Short answer: Linux on x64 CPUs from Intel and AMD.
I worked with HP-UX on PA-RISC and IA-64 (aka Itanium), IBM AIX on Power and SUN SunOS/Solaris on Sparc from 1996 to 2005. All three vendors were touting the advantages of RISC over CISC, i.e. Intel x64 architecture. The reality was that Intel / AMD CPUs were being manufactured in huge volumes whilst IA-64, Power & Sparc were niche products with correspondingly high prices.
It is also interesting to note that HP had its proprietary RISC CPUs in the form PA-RISC family. But due to the enormous cost of designing and building CPUs they formed an partnership with Intel to co-develop the IA-64 family.
Of course, Intel was making many multiples more money from x64 family so they didn't put the same level of commitment into the IA-64 product line. This in turn meant that for any given performance level a multi-socket x64 board was price and performance competitive with the HP-UX based systems. Systems, e.g. Integrity series allowed for vPars and other partitioning tricks along with using FC connected SAN. We are talking multi-million $ systems. It was during this time that Google and Amazon were publicising their commodity hardware based massively parallel cluster architecture.
As well as the other answers here, an additional reason was that during the time from announcement to shipping silicon the performance of x86 had improved such that the Itanium chips were lackluster performers in comparison to a top end x86 being produced by Intel at the time Itanium shipped silicon.
Couple lackluster performance compared to x86 with a significantly higher price than the equivalent performance x86 CPU and you get an additional damper on excitement and corresponding sales.
Two things come to mind: poor backwards compatibility and reliance on compilers to emit optimally ordered code (inherent in the VLSI system philosophy), which they proved unable to achieve, resulting in comparatively poor price/performance. Eventually x86_64 came along and provided an alternative that provided backwards compatibility, horizontal (PC/server) universality, and most of the purported benefits without the hassle, and the Itanium came to grief.
The C compiler on HP-UX was an extra cost item and it wasn't cheap. Porting programs to HP-UX often required some tricky changes because the C compiler wasn't as mature as GCC on x-64 systems of the time.
I recall that once Oracle acquired Sun, they gave priority to getting latest releases of Oracle RDBMS and JVM running on Sparc before HP-UX and AIX. In the corporate space Oracle and JVM were pretty much the main game.
Ivan Godard discusses the various positives and negatives of IA-64 in his description of the Mill Computer spread across many videos. (If you really want to dig deep at a architectural level) https://millcomputing.com/docs/pipelining/
Not exactly answering the question, but was not IA-64 also called as Itanium architecture? The architecture which promised taking base clock frequency to 10GHz?
Short answer: Linux on x64 CPUs from Intel and AMD.
I worked with HP-UX on PA-RISC and IA-64 (aka Itanium), IBM AIX on Power and SUN SunOS/Solaris on Sparc from 1996 to 2005. All three vendors were touting the advantages of RISC over CISC, i.e. Intel x64 architecture. The reality was that Intel / AMD CPUs were being manufactured in huge volumes whilst IA-64, Power & Sparc were niche products with correspondingly high prices.
It is also interesting to note that HP had its proprietary RISC CPUs in the form PA-RISC family. But due to the enormous cost of designing and building CPUs they formed an partnership with Intel to co-develop the IA-64 family.
Of course, Intel was making many multiples more money from x64 family so they didn't put the same level of commitment into the IA-64 product line. This in turn meant that for any given performance level a multi-socket x64 board was price and performance competitive with the HP-UX based systems. Systems, e.g. Integrity series allowed for vPars and other partitioning tricks along with using FC connected SAN. We are talking multi-million $ systems. It was during this time that Google and Amazon were publicising their commodity hardware based massively parallel cluster architecture.
As well as the other answers here, an additional reason was that during the time from announcement to shipping silicon the performance of x86 had improved such that the Itanium chips were lackluster performers in comparison to a top end x86 being produced by Intel at the time Itanium shipped silicon.
Couple lackluster performance compared to x86 with a significantly higher price than the equivalent performance x86 CPU and you get an additional damper on excitement and corresponding sales.
Two things come to mind: poor backwards compatibility and reliance on compilers to emit optimally ordered code (inherent in the VLSI system philosophy), which they proved unable to achieve, resulting in comparatively poor price/performance. Eventually x86_64 came along and provided an alternative that provided backwards compatibility, horizontal (PC/server) universality, and most of the purported benefits without the hassle, and the Itanium came to grief.
The C compiler on HP-UX was an extra cost item and it wasn't cheap. Porting programs to HP-UX often required some tricky changes because the C compiler wasn't as mature as GCC on x-64 systems of the time.
I recall that once Oracle acquired Sun, they gave priority to getting latest releases of Oracle RDBMS and JVM running on Sparc before HP-UX and AIX. In the corporate space Oracle and JVM were pretty much the main game.
Ivan Godard discusses the various positives and negatives of IA-64 in his description of the Mill Computer spread across many videos. (If you really want to dig deep at a architectural level) https://millcomputing.com/docs/pipelining/
I really miss Mike Magee ripping on this architecture regularly, it was great:
https://www.theinquirer.net/inquirer/news/1039092/how-intels...
Not exactly answering the question, but was not IA-64 also called as Itanium architecture? The architecture which promised taking base clock frequency to 10GHz?
NetBurst was heading for 10 GHz; I don't know that Itanic ever made that promise.
yup