When I practice using a VST, when I switch to my internal sounds I can play faster and cleaner with better touch. I have very little audible lag with VSTs but I'm guessing there's some latency you can 'feel' rather than 'hear' which makes playing just a little more difficult.
Anyone else experience this?

My experience is the opposite. I have an RME internal soundcard and when i play the best piano vst’s like both Synchron Bösendorfers from VSL or the Garritan CFX it totally outperforms the internal piano sounds of the Kawai MP11SE. Apart from superior sounds the playability is a level above too. I think the latency of my set up is neglectable and the superior scripting, dynamics and deep details of the vst’s mentioned surpass the internal piano sounds by quite some margin. And that is being nice to Kawai for providing me with such a fine controller.

IMO this is likely due to latency 'jitter', i.e. variation in latency, which could be due to any variation in timing of processes that take place before the audio buffers are filled. The audio buffer, which receives a lot of attention in latency discussions, adds a fixed latency, without jitter. But as far as I am aware anything that needs to happen before that (receiving/processing midi, computing the sound) can be variable on a regular computer. Studies have shown that even small amounts of latency jitter lead people to judge a midi instrument as inferior (+- 3 ms max in a midi drum instrument). Fixed latency is less of a problem as long as it is not extravagant.

The irritable nature of latency jitter probably is that it makes the time connection between action (playing a key) and consequence (hearing the sound) unpredictable, whereas fixed latency doesn't.

Dedicated computers (in your digital piano) do not have to deal with all the non-piano processes that are also running on your vst computer, and hence have no appreciable jitter. I have the impression that Pianoteq runs more nicely on a Raspbery Pi than on a fast laptop, for the same reason. Although Raspberry Pi Linux is still multitasking, on can easily remove all processes that are not necessary for playing piano, and end up with less jitter.

What pianogabe wrote is truth, but it's '2nd level of initiation' I would start with checking basic stuff like:
- audio driver used (is it ASIO if you use extrenal audio interface or WASAPI in case of PC built-in soundcard)
- sample rate (provided in Hz)
- sample buffer/buffer size

You can calculate your latency thanks to math: buffer size/sample rate in kHz.
Some examples:
- sample rate 44.1kHz, buffer size 128: 128/44.1 = ~2.9ms of latency
- sample rate 48kHz, buffer size 256: 256/48 = ~5.33ms of latency

Of course that's just theoretical, because soundcard drivers by themselves intruduce additional latency (usually minimal). As long as you're having something like: 44.1/48kHz and 128 buffer or the multiplies of those you should be good and don't percieve any latency.

There is another variable though. Piano VIs differ one from another and in some cases they introduce additional latency. Even how the sample was recorded matters (eg. if it has long 'pre-attack') it might give a feeling of disconnection (Embertone Walker 1955).

Another explanation (that doesn't exclude my previous one or AlphaBravoCharlie's one) could be that the fixed action-sound latency of your internal sound is simply more than a bit lower than the fixed (or mean) latency of your VST, even when you do not 'hear' the latter. I have measured such things in the past and even with very low buffer settings, dedicated sound card/ASIO drivers etc, the overall difference between internal engine and vst can be 10 ms. With one or two steps larger buffer settings this becomes disproportionally more. Incidentally, on my silent piano the internal sound engine is 10-25 ms faster than the acoustic sound (!), depending on key velocity. So an added 10 ms as compared to internal sound would not be 'unnatural'. And no one complains about the latency of an acoustic , which suggests me to that latency jitter in VST produced sound may play a role.

Speaking of Embertone Walker 1955: I installed it on a SSD drive. I get double strikes in the treble register. Is there an adjustment I can make? I've played around with buffer size and the pre-attack dial, but I haven't found the solution yet.

Could you elaborate a little bit more? Maybe share some recording of the issue? I'm not sure what you mean by 'double strikes'. One thing I can advise for EW1955 (I'm not sure anyone else experienced it) is to set DFD in Kontakt for that particular instrument to the lowest possible value, 6kB. It's strange, because usually the higher DFD the more samples are loaded into RAM and the smoother experience. I got completely the opposite effect with Walker. In anything above 6kB it was basically unplayable. Cracks and pops all the time, everywhere.

With friends like you . . . .

I think a lot of it depends on the VST and the system. Latency is bad, velocity layers are good, and some of us are more sensitive to both than others. Latency makes me feel disconnected from the experience, and velocity layers encourage me to play with more expression and touch. I'm sensitive to latency so I built a system that minimizes it. Note that there's some natural 'latency' in an acoustic piano since there is space between the key/hammer and string. A DP with good electronics should be able to play faster than an acoustic, and therefore many have some delay/latency intentionally built in. If I play my DP with onboard tone and no delay, its fast and responsive, but also encourages me to play differently than I do on an acoustic - more like I'm playing a keyboard.

This makes me wonder. I play most days both my acoustic Yamaha U3H and, mostly on late hours, my VPC1 with several VSTs. My PC is not a beast, is just a 2014 i5 with 16GB RAM and several SSDs (totalling 3.5TB). I use a 128 buffer on the Yamaha ASIO driver which runs the audio on the Yamaha MG10XU mixer. Nothing specially fancy.

Well, I don't have any problem playing the VSTs nor going back and forth to the U3. Is it just that my playing level is too basic to notice the difference?. Is my brain adapting to the circumstances and fooling me?

I guess my playing is at fault but, well, if that means I can use my current setup for some years, while my level hopefully increases, it is not that bad.

I think your DP and upright should be pretty similar. 16 gig i5 with a good ASIO driver should be plenty fast.

I have a similar setup for playing VSTs on my Kawai silent piano, and found that audio buffer 128 (also Yamaha ASIO) is the 'sweet spot' in the sense that the resulting overall latency matches that of acoustic playing best.

Thanks for the comments!. So it seems I found a good enough setting. It is confusing when so many people insists on 64 or smaller buffers. My system, with some VSTs, could manage 64, but I didn't find it specially better and, as several VSTs didn't work nicely, I went up to the current 128

Yes, it seems so. Great

I'm sorry, I just saw your post. I will try it today with the lower DFD, setting. It's odd to me that I did not have this behavior running it on a Mac with an external SSD drive. I am now using a MacBook Pro with an internal SSD drive and 16 ram. When I say double strike, I mean when I play a key in the upper register, the sound of each note repeats rapidly. This also happens with WholeSound Baldwin with runs in UVI. Other VST's, IVLabs and Garritan do not this problem.