Madison IA64 Released -- Fastest CPU Ever - Page 3

Asher · July 1, 2003, 20:46

The difference being, I have some idea what I'm talking about.

panag obviously even has trouble reading, let alone understanding what he's saying.

He's regurgitating marketing bullshit from Apple, from years ago mind you, thinking no one knows enough to call him on it. That's probably the case, in the Mac forums he hangs out in.

But here in the real world, people will probably correct you.

monolith94 · July 1, 2003, 20:50

OK, so apple has an L2 cache in the G5, at 'the norm' of 512KB... that's not TERRIBLE. You made it sound as if the G5 had no cache at all...

Asher · July 1, 2003, 20:50

Quote:

Originally posted by monolith94
OK, so apple has an L2 cache in the G5, at 'the norm' of 512KB... that's not TERRIBLE. You made it sound as if the G5 had no cache at all...

The G5 has zero L3 cache.
It has 512KB of L2 cache, at a time when Intel/AMD are moving to 1MB.

512KB isn't bad, it's just half of what it should be.

monolith94 · July 1, 2003, 20:58

well, hell, I want 60GB L3 cache now!

Wraith · July 1, 2003, 21:03

--"IBM has announced 8+-way Opteron servers?"

Duals, for now. AMD only just launched the 4+ line, though, and IBM tends to be conservative about that sort of thing.
http://www.infoworld.com/article/03/...html?platforms

Newisys has announced a 32 way Opteron server, and they're mostly made of up of ex-IBM people.

Not to mention Cray's little order of 10,000 Opterons for Red Storm.

--"So it's rather convenient, then, that all the motherboards chose 16GB as a limit?"

You do realize that not many chipsets are out yet, right? The stuff that's out is all pretty much focused on the single or dual workstation/server market. There's only so much RAM you can cram on one of those boards, especially if you're after high speed and ECC.

--"What's the address bits on the Opteron? 42?"

40 (I think) in the first gen, but it was designed so they can increase it later if necessary.

Wraith
A book a day - that's all we ask

Mordoch · July 1, 2003, 21:27

Quote:

Originally posted by Asher

It's not just that, but the Opteron can only work in systems with up to 8 CPUs (not to mention that only 3 vendors support 4+ CPUs for Opteron, and I've never heard of any of them, they're small-time Texas companies).

Opteron also can only address 16GB of data. Which is rather small for a 64-bit chip...

On point one, this is wrong, the cpu's can easily be linked together in numbers over 8. The 10,000 chip Redstorm from Cray is a good example. The Opteron is far more effectively scalable than the The reasons you 16gb of ram claim are bogus have already been covered. Why don't you just make things up, your current claims are equally accurate.

Asher · July 1, 2003, 21:30

Quote:

Originally posted by Mordoch
On point one, this is wrong, the cpu's can easily be linked together in numbers over 8. The 10,000 chip Redstorm from Cray is a good example. The Opteron is far more effectively scalable than the The reasons you 16gb of ram claim are bogus have already been covered. Why don't you just make things up, your current claims are equally accurate.

How are the 16GB RAM claims bogus when there's not a chipset out there that will address more? A system is only as strong as its weakest link.

And as for the linking them up in numbers over 8, isn't that clustering?

Wraith · July 1, 2003, 22:07

--"How are the 16GB RAM claims bogus when there's not a chipset out there that will address more?"

How much could you put in the systems that were out at Itanium launch? I mean, if you want to keep this argument up, I'd be happy to take the other side and restrict the Itanium side of things to what was available at the original launch...

--"And as for the linking them up in numbers over 8, isn't that clustering?"

Depends how it's done. If you've got seperate boxes of 8 or less that are linked by some kind of network, it'd be a cluster. It's a cheaper way to go about it, but isn't quite as efficient as putting everything in the same box (ie. on the same switching fabric/backplane/what have you).
Newisys isn't taking that route, at any rate. They're doing the HT switch it'll take to link 32 processors on the same board.

Wraith
I was born weird-- this terrible compulsion to behave normally is the result of childhood trauma

Asher · July 1, 2003, 22:10

Quote:

Originally posted by Wraith
How much could you put in the systems that were out at Itanium launch? I mean, if you want to keep this argument up, I'd be happy to take the other side and restrict the Itanium side of things to what was available at the original launch...

RAM densities were lower then.

Quote:

Newisys isn't taking that route, at any rate. They're doing the HT switch it'll take to link 32 processors on the same board.

Cool.

Panag · July 1, 2003, 23:56

Quote:

Originally posted by Asher

You do realize that that's certainly not a G5 only restriction.

You're like an Apple marketing droid.

hi ,

thanks microsoft droid

bye

monolith94 · July 2, 2003, 00:01

Oooooh he didn't wish you a nice day...

You've pissed him off now!

Drake Tungsten · July 2, 2003, 00:02

Asher · July 2, 2003, 00:03

Is that a first for him?

monolith94 · July 2, 2003, 00:05

I think so...

Yes I'm pretty sure...

...have a nice day...

Asher · July 2, 2003, 00:07

I feel so special now.

Panag · July 2, 2003, 00:14

Quote:

Originally posted by monolith94
Oooooh he didn't wish you a nice day...

You've pissed him off now!

hi ,

who pissed who off , ....

maybe its time for something else , .......

adios

Panag · July 2, 2003, 00:16

Quote:

Originally posted by Asher
I feel so special now.

hi ,

you should , all them dryboned grumphy young men get a nice day , ..... you get something special , .....

see ya

Asher · July 2, 2003, 00:24

Quote:

Originally posted by panag
hi ,

you should , all them dryboned grumphy young men get a nice day , ..... you get something special , .....

see ya

Oh panag, I didn't know!

Skanky Burns · July 2, 2003, 02:21

Cruddy · July 2, 2003, 03:08

When's the wedding? Is Apolyton invited?

Seriously, I can't see much point comparing IA64 and G5. They're aimed at totally different markets (and no, I'm not about to start slinging mud on either side).

As I recall, part of the deal with IA64 was that the actual language compilers would need rewritten to deal with the new coding conventions... it's takes time to bug test these, and then you can start writing/recompiling the IA64 apps. Not that this is new, it's been going on for a while.

To be honest, I couldn't give a stuff what servers use, be it Opteron, IA64, DEC or a giant mouse. So long as it gives better Internet bandwidth at reduced cost, who cares?

Panag · July 2, 2003, 03:15

Quote:

Originally posted by Cruddy
When's the wedding? Is Apolyton invited?

Seriously, I can't see much point comparing IA64 and G5. They're aimed at totally different markets (and no, I'm not about to start slinging mud on either side).

As I recall, part of the deal with IA64 was that the actual language compilers would need rewritten to deal with the new coding conventions... it's takes time to bug test these, and then you can start writing/recompiling the IA64 apps. Not that this is new, it's been going on for a while.

To be honest, I couldn't give a stuff what servers use, be it Opteron, IA64, DEC or a giant mouse. So long as it gives better Internet bandwidth at reduced cost, who cares?

hi ,

héy , ...... we are just good friends okay

well the g5 is in stores now , with its 8 gig or ram , ......

and that ia64 not yet , .....

and as for your last line , thats exactly why we have macs , .....

have a nice day

Atahualpa · July 2, 2003, 04:21

Quote:

Originally posted by Asher

It's not that 64-bit is much different than 32-bit, it's that VLIW is completely and totally different than x86...

Algorithms that worked efficiently on x86 may run like crap on VLIW, and vice versa.

There's a lot more work to getting programs working on a whole new architecture aside from changing the compiler.

And in modern terms, C/C++/Fortran are certainly not HLL. HLL are C#, Java, etc.

C/C++/Fortran are mid-level.

Asher man grow up an Algorithm is an Algorithm is an Algorithm. The compiler does the reordering stuff and not the HLL.
And besides C# is C++ with some "syntactic sugar" as our Lecturer constantly called the features of C# (delegates, properties, ...) and he explained how you could do that in Java as well.

Asher · July 2, 2003, 09:14

Quote:

Originally posted by Atahualpa
Asher man grow up an Algorithm is an Algorithm is an Algorithm. The compiler does the reordering stuff and not the HLL.

That's only part of it.

If it was truly that easy, why can't people just take regular C++ code for x86 and compile it and magically have IA-64 code?

Quote:

And besides C# is C++ with some "syntactic sugar" as our Lecturer constantly called the features of C# (delegates, properties, ...) and he explained how you could do that in Java as well.

C# is more than C++ with some "syntactic sugar", it runs through a virtual machine (hence it is true HLL, not mid-level like C++), has stuff like garbage collection, and is truly object-oriented (no primitive types ala C++)

Asher · July 2, 2003, 09:15

Quote:

Originally posted by panag
hi ,

héy , ...... we are just good friends okay

well the g5 is in stores now , with its 8 gig or ram , ......;

The G5 doesn't go for sale for months...

Quote:

and that ia64 not yet , .....

IA64 chips have been for sale for years, and Madison is for sale right now.

Atahualpa · July 2, 2003, 09:18

definition of a HLL includes "runs through a virtual machine"

HLL != everything is object oriented

Asher · July 2, 2003, 09:21

Quote:

Originally posted by Atahualpa

definition of a HLL includes "runs through a virtual machine"

It's true, that's a true high-level language.

C/C++ are both mid-level.

Quote:

HLL != everything is object oriented

Um. No kidding.
It's a HLL because it runs through a virtual machine. It truly doesn't care about which architecture it's on, and C++/C certainly does care what architecture it's on.

Atahualpa · July 2, 2003, 14:02

Maybe you mix 4th Generation Language and HLL

monolith94 · July 2, 2003, 14:22

"well the g5 is in stores now , with its 8 gig or ram , ......;"
"The G5 doesn't go for sale for months..."

The truth is, as usual, somewhere in between the extremes - the G5 is slated to come out late next month, although delays are still possible (not probable).

Asher · July 2, 2003, 14:40

Quote:

Originally posted by Atahualpa
Maybe you mix 4th Generation Language and HLL

No, maybe you don't understand what a HLL is?

A HLL is completely abstracted from the hardware beneath this.

Quite obviously, C++ doesn't count, because you can't simply switch the compiler and have it work on, say, IA-64.

However, with a language like Java or C#, this is entirely the case. Because it runs through an interpreter, it is truly a HLL.

C and C++ are not HLL, they used to be regarded as such a long time ago because, compared to machine code and assembly, it was "high" level. But it's not high anymore, it's just mid-level.

High-level languages are completely abstracted from the hardware, like Java/C#.

Panag · July 2, 2003, 15:42

Quote:

Originally posted by Asher

The G5 doesn't go for sale for months...

IA64 chips have been for sale for years, and Madison is for sale right now.

hi ,

it seems you can order one true the apple store , .....

have a nice one