I know I'm just a newbie here, but the engineer and pc enthusiast in me have to reply here. I'm very happy with the replies from Swiftech reps here, not just for the insight they provide, but also that they take this much of an interest in Tom's review.
About:
Are you asking me?
Read my post please. I never said that the data wasn't valid. I suggested explanations for the variations in temperatures between Tom's data, and our own. I do stand behind my recommendations as two publishing two separate data sets (one with supplied thermal compound, and one with same thermal compound for all items compared in the article, as it provides users a more comprehensive view of the products.
I would love to see results for both, heck, all products with both 'supplied thermal paste/solution' results and results after refit/with identical thermal paste. I'd go so far as I would like to see those result sets with identical fans too (ie four sets of results and probably the average of three runs per test for averages: I know, it's a huge time sink and very hard on reviewers, but I'd still like to see it
).
That said, it will be difficult to do if the product only has pre-applied paste (i.s.o. some included paste that the user has to apply themselves), since obviously no company would provide reviewers a new, previously reviewed, product every time a new product comes out, that's just silly. So it might be tricky to get the first set of results to be on the same exact rig/driver/software set as all the others. The second set will be able to (provided the reviewer takes the time to retest everything) include all this, so people like me would be more inclined to look at that anyway.
Then there is the whole bonus of the H220: you can mod/expand the set, which is a great thing for starters. But you wouldn't always do that with the set installed on your cpu, you'd take it off and swap paste to either your personal favourite, or a new application of the provided paste (if you have enough left). When you get bad heat results you might want to refit your cooler (I know I have done this quite a few times with my air coolers) just to be sure you may have drawn a bad ticket in the silicone lottery, which again means new/other paste.
So I think it is fair to say that the combination of using identical thermal paste (that might be of lesser quality than the H220 provided paste, although that too could be tested on both units) combined with the fact that the corsair fans in this case might be running at (way?) higher RPM and possibly higher noise levels (I think Tom commented on the low noise of the Swiftech H220) could very well cause the H100i in this case and set up to provide better thermal results than the H220, as the Swiftech reps have done. So kudos to both here. I would like to see Tom clarify this more rather than just saying the response validated his results (which I agree it did), to include some info like this rather than just mentioning it, but I guess he's pretty busy anyway
It is of course a bit impractical to try and review every unit with pre-applied/provided TIM since they don't all come out at the same time, making any results far enough apart to be compared fairly and I think the Swiftech guys kinda know that too, but I understand that their benches will include that. The whole RPM levels vs Corsair Link 'load levels' is something that will always be tricky, yet easily solved by switching fans (hey, it's not like reviewers need a life or anything
) and redoing all the test runs.. Yeah..
That Thermal Engineer's opinion sounds dodgy at best btw, I'll wait for actual tests on that rather than accept it now.
After this nice wall of text I'll just add that with Tom's bump up to gold (and thus due to Swiftech's more aggressive pricing) I've regained interest in the unit (slightly pending pricing in NL) and will wait to see results for expanded solutions tested
