>Didn't support x86 software>Less power efficient than ARM, PowerPC and even x86>Constantly ran hot>Cache misses and branch prediction failures were more costly than on NetBurst>Overengineered and too complex for anyone to want to consider writing software for (needing to manually schedule the hardware pipeline and attempt to predict instructions before running them)Why?
>>109409130>64-bitIt ticked a box.
>Couldn't handle standard serial code efficiently>Used way more juice than standard PowerPC and x86 architectures>Generated enough heat to make the early PS3 sound like a jet engine>SPU pipeline stalls were more punishing than standard CPU cache misses>Overengineered and too complex for third-party studios to want to consider writing software for (forcing programmers to manually write the compiler scheduling and loop unrolling loops just to get parity with the Xbox 360)Why didn't they learn from Intel?
>>109409166But Sony did learn from the PS3 and they switched to a mostly standard x86 SoC for the PS4? Meanwhile Intel kept trying to push Itanium until 2019. Though, it might have been more on sunk cost fallacy, than anything..
>>109409130it was major success for intel, they managed to convince almost every independent cpu maker to stop investing into their own risc-based silicon and effectively killed them all in the end
>>109409187This! Even if it was a money sink it didn't matter if the competition was killed. Having competitors is the bigger risk. Eliminate them and you rule for decades.
>>109409166xbox360 had the exact same shitty PPC core as the Cell did. So two consoles ended up with the exact same worthless core.
>>109409130>Intel could close down the platform again and sell crap rambus memory to go alongside itit ticked a box.
>>109409130It didn't have any of the security problems which have since bitten every out of order RISC / CISC CPU ever made. The guys making this shit know it can never be reliable.>>109409650Also this.
>>109409187they even got HP to hold the bags.
>>109409422ya but xbox didn't go full retard like Sony did. they just treated it like another shitty netburst, not as some next level SIMD super CPU of the future.
>>109410528Referring to it as a "shitty netburst" is an insult to netburst. PPE is way worse than netburst. An in-order cpu that has a longer pipeline than out of order cpus.
>>109409187Did it really though? They bought DEC, POWER remained the best funded CPU in the world, ARM and SPARC persisted. MIPS died of mismanagement more than anything else I think, but Itanium could've had something to do with it. Access to top tier wafers seems to have been the biggest issue. Remember this was before everything outside Intel folded into TSMC.
Later iterations of Itanic were good, but by that time it was too late. AMD already had Opteron/Athlon 64 and x86 won again.
>>109410589It was trying to be faster than $100k POWER CPUs not x86 though.
>>109409130it can scale cache better than x86/64
>>109410565>MIPS died of mismanagement more than anything else I think, but Itanium could've had something to do with itItanium directly killed it, MIPS was a Silicon Graphics in-house business at the time and the siren song of Itanium led them to immediately shitcan all big iron MIPS research and divest the entire unit to whoever was willing to pay for it. This backfired almost immediately and would later play a major role in killing the entire goddamn company, they had to go crawling back to MIPS but ended up just reheating the same design over and over again instead of ever making anything truly new again.
>>109409130>Constantly ran hotDue to emulating x86 in hardware>Less power efficient than ARM, PowerPC and even x86Bullshit>Cache misses and branch prediction failures were more costly than on NetBurstBecause nobody wrote predicates and everyone just used more branches.>Overengineered and too complex for anyone to want to consider writing software for (needing to manually schedule the hardware pipeline and attempt to predict instructions before running them)Seems like a skill issue.Itanium was the GOAT. It was everything a systems programmer could wish for. Effortless hardware parallelism, easy branchless code, 128GPRs, register rotation engine.
>>109409130all false after itanium2
>>109410565>They bought DECNot Intel. Compaq did prior to the HP merger, and whilst Intel struck a deal with HP for Itanium, whatever parts of DEC are left are still in HP's hands. (HP Enterprise is its name nowadays)
>>109412118Itanium really was the CPU for people who understand CPU. Which almost nobody does. Cursed marriage of Soviet paternalist design and American rent seeking.
>>109409130>Why?It was a different era, one in which new things could still be tried.
