Strange observation about "steinberg power scheme"

Hi there people, long time no see :slight_smile:

I recently built a new music computer and I am mighty pleased.I basically upgraded everything including audio card and os (went from winxp 32 to win7 64)During this process I’ve encountered some strange things that I wanted to share in case some of you are stuck with the same “problems” or might be in the near future when you upgrade your machine/os.BTW the only os tweaks I’ve applied is “optimize for background services” and turning off hpet and eist (powersaving things) in the bios.

So here goes:
Weirdness 1: Activating Steinbergs suggested power scheme in Cubase 6 (both 32 and 64 bit versions) causes audio spikes and a lot of them. Even on small cpu easy projects at 128 samples of latency.
Turning it off and choosing the standard “high power” power scheme in win7 64 takes care of the audio spikes even on huge projects with latencies as low as 32 samples.This has been giving me headaches since I had the machine up and running because I trusted that this “steinberg power scheme” was something I would want to have turned on and that it in someway must be better tweaked for music than the standard win7 power plans/schemes.Well it is certainly not so on my machine, so be sure to check this tip out if you have trouble with audio/cpu spikes.
Weirdness 2: Turning HT on or off has no effect on performance/spikes whatsoever even on low latencies. I understood it like “HT is not supported so don’t use it”, but it seems to me like it doesn’t matter much, and believe me I know how to stress my Cubase system, hehe.With HT on I see 8 cores with load distributed evenly between the cores.Same with HT off, exept then I/win7 only see 4 cores.Other than this I am a really happy camper now.
64 bit os/Cubase and 16 gig of ram REALLY outshines 32 bit os/Cubase and 4 gig of ram where 2 gig has already been stripped away by the os and the graphic card LOL

Kim :ugeek:

Welcome back and thanks for sharing :slight_smile:
Concerning 1: I’m not too sure, but I think someone here recently said that Steinberg power scheme takes the current active powerscheme and only adjusts the CPU power sleep function to off. No idea how that would cause so much trouble. :confused:

I have an idea but I will admit it is really just a guess. The power scheme may prevent the processor from “stepping down” when it is not needed as much, which leads to more heat, which leads to throttling when the processor IS needed. With the power scheme off, the processor steps down when it can and stays cooler so that it can absorb some extra heat when it is needed for a burst without overheating. As I said, it is a guess.

Coooooooooool puter! :sunglasses:

DPC latency checker reports reduced latency when using the Steinberg power scheme. I think. :confused:

Hey Kim! I was thinking about just today! I was wondering if you ever put together a CD or download

HT was an issue with the older OS versions because the scheduler was using a typical “round robin” scheme to put threads on various cores. There was actually an API in Windows that allowed an application developer to specify that a particular thread was supposed to run on a specific core and that core only.

I can’t say definitively but my guess is that Windows 7’s scheduler is a lot more intelligent when it comes to determine thread distribution. Not only that, but if you are on 64-bit then you have less limitations in terms of process resource usage and contention so the ability to run successfully on a HT enabled CPU is going to be better.

This is just my speculation, however.

Pretty good one. It changes the min CPU clock setting to 100% in the power scheme. I think that’s all it does and it may be irrelevant because you can set it permanently in BIOS, iirc. If you go into Windows Resource Monitor and look at the CPU graphs you will see a blue line in the top one that traces this activity. When SAP is running it will be anchored to the top. There may be other stuff in BIOS which it sounds like you’ve already found.

If it still spikes, look for a plug-in to blame!

There was a recent thread about this I contributed to:

Why doesn’t Steinberg explain how the power scheme works? :question: