Abysmal efficiency with TPS63001

[Artlav]

Greetings.
In short, i'm trying to use TPS63001 to raise 2-3V to 3.3V, and the efficiency i get is very low at any load.

Long story - i want to make a phone charger (couple of USB 5V, 500mA outputs) from a single huge LiFePo4 cell, and TPS6300(0/2) seems to be the chip for the job.
But only TPS63001, the 3.3V version, was immediately available here.
So i got one to try it out and see what it can do before i order&wait for the '2 version.

Here is the test board:
Schematic: http://orbides.1gb.ru/img/tps_sch.png
Layout: http://orbides.1gb.ru/img/tps_lay.png
Appearance with 4A inductor: http://orbides.1gb.ru/img/tps_conv.jpg
Appearance with 1.2A inductor: http://orbides.1gb.ru/img/tps_conv2.jpg

The buck mode sort of works - i can get over an amp out of it at 3.8 to 5V input, at up to 70% efficiency. Still below the datasheet info.

However, the boost mode i'm more interested in almost does not work.
From 2.8V it can give up to about 100mA output, heating like crazy. Efficiency below 20%.
Lower it gets even worse, higher - not much better.
Far cry from 70% and 1A datasheet claims at 2.6V.

So, i figure i did something horribly wrong somewhere.
Layout is my biggest suspect. The big 4A inductor might not like the 1.5MHz frequency or something.
But with a smaller 1.2A inductor the difference is negligible.
Connections on the IC seems to be fine. I reflowed it one more just in case, with no effect.

That leaves the actual layout or something i've missed.
Is there anything obviously wrong with it?

More generally, what can be causing the losses?
It's the IC itself that heats up as far as i can tell.

 

[Artlav]

Ok, turns out the IC was bad, or more likely became bad through some sort of mishandling between unpacking and first power up.
Anyway, i soldered in a spare one, and now everything works like a charm.

Sorry to bother.

- - - Updated - - -

One more thing - TPS63001 can not be made to produce 5V output by fooling it's FB pin with a resistor divider.

It would work nicely for 10-20 seconds, then start deregulating, go chaotic and die in a manner described above.
That is what got my first IC (i initially had the divider there, but removed it when it started acting up 10 seconds later), and i reproduced it exactly with the second one after having it do 3.3V nicely for a while in all ways and orientations.

Just in case someone googles up this thread.