If you are talking milliseconds then my playing is way too slow (or just plain bad) to notice any problems.
 
@

brundlefly
I was running an E-MU 1820m at the time of this thread as well. It did have excellent MIDI performance, but the x64 beta driver proved very unstable for me, and a hardware fault eventually forced me to dump it. I'm now back to using my MOTU MIDI Express XT USB which is much slower, but jitter continues to be a non-issue for me now as it was at the time.

What do you mean by slower. Slower what? It either works or not I suspect. The reason I ask is I have just built a new machine and I am needing a new midi interface as I am running Win 7 64 bit. There are not many drivers out there for an 8 IN/ 8 OUT interface except for the MOTU. So for me this is one of the most obvious choices.

Hey Dave , yes that's the part I guess I meant buy playing soft synths live, in other words real keyboard people. I'm talking about us guitar players who DO have to use quantizing or you'd certainly not be able to use my tracks :)
This is say, a simple organ or string part. If I play piano parts they never get beyond 4 notes at a time, nothing faster than a 16th note.
I've recorded and edited the midi from real piano players and I would not even attempt to quantize that!  I will manually correct the odd clunker in those cases. So I totally follow what real keyboard players would be dealing with.
When I record ( real) bass I have to use headphones or stick my head close to the monitors because I can't stand the delay it takes the sound to travel 3 feet!!! Just kiding, but it sort of true if your a timing freak. 
And I've stopped quantizing my drum parts since I got my Yamaha DTX kit. I only quantize the Kick now. I find my snare and hi hats are just a hair ahead of the beat and that seems to be where they belong to my ears.
 
I'm finding my midi set up is now as good as it was on my Atari so I'm happy again.  
 

Human keyboard players (or any other musicians) aren't entirely accurate in the first place, nor do they repeat the same note pattern with exactly the same timing every time.

A few years ago I saw a documentary that was primarily looking at reaction times, "muscle memory" and transmission of commands via our nervous system.

One of the things they did was high-speed video concert pianists playing the same music multiple times. When the video was slowed down it turned up a couple of relevant things. Taking a five-finger arpeggio as an example, once the thumb hit the first note then the timing of the other four fingers hitting the top of the keys in turn was consistently within a 3-5 millisecond bracket. Which the researchers thought an amazing level of repetitive accuracy. That first note of the arpeggio however varied in its timing placement by as much as 15ms.

The conclusion was that once a well-learned muscle movement pattern starts it can have very tight and consistent timing within the pattern, but outside such patterns things get sloppier. Which is why I guess we practice, to make as many muscle patterns "automatic" and cut down the "what do I do next" bit to as short as possible.

In other words, sweating about a few milliseconds of MIDI jitter when playing live is probably pointless because the human part of the equation isn't that accurate in the first place. After all, how many guitarists can play in time when five feet from their speaker cab but find their timing goes to pieces once they get eight feet away?

Midi jitter is only one aspect of midi performance.  And on its own with just one instrument involved it may be insignificant and not affect the performance unduly.
 
But there are other aspects of midi especially when a larger number of external devices are involved and you are creating sequences where say 8 to 16 things are playing at once.  As a composer coming from background as a drummer I am rather over sensitive to timing issues so for me it has always been about all these things coming together very well in groove and time and feel.
 
If you have a midi interface with 8 OUT ports then you should be good to run at least one instrument per port to minimise and midi latency due to multiple channels being on one port. With this setup only one synth is on a port so it should take 1 mS to send a note on from the midi interface to the hardware.
 
Some parts in your sequence may be quantised but others may be completely live against the click.  The DAW has to record and playback the live performances perfectly and accurately and to multiple sound generators and keep all that timing intact too.
 
So there is the issue then of how well the whole midi system performs driven by a computer that has also got many audio tracks under its control and a CPU that may be working hard doing it all while managing many plugins.  The timing between the midi and audio should be strong and relentless and not vary in any capacity.  That is the heavy stuff the total system has to keep under its control.
 
Some DAW's do this better than others and also some hardware configurations also work better I believe.  For a long time I had audio being controlled over the PCI port and the midi over the serial port.  That had been a brilliant combination and timing is rock solid especially when the audio side of the machine is working ultra hard.
 
But a new build for me involving Win 7 and 64 bit has left many of these midi interfaces out in the cold.  So that is why I am keen on the MOTU Midi Express XT in order to make my newer setup work.  It is the main option for me now.  Currently I am using a nice RME PCI sound card now but the midi will be over USB instead.  I am hoping that is going to work.
 
I have found Logic and Studio One to be strong in keeping midi timing not effected by heavy audio CPU usage. (but that was midi over the serial port so not sure yet about midi over USB) They also seem to record and playback midi performances accurately too.  That is a critical factor I feel in the whole process too.  I still feel it is better to keep audio and midi on separate ports too.  It makes sense, just less for each port to do.  Means better performance.
 
Midi on virtual instrument tracks driving VST's poses other issues.  The fact the audio side of your DAW has to be allocated to making the sounds is one of them.  VST latency can affected more on what the rest of the audio system is doing too compared to external midi devices running off its own port.  Live input I mean in this sense.  Fast computers really help in this area while audio processing can be done using external DSP such as the UAD system backing up your own main system.

Jim Wright

There needs to be a MIDI 2, but I don't think there would be enough interest now.

Oh, there's definitely interest. The standards folks are still working on "HD MIDI" (Google it) -- these things just take time to percolate. If HD MIDI doesn't happen, something else will come along eventually. Don't give up! (Not that this helps any musician struggling with a DAW today, unfortunately....)

- Jim

I just got to to this post by Jim so I googled HD Midi.  It looks like it's still alive: http://www.midi.org/aboutus/news/hd.php
I wonder when or if we're going to see anything that supports it.
 
I also wonder about the performance of Thunderbolt audio/midi interfaces.  In this thread it seems like the concensus was that the old parallel port and PCI interfaces offered the best performance in terms of latency and jitter followed by firewire.  USB was the worst but still able to get under 1 msec of jitter if everything was right.  I wonder how Thunderbolt stacks up.  Is it just a faster USB or does it allow some kind of synchronization like firewire does to help line up midi and audio?
Dan

brundlefly

[...] I'm fairly confident that the vast majority of users who believe that SONAR does not capture and/or playback their performances with sufficient precision are deluded about the precision of their playing. [...]
 

I am blinded by such a beacon of truth, but it is still far simpler to blame my computer rather than accept my own flaws!!!