Windows 7 Reduces Audio Latency

[braincell]

The Windows 7 versus Vista and XP tests are not made for musicians so don't believe it when they say Win 7 is the same as Vista.

Some time ago I saw a video with a Microsoft engineer talking about how they were working to reduce audio latency in Windows 7. I am getting ten times faster audio than in Vista. Audio latency has always been a problem for me from Creamware, Sonic Core and more recently Echo Audio. I thought I had solved this problem by using ASIO4ALL WaveRT drivers in Vista but then I found out that I had been recording at 11 Khz! I went back to Echo Audio WaveRT in Vista and I had to settle with 40 MS latency which is about what I got with SC in XP. Now I am happy to say with Windows Vista installed, I can use 4 MS latency have have no dropouts with WaveRT drivers.

[astroman]

you know... they talk a lot in that corporation...
but if you were unable to achieve anything lower than 40ms of Asio latency in XP then it's your inability to configure your system properly.
But most likely you referred to some WDM rubbish, so it would be nice to make that a bit clearer in your post.
A Windoze 98 box can do the 4ms (or less) Asio trick without problems, but the maker of your favourite OS (unlimited resources apply) needs > 10 years to implement a similiar functionality...
And they even dare to talk about it... LOL

cheers, Tom

[Immanuel]

40ms is not even an option in Scope - and never was. 34ms is but you need a realy old PC or an incompetent installation to need that much latency from Scope.

[braincell]

.

[dawman]

I said similar to Scope not the same nimrod. You need to read my post. I tried configuring and asking questions out the anus nobody could help me not Creamware support nor anyone else in the group so you can kiss my ass f*cking astrocrap bitchboy!

Ladies & Gentlemen...............Braincell....ankyu...
I was wondering how long it would be....

[garyb]

you know... they talk a lot in that corporation...
but if you were unable to achieve anything lower than 40ms of Asio latency in XP then it's your inability to configure your system properly.
But most likely you referred to some WDM rubbish, so it would be nice to make that a bit clearer in your post.
A Windoze 98 box can do the 4ms (or less) Asio trick without problems, but the maker of your favourite OS (unlimited resources apply) needs > 10 years to implement a similiar functionality...
And they even dare to talk about it... LOL

cheers, Tom

yeah, i thought something similar...

[garyb]

I said similar to Scope not the same nimrod. You need to read my post. I tried configuring and asking questions out the anus nobody could help me not Creamware support nor anyone else in the group so you can kiss my ass f*cking astrocrap bitchboy!

Ladies & Gentlemen...............Braincell....ankyu...
I was wondering how long it would be....

i'm not sure what the intense anger is from. intense mood swings....

[braincell]

Certain people in this group are not objective. Don't pretend to know everything when you clearly are biased.

[garyb]

but windows WDM drivers have been crappy for quite some time, while ASIO has always worked well....

[astroman]

that's what I meant...
Braincell is a little overreacting, but that's ok with me.
I'm just not willing to accept the 'biased' attribute...

towards SC ?
indeed, I second a small company that's got the balls to go contra mainstream
and I value 'Made in German' as much as the folks at Accutronics label their spring reverbs "Made with pride by beautiful girls in Green Bay Wisconsin"
but other than that ? I even posted a track made only from Scope sources is (most likely) boring to my ears. Didn't invest in SC stuff for quite a time - it's no secret that I aquired half a dozen 'real' string instruments meanwhile.

Yet the Scope system will remain my tool of choice
Because it's so simple
I never had problems to set it up, given I used reasonable hardware
and I really tried a lot of stuff from mini-ITX to HP CAD-Workstations
it even worked in a (officially unsupported) 160 MHZ Powermac with full remote desktop control from a Windoze box... just to prove it could be done.

I'm longer in this business than there is Windoze - I know it's origins, all of them.
My living depends on IT service and sales, not on funky tracks

You cannot call experience a 'bias', I really appreciate that M$'s and Apple's sh*t keeps my business going, but OS and apps remain poop - whatever you call it.
I really can estimate what could have been 'possible' in > a decade
Sorry Brain, you just got fooled again
They will repeat this over and over, ka-ching, can you hear it ring ? as Miss Shania would have put it...

cheers, Tom

[braincell]

I have *never* used WDM drivers to make music. Who would do such a thing? WaveRT is faster than ASIO.

[Janni]

WaveRT is faster than ASIO.

sure...

I went back to Echo Audio WaveRT in Vista and I had to settle with 40 MS latency

???????

[next to nothing]

I welcome any alternative to a current monopoly. It's not something i miss though, since I've been working on 3-4ms for the last 9 years on 4 different systems; one VIA/AMD, one nForce/AMD, one nForce2/AMD and now an nForce 650i/Intel setup. Never had any latency issues.

And mind the language.

[braincell]

Perhaps you can't play as fast as I do and you don't use HD streaming.

[dawman]

Perhaps you can't play as fast as I do and you don't use HD streaming.

Hey, you'll even need faster more powerful solutions when you start using both hands...

[next to nothing]

Perhaps you can't play as fast as I do and you don't use HD streaming.

HD steaming would be more strangled by CPU/RAM combination than ASIO AFAIK. If you mean you play at a tempo where 3-4 ms in 44.1 is too much you live in a world with no other inhabitants. When you play a classical piano you have approximately 3ms latency from the hammer hits the string to the sound hits your ear.

Then again, if your incredible fast hands are triggering too many notes, it still is rather a problem of CPU/RAM.

But again, as i said, I welcome every good alternative that might pop up.

So, again, rather than calling people names and claiming 20ms as normal for a CW/SC card, you should get someone to help you set up your computer properly or read up on it.

[braincell]

I use a Quad. Yes CPU is important but your guess is wrong. Each note played is akin to another track of audio. I can clearly hear dropouts when I lower the latency setting too much on my card and I play as fast as humanly possible with the sustain pedal down. Like I said, now it's 4 MS which for me is incredible considering the dismal performance I have had in the past.

[astroman]

maybe a little number game can put things into context...
a single sample last for 1/44.100 second
during that time of a single sample a 2 GHZ CPU can perform 45.000 cycles
Midi is a 33 kbit protocol, which means you might squeeze about 1000 note events into 1 second of datastream (if you're lucky ?)
which means for each note event you'd have 2 million CPU cycles available
and in the duration of a 4 ms buffer a modern CPU does half a million cycles
so where's the problem ?

of course this is rubbish cause the CPU has a bit more to do than just wait on single samples and notes (and the figures are just estimations anyway), but still it illustrates the dimension of processing power quite well.
Or the waste thereof ...

I remember Gigasampler 1.x streaming the Gpiano from a network drive on a Celeron 333 with 128kB of Ram under Win98. Only tried it for curiosity as it was mentioned in the Docs somewhere. Sure it wouldn't have passed Brain's hardcore-pedal-down-test, but for a more regular playing it really seemed to work.
You could hardly install XP on such a box, let alone it's successors...
Ask yourself (and your supplier) what the heck the damn thing is doing, while you want it to perform music, and wtf you aren't allowed to shut THAT off

cheers, Tom

[valis]

It could also be that when Win7 read the ESCD data from your BIOS & APIC 2.0 controller it shuffled around resources and virtual irqs in a way that hadn't been done before, easing some conflict...or perhaps a particularly nasty driver was replaced with a generic one (graphics card or wireless networking stack) that isn't hogging the huge time window it was before.

With a modern PC, if you can't handle low latency ASIO its because somewhere something is vying for that bus bandwidth, starving the cpu for the cycles it needs to service your amazing 450 voices of lush piano. When your computer serves the i/o handles (whether its for ASIO, Wavert or w/e) the primary cause of overhead isn't the 'cpu power' involved. It's the wait states that result from tying the cpu up shuffling data and waiting on the release of the interrupt request so that it can go back to other tasks. This is where the cpu rise comes in, whether the cpu is busy as hell working on some dsp magic for a plugin or sitting there waiting for an interrupt to release, the same amount of cpu time is taken. One gets you more 'voices' while the other 'steals' from your voices, so you might think that somehow it's a computationally bound task. Wrong, it's an i/o bound task (low latency) and inability to serve it is a problem of congestion. It seems like you're running out of cpu cycles but really you're hogging so much cpu time that by the time it's released and the number crunching can continue there isn't enough time to get the work done before the next wait state. A lengthy way of saying that it does no good to have fast ram & a 'quadcore cpu' (which doesn't say as much as you think it does) if you tie its hands up.

Have you ever looked at DPC latency checker (do you use wireless?) or checked to see how much your latency timings change when you change modes for your HD subsystem (aka cpu-bound onboard RAID)? Try it under Vista if you still have that partition.

Also, my understanding is that WaveRT made some moderate changes to what was once called the 'streaming kernel mode' WDM Wave drivers to 'improve' MS's low level audio stack, but remember the main reason they needed to create the ability to have 'low level' is because the windows audio stack(s) is(are) loaded with tons of crap. And most of the changes touted for Vista and then Win7 have more to do with adding little sliders to every app that might have audio (like aol instant messenger, msn messenger and your web browser) and 'improving' the default windows mixer & control panel, rather than creating low-level efficiency needed for musicians and audio professionals. There are so many layers of mixing & resampling in the windows audio stack that ASIO is a de-facto standard for audiophiles as well as musicians, so that they can bypass the numerous non-transparent processing layers between the soundcard and the source application. WaveRT & 'streaming kernel drivers' basically attempt to do this too, but with mixed results (based entirely on application & hardware support, or lack thereof.) If MS *truly* cared as much about professional audio as they do about directx gaming, outlook calendaring and chat clients we would all have dropped ASIO some time ago (at least imo). BeOS did it fine on an early pentium...

[braincell]

As Microsoft points out, for best results WaveRT must be implemented at the hardware level. They may not have been thinking of pro musicians when developing WaveRT but the bottom line is that latency creates dropouts in audio. When streaming audio, such as from internet radio stations, dropouts are not acceptable. The competition from Apple could also have been a consideration.

I don't have any wifi in my DAW and I am using the same graphics card driver as I used in Vista. It had to be manually installed. The difference is clearly Windows 7.