Mini 2012 Base vs Mid - Thermal / noise performance

[Maxx Power]

Originally Posted by Snowcake

The cpu not using Turboboost is not throttling.
Throttling is: Working at a lower speed than the standard speed. TurboBoost is above standard speed.

Not to be pedantic to argue what the "standard speed" is, the idea is fine as described.

"Throttling" as used by me in the last post describes a down-binning of CPU clock due to (as was the context) thermal restrictions. If you want to get down to the nitty bitty details, then consider what Turbo really is: extra few unlocked multipliers. The algorithm to use the Turbo is two part, the CPU and the motherboard (logicboard) decides the current multiplier. On most laptops, this means the Turbo is typically viewed as a few "extra" bins of speed. In that case, backing off the CPU speed from top Turbo bins is a little different than on most current desktops, where the CPU+Board decides that the Turbo should/can be the max operating frequency (sometimes not of only 1 core, but ALL cores). In the former case, thermal constraints dictates the availability of Turbo bins, but it should be mentioned that the idea of Turbo is to make use of all TDP available, since one core loaded at default frequency dissipates less than all cores loaded, so it was never completely "extra" anyway, since you pay for a specific TDP laptop that should be able to dissipate all that heat anyway by design. Any laptop that holds a 35W TDP CPU can't sustain Turbo modes have poorly designed or under-specced cooling subsystems. And in the latter case, the term "throttle" maybe more appropriate. However, since it is the same phenomenon that is occuring regardless of CPU+board interpretations, the choice of wording is a bit pedantic.

In fact, the term throttling is just another generic term for dynamic frequency scaling was coined long ago:

http://en.wikipedia.org/wiki/Dynamic_frequency_scaling

Whether the factor involved in the throttle is thermal, battery, or whatever is irrelevant. Turbo Boost is just an euphemism for a more clever dynamic frequency scaling. As far as ACPI and Intel is concerned, the Turbo bins are just P-states. The top bin is still P0, which was historically the "standard speed". Just that the P0 states is now a function of core load and temperature.

[Snowcake]

Originally Posted by Maxx Power

Not to be pedantic to argue what the "standard speed" is, the idea is fine as described.

"Throttling" as used by me in the last post describes a down-binning of CPU clock due to (as was the context) thermal restrictions. If you want to get down to the nitty bitty details, then consider what Turbo really is: extra few unlocked multipliers. The algorithm to use the Turbo is two part, the CPU and the motherboard (logicboard) decides the current multiplier. On most laptops, this means the Turbo is typically viewed as a few "extra" bins of speed. In that case, backing off the CPU speed from top Turbo bins is a little different than on most current desktops, where the CPU+Board decides that the Turbo should/can be the max operating frequency (sometimes not of only 1 core, but ALL cores). In the former case, thermal constraints dictates the availability of Turbo bins, but it should be mentioned that the idea of Turbo is to make use of all TDP available, since one core loaded at default frequency dissipates less than all cores loaded, so it was never completely "extra" anyway, since you pay for a specific TDP laptop that should be able to dissipate all that heat anyway by design. Any laptop that holds a 35W TDP CPU can't sustain Turbo modes have poorly designed or under-specced cooling subsystems. And in the latter case, the term "throttle" maybe more appropriate. However, since it is the same phenomenon that is occuring regardless of CPU+board interpretations, the choice of wording is a bit pedantic.

In fact, the term throttling is just another generic term for dynamic frequency scaling was coined long ago:

http://en.wikipedia.org/wiki/Dynamic_frequency_scaling

Whether the factor involved in the throttle is thermal, battery, or whatever is irrelevant. Turbo Boost is just an euphemism for a more clever dynamic frequency scaling. As far as ACPI and Intel is concerned, the Turbo bins are just P-states. The top bin is still P0, which was historically the "standard speed". Just that the P0 states is now a function of core load and temperature.

Indeed, you're right.

But as far as i know, the Mac Mini never throttles because the temperature never goes above the tjuction of 105C
Intel quadcore spechttp://ark.intel.com/products/64900/...up-to-3_30-GHz

Intel says this: http://communities.intel.com/thread/29957

Why do you think the Mac Mini throttles, how do you know?

[motrek]

Originally Posted by Snowcake

Indeed, you're right.

But as far as i know, the Mac Mini never throttles because the temperature never goes above the tjuction of 105C
Intel quadcore spechttp://ark.intel.com/products/64900/...up-to-3_30-GHz

Intel says this: http://communities.intel.com/thread/29957

Why do you think the Mac Mini throttle?

These recent Intel chips (i-Whatevers) only turbo boost when temperature allows. So they won't boost if it puts them over their max temperature. I think it's probably safe to say if the chip is running at almost exactly its max temperature then it could likely boost faster but isn't, to keep the temperature down. It's a semantic issue if you want to call that throttling or reducing the turbo but it's the same thing.

Although now that I think about it, maybe Apple designed the Mini cooling system so that the chips could boost the maximum amount and the chip would be at its maximum temperature. It makes sense.

[Maxx Power]

Originally Posted by Snowcake

Why do you think the Mac Mini throttles, how do you know?

I just remembered from when the Sandy Bridge Mac Minis came out, there was this work around in Terminal to allow Turbo Boost at all times instead of thermal-regulating it. That's about it. Anandtech tested some older MBP (non-Retina models) and determined that they do lightly throttle and posted temperatures (I say lightly, because you should see the Dell XPS and Samsungs...). I also have a MBP 13" non-retinal Sandy Bridge, and I know it can't stay on Turbo for more than a few minutes at a time, and it runs VERY HOT (high 80's) on single core load for code crunching. Therefore I concluded that if the Mac Minis have similar temperatures and CPUs, it MUST be backing off the highest bins. Apparently according to Anandtech reviews, throttling starts when the CPU transitions from about 70 to 80 degrees Celsius, and nearly all Mac Minis with Sandy or Ivy runs over that on a single core full load... I know you can program the EFI/BIOS to force speeds to whatever, so I have no idea at what EXACT temperature does the Mac Minis start throttling, just that from what I reasoned, they (by all likelihoods), should...

Oh, and I just remembered that if the Intel CPUs goes above the Tjunction temperature, the states it will enter are T-states (deep deep throttle states, past P-states). In that case, the core is left to idle for cycles between workloads.

[Snowcake]

Originally Posted by Maxx Power

I just remembered from when the Sandy Bridge Mac Minis came out, there was this work around in Terminal to allow Turbo Boost at all times instead of thermal-regulating it. That's about it. Anandtech tested some older MBP (non-Retina models) and determined that they do lightly throttle and posted temperatures (I say lightly, because you should see the Dell XPS and Samsungs...). I also have a MBP 13" non-retinal Sandy Bridge, and I know it can't stay on Turbo for more than a few minutes at a time, and it runs VERY HOT (high 80's) on single core load for code crunching. Therefore I concluded that if the Mac Minis have similar temperatures and CPUs, it MUST be backing off the highest bins. Apparently according to Anandtech reviews, throttling starts when the CPU transitions from about 70 to 80 degrees Celsius, and nearly all Mac Minis with Sandy or Ivy runs over that on a single core full load... I know you can program the EFI/BIOS to force speeds to whatever, so I have no idea at what EXACT temperature does the Mac Minis start throttling, just that from what I reasoned, they (by all likelihoods), should...

Oh, and I just remembered that if the Intel CPUs goes above the Tjunction temperature, the states it will enter are T-states (deep deep throttle states, past P-states). In that case, the core is left to idle for cycles between workloads.

So The 2012 Mac Mini at 104C is throttling?
Does that just mean it has no turbo boost anymore at that temp or wil the cpu frequency go down from the standard 2.3ghz?

If that is the case, i buy a slow computer, because i can't use full performance under heavy loads because the cpu is downclocking...

[motrek]

Originally Posted by Maxx Power

I just remembered from when the Sandy Bridge Mac Minis came out, there was this work around in Terminal to allow Turbo Boost at all times instead of thermal-regulating it. ...

Such a shame that these computers run so hot. If you look at a teardown, the heatsinks and fans are SO tiny. Basically they are laptop parts. If Apple just increased their size by half an inch or so, I bet it would double the amount of cooling. I bet the Mini could run under full load and hardly need to spin up the fan. But instead this is what we get. Oh well.

----------

Originally Posted by Snowcake

So The 2012 Mac Mini at 104C is throttling?
Does that just mean it has no turbo boost anymore at that temp or wil the cpu frequency go down from the standard 2.3ghz?

If that is the case, i buy a slow computer, because i can't use full performance under heavy loads because the cpu is downclocking...

At 104C it's certainly not doing any turbo boosting, no. But that is probably not a big deal since to get it to that temperature, you are probably beating on all four cores, and the turbo boost when four cores are active is usually very small, maybe around 0.1GHz. So you'd hardly notice the drop in performance.

At that temperature it's *possible* that the chip is doing some failsafe measures to keep from overheating. Basically cutting the clock speed in half, or more. The way it does this makes it hard to measure since it probably still reports that it's running at its nominal 2.3/2.6GHz. A good way to tell would be to run Prime95 in the background and run a CPU benchmark before and after the temperature gets close to its maximum.

But I personally find it hard to believe that Apple would design a cooling system such that the CPU has to take fairly extreme measures to prevent itself from overheating.

[Maxx Power]

Originally Posted by motrek

Such a shame that these computers run so hot. If you look at a teardown, the heatsinks and fans are SO tiny. Basically they are laptop parts. If Apple just increased their size by half an inch or so, I bet it would double the amount of cooling. I bet the Mini could run under full load and hardly need to spin up the fan. But instead this is what we get. Oh well.

----------



At 104C it's certainly not doing any turbo boosting, no. But that is probably not a big deal since to get it to that temperature, you are probably beating on all four cores, and the turbo boost when four cores are active is usually very small, maybe around 0.1GHz. So you'd hardly notice the drop in performance.

At that temperature it's *possible* that the chip is doing some failsafe measures to keep from overheating. Basically cutting the clock speed in half, or more. The way it does this makes it hard to measure since it probably still reports that it's running at its nominal 2.3/2.6GHz. A good way to tell would be to run Prime95 in the background and run a CPU benchmark before and after the temperature gets close to its maximum.

But I personally find it hard to believe that Apple would design a cooling system such that the CPU has to take fairly extreme measures to prevent itself from overheating.

I love the ifixit tear-downs. I noticed that Apple changed the fan on the new Minis but not the heatsink assembly. Personally, I think Apple is going in the right direction in a slow fashion. Their new retina MBP's are MUCH cooler and sustain top speeds of the CPU used.

Maybe the next revision of the Mini we'll see some similar accolades.

----------

Originally Posted by Snowcake

So The 2012 Mac Mini at 104C is throttling?
Does that just mean it has no turbo boost anymore at that temp or wil the cpu frequency go down from the standard 2.3ghz?

If that is the case, i buy a slow computer, because i can't use full performance under heavy loads because the cpu is downclocking...

There was quite some debate about whether or not the sensor data from the CPU is the same as Intel's Tjunction. I am not sure myself. It would seem that the Tjunction is higher than the CPU temperature as measured by the sensor. In either case, what I have observed in the past is that the CPU in Macs doesn't throttle below "base frequency" (Intel doesn't use the terminology standard, just base) UNLESS you live in a hot environment, at high altitudes (where air is thinner) or hack it some how.

[Snowcake]

Originally Posted by motrek

Such a shame that these computers run so hot. If you look at a teardown, the heatsinks and fans are SO tiny. Basically they are laptop parts. If Apple just increased their size by half an inch or so, I bet it would double the amount of cooling. I bet the Mini could run under full load and hardly need to spin up the fan. But instead this is what we get. Oh well.

----------



At 104C it's certainly not doing any turbo boosting, no. But that is probably not a big deal since to get it to that temperature, you are probably beating on all four cores, and the turbo boost when four cores are active is usually very small, maybe around 0.1GHz. So you'd hardly notice the drop in performance.

At that temperature it's *possible* that the chip is doing some failsafe measures to keep from overheating. Basically cutting the clock speed in half, or more. The way it does this makes it hard to measure since it probably still reports that it's running at its nominal 2.3/2.6GHz. A good way to tell would be to run Prime95 in the background and run a CPU benchmark before and after the temperature gets close to its maximum.

But I personally find it hard to believe that Apple would design a cooling system such that the CPU has to take fairly extreme measures to prevent itself from overheating.

Does anyone know for sure that the Mac Mini CPU is not going to lower its frequency (Like 1.6ghz) below its standard 2.3ghz at temperatures at 80C or 104c?

[Maxx Power]

Originally Posted by Snowcake

Does anyone know for sure that the Mac Mini CPU is not going to lower its frequency (Like 1.6ghz) below its standard 2.3ghz at temperatures at 80C or 104c?

Sorry, I wish I had something more concrete to give you on this. My best guess based on available information is that at room temperatures up to say, short of a really hot summer day and you are running Linpack, with the Mini well ventilated on a flat surface, you should not expect it to go below the base frequency.

[dasx]

Originally Posted by Snowcake

Does anyone know for sure that the Mac Mini CPU is not going to lower its frequency (Like 1.6ghz) below its standard 2.3ghz at temperatures at 80C or 104c?

Well, I was gonna try it yesterday with the MSR Tools you said, but…
Downloaded it and after launching the app it asked for my admin password. Before putting it in, I tried to get to the developer's site with no luck, plus I wasn't able to find another link for its download aside from that in Yahoo Answers.

So I have difficulties trusting that app. This is my primary machine which I use for everything, I just don't wanna mess it up. :/

Any other software I could use?

[motrek]

Originally Posted by Snowcake

Does anyone know for sure that the Mac Mini CPU is not going to lower its frequency (Like 1.6ghz) below its standard 2.3ghz at temperatures at 80C or 104c?

My information may be way out of date here but I believe the Pentium 4 would reduce its temperature if it was on the verge of overheating by multiplying the duty cycle of its clock input signal, meaning that a 2.3ghz chip would go down to 1.15ghz, then 575mhz, etc. Basically keep halving the frequency until it got to a point where it wasn't about to overheat.

I imagine Intel is probably still doing the same thing, because otherwise advertising these base frequencies (like 2.3ghz) is relatively meaningless, i.e., why not just make all the chips 500mhz with a turbo boost that gets it up to 3.whatever ghz?

But to answer your question, given the various benchmark results for the new Mini running Handbrake, etc., it's really fast, and it would be MUCH slower if it divided its clock speed in half, so it's not a guarantee that it never steps down the clock speed but I doubt it does.