Performance Predictions

[Wombat]

I'm very much in 64-bit country. You've got a while to wait before your home machine is a (true) 64-bit architecture. Don't forget that the 486 was faster and cheaper than the Pentiums for quite a while. It all depends on what you're using them for.

[rugger]

Bzzzt

Wrong answer Midnight

The correct answer is: While 64bit cpus may improve the performance of some algorithms (the same way AltaVec, SSE and MMX improve those algorithms today), most algorithms and programs will not be strongly affected by 64bit CPUs. 32-bit CPUs are already expressive enough for the vast array of normal tasks done on computers. In fact, programs recompiled for 64bit cpus may be larger and slower than their 32bit companions due to extra memory space required to store the instructions and data, and the associated crowding out effect that has on the CPU caches.

[midnight]

Re: Bzzzt

Originally posted by rugger
Wrong answer Midnight

The correct answer is: While 64bit cpus may improve the performance of some algorithms (the same way AltaVec, SSE and MMX improve those algorithms today), most algorithms and programs will not be strongly affected by 64bit CPUs. 32-bit CPUs are already expressive enough for the vast array of normal tasks done on computers. In fact, programs recompiled for 64bit cpus may be larger and slower than their 32bit companions due to extra memory space required to store the instructions and data, and the associated crowding out effect that has on the CPU caches.

Remember that the 386 was subpar but the 486 was trully a leap forward only a short while later. Also the whole 3.1/16 bit to 95/sorta 32 bit paved the way for new bus, memory, and cache designs. With Intel rumored to be able to put 6+ megs on die with their .1 process and AMD with alpha, mips, and other 64 bit salties on board it is no stretch to think the hardware end is ready to move forward. We have only seen what 64 bit can do for the niche markets like cad visualization, datacenters and war rooms. With ubiquity of application comes the ability to introduce more r&d into the problems you mention.<p>
I think the next specialized set of instruction either on processor or on vid card is going to be for a physics engine. Didn't microsoft allude to something in the upcoming directx standards, I may be wrong.

[rugger]

Not really,

You are confusing the word size of a processor with the archetectural design of a processor. The far better "design" of the 486, rather than its wordsize, is what fed its performance over eariler 16-bit only designs.

I don't understand most of your paragraph though. I simply don't see (being a programmer myself) how general 64-bit processing is able to help most programs. 64-bit processing only helps in problems where:

1) You need to perform calculations on massive numbers. (some encryption algorithms like RSA)
2) You need a particular program to address more 4 gig of memory. (32 bit programs can access more than 4 gig, but not very effeciently, think EMS on old computers)

Otherwise, you are just wasting memory by having longer code (each instruction my require more space, especially on strict RISC machines) and bigger data sizes. (most processors don't like unaligned accesses, so you MUST use more space) Larger programs cause more cache contention, reducing the effiency of caches.

Not that moving to 64bit won't provide any benefits for the end user. I am simply suggesting that the move to 64bit won't change the world, as opposed to the change to 32bit, which brought many other features (real memory protection, better archetectured CPU's, real operating systems)

[Wombat]

It certainly can change the computing world. The real benefit comes when you stop programming in x86-land, and start using something that's actually designed for the space.

[rugger]

Originally posted by Wombat
It certainly can change the computing world. The real benefit comes when you stop programming in x86-land, and start using something that's actually designed for the space.

Please elaborate? Why can changing to 64bit processors (or changing CPU instruction sets) change the computing world?

Everybody buys this hype, yet nobody supplies real reasons.

[superfly]

Originally posted by Wombat

and really isn't that fast anyway.

right...If you keep repeating that enough times,you might end up believing it yourself.....


Greebe...

Not to worry,....I'll still get a little more out of it yet and even if there's still some that are faster clocked still....Well,such is life....



And as far as 64 bit computing goes,it's main advantage(at least right now) is the amount od addressable memory space it allows(4 TB i believe),since 32 bit processors(execpt for Xeons) can only address a max of 4 gigs of memory....


Does anyone here think they'll need more that 4 gigs anytime soon?...

[Electric Amish]

I predict that it will kick ass and chew bubble gum, but will be all out of bubble gum.

amish

[ravalox]

Originally posted by superfly

right...If you keep repeating that enough times,you might end up believing it yourself....

I did some guesstimations about your query last night, and I came up with under 18K. Remember, AMD is still superior to Intel for 3D gaming.

[TdB]

i always thought the 4 gigs was an OS limitation, afaik newer version of OS/2 could address 64 gigs on a pentium2, I think OS/2 used both paging and segmentaion for each process, as opposed to windows and unix-clones that could only handle 1 segment per process, thus the 2 gigs limit for a process(and another 2 gigs for the OS).

or maybe im just confusing myself.

[rugger]

Originally posted by superfly


And as far as 64 bit computing goes,it's main advantage(at least right now) is the amount od addressable memory space it allows(4 TB i believe),since 32 bit processors(execpt for Xeons) can only address a max of 4 gigs of memory....


Does anyone here think they'll need more that 4 gigs anytime soon?...

Actually, in segmented 32-bit mode, a process on an x86 can access up to 48gig of memory space (4 code, 4 stack, 32 data). The operating system itself however, can change the segment registers and access a LOT more. However, most 32bit x86 OS's use a flat memory model (all segments are set the same) allowing processes to only access 4 gig.

While I beleive we will need 64bit processors within the next 2-5 years for the memory access problem, it won't result in a revolution in operating systems and programs that the introduction of protected memory 32-bit processors did.

[superfly]

Originally posted by Electric Amish
I predict that it will kick ass and chew bubble gum, but will be all out of bubble gum.

amish

Now where does that quote come from?....Duke perhaps?....



Which reminds me....Where's DNF?...

[superfly]

Originally posted by ravalox


I did some guesstimations about your query last night, and I came up with under 18K. Remember, AMD is still superior to Intel for 3D gaming.

Welli guess i'll have to bust out some extra cooling gear and use on the card to "motivate" it....

[rugger]

Duke nukem forever.

I beleive it is waiting release with the G800.

Any day now

[yakko]

Originally posted by rugger
Duke nukem forever.

That would be hilarious - Parhelia shipping with DNF (Nope, not seriously expecting it)