My next processor going to be 64 bit??

[akcorr]

I haven't been keeping up with the trends lately and was wondering if the next processor I buy should be a 64 bit processor like the ClawHammer or Intel's 64 bit chip? I was looking at the AMD roadmap and was really waiting to see what the barton is going to offer but it looks like the clawhammer(64bit) will be coming out a short time after that! Are the 64 bit flavors from AMD and Intel just for servers and not for the desktop or are we all eventually supposed to move to 64bit in the next two years?? Thanks!

 

[Athlon4all]

In short, yes, the Desktop ClawHammer CPU will have x86-64 support among other things, however, you will need OS support, and I know that MS recently announced that it would be supporting x86-64 in its OS'es, but I am not sure when that support will come. I would estimate that in XP SP1, MS will provide x86-64 support. I would not be that concerned about it really. Its not as big as people make it out to be. Intel's 64-bit CPU however wll not be in Desktops for a very long time, if ever.

 

[imgod2u]

Moreover, you'd need software that actually uses 64-bit integers.

 

[dullard]

It really depends. I'm going to assume you have no 64-bit programs, so you don't need 64-bit. Lets focus on the 32-bit programs that you do have and will be buying. Are you willing to pay double or triple for the processor just to get a 20% speed boost? If so, then I would get the 3400+ Clawhammer instead of the 2800+ Barton. The Clawhammer is meant for high-end desktops/low-end workstations, the sledgehammer is meant for low-end servers/high-end workstations.

Intel will eventually have a desktop 64-bit processor. It just might take a decade or so until 64-bit processors are really mainstream. Intel has been 64-bit for non-destop computers for quite some time. These are just too expensive to be worth buying. So at the moment, if you go Intel you will stay 32-bit.

"we all eventually supposed to move to 64bit in the next two years"
Change that to ten years and I'll agree. Just think how long the change from 16-bit to 32-bit took. And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick.

 

[ST4RCUTTER]

And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick.


This is a misleading statement. The preliminary numbers have been 20-30% faster at the same clockspeed as current Athlons. Current Athlon's max out at 1.73Ghz. The Clawhammer will debut at 2.0-2.2Ghz. Combine that with the fact that they are comparing 32-bit operations between Clawhammer and Athlon, not 64-bit. At 64-bits, scientific and other highend apps are significantly faster than the same programs handling 32-bit operations. Of course those of us buying a 64-bit processor for the desktop will find a dearth of 64-bit applications that will benefit from x86-64. How many home users do you know that run programs requiring up to 4 billion times the amount of data handled by 32-bit applications?

 

[JellyBaby]

The transition from 16 to 32-bit desktop PCs took forever and that was in the pre-internet days. Back then Ziff Davis pretty much controlled what was "truth" and what wasn't. They were the mouthpiece for Intel and promoted 32-bit computing to no end yet still people didn't buy into 32-bit computing until much later.

Today with consumer software not really needing the extra bit width, the move to mainstream 64-bit computing may take even longer. Also factor in all the 'net sites that will decry 64-bit for some time. And as others have said you need the chip, OS support, 64-bit compilers and application support.

So, no your next PC will still be 32-bit unless you fall for the hype. Perhaps even the machine after that will still be happily serving you 32-bits at a time.

 

[winterlude]

Barton will just be an extention of Palimino, but with double the cache, and possibly, but not initially, a bus increase to 166x2. Barton and thoroughbred have been chronically delayed because of some problems Motorola was having (at least, that's the way I remember it), so AMD broke off with them. Originally, Barton was to have SOI (silicon on chip) licensed from IBM which accounts for some of Clawhammers superiority.
As far as first gen hammers are concerned, you probably won't use the 64 bit, because by the time a useful app comes out, processors will be far more powerful and it would be "silly" to use the "slow" processor instead of a new one. Remember MMX? It was supposed to be such a big deal, but by the time MMX optimized programs started to crop up, who was going to use their 166 pentium MMX to run them while 500 mhz Celeron were already out?
So all that matters for now is the 32 bit performance vs price. None of us knows how much a hammer platform will cost. AMD says that pricing will be competitive, but competitive with other 64 bit servers, not necessarily with standard fare desktops.
If the price is reasonable, it's a no-brainer to go with the hammer if you need a new system, but if you already have a capable system, you might as well wait for dual CPU hammers which will be "necessary" for MS's (if you buy it) next OS (Longhorn, I believe it's called which is 3D textured and multithreaded).
I'm still using win 98SE since I haven't come across a convincing enough argument for why I should upgrade to XP.
The hammer architecture is elegant, even though it is an extenstion of the messy X86 code. Hammer has more to offer than just the 64 bit that people are always fixated on. But will it be selling retail in the 4rth quarter? I wouldn't hold my breath for that. For anyone who has been following AMD roadmaps for the past few years, you'll remember that the hammer was first expected to enter the market in 2001.
The longer AMD waits, the more necessary the hammer becomes. However, 64 bit on desktops is a far cry from necessary for now and for the next few years.

Well, that's just my 2 cents.

 

[Athlon4all]

<< Hammer has more to offer than just the 64 bit that people are always fixated on. But will it be selling retail in the 4rth quarter? I wouldn't hold my breath for that. For anyone who has been following AMD roadmaps for the past few years, you'll remember that the hammer was first expected to enter the market in 2001.
>>

Well put on the first part. And I am not excited about x86-64, I'm excited about 2GHz and above clock speeds, Integrated Memory controller, etc. On the latter, well, we'll see. Maybe it will hit in Q4, maybe it won't

 

[dullard]

<< This is a misleading statement. The preliminary numbers have been 20-30% faster at the same clockspeed as current Athlons. Current Athlon's max out at 1.73Ghz. >>


I think your statement is the one that misleads. By the time Hammer is out, Athlon will NOT be at 1.73GHz. The Athlon will be at about 2.0 GHz at the same time the Hammer is available. I'm going off of AMD PR ratings. A 2800+ Athlon and a 3400+ Hammer will be sold at the same time. 3400/2800 = 1.214. Thus unless AMD is lying with its PR ratings, the Hammer will be about 20% faster than the fastest Athlon. So my statement was just repeating what AMD is saying. How is repeating AMD misleading?

 

[Diable]

<< Are the 64 bit flavors from AMD and Intel just for servers and not for the desktop or are we all eventually supposed to move to 64bit in the next two years?? >>



The 64bit processors from Intel are definitely intended for servers. I have never seen a roadmap from Intel that talked about IA-64 desktop processors in the next two years. AMD on the other had is aiming their 64bit chip at the small server, workstation and highend desktop market.

It will take a lot longer than two years for the vast majority of computer owners to migrate to 64bit machines. I also don't see a killer app that will make everyone "need" a 64bit processor unles geographic and seismic modeling become hot

 

[CTho9305]

<< The transition from 16 to 32-bit desktop PCs took forever and that was in the pre-internet days. Back then Ziff Davis pretty much controlled what was "truth" and what wasn't. They were the mouthpiece for Intel and promoted 32-bit computing to no end yet still people didn't buy into 32-bit computing until much later.

Today with consumer software not really needing the extra bit width, the move to mainstream 64-bit computing may take even longer. Also factor in all the 'net sites that will decry 64-bit for some time. And as others have said you need the chip, OS support, 64-bit compilers and application support.

So, no your next PC will still be 32-bit unless you fall for the hype. Perhaps even the machine after that will still be happily serving you 32-bits at a time. >>



disclaimer: i was like 6 when this occured

I thought 32-bit processors became popular pretty quickly (386s) - just that MS DOS was 16 bit along with win3.1 (until "win32s"), so the benefits were not actually realized (as in, made real ) for a long time

 

[Idontcare]

<< I thought 32-bit processors became popular pretty quickly (386s) - just that MS DOS was 16 bit along with win3.1 (until "win32s"), so the benefits were not actually realized (as in, made real ) for a long time >>



You are correct. I think there is another point that people are neglecting when they waive-off a rapid uptake of 64 bit cpus in the desktop market -> namely the 4GB RAM limit. When the transition from 16 bit to 32 bit was going on most people weren't pushing the max ram limit of a 16 bit cpu so the only reason to upgrade to a 32bit cpu was if that cpu offered a huge performance advantage or cost the same. However, within about three years most desktops will be pushing the 32bit ram limit and I really think that this is going to push people to use 64 bit processors in their desktops a helluva lot sooner then ten years from now.

Remember, why do people who buy workstations care about 64 bit cpus? They cost a helluva lot more than your desktop 32 bit cpu but perform only marginally better (and sometimes worse!, check the SPEC scores). For them, having >4GB ram is far more important then having a speedy processor (except for alphas which kick arse). I think this will become true for the desktop market sooner than most people are ready to admit. Five years ago I maxed my computer out with 128 MB of ram, currently I'm maxed out with 2GB (trust me you need this much for photoshop and office apps when working with >400pg documents).

My 2cents, -Phil

 

[Mrburns2007]

The clawhammer will have more then just 64 bits to entice users, the chip itself is more advanced and it will perform better with 32 bit apps.

Future setup will look like this:

64 bit hardware
Windows 2000/XP 64 bit OS
32 bit apps mainly with a couple of 64 bit apps

And of course the million dollar question "How well will they overclock ????"

Overall should provide better performance then Thourghbred or barton.


Bottom line: Does anybody really need a super fast computer for surfing the net at home, of course not but we want the POWER anyways!

 

[ST4RCUTTER]

And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick. -dullard


This is a misleading statement. The preliminary numbers have been 20-30% faster at the same clockspeed as current Athlons. Current Athlon's max out at 1.73Ghz. The Clawhammer will debut at 2.0-2.2Ghz.-STAR


I think your statement is the one that misleads. By the time Hammer is out, Athlon will NOT be at 1.73GHz. The Athlon will be at about 2.0 GHz at the same time the Hammer is available.-dullard


My mistake dullard. When you said "...with the Clawhammer starting at just 20% faster...", I thought you were comparing it against current Athlons...not future ones.

 

[JellyBaby]

CTho9305, not really. Intel's Pentium Pro was a big flop. It took a while for the chips to gain market share and much longer for everything to be in place on the software side.

FYI, I'm not disagreeing with you but it may all depend on how costly that 4+ GB of RAM is to us. Memory technology isn't keeping pace with processors, memory designers don't exactly enjoy tons of R&D surplus right now and the market itself shaky. This could lead to people living with less RAM than they may prefer even years from now.

 

[Nemesis77]

Clawhammer will run ALL 32bit apps and OS'es perfectly (assuming they are x86 compatible) and fast. Using 64bit OS and apps will give you about 10% boost in performance. So having 64bit software is not required but it has it's benefits.

How to get 64bit software? Well, in Linux it's not a problem. You just need x86-64 support in the compiler (coming up in GCC 3.1 due out any time now) and in the Kernel (standard in 2.5/2.6 kernels, patchable in 2.4-series). So Linux is taken care of. Clawhammer already runs in native 64bit mode in Linux without any problems.

Windows is another matter. MS will support it, but I have no idea when they will release it. Propable when Clawhammer is released (that seems to take place in october).

 

[Bluga]

<< . And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick. >>



It's still better than Itanium. Itanium performance is horrible. 800MHz

 

[dullard]

<< However, within about three years most desktops will be pushing the 32bit ram limit and I really think that this is going to push people to use 64 bit processors in their desktops a helluva lot sooner then ten years from now...Five years ago I maxed my computer out with 128 MB of ram, currently I'm maxed out with 2GB (trust me you need this much for photoshop and office apps when working with >400pg documents). >>



I buy a new computer for home every five years. Each one is just slightly above average in price. Lets look at what was available in 5 year increments:
1) Fifteen years ago, the average computer had about 256 kB of memory
2) Ten years ago, the average computer had 4 MB. The expensive computers had 8 MB and the cheap ones had 2 MB. Thus in five years the average computer had 16 times as much memory.
3) Five years ago, the average computer had 64MB. Sure some top-of-the-line models had 128, but it wasn't typical. 32MB only existed on the rock bottom clearance computers. Thus in five years the average computer had 16 times as much memory.
4) Today, the average new computer has 256 MB. Dell and Walmart still sell a lot with 128MB, but that is getting less typical. Thus in five years, the average computer had 4 times as much memory.

Now it can be dangerous to extrapolate, but in 2007 we will probably see the average increase in a similar fashion. Thus I'd expect the average computer to have 4 to 16 times as much memory as the average computer today. Thus the result is somewhere in 1GB to 4 GB range. Notice that in five years, the average computer will still be just fine with 32-bit memory! So I have to disagree with your "3-year desktop limit".

However, there are rare people that need much more than average (at work I use computers with 1 GB of memory, you need 2 GB of memory for your work). So the extreme people will not be satisfied with a 4GB limit - I agree with you. However there is a solution. Current top-of-the-line motherboards already use 36-bit for memory access. That leads to 64 GB of memory available TODAY on a 32-bit processor. All that needs to happen is to have MB manufacturers put 36-bit memory on the average computer (and then you also would want Windows support - since Windows currently craps out around 3.5 GB - but Linux is already able to handle it). I think 64 GB will handle nearly everyone for the next five years (of course there will be 64-bit solutions available for people needing more than the current 64GB limit).

 

[acejj26]

36 bit memory....the extra 4 bits are for ECC....it still is on a 32 bit bus...which equates to 4 GB limit

 

[Diable]

<<

<< . And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick. >>



It's still better than Itanium. Itanium performance is horrible. 800MHz

>>



No one that buys a Itanium will run 32bit apps cause Intel will help you rewrite your old 32bit software so its runs in glorious 64bit goodness.

 

[andreasl]

<< 36 bit memory....the extra 4 bits are for ECC....it still is on a 32 bit bus...which equates to 4 GB limit >>



No.. it's not ECC. ECC is invisible to software and done in hardware. With the 36-bit extension you can have up to 64GB of memory, BUT each process (or thread?) is limited to 4GB still, which means some recoding is necessary to take advantage of it. Which is one of the reasons I think that 64-bit will be used after 32-bit with no 36-bit step. The other (primary) reason for that is that 64-bit will simply be availible at the time it will be needed. In fact it will be availible from AMD long before and possibly from Intel (Yamhill 2004/05 or so) as well.

(Note to Dullard, I'm talking about desktop systems only here)

 

[Sohcan]

<< << . And with Clawhammer starting at just 20% faster, and a whole lot more expensive, the switch to 64-bit won't be quick. >>

It's still better than Itanium. Itanium performance is horrible. 800MHz >>

Why are you comparing an MPU that was intended to be released 4 years ago to one that isn't due out for around 6 months? Be prepared for the 1GHz McKinley to blow well past the current leader in SPECfp (1.3 GHz Power4) and silence the critics that said IA64 could never yield respectable integer performance when it is released in a month or two.



<< << 36 bit memory....the extra 4 bits are for ECC....it still is on a 32 bit bus...which equates to 4 GB limit >>
No.. it's not ECC. ECC is invisible to software and done in hardware. With the 36-bit extension you can have up to 64GB of memory >>

Hehe, yep, addressing space and data bus width are two entirely different entities...not to mention SECDED ECC for 32-bit data would require 7 ECC bits, not 4 (# of SECDED ECC bits for n data bits = (log (base 2) n) + 2). The Windows Server family already supports the P3/P4 (and Athlon IIRC) 36-bit segmented addressing for 64GB of addressing space, and the E7500 P4 chipset supports 12GB (or maybe 16GB?) of DDR SDRAM.



<< BUT each process (or thread?) is limited to 4GB still, which means some recoding is necessary to take advantage of it >>

Well, I hate to sound like Gates in the spirit of the "640K is all we'll ever need," but by no means is a 4GB limit for processes being stressed, and many question how long before that will be the case.