Timing.
Benefits compared to upgrading costs.
>>54497853
incompatible with x86 code
Poor performance compared to competition, high power consumption, high heat output.
Then it got killed Redhat and Microsoft who dropped support for it.
The only OSes working on them are HP-UX and SLE for legacy reasons.
>>54497853
Dead.
>>54497959
>Poor performance compared to competition
Mainly this
>high power consumption, high heat output
>he hasn't seen power8
engineers actually believe rearranging silicon gates will somehow be better _this_ time around
fucking idiots
>>54497853
Muh CISC.
>>54498577
More like double-CISC, called WLIW.
>>54497853
> legacy SW install base trumps moderate architectural gains
> static scheduling requires very good compiler, early compilers produced mediocre output
> cost a shit ton
> nobody really likes vendor lock-in
fair number of reasons, and I'm sure I forgot a bunch
>>54497853
Delay after delay.
It was an incredible chip when it was due for release in 1997
or 1998
or 1999
or 2000
or 2001
and by the time it finally went commercial, it was very lackluster, the x86 emulation was horrid making a transition implausible for many companies and consumers alike, and Intel's magic compiler didn't perform to expectations.
That's not even mentioning the systems built on it, which were just absolutely half-assed. SGI basically committed gruesome suicide, binning IRIX and MIPS before the Itanium was even ready, and by the time it was, they hadn't even gotten to work on an IRIX port, choosing instead to dangle a bunch of overpriced commodity GNU/Linux shitboxes in front of their disgusted customer base. IBM paid lip service to it, released a couple models and then gave up and withdrew everything from the market less than a year later.
HP was probably the only one that handled it somewhat well, and even then, they destroyed both PA-RISC and Alpha to do it. At least they ported HP-UX and NonStop to Itanium instead of trying to shove GNU garbage down their customers' throats.
In the end, the world really just wasn't ready for such a radically different architecture. However, when the compiler did work, IA-64 chips were very fast, and for that reason it did find a very comfortable niche in supercomputers, at least until Intel gave up on it too. Now pretty much all it has left is RAS features, but Xeons are creeping up on that too.
>>54498577
The RISC/CISC wars are horse shit, and the Itanium was neither.
>>54499013
You know, when I really think about it, the Itanium was probably about as bad on non-x86 Unix platforms as the dotcom bubble.
Why bother investing in and updating software for an alternative platform when not even the manufacturer wants to support it when Intel finally came out with their magical 64-bit architecture?
VLIW/EPIC depends heavily on the compiler being smart enough to handle basically everything. Compilers were not smart enough and being a completely alien architecture excluded it from being used with existing software(outside emulation which resulted in horrific performance penalties)
VLIW is cool stuff but I have no idea if we'll ever see it being used.
>>54497959
>high power consumption, high heat output
That is the same thing
>>54499013
Itanic was pre-Compaq merger.
Alpha was the fucking shit back in the 90s
>>54500619
vliw was all the rage in the 80s and it got memed out of proportions that even intel hooked that bait
>>54500637
A little bit before, you're right, but HP did take over Alpha production and were ultimately the ones that killed it in favor of Itanium products.
