Hi Tri,
What you're looking to accomplish is *the* most timing sensitive objective.

It can certainly be done, but you need to get round-trip latency down as low as possible.
That means using an audio interface that goes down to a 32 or 48-sample ASIO buffer size... and one that uses a small hidden safety-buffer.
The OctoCapture is a good audio interface. It yields round-trip latency of 7.4ms at a 48-sample ASIO buffer size/44.1k. That's not bad... but it's not low enough for your purpose.
The only way to mitigate the 7.4ms RTL is to increase the sample-rate at which you're working.
Double the sample rate... and you cut the latency of the ASIO input/output buffers in half (albeit at the expense of higher CPU use).
If doubling the sample-rate isn't possible (for performance reasons), then the only solution is to use an audio interface that yields lower round-trip latency. RME (although not cheap) is particularly good.
ie: The RME USB units yield 4.9ms round-trip latency at a 48-sample ASIO buffer size/44.1k
Increase the sample-rate to 96k and the round-trip latency drops to 2.2ms (96-sample ASIO buffer size/96k).
For your purposes, I'd want round-trip latency to be in the 3ms range.
I know some folks doubt this, but as someone who's worked a lot with triggering of drum samples, there's a huge difference in feel between triggering with 3ms vs. 6+ms of latency. It's night vs. day when it comes to feel...