I had been using WP Super Cache for a while and switched to W3TC. I saw a major performance hit compared to W3TC.
HOWEVER, just to be clear here, I use a few separate performance plugins and get pretty decent performance.
WP Super Cache
DB Cache Reloaded
WP Widget Cache
Here is what I found by using each of these plugins separately RATHER than W3TC (which I commend as a great product).
1. By using the separate plugins each of them is separately supported and each developer focuses on their niche products like a laser. Fixes, improvements, etc., come much quicker and are more narrowly focused.
2. By using the Widget Cache as a separate plugin the widget cache, CSS/JS combine-minify are not processed for each page that is cached. This allows for better performance overall because after the cache is cleared any single page can regenerate any of these objects. Also, each of these items has separate expiration / regeneration times so that unless I manually clear all of them they are not re-generated at the same time. Now the architecture of W3TC may allow for some of this but I am pretty sure the widgets, and the separate expiration timings are not accounted for in W3TC (although they may be).
However, one other thing I have done is to manually adjust several of the HTACCESS rules by hand. I'm sure I could have done a lot of that with W3TC but I'm pretty familiar with these plugins and am happy with them.
Results: About a year ago my site would take anywhere from 8 - 15 seconds to 100% load even from a location close by.
After many, many changes, tweaks, adjustments, and enhancements around the plugins and some hand coding I now have it consistently under 2 seconds from nearly anywhere in North America. It is also under 3 seconds from most anywhere in the world except for a few 3rd world locations with REALLY bad communication infrastructure.
Many times my site will load in close to 1.5 seconds for locations fairly close by in the U.S. As an added bonus, after the first visit, repeat views of the site load in about .5 seconds because of a bunch of the hand coded HTACCESS rules. That's right, 1/2 of one second!
I'm sure I can do more but my goal was to get consistently under 2 seconds. If I play around with cloud flare I should be able to get that globally but will wait on that for a while.
Overall I am happy with the current setup and performance is great. For me, even IF W3TC is a superior product the risk of experimenting with it for any small additional gains I might get are too high.
That is my analysis.
Check out the site yourself at http://www.R3Now.com
You can test the page speed from WebPageTest.org
IF it is still there, check out today's performance result (retrieved Jan 8, 2012):