Unlike Tibetan, where the ga-kha-ka sequence can also be rendered ka-kha-ka in some grammars (though really, it's more like ga-ka-ka, where the third variant can end up voiced if certain superscript, subscript or prefix consonants are attached), the Mandarin distinction is very clear, to the point that no one even understands why the Wade-Giles system tried to use the p-p' distinction for b-p, etc.

Given that Cicero noted how Ennius called Pyrrhus "Burrus", and how κομμι > gummi, κραββατος > grabbatus, αποθηκη > Vulg. botteca, πυξος > buxus, and κυβερνω > guberno, as well as the old Latin "Antioco" and "Corinto" for Antioch and Corinth (see Vox Latina at 12 and 26), it may be that languages lacking the distinction just don't hear it that way. Since all of our voiceless stops are aspirated as well, we can't distinguish between t and th when both are voiceless stops, one aspirated and the other not.

However, with Mandarin I'm very, very comfortable saying that the Chinese themselves don't see things like that. With Tibetan it's probably a situation where I'm just not hearing it "right" (like when I try to explain to my mother palatization of consonants in Russian - she can't hear the difference in the "L" no matter how hard she tries).