Hacker News new | past | comments | ask | show | jobs | submit login
Intel Power Gadget (intel.com)
62 points by jhack on July 21, 2018 | hide | past | favorite | 53 comments

I’m going to predict that they release a new version for macOS that includes a bug fix for the latest generation of processors. It will fix inaccurate readings. The new version will “correctly” show the flat line instead of the spikes.

Then again it could also just be related to reports of it causing some macs not to be able to reboot after install. The only fix was booting safe mode and manually removing the kernel extension.

I believe that was fixed in the most recent Mac version already - see the devs comments on the page. I have it installed on my MBP and have had no issues.

bizarrely though they have pulled it for some reason.

> Then again it could also just be related to reports of it causing some macs not to be able to reboot after install. The only fix was booting safe mode and manually removing the kernel extension.

Yikes, If that's true I'd bet that's the case. A bad kext that prevents booting would be a horrible experience.

Why does this need to be a kernel extension? Is there really no access to MSRs in macos as shipped by Apple?

Generally no, the instructions that get access to many MSRs are privileged since they can affect the way things execute.

I understand they are privileged but the way Linux handles this feature seems acceptable.

Why? MSR are not userspace registers. You don't want the userspace messing around with them either because they won't know 100% what they do. But more importantly for security.

Even Linux requires this be done in kernel space, it's a requirement set by the architecture that the instructions that access them are run as part of the kernel. This is just like OSX, no different at all.

In Linux root can open, read, and write /dev/msr.

I haven't seen a /dev/msr before, but there is /dev/cpu/*/msr [1] provided by a kernel module. So there is a generic interface for them, but it's still all happening in kernel space. It's also not loaded by default either because it can cause system panics and corruption.

[1] http://man7.org/linux/man-pages/man4/msr.4.html

There are reports in the comments of the kernel extension causing boot failures that require going into recovery mode and deleting the file. I would guess that when the freezer video came out and this gadget became more popular than ever before, somebody disabled the link while the one developer figures out what’s going on.

Luckily it's still available via homebrew:

  brew cask install intel-power-gadget

Or actually from Intel too, they just removed the link


The regular link seems to be back now.

Need one more hyphen: intel-power-gadget

That's what I get for forcing homebrew onto Ubuntu to check, thanks!

LTT stream relevant part https://youtu.be/jtHhvcdjNWw?t=19m52s

Actually you can still download the app

https://software.intel.com/file/641033/download (direct link)

Is this related to the i9 Macbook Pro overheating?

If it is, maybe related to LTT's live stream yesterday that used it? https://youtu.be/jtHhvcdjNWw

Counter argument is the suite of tests he used were optimised for the i7 and not the i9 [1].

[1] https://www.computerworld.com/article/3291760/apple-mac/abou...

While it's true optimizing the software could increase render times, the fact that the CPU does run regularly below the base clock with heavy workloads (ie: throttles) remains true.

I know everyone likes to bash Apple at every opportunity but stating the fact that the i9 model throttles is _not_ inflammatory.

The better question is, how much does it matter? The machine still has 6-cores. MKBHD was still noticing a 15% performance increase (vs. last gen) without new, updated builds of Final cut etc.

This doesn't get posted enough. But the issue seems to be that Apple favours a quieter and cooler machine over performance. As such they run the fans a lot slower and later than they could.

So if you manually adjust the fan speeds (plenty of apps that do this) you can get significantly (10-15%) better performance.

Exactly, and that's why the freezer test mentioned in the article the grandparent linked to was done. To try to get an idea of how much the throttling was affecting.

It doesn't matter if it is faster until it starts throttling. i9 is intended for pros who need that CPU power for longer periods of time. And if the issue appears after a minute of usage and performance becomes worse than i7 (as shown in initial Dave's video), then it's a bad product (Macbook, because of not being able to cool it, not the i9 itself).

And benchmarks should be changed to run for longer time periods or else it's not showing actual performance indication.

The optimization argument is that even after throttling, if the software was optimized for the 6core machines it would no longer be slower than the i7 after one minute.

But that's yet to be seen. As of this moment in time you're correct.. buying the i9 model makes zero sense.

It doesn't matter if it is faster until it starts throttling. i9 is intended for pros who need that CPU power for longer periods of time

What's the evidence for this? I'd love a faster multi-core processor, but only to accelerate tasks which complete in a relatively small amount of time already.

Intel's own marketing talks about it being used by these kinds of people https://newsroom.intel.com/news/intel-core-i9-processor-come...

That is not a counter argument! If software is optimized for multiple cores then it is optimized for multiple cores. It is not like they tested it on a single core CPU with higher frequency first. It could be that i9 might even have new instruction set with faster instructions but it should still have the same ones that i7 have as well.

The problem with that is that, while yes it means that performance can be improved (faster rendering times, more computations per second etc.) it doesn't change the thermal issues at all with the particular implementation from apple.

How does changing the benchmarking software in any way not invalidate the comparison, when trying to compare two compatible hardware implementations? To see if the hardware has gotten faster, you must run the same software on it.

IMHO this "grasping at straws" is a sign of the end of actual performance improvements. From the start of the PC through perhaps the mid 2000s you could run the same software, and each new generation of processor would show a measurable (and often dramatic) improvement.

What counter argument is this? The i9 is a 6-core i7. That's basically all.

The Mac download link has been removed from the page.

They do pull out battery life estimations when it becomes an issue.

ltt did a stream yesterday where they used power gadget to show it throttling... https://www.youtube.com/watch?v=jtHhvcdjNWw

Or, on Linux:

    $ perf stat -e 'power/energy-cores/' sleep 1
More examples here: https://developer.mozilla.org/en-US/docs/Mozilla/Performance...

There is a link to version 3.5.3 on the web page now


We've updated the title from “Intel Pulls MacOS Version of Power Gadget”. It's fine to mention what might be of interest to other readers, but in the comments please.

I’d say the previous title was good since it’s probably what the submitter wanted to show and not the existence of the tool itself. I didn’t even click the link because I knew about it, but if I saw the previous title it’d have piqued my interest.

Edit: typo

If the intent was to focus on the news, linking to the download page which doesn't even mention the need except that one can kind of Uber it from the dead link and updated date, is a mistake. Looking to a news article about them pulling it threat provides context would be a better idea, like: https://9to5mac.com/2018/07/21/intel-power-gadget-mac-remove...

Maybe a typo, the word you want is piqued:

pique - verb (piques, piquing, piqued) 1 [ with obj. ] arouse (interest or curiosity).

Fixed, thank you!

I hear you, but that's just not how this particular site works. Here it's “original title, unless it is misleading or linkbait”, which is a fundamental quality. A link aggregator with editorialized titles would simply be a different thing.


But in this case the original title "Intel Power Gadget" is more misleading than the "Intel Pulls MacOS Version of Power Gadget".

It's Intel's page about Power Gadget. Its title 'Intel Power Gadget' is not at all misleading. When you click on it, you don't end up on a page about beekeeping.

It’s misleading since the newsworthy context has been removed for the sake of pedantic rule adherence.

'Misleading' is not a synonym for 'I don't like the rules'. Using it like that is misleading.

I don’t understand what relevance your definition has to this discussion. If I wanted to comment that I don’t like the rule, I would have said so. Please don’t lazily impose your lack of understanding on other people’s comments and listen in good faith to accept they genuinely have a point and try to understand it.

You did comment that you don't like the rules.

I said I don't like pedantic rule following… the rules are what they are, I'd leave if I found them insufferable. Am I aloowed to point out where they fail or is criticism indistiguishable from dislike?

I wonder why.

You guys can still find it on reddit. If you guys still want it to do tests.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact