Intel's Itanium (64 bit CPU) has an official "permanent" release date of March 1st 2001. However. it's useless without a 64 bit OS, such as Windows 2000 64 bit. Currently, M$ has not released a release date for Windows 2000 64 bit. There are versions of Linux on the web that are 64 bit though- for you Linux nuts.
well, intel's 64 bit itanium is due next year, so is the AMD's "hammer"! apart from cost issues on which nobody needs to say anything, both will be tremendous in performance, complexity and sheer power. remember how 386 was to 286, when the world moved to 32 bit from 16 bit? this time its much more complex and much more is at stake.
we will have a brand new architechture (maybe still crippled with x86 legacy), 64 bit integers, more instructions, maybe longer and more accurate floating point math, and a whole lot of code to port!
but what has the cpu to do if the os or any app cannot take advantage of its 64 "bitness"? a lot of code is being ported to 64 bit, especially in linux, and M$ is working on win64 as well.
64 bit chips might run "older" 32 bit code, but might not be as fast as the "original" 32 bit ones. remember 16 bit code suffers on most 32 bit cpus. it will suffer still more on 64 bit ones!
for a programmer it would be more and larger registers, larger address space, and more instructions. and for the first time we may have a close competition between the first 64 bit x86 processors, itanium and hammer!!! alpha and some others already have 64 bit procesors.
It's quite simple really, the main reason to go from 32-bit computing to 64-bit computing is to gain the ability to address more than 4GB of RAM. You may ask why anyone would want to address that much memory when the average computer sold today has 46-128 MB of RAM. The question comes down to servers. If you look at the likes of Amazon who get millions of hits a day, and store details of the browsing/buying of all the people who visit, that's a lot of data to have ready for access. When you go in to buy a new book/CD etc. you don't want to wait for Amazon's servers to dredge up your data from a dirty big file somewhere on disk, you want Amazon to get your details from storage and have it readily available for your entire visit. That kind of data throughput, especially when you are talking about thousands of people logged on at the same time requires substantially more than a measely 4GB. That's where 64-bits comes in handy; you are able to address 16 Exa Bytes (that's approx. 16 000 000 000 000 000 000 bytes), though most servers will have a top address of around 512GB physical RAM because of motherboard/cost constraints. That's a lot of memory! Of course, that won't be all physical RAM, a lot of it will be implemented as virtual RAM (remember, with 32-bit computing you simply can't address above 4GB, even with virtual RAM). If you look at any of the servers for the big e-companies you'll certainly not find "Intel inside", they all use 64-bit unix servers with the likes of Sun's 64-bit UltraSPARC processors, Digital's (now owned by Compaq) 64-bit Alphas, and Silicon Graphics' Mips based 64-bit servers. 64-bit computing has been around for quite a long time (relatively!), just we haven't got it yet because of the lack of a consumer 64-bit OS. Look how long it took for Microsoft to get a good 32-bit OS (windows 2000), we had decent 32-bit processors in the 80's but we had to wait until just reasently to get a truely reliable and versatile OS from Microsoft. (That's why you only really see Pentium pros in business machines; you could only really run NT4 on them.
Other benefits are 64-bit Integer calculations, 64-bit single and 128-bit double precision floating point calculations (that's damned accurate folks!)
If you want to know more about some of the 64-bit processors that have been floating about, try:
<A HREF="http://www.sun.com/microelectronics/UltraSPARC/" target="_new">http://www.sun.com/microelectronics/UltraSPARC/</A>
<A HREF="http://www.mips.com/products/s2p2.html" target="_new">http://www.mips.com/products/s2p2.html</A> (the MIPS architecture used in Silicon Graphics servers)
<A HREF="http://www.support.compaq.com/alpha-tools/documentation/current/chip-docs.html" target="_new">http://www.support.compaq.com/alpha-tools/documentation/current/chip-docs.html</A> (this is the best I could manage for the alphas)
<A HREF="http://www.sgi.com/servers/" target="_new">http://www.sgi.com/servers/</A> (this has a great wee table with info of the specs. for Silicon graphics servers, note the maximum memory on the high end Mips servers)
Here's some links to some of the up and coming new 62-bit architectures:
<A HREF="http://www.intel.com/eBusiness/products/ia64/index.htm?iid=prodinfo2+bizprocitp&" target="_new">http://www.intel.com/eBusiness/products/ia64/index.htm?iid=prodinfo2+bizprocitp&</A> (Intel's Itanium processor that'll be using the new IA-64, I wonder if they'll ever get this processor released! It was origonally set to be released a year ago!)
<A HREF="http://www.amd.com/products/cpg/64bit/overview.html" target="_new">http://www.amd.com/products/cpg/64bit/overview.html</A> (AMD's 64-bit extension to Intel's x86 architecture, Hammer)
AMD's hammer chips will not run 32-bit code slow at all. It doesnt need to emulate X86 at all as it's a 64-bit extension of X86. Also, it will be pretty inexpensive with a price comparable to the Durons (Clawhammer) and Athlons (sledgehammer). Now everyone will be able to have a 64-bit machine at home Thanks AMD
64-bit processors will not be out until mid to late next year. there is no point in it just yet, seeing in that we don't have any software that will run 64-bit. The software implementations are endless, but the support just isn't there yet.
"upgrading is no longer an option...it's a necessity"
Visit www.elitehunters.com --SoulReaper =)
Great links.. the old assembly programmer in me is silently drooling over all the new registers. LOL. Down at that level, more registers = more speed. Even more so than cache, because cache you still have some latency. At least it is when software is written to take advantage of it.
"The answer is not in your hair."
"I'd rather jump in the lava than be fragged by you."
i have a stromg belief that assembly language is the BEST. i even do many of my windows apps in assembly, but these days its difficult to do so, due to time constraints et al.
now, did you hear that itanium (slated to be the first x86 64 bit processor) will have 128 GPRs? thats general purpose registers. it was fun juggling with 14-18 16 & 32 bit registers, but ONE HUNDRED AND TWENTY EIGHT? its too much for a assembly language programmer! i hope its the 68k way, all true general purpose registers, dedicated address regs and like. hope registers do not contend for the "execution chamber" like it is in P6 architecture. hope now that the compiler does well to use all of them.
If we people still used assembler, we would not need high mhz parts... It's funny we keep creating CPUS designed to run some of the worst code on the planet such as code generated by a C compiler.
As for the 64 bit thing, even with an optimized OS, the reality is that the average user here is going to be lucky to see a 10% performance increase from it
Reasons:
1 > formost / The law of dimmenising returns, data is rarly processed in 64 bit sizes, there is still lots of 8 & 16 bit operations going on, in most cases, the extra 32 bits will go wasted on address ranges that none of use will be using anytime soon, as well the extra 32 bit data sizes will rarly get used as 4 billion(32 bit) states usually handles mosts tasks just fine
2 > the 64 bit memory model will not benifit any average user for a long time, probabbly another 15 years before a user would need such a memory range, including virtual memory, If memory serves me right, the P2/3 can handle some 64 terabytes of virtual+physical memory already, and in fact the memory bus is 36 bits wide, not 32.
The biggest advantage that the next cpu's could present to average users is more registers ( as some here have pointed out already ), more data registers, more address registers, ect would provide much more performance than moving to the 64 bit memory/data model
My somewhat negitive post on 64 bit is some what mute though as it looks like the 64 bit units coming will indeed have the extra registers, but I will bet any performance increase seen with them will be due to increased register size, not the 64
With 256-bit graphics processors, doesn't it seem kind of silly that we have had to wait this long for 64-bit CPU's? Would we really NEED GHZ-level processors if we already HAD 64-bit processors?
Software will catch up with hardware. If designers knew otherwise, we would have GHZ level 386's instead of PIII's and Athlons. If the original Pentium would have been 64 bit, as it should have been, we would be happy with the performance of a PIII 500 at 64-bi'ts instead of needing one at 1GHZ at 32-bits. So why couldnt designers skip 64-bit and go directly to 128-bit? Couldn't they just use a compiler to divide by 4 to run 32-bit applications?
hoho~~, IMO, speed is not everything. to average users (actually, not just average users), legacy software/hardware support and compatibility and price/performance ratio both come ahead of speed.
if a 32bit CPU runs 90% of current software applications and is compatible with 90% of current hardware components in the world, will an average consumer buy a 64bit CPU that can only live with the rest 10% but is superfast? no.
if a 1GHz 32bit CPU runs about the same speed of, say, a 400MHz 64bit CPU but costs only 1/3 of the latter, will an average consumer buy the latter CPU for its superior design? no.
technology is not for technology's sake. if we even haven't depleted the resources in a 32bit architecture, is there any reason for us to skip directly to 128bit?
-----------------------------
Some are ignorantly happy,
While some, happily ignorant.
64bit alone will not give you much a performance boost compared to 32bit. Only when you need more than 4GB of memory does the bit size really matter.
Just compare a 4CPU 32bit machine and a 64bit Alpha with 2GB of memory running say SAP/Oracle with the same number of users you will not see any difference other than the price of the box (32bit generally cheaper) however as the number of users and transactions increases and more memory is required only then does then 64bit comes into its own.
performance on 64 bit processors would come only with carefully optimised software. increasing number of registers and widening them will be helpful - for programs those use them. more registers mean larger scratch area, at register speed! larger registers mean faster calculations. newer architechture means better throughput of both instructions and data. but what about current applications that dont know 64 bit?
these wont run any faster on given clock speed, only a wee bit better due to architectural improvements.
more memory will only mean the operating system wasting it away and performing even badly. btw who wants more than 64 TB that 32bit chips already support? 4 GB physical memory capacity itself isnt used up well, why ask for more?
The majority of users will not run a 64 bit processor for at least 5 years, and that's a very libral estimation. (IF it is pushed like the 386, IF the cost is right, IF mainstream hardware and software designers take advantage if it) The largest impact will most likely be in large corporate systems. With Internet 2 coming along, my guess is we will see these 64 bit monsters become the backbone of many large directory sites and file servers. (Digital will have them, probably AOL and MS. Of course, most of these already have Alpha Servers driving them.)
The other group which will probably see an advantage are educational and research facilities. People that have a need to run very large simulations (such as complex gravametric formlae, simulations of the big bang and the like) will benifit from a (hopefully) cheaper solution than the current favored Alpha chips. Also, as these organizations generally prefere Unix builds, and use almost exclusivly in house software, they will not need to rely on main stream support.
One great advantage that I see is the ability to access a terabyte of main memory. This will certainly come in handy for the newest VR simulators at popular theme parks, and perhaps sooner than later, the driving force behind some of the hottest arcade games to hit the market.
64 bit is a long way from mainstream production. I believe it WILL find a quick acceptance from some larger companies, as well as provide a lower cost solution for smaller organizations. (My university spent 10 million on an Aplha Server with 10 GB of RAM, used mainly for theoretical programming and large scale engineering fault tolerance simulation)
One thing I can promise you. Even with cheap ram, you had better have pretty deep pockets to purchase one of these babies. Heh. Tom will probably test one the day it comes out.
Me, I'm holding out for ClawHammer and SledgeHammer rather than Itanium. There are a few problems with 64-bit hardware, however, whatever platform you eventually go for. For example:
1. PCI Bus is 32-bit in desktop machines, and so all cards are too. To take full advantage of a 64-bit processor, 64-bit PCI will be needed, and manufacturers will start having to produce 64-bit cards in bulk (not just for servers) to take full advantage of the processor.
2. We might need a new memory archatecture, or start plugging DIMMs in in two's (like ppl did with SIMMs before DIMMs hit the market).
3. New modes on the processors will need new thinking, like people had problems initially contending with real and protected mode on older processors.
4. Emulation of 32-bit software will probably be slower than running it on a 32-bit processor (except with Hammer processors which will be 32/64 bit x86 compatable)
Yeah, the good thing about the Hammer is it won't emulate 32-bit, it's supposed to be like 2 processors in one (1 for 32 and 1 for 64) or something. Sounds cool.
You are about to answer a thread that has been inactive for more than 6 months. If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.