum performance for the vast majority of use cases.



More sensitive to electromigration, yes, definitely. Heat is universally entropic no matter how small you are :p



Yes. For which they would be justified because for at least the last decade (more really), CPU failure rates have been extremely low, even in 24/7 environments.



Yes, again though, context is important. "Much longer durations" does not usually mean 100% load 24/7 for 10 straight years. It can certainly mean potentially weeks or months per year where the CPU is nearly fully loaded, but almost never is it 24/7 for years on end.



Unfortunately no (at least none that I know of that are in scientific journals).

--

To round this out, I can provide an anecdote if you're interested. One of my businesses puts machines down in the Bahamas. We frequently use consumer grade hardware (to wit, not Xeons and ECC RAM) to save costs, but they are i7s. One of the problems with the Bahamas is that the scattered islands make service times very long (the company we deal with can take weeks to get to a machine sometimes - both by negligence and the physical difficulty of getting to places after heavy storms).

Over the last five years that we've had machines down there, several times now we've gotten calls about "the system running slowly." A quick scan of the logs shows the CPU fan died MONTHS before they started to notice "something." Note that every one of these machines runs 24/7 with system load averages around 100% (load average =/= CPU load, but the point is they aren't idle) in pretty hot ambient conditions. The system logs are megabytes worth of line-after-line kernel temperature throttling messages.

I'm talking CPUs that have been bouncing off TJmax for months like a tachometer's needle bounces around when you hit the rev limiter.

After they eventually get around to replacing the CPU fan, the system immediately returns to normal and continues to run just fine for many years after. This has happened several times now.

Just an anecdote, I know. Not empirical to a standard I would remotely consider scientific. But it is interesting none-the-less how much abuse these consumer grade CPUs can actually take.
You’re the man. You obviously know what you’re talking about. This and your previous posts in this discussion are the posts I like to read on this forum.
 
Every time there’s a high-temp discussion, people talk about lifespan of the CPU. I’ve never read of a CPU burning out on a Mac. Seems like it’s all anecdotal worry (if that) if you ask me, - all fear-mongering.

It seems that GPUs have had a much more pronounced failure rate for Apple (NVIDIA 8600M GT, Radeon 6750M/6770M, NVIDIA GT 650M, ATI Radeon M290/M295X).

It is possible that CPUs have failed and the whole motherboard has required replacement, but Apple has not issued a motherboard repair program that I can recall, even though they have instituted several GPU/Video Issue related Repair Programs in recent years.

I think if Apple is using 5-7 after the end of manufacturing as a general guide before ending formal support on a particular computer, you are seeing anywhere from 7-10 total years of life expected from a Mac. Some may live longer lives, some may live shorter lives, but I think they have lived full lives at this point, because either the parts are now harder to obtain, especially from Intel or the technology itself has moved on.

Some users would call that planned obsolesecense, but it is no different than what we experience for other items purchased in our lifetimes, such as a car, a refrigerator or a TV. Technology advances and companies cannot keep manufacturing parts for obsolete tech. Example: Try to find a brand new IDE hard drive for an older Power Mac G3 or G4. Pretty much impossible, although converters do exist for SATA to IDE. The next question would be why would someone do that anyways other than for an application that has never transitioned to Intel and runs some sort of critical equipment to which an alternative was never built or a vintage computer collector. In other words, edge cases.

In 4-5 years, finding a 300-Series motherboard that would work with a Coffee Lake CPU will be impossible to find new and a tad more difficult used. Anecdotally, I think Intel only expects 5-7 years total on it consumer CPUs, maybe longer for the Core X-Series and I think the minimum for the Xeon is 10 years.

My long-winded point is that all of the gnashing of teeth regarding thermal throttling and Apple not building properly cooled enclosures ignores quite a bit of market reality, the realities of simply engineering anything beyond a beige or black mini-tower and the expertise that we as the end user lack compared to an Apple engineer with a degree in Electircal Engineering, Computer Science, thermodynamics, et al., but yet seem to think they can engineer better than Apple’s or for that matter, Intel’s personnel.

The Monday morning quarterbacking that happens here and on other forums is overwrought and exhausting.
[doublepost=1541870228][/doublepost]
I'm not expecting the i5 to be faster. I'm expecting it to be a little cooler than the i7 based on these findings and hoping for fans to spin up a little less often with it.
I really care about quiet operation, just not enough to get an i3. ;)

I'm well aware of the fact that the i7 will be a little faster in cases that make use of hyperthreading.

So the i7 is stable at 3.6 to 3.7ghz under load, depending on who you ask.

And the i5 is stable at 3.9ghz.

I guess that proves that the i7 runs a little hotter due to hyperthreading. More and more convinced I should go for the i5.

The delta between the Core i5 and the Core i7 is only 200MHz, which is really inconsequential considering the potential that multi-threaded operation may bring to your particular workload. If the Core i5 was 8-cores/8-threads, I think you would have a better argument for going with that particular CPU if your workload benefitted from more “real” cores than from more “virtual cores” in the i7. Bearing that out, clock speed almost always ends up ruling the day in some way, and that does not just apply to synthetic benchmarks.

Also, while Apple does not have the absolute fastest systems in terms of performance, there are a multitude of YouTube videos showing that Apple does have the most “balanced” systems. Meaning, that what I tend to see is the performance of many (most?) mass market PCs and quite a few bespoke PCs end up having higher peak performance over a short duration, but end up at the bottom when it comes to sustained performance needed to accomplish a particular task.

The most popular lament has been the lack of a discrete GPU and user-serviceable and user-replaceable storage, but the GPU would generate additional heat and take power that would not fit within the budget of the integrated PSU unless Apple chose a lower wattage CPU. The soldered storage, while being controversial to say the least, ensures a consistent experience for all users, while allowing the end user to replace the storage would guarantee that someone choosing a bargain basement NVMe stick, with barely better that SATA speeds. While faster than an HDD, almost al of the currently sold solution does not come close to matching Apple’s solution. Not without going Samsung 970 Pro or an Intel 905p, which sadly would burn a hole right through the mini’s chassi, I think. But it is sweet!

Personally,I would go for the Core i7, all day and twice on Sunday.
[doublepost=1541871259][/doublepost]I have not had a chance to read all of this article, which was just published yesterday, but it might be of some interest or help on this particular thread - https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo

Enjoy (the nap, depending on your interest level)!
 
Last edited:
I installed Mac fan control which gives me some additional temperature readings. One of those readings is CPU proximity. So I'm doing an encode in handbrake, maxing the CPU and it reports 97c on the CPU but CPU proximity is showing 61c. Does someone know where this sensor might be located? It could be indication that the cooler is pulling heat away from the board very well.

Other temps within the mini:
GPU PECI is at 72c
airport card is at 46c
platform controller hub die 63c
CPU Proximity 61c
Memory proximity 53c

Edit - Encoding is done and it's several hours later, cooler in the room. I'm transferring tons of data from my PC to the Drobo now connected to the Mac but not doing much else. I thought I'd update the temps. CPU is clocked at 4.3ghz right now.

CPU temp is 71c
GPU PECI 66c
airport card 46c
platform controller hub die 62c
CPU proximity 53c
Memory Proximity 51c

To me it looks like the CPU ramping up 26c has little effect on the rest of the components of the computer. The room is also cooler now than it was early today so that can have an affect on case temps as well.

In terms of ram usage in case anyone is interested, I'm currently at 12.65gb used. I also have transmission running downloading torrents. Plex server running though nobody is connected atm. Sonarr is running but not doing anything. Safari open with 7 tabs. Activity monitor and Intel power gadget are open. Still setting things up and once the new Drobo arrives I'll get our photos setup and Lightroom going again which will push memory usage even higher.
 
Last edited:
Does someone know where this sensor might be located?

I tried squinting at the board pictures from the tear downs and I couldn't find it. I need to be able to see the markings to be sure and the pics aren't high enough resolution. It could even be somewhere in the heatsink or fan assembly. Maybe a bored and slightly curious EE could spend five minutes with their Mini and a magnifying glass and find it for us.

For reference though, the Airport "card" (hehe) is maybe two inches away from the CPU on the bottom of the board, if that helps you visualize the temperature gradient at all. Again, some slightly bored individual could take a FLIR picture of the logic board after running the Mini hard for half an hour if they are quick enough to get it out of the case after shutdown to get a reasonably useful picture for you to see how abrupt the gradient actually is.
 
I wonder if replacing the thermal paste with something good would change the results.
I am doing this on day one.
I have done it on about 35 Mac minis so far from the 2011 to mostly 2014 models and I get fairly decent results, meaning normally a high 30-to mid 40C idle.

I enjoy tinkering, and just did 6 more minis this weekend along with some older iMac 27" CPU swaps.

For the older, even base model Mac minis. Once you swap out that absolutely terrible HDD with even a cheap PNY SSD from BestBuy it makes the thing quite snappy.
You're not going to be doing anything substantial but for everyday normal word processing, general office type workflow it is a very good small, low power option. again with an SSD swap.

But back to the subject. I have 32GB of RAM on the way with the Mini which says it will be delivered between the 19th to 21st of Nov.
So once it is opened up, I'll do the RAM change and some new paste and see how she performs.


****
Now a delid thrown in might be fun.
 
Last edited:
  • Like
Reactions: Trusteft
Some cinebench scores of the i7 have turned up in other threads, showing multicore results in the 900 to 1100 range. This isn’t extremely low but again this seems to confirm underwhelming cpu performance. Those results are in the same range as those from the 2,6 ghz 2018 macbook pros which are 45 tdp cpus whith a lower base clock in laptop enclosures. Performance on desktops would be expected to be higher. You can find results around 1400 for the multicore cinebench test run on other intel 8700 desktops. We need more tests with multiple cinebench runs to see more clearly what happens, but it seems the cpu throttles under load (expected even below base frequency). Anyhow it doesn’t seem typical desktop performance is sustained.

Maybe we need a mac mini update from apple as they did for the 2018 mbps? Anyhow we need more indepth tests. Perhaps some DAW users with an i7 mac mini wanting to run logic pro benchmarks?
 
I am doing this on day one.
I have done it on about 35 Mac minis so far from the 2011 to mostly 2014 models and I get fairly decent results, meaning normally a high 30-to mid 40C idle.

I enjoy tinkering, and just did 6 more minis this weekend along with some older iMac 27" CPU swaps.

For the older, even base model Mac minis. Once you swap out that absolutely terrible HDD with even a cheap PNY SSD from BestBuy it makes the thing quite snappy.
You're not going to be doing anything substantial but for everyday normal word processing, general office type workflow it is a very good small, low power option. again with an SSD swap.

But back to the subject. I have 32GB of RAM on the way with the Mini which says it will be delivered between the 19th to 21st of Nov.
So once it is opened up, I'll do the RAM change and some new paste and see how she performs.


****
Now a delid thrown in might be fun.


If you remember, please let me know.
 
  • Like
Reactions: mmomega
But it's only throttling when it hits the limit intel set. This would be the same if you had it in a desktop you built. Intel set the limit. Intel's stock coolers are only designed for the 8700 to 65w TDP at 3.2ghz. Not for turbo boost frequencies. Anandtech has a nice write up about this.
This thing is going all the way to the thermal limit without any additional throttling from apple.
So let's say you could put in a better cooler and drop those temps 10c to say, 87c. It's just going to ramp further and temps go right back in to the 90's. That's what it's designed to do. And you can keep putting on better coolers and it's going to keep ramping up until it hits 100c and throttles or hits the turbo boost upper limit. And then you'd need an even better cooler to back it off from that.
My 5820 6 core on my desktop is in the 80's while encoding and I have to use water cooling just to achieve that! Short of putting a water cooler in the mini I don't know what else apple could have done.
Apples design is very good for getting heat from the CPU out of the case. My mini's case stays cool under load. Only the air coming out is warm. So it's not holding a lot of hot air in the case. I have opened it up to upgrade the ram. I'm not an engineer(I only work for them) but it looks like a good design overall. The heatsink and fan take the heat directly from the CPU straight out of the back of the case.
The only concern is the heat on the board around the CPU itself. Is there much heat there? Did apple design the board so more heat sensitive components are further away from that area? It would be easy enough to test some of this.
The mini is a good design , the 4 core should be fine , the 6 core not so sure. We will see . At least Apple did not try to make the mini thinner
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.