TWEAK: Enable HPET (in BIOS and OS) for better performance and FPS

By hardbag
in Tips, Tweaks & Customization

 Share

Recommended Posts

Grand Master

mattmatik

Posted
Posted

Hi guys, I could not resist joining this forum to add my two cents. HPET made me go nutz, too.

 

 

HPET on/off making a significant difference is definitely NOT snake oil.

 

 

On my previous Win 7 build (DAW only), disabling HPET in BIOS & OS resulted in a DPC latency of comfortable 2 - 4 micro seconds. Before that, the latency rarely ever dropped below 130 micro seconds, which already has been quite good regarding the 270 which I?ve had before I got rid of my nvidia card and its crappy drivers.

 

Not only there were no more drop outs (at reasonable buffer sizes), the whole system was much more responsive. The negative result was that my ASIO load meter got more "nervous/sensitive". Means that the meter in Cubase jumped up and down more than before when it was slightly higher but also more stable.

 

I?ve dealt quite a while with the HPET thing and have come to the conclusion, that HPET is neither the "best timer" nor the worst. Computers are not interested in having the timer "ring" every 1000 or 976 or whatever when. They have no personal preference and don?t care about the output. Just about the processing itself. So the important thing is to have as many timers as possible as this means more potential moments to generate interrupts.

 

["...interrupts can be missed if the target time has already passed"]  the wiki link

 

Enabling HPET & using "bcdedit..... true" to force Windows to use only one timer seems absurd. More

 

Just think of it as drinking in a bar where two out of three waitresses have been fired and the one who?s left may be extremely punctual at appearing only once per hour. One of the former waitresses (which meanwhile are unemployed and hungry) could have come your way already 45 minutes after your last beer order. That?s when you were supposed to generate that one interrupt to quench your thirst.

 

I am aware that there has to be some kind of down side to it. Otherwise, nVidia would not set their drivers to strictly induce a latency of 255 micro seconds instead of matching it with the PCI Latency Setting found in BIOS. Maybe this allows a more stable framerate (not a faster one!), like slaves on a galley working their paddles to the decent but steady beat of the drummer. Less strokes, but more load per stroke.

 

While its benefits/disadvantages depend on each single person?s needs, its influence on the computer at the bare system level has to be clearly explainable. There is no place for "every setup is different, so this function does whatever its wants to do", computers don?t work this way.

 

For me, disabling it much more serves my needs as constantly keeps my snake oiled.

 

 

Cheers from Germany

 

'Nuff said.

Grand Master

Praetor

Posted
Posted

thread is 2 years old, still not a single proof or evidence that this works.

 

on the contrary, all the claims that this works makes me wonder what's the purpose of deceiving people into this?

Mentor

snaphat (Myles Landwehr) Member

Posted
  • Member
Posted

thread is 2 years old, still not a single proof or evidence that this works.

 

on the contrary, all the claims that this works makes me wonder what's the purpose of deceiving people into this?

Personally, I think it is just the placebo effect instead of subterfuge in many cases. The wintimertest results are enough to convince folks that it must improve speed objectively everywhere. We aren't immune to such things by any means either. We might defrag an HDD and perceive the same thing to be honest. Tricks of the mind. It's similar to people who listen to uncompressed audio (versus XYZ compressed audio) and claim they can hear a difference, yet double blind studies show otherwise. Expectations are everything, huh?

Grand Master

Praetor

Posted
Posted

It's similar to people who listen to uncompressed audio (versus XYZ compressed audio) and claim they can hear a difference, yet double blind studies show otherwise. Expectations are everything, huh?

 

well, i can listen to 128kbps audio file and find differences from the same uncompressed format; 320kbps, not so well. Buy i agree with you, even a defrag can make a subtle difference so this is more a placebo effect then anything.

  • snaphat (Myles Landwehr)
  • Like 1
Mentor

snaphat (Myles Landwehr) Member

Posted
  • Member
Posted

well, i can listen to 128kbps audio file and find differences from the same uncompressed format; 320kbps, not so well. Buy i agree with you, even a defrag can make a subtle difference so this is more a placebo effect then anything.

Well 128kbps is the questionable line afaik (or 192kbps -- depending on the compression algorithm). It's the higher bit rates where you can't perceive in double blind tests, yet, people claim otherwise.

Grand Master

Praetor

Posted
Posted

Well 128kbps is the questionable line afaik (or 192kbps -- depending on the compression algorithm). It's the higher bit rates where you can't perceive in double blind tests, yet, people claim otherwise.

 

well, considering that when i was learning Java i made a small audio program that i could insert two filters:a low-pass and a high-pass; learned pretty quickly i was going to become deaf more faster then i expected lol

Mentor

Joe User

Posted
Posted

How do I disable it?

Can't find it in my bios....

I have a hp laptop.

Help!

 

You don't, If your computer came with Windows, it's already optimized for the hardware.

Mentor

snaphat (Myles Landwehr) Member

Posted
  • Member
Posted

But it should be in bios!

Why is it not?

Your laptop is probably too new (pretty sure mine doesn't have it either). The only good reason to have the option in the bios was during the early years when the actual hardware implementations weren't well tested or supported.

Neowinian

Overlord1

Posted
Posted

Your laptop is probably too new (pretty sure mine doesn't have it either). The only good reason to have the option in the bios was during the early years when the actual hardware implementations weren't well tested or supported.

Do you know any other method that reduces lag and improves gaming?

Mentor

snaphat (Myles Landwehr) Member

Posted
  • Member
Posted

Do you know any other method that reduces lag and improves gaming?

Well you can still force only HPET in Windows (as was outlined here in the first post) using step 2 if your laptop supports it (you just can't turn it off/on in the bios). Note, I am not saying this will improve your performance (I don't believe it will).

Neowinian

Overlord1

Posted
Posted

Well you can still force only HPET in Windows (as was outlined here in the first post) using step 2 if your laptop supports it (you just can't turn it off/on in the bios). Note, I am not saying this will improve your performance (I don't believe it will).

Can't find that 2nd step... :(

Grand Master

Krome

Posted
Posted

For many years already my HD never scream.  After turn this on:

bcdedit /set useplatformclock true

 

I heard my hard drive screaming.  I mean it has never make any noise before.  So yeah I think I will need to get a new hard drive to back up my data soon and then get back to it's original setting... not turning on the HPET that is.

  • 3 weeks later...
Enthusiast

Eric-G.

Posted
Posted

HardBag (or anyone else here),

 

How do I check the BCD on what's the current setting of useplatformclock, before making any changes to the BCD?

 

BTW, I have checked my system to see if Invariant TSC is supported or not. Says it's supported.

So does that mean I don't need to edit my BCD?

 

For those who don't know how to check, here's how:

 

1. Open AIDA64 -> Motherboard -> CPUID
2. Look for the field "Invariant Time Stamp Counter", and check whether it's supported or not.
Mentor

snaphat (Myles Landwehr) Member

Posted
  • Member
Posted

 

HardBag (or anyone else here),

 

How do I check the BCD on what's the current setting of useplatformclock, before making any changes to the BCD?

 

BTW, I have checked my system to see if Invariant TSC is supported or not. Says it's supported.

So does that mean I don't need to edit my BCD?

 

For those who don't know how to check, here's how:

 

1. Open AIDA64 -> Motherboard -> CPUID
2. Look for the field "Invariant Time Stamp Counter", and check whether it's supported or not.

 

For the first question: type 'bcdedit' on the command line. 'useplatformclock' will appear under the Windows Boot Loader section if it set.

 

For the second question: the TSC is completely separate from the HPET timer. See: http://en.wikipedia.org/wiki/Time_Stamp_Counter. The invariant version of the TSC is an updated version of the TSC on newer processors:

 
17.13.1 Invariant TSC

The time stamp counter in newer processors may support an enhancement, referred to as invariant TSC. 

Processor?s support for invariant TSC is indicated by CPUID.80000007H:EDX[8]. 

 

The invariant TSC will run at a constant rate in all ACPI P-, C-. and T-states. This is the architectural behavior 
moving forward. On processors with invariant TSC support, the OS may use the TSC for wall clock timer services 
(instead of ACPI or HPET timers). TSC reads are much more efficient and do not incur the overhead associated with 

a ring transition or access to a platform resource.

 

--

http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-software-developer-vol-3b-part-2-manual.html

Rookie

acemod13

Posted
Posted

Look what I found:  useplatformclock [ yes | no ]

Forces the use of the platform clock as the system's performance counter.

Note  This option should only be used for debugging.

http://msdn.microsoft.com/en-us/library/windows/hardware/ff542202(v=vs.85).aspx

  • 1 month later...
This topic is now closed to further replies.
 Share
Go to topic listing

  • Recently Browsing   0 members

    • No registered users viewing this page.