Are We Not Ready For 64-Bit?

Q3vi1 writes "The Inquirer posted an intriguing article about how Intel doesn't think that we'll be ready for mainstream 64-bit computing until 2007. Coupled with the fact that MS isn't supporting the Opteron yet for their Windows 2003 Server, we may see a delay in consumer applications for 64-bit computing. However, as this article states, some people don't really care and will just go for Linux and AMD as a nice marriage."

Comments

Opposite speculation

[certron]

Here's something in German that you might want to run through: http://www.heise.de/ct/03/07/026/

Yes, go translate it, unless you can read and understand German, or just don't care to read it. :-)

4th paragraph under what babelfish translates as "Imbedding"

"Nevertheless one will not only be able to select to the planned Launch between different 64-Bit-Linuxen. Microsoft announced in the meantime, one day before the planned launching of a vessel, thus on 21 April to bring the Windows-XP-Server-2003-Version out for AMDs 64-Bit-Prozessor officially."

Looks like the story is still up in the air...

I fail to see what the big deal is...

[Rorschach1]

I've got two 64-bit machines at home, myself - an SGI Indigo2 and a DEC AlphaStation 200. Yeah, they're seriously out of date now, but they're still nice little workstations. *nix has been doing just fine on 64 bits for some time now. I do have to put up with all sorts of 'cast to pointer from integer of different size' warnings when I compile stuff, but I'm able to run 99% of the stuff I'd run on an x86 box on the Alpha.

Mandrake..

[Bush_man10]

I wanted to find some more information myself about Linux supporting 64-bit processors and this is what I found. Mandrake will have support by early 2003, I'm not sure if it's done now or not but it should be nearly done. Redhat is also offering support for the X86-64, check out the news release. Personally I think this is a great oppertunity for Linux to catch up to MS in market share. I look forward to upgrading...

Were we ready for 32bit in 80s?

[Anonymous Coward]

32bit 386 was launched in 1985, 32bit was Windows 95 launched in 1995. However, this didn't slow the sales of 386.

64bit won't be any different. You don't need 64bit to run a browser or a word processor, however 64bit will be useful for high-end PCs pretty soon. Epic is already working on 64bit port of UT2003, and many 3D applications use a lot of memory.

Re:Why would I want to move to 64 bit computing?

[larien]

It's actually fairly true. If you move to 64-bit computing, registers take up twice as much space, as do some instructions. As a result, you need more memory to cope with the increase in space required. It's not always a doubling; experiences with SPARC (it can run 32-bit & 64-bit natively on the same CPU with backward compatibility to 32-bit programs) show there is an increased memory requirement for running processes.

People might say that memory is cheap right now, but that's not the problem; the main limitation is the L2 cache; if the core of the process increases in size sufficiently to be larger than cache sizes, performance will suffer. This is partly why Intel is ramping up the L2 cache on Itanium 2; it needs it to keep performance up. The other reason is that it needs to compete with SPARC, Power-4 and PA-RISC in the server space which all have at least 4MB L2 cache, with 8MB being common. IIRC, newer PA-RISC CPUs have 32MB L2 cache (although they are dual-core, so it's really more like 16MB/CPU).

Fact is, most normal users aren't pushing the envelope of 32-bit computing yet, so consumers don't need 64-bit. It is desperately needed in scientific computing & servers where the 4GB hard limit is becoming a problem, but these are not "normal" users.

Personally, I'll go to 64-bit (well, other than the Ultra 30 I have) when it's a good idea for me to do so, either because I need the extra address space (unlikely in the short term; I'm hardly using my 768MB at the moment) or the price/performance is right.

Re:Well

[D4Vr4nt]

Actually.. Xeon 2.0 512K $359.00 Xeon 2.4 512K $384.00 CDN $ too..

Re:Well if history is any guide...

[coult]

2. Businesses arent going to buy iLamp servers: something with pretty colors but without RAID-5 isnt really a server, sorry.

Two words: Xserve RAID.

Re:neither has my grandmother. she also doesn't ca

[steven_r]

Windows Xp Pro can be set up to allow 3GB for the user address space. This gives some applications a little more breathing room, although it doesnt move the ceiling that far.

The main problem I'm coming across with the limited address space is fragmentation of the free space. You can easily get to a point where the is still large amount of free address space available but there are no spaces large enough to allocate the chunk of memory of the size you want.

See http://www.microsoft.com/hwdev/platform/server/PAE /PAEmem.asp for info on the /3GB switch.

Look up a combination of "LARGEADDRESSAWARE, boot.ini, bootcfg, /3GB" on msdn to find out more...

cheers,

Steve.

Floating Point

[Detritus]

Most floating point calculations are already done in 64-bit (double precision) or 80-bit (extended precision). 32-bit (single precision) floating point is of limited usefulness on commodity Intel hardware.

Re:neither has my grandmother. she also doesn't ca

[coats]

Consider, of course, that the average amount of RAM in use in deskop PCs 256MB. Few people go above 512MB, even for 3D rendering. And, very, very, very few people have 1GB.

Your numbers are a year out-of-date, I think.

My wife is an attorney (corporate health care law; if you want to buy an HMO, she's your girl!), and found that to get decent performance with her day-to-day work (given the characteristics of MS -- and adobe! -- bloatware), she needed to go with 1 GB RAM; 512K was not enough.

And for what I do (serious environmental modeling), having to live within the limits of dinky little 2 GB files is ridiculous. And we wind up doing the analysis on desktops instead of directly on the supercomputers... Thank God (or Linux, or Alan...) for Large File Summit support in the recent Linux releases!

Re:POWER4

[mmol_6453]

There's a really interesting article over at Ars Technica that discusses performance issues and what AMD does about them in their x86-64 architecture.

Re:microsoft have power of life an death over opte

[mitcharoni]

Microsoft was the primary developer of the x86-64 instruction set along with AMD. They basically told AMD "Here's what we want in a 64-bit instruction set" and AMD produced. They have a vested interest to see Opteron/Hammer succeed. They only developed a version of Windows for Itanium because they felt they had to, not because they wanted to. Itanium is just a weird fit with Microsoft.

Even if that were not the case, x86 is not going anyhere anytime soon. So Hammer will run anything Microsoft puts out just fine. Hammer will kick the crap out of ANYTHING Intel puts on the market and is doing so right now.

And Microsoft will be coming out with x86-64 versions of Windows, rest assured. That is not speculation.

If there's ever anything Microsoft understands, its "volume" and "price".

Re:Honestly....

[Anonymous Coward]

Don't forget, sir, that modern x86 processors use register renaming and already have several times as many registers as are visible to the programmer - something like 60 or 80 or something total in the Athlon; I don't remember the exact number. Because of this, x86 chips won't benefit nearly as much from having more registers as would RISC chips. 16 is probably going to be more than enough for most applications.

I do however agree that it would have been a good idea to support the extra registers in 32 bit mode. 8 byte pointers, with most of the top four bytes zeroed out in each one, are a big waste of space in most cases.

Classic Intel

[TheNetAvenger]

They will ride the 32bit horse as long as they can get away with it. This isn't a case of available technology, but what they milk out of the consumers.

I think AMD is taking the right road to the 64bit transition.

BTW, Don't for a second think Microsoft is not going to fully support the AMD 64bit platform. It is already there, just not officially released. So don't yell at Microsoft for crushing this technology.

Windows 2003 is already filled with AMD64bit binaries and once the AMD chips hit, there will be 64bit replacements for XP and add-ons to Windows 2003 Server to fully support it.

With an NDA, I cannot say anymore. :)