Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Bukkake, Sep 18, 2012.
Hmm. I just remembered there is an APIC PM Timer in most x86 CPUs which usually runs at 3-4 MHz (rather than the TSC which runs 3-4 GHz), just like the default Windows 8-10 QPF used to (before the recent Win10 changes). It's actually older than HPET. So, may be QueryPerformanceFrequecny was the APIC PM Time frequency. And now I have a dejavu. I think I already arrived at this conclusion some years ago and just forgot it. LOL.
Bruh. Learn this: You don't need to "force" it on or off. Windows as stock does not use it and uses it when needed, simple as that. Or it even may not use it anymore, stop "tweaking" this for the love of god, there is other stuff that can help a bit, but just leave this alone. Some programs may call it, but that's it, as it comes out of the box, Windows uses TSC or whatever and then uses HPET if needed, if ever. And that's it.
Thank you, but I know this and also know how Windows handles this (QPC). Still no damage in reporting the results of my own tests. And again, not everyone is only playing games with their computers.
Care to elaborate what happens when TSC is used with some critical audio software?
Quit forcing times in your O.S, that is all I can recommend. It won't "boost" anything.
Nothing bad on modern hardware, likely on the contrary, but I cannot test that on this platform (will rebuild my Ryzen rig soon). Audio software usually uses QPC or offers both as alternative (Cubase offered that for quite a long time), I will have to query what audio/midi drivers use nowadays.
The question is rather what happens when software uses HPET more or less involuntarily. This is why I tested what happens on the 9900k rig when HPET is used/forced. In essence the system becomes unusable. That's progress nowadays...
Why do you care when this is not a real occurrance?
I've btw. been disabling HPET in the bios of my ASRock mainboard for ages. Not sure though if every ASRock board offers it.
I care because among other things I test hardware, software and drivers, and generally I like to know how things in practice vs. theory. People are forcing the use of HPET, Windows uses "heuristics" to determine which timer to use and in practice things just do not always work as written in the book. HPET testing and comparison with other timers is one thing I do with every computer and major OS revision.
In audio/midi time-stamps are used for incoming and outgoing audio/midi streams. Tracks are also shifted in time versus each other to align them to various hardware and software (buffer) latencies. As mentioned in another thread already, at 44.1 Khz a single audio sample is only 0,0227 ms long. So every bit of extra timer resolution helps, if it works in practice that is.
New boards have it enabled by default, O.S sets it to auto and when it's needed, its used, if not, it isn't.
Force HPET On. Runs benchmarks, games, play audio.
Force HPET Off. Same as above.
Set HPET to Auto. Same as above.
No one can know what works best for your configuration except you, all computers work differently. Some report HPET helped them a lot, while others report it destroyed their performance. I'd just stick on keeping it to auto, if you really think it will literally give you such a performance boost like if you had faster hardware, that's just unrealistic.
This is not logical. If TSC is good enough for a particular task on one modern system, it is for all.
And where is HPET ever picked by anthing? This is often claimed, but where is the proof?
Apparently some programs and games need it and they "call" for it, but the experiment would need to be done with HPET off and then see system behavior.
Never seen any program not working since I turned it off in bios, which was years ago.
HPET was invented as a countermeasure against faulty TSC implementation, and has been used until Invariant TSC came. It can be that some of those old CPUs are still in use, and Windows can use HPET on those rigs.
PS HWiNFO has a setting to use HPET. I don`t know why to not use TSC but since I believe in programmer`s intelligence there can be good reason to utilise HPET. For example, IIRC, HPET can be programmed to fire interrupt over a specified interval - a feature which TSC lacks.
PPS I don`t believe a "simple" game can use HPET, because I believe only kernel mode code can interact with HPET. HWiNFO loads a driver which can be a kernel one.
By the way, I am not using an enforced HPET. I tested it and shared the results.
The thing is, why would you like to force it or do tests with it on? In my case, Windows does not use it, I don't even know if any program would use it, but it causes no harm. I will force if Off and when I can I will reply back to see if it made any changes in terms of how smooth things feel. By the way, I recommend you only you disable it on Windows, because if you also disable it on BIOS, apparently it can de-sync some things. Newer motherboards don't allow you to disable it anyway, if I recall correctly.
I wonder if even Haswell needs it. My wild guess is that it doesn't since tis relatively new.
And? Does it say why HPET and not TSC? Any details?