Last I heard, LTE hasn't set a standard for 4G voice, but it most likely will. Or they just use VoIP, either way. Thing is, everyone but Sprint is going to LTE, so all future iPhones and DROIDs will be in the same 4G boat. Which, for some years, will probably mean "still using 3G"... or in the case of the iPhone, probably EDGE.
Actually all of functionality of the LTE network will be IP based. That is what will make is so blazing quick vs. todays CDMA technology. Currently while talking, the device is using a more advanced version of TDM (Time Division Multiplexing) which uses a rather different technology than IP based technologies.
You're confusing different layers of the network stack. LTE is fast because it's doing nearly 100Mb/s per radio per cell... 326.4Mb/s using 4x4 MIMO. LTE is also very flexible, supporting many bandwidths between 20MHz and 1.25MHz. WiMax has a few options from 10MHz down, like HDPA/UMTS is 5MHz-only.
LTE actually supports both FDM and TDM. The higher level layers of the stack should not be a significant drain on performance (ideally.. but in practice, TCP/IP itself is well proven, so there's no reason to object). But do keep in mind, TCP/IP is still a higher layer protocol... it runs on top of LTE-SAE, which incorporates EPC (evolved packet core), basically LTE's answer to GPRS. There's nothing that dramatically different in the network architecture... the performance comes from higher throughput radios.
There currently is a group within LTE lobbying for a separate voice protocol, which runs at the EPC layer, rather than VoIP layered over TCP/IP. So it's correct to say that LTE is
currently IP based.
WiMax can offer about approximately 75 Mb/s per radio per cell on a 20MHz link, though of course, they can employ the same MIMO techniques as used on LTE, HSPA, and 802.11n.
I'm not sure about EvDO.. it's everywhere, but limited to 1.25MHz channels and all, and it hasn't really seen much work. Both Verizon and Sprint are looking to 4G, and it's just the case that EvDO didn't have many other proponents.
HDPA (UMTS, W-CDMA, whatever of the many names you'd like to tag it with), on the other hand, has continued to evolve. It helps, performance-wise, that the protocol allocated 5MHz per cell. But is has artificially limited the rollout, based on spectrum ownership.. many locales didn't support 5MHz channels at 850/900MHz. The new frequencies allocated (2100MHz) have shorter range, required a whole new set of agreements. And this is crippled a bit at 850/900MHz, as they're usually only one 5MHz channel available, so the protocol is half duplex, versus GSM voice and all CDMA, which runs full duplex (separate channels for transmit and receive). You get these separate channels at 2100MHz, of course.
They started with a per-cell max of 14Mb/s with generic HSDPA, 21Mb/s with HDPA+, and 42Mb/s with MIMO. In some locales, they offer dual-cell coverage with MIMO at 2100MHz, so it's actually possible to see 7.2Mb/s on your phone (peak), if you're on one of these dual-cell 4x4 MIMO towers.. and you have the latest baseband processor on your device. That's awfully close/better-than the promises with Wi-Max. Sprint is saying 4-5Mb/s, but maybe that's considered typical, not sure. Clear is saying they'll get to 10-12Mb/s peak, Real Soon Now. EvDO peaks at around 3.1Mb/s, but most devices support this. LTE folks are also claiming similar, 8-12Mb/s to start with. Both LTE and Wi-Max have future plans leading to a gigabit per cell.
Verizon's big advantage is their ownership of the 700MHz C-Block.. no, not the German rap group, but 22MHz of sweet, sweet lower-UHF bandwidth. They have this everywhere... this is part of the old NTSC spectrum not permitted for ATSC use. Not as much total bandwidth as the WiMax people (see below), but they'll get much better coverage nationwide. One of the big reasons Verizon had better 3G coverage today is that they were one of the two possible 850MHz licensees in just about every market in the country.
Sprint + Nextel (along with Clearwire, Comcast, Google, Intel, and a few others, pooled together to form Clear) have quite a bit of spectrum at 2.5GHz, at least 90MHz in many areas. However, the Wi-Max standard only allows channels up to 10MHz wide. And 2.5GHz has a much shorter range, as well as taking much more attenuation through buildings and, especially, foliage (I design radio systems for use in robotics... our current 430MHz technology, based on 5MHz highly-modified Wifi with optional TDM mode, completely destroys the more common 2.4GHz radios, particularly through forests and in buildings).
I think I'd worry if I weren't either of these guys, come 4G. AT&T has a competitive change... they won 700MHz block B, which is 12MHz... jsut over half the bandwidth that Verizon won. But in some places, that won' matter.. the low frequency will be the deciding factor, not the top network speed. Otherise, there's not enough channel space on today's cellular frequencies to compete. Some smaller WiMax companies are going to higher-still frequencies... plenty of bandwidth there, but the range stinks, and you have to deal with increasingly less penetration though forests and buildings, rain fade, etc.
I would also worry if I wasn't backing LTE in the long term. You can look and see how far the GSM folks have pushed W-CDMA/HPSA, which started out as a poor alternative to EvDO. There's every reason to believe they'll do the same moving LTE forward. There are certainly some powers behind WiMax, but it's not the whole world. Things improve faster with it's essentialy the whole world doing the push.