Well, I'm finally back from my few weeks of W2K-XP Upgrade Hell. I found out something new about why a Windows XP / ATA/ATAPI machine can have dreadful DAW performance. (As a quick precursor to this post, I did a search for 'PIO' in this 'Problems' forum, and didn't get any results from the search. I honestly think that this could very possibly be the root cause behind many DAW problems experienced by Windows users.)
I finally decided to join the 21st Century by upgrading my system from W2K to XP Pro SP2.
As I have a considerable amount of both VST and Scope plugins, I really wanted to upgrade my system without doing a completely fresh install if I could at all avoid it. This is not a good idea at all; if you're going to change operating system, it's for the best to format the drive and start from scratch regardless of how painful a procedure this may seem to be.
I observed lots of strange problems after performing the 'over-the-top'upgrade: system would not reboot, system sometimes wouldn't shutdown cleanly, USB would only be enabled on every other reboot and other minor issues (and this was after I'd completely patched-up the OS with every available Microsoft update).
So, on to the most important bit, the audio system...did it still work as flawlessly as before?
=>NO<=
In fact, the audio & VST performance of my system were the worst I've ever experienced.
I tried everything that I could think of to resolve the issue:
o install as Standard PC
o install as ACPI Uniprocessor PC
o BIOS settings ('P&P OS', 'PIC/APIC', 'IRQ assignments', etc.)
o IRQ steering on/off
o IRQ sharing on/off
o XP Performance tweaks (a great many!)
o Cubase VSTMultitrack setttings
o etc. etc.
Nothing, but nothing I tried changed affected the performance. Then I noticed something which seemed very odd; in Cubase, both the VST Performance meter and the Transport Panel show CPU usage. With Cubase not even playing anything (i.e. transport stopped), the CPU usage was showing something like 20-25% but then every couple of seconds, it would show 'CPU Overload'! Very strange. When playing, the same thing happened and every occurence of the overload would result in a complete breakdown of the audio system.
So I started doing some research on this phenomenom.
It seems that if XP detects 6 (over all time, not consecutive) CRC errors on an IDE channel, it starts ramping-down the speed at which the bus runs at, from UDMA5, to 4, to 3 ... and eventually the bus will be set to run in 'PIO' mode, which stands for 'Programmed I/O'.
This is a VERY BAD thing, especially in a DAW. It means that the CPU has to manage all data transfers to/from the bus, rather than allowing the DMA controller to acheive this.
OK, it's true that the CPU does need to give up some of its operation cycles for the DMA controller to do its thing, but the DMA controller is much more efficient at moving data about than the CPU is.
Just sticking a badly scratched CD or DVD into your drive can easily cause 6 CRC or timeout errors to occur, changing the IDE channel from UDMA mode to PIO.
Most problems I've seen posted to this forum seem to concentrate on IRQ issues, which I'm not personally convinced are much of a problem really; I'm not saying that no problems can be caused by interrupt sharing but if an interrupt fires, then its ISR (Interrupt Service Routine) needs to run and will be called regardless of what IRQ number it's on, which will take up CPU cycles.
There are a number of ways in which this PIO mode problem can be checked, and potentially corrected.
Firstly, go to Control Panel->Admin Tools->Computer Management->Device Manager, expand the 'IDE ATA/ATAPI Controller' node, right-click on the Primary (and afterwards, Secondary) IDE Channel icon and select 'Properties', tthen select the 'Advanced Settings' pane (should it exist; it might not always be there).
Check the 'Transfer Mode' and 'Current Transfer Mode' settings. If they indicate 'PIO' mode, then this will certainly mean very poor DAW performance.
Another thing worth trying as a test (if the 'Advanced Settings' tab is missing from the above channel properties dialog) is to run a hard disk speed checking application.
I used a freeware application called DiskSpeed:
http://home.earthlink.net/~alegr/download/diskspeed.htm
The page also has some useful information on the sorts of speed and CPU usage you should be seeing if the IDE channel the disk is on is working in UDMA or PIO mode.
There are some registry settings which you can try changing/deleting/adding which affect the IDE channels. These are located under the following key:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E96A-E325-11CE-BFC1-08002BE10318}\
Under this key there are entries such as '0001', '0002' etc. which contain settings for the IDE Channels.
Of particular note are entries called 'MasterIdDataChecksum' or 'SlaveIdDataChecksum' - if these exist, then this means that errors have been detected on the IDE channel and the operating speed has almost certainly been affected.
This article describes things much better than I and suggests many solutions:
http://winhlp.com/WxDMA.htm
Miscrosoft also know about the problem:
http://support.microsoft.com/kb/817472
About 4/5ths of the way down the following page, there is a good description of the mask values used for the MasterDeviceTimingModeAllowed and SlaveDeviceTimingModeAllowed registry entries: http://snoopdos.com/blog/index.php?skin=clean
I found that information useful when changing the registry settings to try to get UDMA mode back on the IDE channels.
I tried pretty much everything, removing registry entries, changing others, uninstalling and reinstalling IDE channels etc. etc. This did improve things to the point that I could play back & record some music, but the performance still wasn't quite as good as before.
In the end, I bit the bullet, re-formatted the hard disk and reinstalled Windows 2000 in 'Standard PC' configuration. I probably should have at least tried this approach with XP, but didn't want to have to go through all that twice should a clean XP install not have worked. I now have a rock-solid DAW system again
Another thing which is perhaps worth mentioning about IDE-channel based drives is that, should a channel have 2 devices on it, only one of these devices can use the bus at any one time; it's transaction needs to complete before the other device may use the channel. For this reason, to get the best possible performance out of an IDE channel, it's probably best to only have one device installed on it.
As far as I'm aware, the above does not apply to Serial ATA (SATA) based hard disk drives.
Hopefully somebody will find this information useful, especially if you are experiencing poor (or very poor) DAW performance!
Royston
