W2k vs XP ACPI performance

[mgu2]

For all W2k advocates:

http://service.steinberg.net/knowledge.nsf/show/acpi_kills_audio_performance

XP seems to have better ACPI performance compared to W2k. So, is it really true that XP = W2k + eyecandy?

Quote:

The ACPI problem is not as severe with Windows XP, as Microsoft has introduced a number of beneficial changes.

[Ryu Connor]

FUD.

[mgu2]

Anything to back your insightful comment? Steinberg is a renowned audio software and hardware house. Certainly, I would not call them FUD spreaders. Either bring some facts or your comment is just as you put it: FUD. BTW, I have tested 3DMark 2001SE performance of ACPI and non-ACPI WinXP installation (Marvel G400, AMD XP 1600). Results: ACPI 2344, Standard 2294. Anyone compared W2k in this way?

[Ryu Connor]

There are various results to be had out on the internet. None will show a benchmark difference larger than the margain of error.

[mgu2]

The results I had are consistent and can be repeated, no error here. It looks like MS puts more work to ACPI kernel than to the standard one. I am still waiting for anything supporting your claim that 'stuttering' performance of W2k audio under heavy load is just FUD. Steinberg clearly states that remedy is to disable ACPI in W2k, XP being not affected that much by this problem.

[Ryu Connor]

You had a 50 point difference. That's a 2% difference. It's within the margain of error.

It's no huge secret that IRQ sharing creates additional latency.

http://www.microsoft.com/hwdev/platform/proc/IO-APIC.asp

So in relaity it would actually kill the performance of everything, because the latency impact ends up being the same upon the whole system as it effects the timeframe in which IRQs are processed. The really funny thing, as outlined by the article above, is that it's not ACPI that's really the problem, but instead the legacy IRQ controller is the weak link. That's the same controller that gets used when ACPI is not on as well.

Of course you're own results just summed up exactly how much impact this latency has 2%. I'm not sure if they develop their own hardware in house, but this just looks like a great smoke screen to protect developers from having to write ACPI complaint drivers.

ACPI does not turn your machine into a slug. You know for a standard that's so old it's amazing how many people still rebel against it. Why is change so scary? I guess that's one for the soap box.

[mgu2]

Even if it is bad hardware/software on their side, the fact remains that W2k has bigger problems in ACPI mode than XP when handling legacy(?) hardware.

[Ryu Connor]

Okay, so you are going to make a blanket statement based on a 2% performance difference and an article that doesn't even back up their statement with fact or benchmarks?

Okay.

Listen, I got this bridge for sale. Real hot deal.

[thop]

[mgu2]

The bench I posted was clearly marked as BTW. The article is no journalist/sensationalist crap but an entry in Steinbergs knowledge base, i.e. real world experience of their engineers/support. I take their word much higher than your typical W2k zealot responses. They tested their products performance on both W2k and XP, something you obviously never did. So to sum up, you have not presented any facts justyfing calling Steinberg FUD spreaders or liars. What new facts do you want them to give? They say the performance of W2k is worse than XP in their experience, that is a fact. It is you who makes blanket statements based exactly on what? Do you think they made up those benefical changes?

Quote:

The ACPI problem is not as severe with Windows XP, as Microsoft has introduced a number of beneficial changes.

[mgu2]

Ok, sorry Mr. Ostrich.

[Dr Mordrid]

The best justification I've seen for using StandardPC is on video editing systems.

Lots of capture boards (save for the very latest ones like the RT.X100) can be more than a bit twitchy under ACPI if there are other high bandwidth cards in the system (RAID, high latency audio [CREATIVE!!] etc.). Some are twitchy even if they're the only PCI card present

When I've run into these problems they are almost always cleared up by using StandardPC in either Win2K or WinXP.

Still; I prefer to install using APCI first and evaluate the systems performance before going StandardPC, especially if APIC is supported by the mainboard.

Dr. Mordrid

[Greebe]

But which version of APIC, 1.1 or 1.4?