Question: Platinum lower latency than X3 or not ?

MacOS doesn't produce a lower latency than a well set-up Windows PC attached to an interface with good drivers. My UFX returns pretty much the same latency on PC and Mac, give or take a millisecond or so. Both can produce under 5ms round-trip latency depending in how hard the DAW is pudhing the cpu.

What the Mac's Core Audio system does is make setting up for low latency much more simple. E.g. many Windows PCs suffer from serious PCI bus/dpc latency caused by the wi-fi driver. Which makes running wi-fi and low audio latency at the same time impossible. Hence the common advice to turn off the wi-fi on a Windows PC. Macs generally don't suffer from that problem.

I'm sorry but it sounds like you're basing your concerns on a lot of hearsay and don't have a lot of personal understanding of how latency affects your process. Excuse me sincerely if I'm misjudging.

You may have been listening to an old fashioned ProTools engineer. For the longest time, PT did not have automatic plugin latency compensation. That means that if you put a plugin on a track and it generates, say, 20ms of latency, that track will be 20ms out of sync with the other tracks. Modern DAWs (including the last few versions of PT) automatically compensate for plugin latency and make sure all your tracks are in sync. Moreover, unless you're streaming live audio into your inputs, this has nothing whatsoever to so with I/O latency which is practically irrelevant for mixing. Sure it is annoying when you move an EQ knob and you hear the change half a second or a second later, but you would be hard pressed to notice even 100ms in such a case.

When you're bouncing a mix, most DAWs (with PT again being late to the game) include an "offline bounce" function which means the computer doesn't stream live audio but simply maximizes CPU use to calculate your final mixdown as quickly as possible. Neither kind of latency comes into play here at all.

If you consider my figures "barely ok" I would suggest doing a good search online and see how many ms you can shave off. I believe it's pretty cutting edge already. Certainly that professional mastering engineer you spoke to worked with worse.

I'm finding my figures from the reported figures in the Sonar driver setup window by the way. Technically I should measure it manually but RME reports pretty reliable figures. The softsynth latency is just the output latency without the input latency, although if you're playing a controller there will be some minimal latency from the midi input, depending on which controller you use.

Sheanes
Iic remember a pro mastering engineer advised to keep latency under 10ms (or so) and he said a mix with 20ms latency is a problem.
So according to him, your roundtrip would still be ok but only just....maybe lower the buffer when your exporting/printing or finishing your mix ?
Can I ask you how you measure or get the VST latency ? 

The thing about latency is it only matters when it causes you a problem. Round-trip latency only matters if you are monotoring an audio source through the DAW. For example, using Sonar's track echo function to play a guitar through an amp emulating plugin. "One way" outgoing latency matters if the gap between pressing a controller key and a software synth emiting the resulting sound is a problem.

How much latency causes a problem varies from person to person. There's a general "rule" that under 10ms round trip plus the time it takes for sound to keave nearfield monitors and arrive at their ears probably isn't an issue for most people. Sound travels about one foot/33cm per second, so 10ms round-trip latency plus say three feet monitor to ear distance totals 13ms, or put another way like playing guitar 13 feet from an amp's speakers.

Some find that a problem, others don't.

Chasing the lowest possible latency often isn't the best way to go. Once you get to a latency low enough not to be noticable or affect your playing getting the latency any lower will make little or no difference in the real world. What pushing the latency as low as possible will do is limit the track and plugin count because the lower the latency the harder the cpu and the rest of the system has to work to keep up.

One solution if monitoring through the DAW (which is how I usually do things) is to keep the latency in the "doesn't affect me" range while tracking and use low-cpu usage plugins only at that stage and freeze tracks once they're recorded. Then when ready to mix increase the latency to whatever it takes, within reason, to get a stable result with no pops, clicks or dropouts when you add in more resource-demanding plugins. Some plugins add quite a lot of latency because of how they work, most plugin makers don't tell you how much latency their plugs add (Waves does tell you) but in general watch put for convolution reverbs and anything that "looks ahead" - compressors and limiters often do this.

As for measuring latency, the only accurate test is to connect a patch cable between an interface output and input then set up a track containing a "ping" or percussive noise with a very obvious beginning. Play the pong and record the return ping from the interface on a new track. The DAW timeline will show you how long it took the signal to make the round trip. Make sure you disable any latency compensation in the DAW first though or the DAW will shift the incoming ping to where it thinks it ought to be.

Many ASIO drivers report latency and Sonar's preferences will tell you what they report. Unfortunately many interfaces contain a built-in 'safety buffer' that the driver doesn't report which contributes to latency. Which interfaces have such a buffer I don't know, but one reason I chose to go with RME is that they do say what the minimum possible latency is due to the time it takes their hardware and firmware to do its thing, while most interface makers don't.

I forget what RME say that figure is, other than it's very small. In practice I find I can handle round-trips of around 10ms OK, but by 12ms (or 16 if we include nearfields to ear distance) it starts to feel like I pick a note...pause....sound which is off-putting

Just for completeness, another factor in latency is that external MIDI-controlled hardware also has a built in latency because it takes a certain amount of time for hardware to process MIDI. And if the hardware is a digital synth (or "virtual analogue" which is code for "not actually analogue at all, but digital), the synth will take some milliseconds to do the sums and make a noise. And that time varies from synth to synth.

So my advice is don't fret over latency unless it's causing you a problem. :-)

And if you think this is complicated just be grateful you don't have to pay for then regularly clean, demagnetise and bias a 24 track tape recorder, respool tapes from time to time to reduce print-through and all the other fun and expensive things multi-track recording required 25 years ago :-)

Sheanes
My RME UC does not let me lower buffers all the way, whenever I have a project open.
(I can before opening a project).
I could/want to lower them more as my projects are small/light on a strong computer.
Guess RME built in a safety feature to prevent crashes.
 

You have to change the buffers settings of your RME UC from inside SONAR.
 
Preferences>Audio>Driver>Settings:
 
Click on the ASIO Panel button, this will launch RME's own Driver Settings panel.
 
Now you can change the buffer settings, close RME's Driver Settings window and accept the changes in SONAR.
You don't have to restart SONAR.
 
I do this all the time and never had a problem.
 
Edit to add:
This is not a "could work or not" solution. It's intended to work this way.