"hence the stunning latest benchmarks."
IBM's recent stunning benchmark have very little to do with the core. It is the memory architecture that is outstanding.
"Cheap knock off? EMT64"
If you have some kind of knowledge on this, you would know that Intel has spent a lot of research on extending x86 to 64-bit during many years. Research has been done, but a release has not been made for other reasons. Now, they had to change their implementation a bit to match AMD.
The Itanium effort has not hampered Intels x86 line as much as you think. They have invested more money on x86 than IA64.
Yes, they delayed their introduction of 64-bit in their x86 line, but that is about it. Also, much of the money which has been put on Itanium is of benefit even for the x86 line.
The "problem" Intel have is that they can't compete in memory latency with AMD currently.
Yes, I agree. I don't understand Slashdot people. I thought they were geeks and were interested in computing. Now it looks like many want ONE OS and ONE architecture. Not a fun future. I on the other hand want many OS:s and _many_ archs. It is a _lot_ more fun to have more archs to code.
Itanium cores are actually quite small. And we have no real small cache Itanium to compare with, since they (the current small cache versions) just have disabled cache and thus very low associativity.
An Itanium design from the start with small caches would be very different from the ones you see today.
", and no one is going down the EPIC path."
That is just plain wrong.
Do you know how much time compiler writers use to schedule instructions on x86 and RISC? Very much and their time on this increase...
"1) The MIPS architecture dead-ended. It couldnt scale far enough to keep up with x86."
Sorry, you are very wrong here. The MIPS arch is not bad, but it was the implementations that could not compete with x86 implementations.
Slashdot Mirror
A Itanium with about the same performance of a current MIPS draw about the same amount of money.
The IA64 design is not broken. With a future good low latency memory architecture it will yield very good results.
"hence the stunning latest benchmarks." IBM's recent stunning benchmark have very little to do with the core. It is the memory architecture that is outstanding.
"Cheap knock off? EMT64" If you have some kind of knowledge on this, you would know that Intel has spent a lot of research on extending x86 to 64-bit during many years. Research has been done, but a release has not been made for other reasons. Now, they had to change their implementation a bit to match AMD.
The Itanium effort has not hampered Intels x86 line as much as you think. They have invested more money on x86 than IA64.
Yes, they delayed their introduction of 64-bit in their x86 line, but that is about it. Also, much of the money which has been put on Itanium is of benefit even for the x86 line.
The "problem" Intel have is that they can't compete in memory latency with AMD currently.
Correct. (although the linux world is not tiny) ;)
Yes, I agree. I don't understand Slashdot people. I thought they were geeks and were interested in computing. Now it looks like many want ONE OS and ONE architecture. Not a fun future. I on the other hand want many OS:s and _many_ archs. It is a _lot_ more fun to have more archs to code.
Itanium cores are actually quite small. And we have no real small cache Itanium to compare with, since they (the current small cache versions) just have disabled cache and thus very low associativity. An Itanium design from the start with small caches would be very different from the ones you see today.
", and no one is going down the EPIC path." That is just plain wrong. Do you know how much time compiler writers use to schedule instructions on x86 and RISC? Very much and their time on this increase...
Interesting... a new OS? I know about the security features in IA64 which are very good.
"1) The MIPS architecture dead-ended. It couldnt scale far enough to keep up with x86." Sorry, you are very wrong here. The MIPS arch is not bad, but it was the implementations that could not compete with x86 implementations.
No, they haven't. They are actually making good progress with their Altix boxes and have some very interesting future plans (UV project etc.)
Do you expect Slashdot people to know what Tukwila is? ;)
BTW. I think the important is the improvements of the system architecture.