Sign in with
Sign up | Sign in
Your question
Closed

Why do we need 64 bit processors?

Last response: in CPUs
Share
September 5, 2002 11:23:12 AM

Now I was talking with a mate of mine and he brought up the question why do we need 64 bit processors?

He reasoning was thus:
The increase to 64 bit will allow more instructions on the CPU (MMX,SSE,SSE2 etc) but on a 32 bit scale we are no where close to saturating the number of instructions possible on the chips today, so do we really need 64 bit.

I said well 64 bit could be useful because:
With the increasing need for precision in todays applications eg mapping the human DNA etc 64 bit calculations are becoming more common, so it makes some sense to move to processors that can do 64 bit calculations in one hit rather than having to perform several operations on it to do the same work.


I wonder if I have missed the point and if my friend has definatly missed the point. I know there are several reasons to move to this new technology but they escape me at this very moment.

Can anyone think of any reason why we need or dont need this new technology?

More about : bit processors

September 5, 2002 2:10:51 PM

If I'm not wrong, most of them are in a similar previous thread. Just make a search about 64 bit processor.

DIY: read, buy, test, learn, reward yourself!
September 5, 2002 2:31:19 PM

bit stuff are pretty unrelevant

At the end i have speak with a horny lady
Related resources
September 5, 2002 3:24:37 PM

umm I did a search and all that came up was threads I have posted in before, none of which gave me any information on why we need them.....
September 5, 2002 4:08:04 PM

The only good reason for 99.9 % useur it the new adressing.32 to 64 bit adressing

At the end i have speak with a horny lady
September 5, 2002 5:09:15 PM

We don't need it....it is worthless to the average joe. Only really need for scientific stuff.

<font color=blue>Unofficial Forum Cop</font color=blue>
September 5, 2002 5:38:31 PM

At it can look Intel can built a VPU feature
in the core and 3X the price.So a P4 5 GHZ with a 1 ghz R300

At the end i have speak with a horny lady
September 5, 2002 6:56:31 PM

Quote:
Now I was talking with a mate of mine and he brought up the question why do we need 64 bit processors?


We as in the general enthusiaste community?

Quote:
The increase to 64 bit will allow more instructions on the CPU (MMX,SSE,SSE2 etc) but on a 32 bit scale we are no where close to saturating the number of instructions possible on the chips today, so do we really need 64 bit.


The "bitness" of a MPU has nothing to do with the instruction length or amount of parallelism the MPU can handle. It's about data size.

Quote:
I said well 64 bit could be useful because:
With the increasing need for precision in todays applications eg mapping the human DNA etc 64 bit calculations are becoming more common, so it makes some sense to move to processors that can do 64 bit calculations in one hit rather than having to perform several operations on it to do the same work.


Somewhat correct. Although most DNA modeling nowadays involves heavy FP work, not integer. FP data is already 64-bit capable on all modern MPU's since the Pentium 2 I think, possibly even before that although my knowledge of Intel processor history isn't that great.
When a MPU is said to be "64-bit" it means the general purpose registers are 64-bit wide. General purpose registers, hereby known as GPR's, are little pockets of memory on the processor that store data that the processor needs (think of it as ultra-high level cache). The difference is, each little pocket of memory only stores 1 data type. Modern x86 MPU's have 8 GPR's. These are always used for integer and memory address data types. So by going "64-bit", you are expanding the capability of your MPU to handle larger numbers at once. A 32-bit integer is capable of handling numbers from -2.1 million to 2.1 million. If you have an application that uses numbers larger than that, 64-bit computing would help speed up things significantly. Then there's memory address. A 32-bit memory address is only capable of addressing up to 4 GB with flat memory addressing. Expanding to 64-bit memory addressing would mean you are capable of using 1.7 x 10^10 (roughly 17000000000) GB of memory. Of course none of the upcomming 64-bit processors (be it Hammer or Itanium) actually uses 64-bit flat memory addressing, they all use 48-bit. No one really needs that much memory capability nowadays. Not even at the ultra-high end.
So, to recap, moving to 64-bit by itself would mean:
1. Able to handle larger integers without significant slowdown
2. Able to use larger memory in flat addressing mode.

It says nothing about expanding the MPU's capability of handling instructions. Those are usually just added anyway but they have nothing to do with the move to "64-bit computing".

Quote:
Can anyone think of any reason why we need or dont need this new technology?


Do you need more than 4 GB of memory? No consumer-level application currently uses integers that even comes close to 2.1 million or negative 2.1 million in any significant percentage of its code. So the 64-bit capability will just pretty much sit there idly until applications get so large and bloated to the point where they are constantly handling integers larger than 2.1 million (I'd give it 2 years minimum).

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 5, 2002 9:22:09 PM

so essentially your saying that we dont need 64bit now but in a couple/few years (when 4GB is more limiting) 64bit will be usefull? if so does that mean that we would virtually never need 128bit processors?, unless 17000000000 GB does become a limiting factor, actually having said that the way newer versions of windows use so much memory.......

i have posted here before but i forgot my password and thgc wouldnt send me a reminder!!
September 5, 2002 10:29:30 PM

Hey I am sure I could use that much memory :-P

So I am guessing that AMD and Intel are going to convince the average joe user that they need 64-bit processors *enter rubbish made up reason here*

Now I am thinking the only real thing people would go 64bit in the case of the Hammer is because of higher clock speeds and an updated core. Also ofcourse the hammer supports upto 8 processors(sledgehammer), so for business users who require this sort of power will be able to get it.

I believe AMD could have got away with yet another 32 bit CPU but with this new updated core, but trust me I believe people will have these processors, if only because of the increased clock speed. I will be annoyed if AMD produces a mediocre 64 bit CPU instead of a stunning 32 bit CPU.

I think the only use for 64 bit CPU like the hammer or the itanium is on database servers or webserver, if only due to the fact you can have 1-8 AMD processors or 2-32 itanium porcessors.

Anyway why would I buy a hammer, 8 processors and that is it. I may get a barton MP (if they release one) as I am sure with the Hammers release the prices will fall, well we can hope.
September 5, 2002 11:28:47 PM

The need for more computing power and memory space will always be there. In industry it's not so hard to find computations that take weeks to finish just one run (and multiple runs are needed). Also in some areas terabytes of data are common. It's short-sighted to claim 'that's all we need'. Even for average consumers, Windows XP has become so big that on 128MB of ram is just barely enough to run the system. The development of computers is still in an early stage, and I don't see an end to it. Technology will continue to develop and more and more once unthinkable things become reality. The only end to it is the end of the human species, such as self destruction or natural disasters.
September 6, 2002 7:37:45 AM

Ups, I tought I had read a thread with a similar title before. I suppose it would be yours. Sorry Hoolio.

DIY: read, buy, test, learn, reward yourself!
September 6, 2002 1:19:36 PM

Quote:
So I am guessing that AMD and Intel are going to convince the average joe user that they need 64-bit processors *enter rubbish made up reason here*

It's just a next step in the technology. Just like any other advancement. 64Bit isn't a marketing strategy, it's where they have to go to continually improve performance.

This sig runs too hot.
September 6, 2002 1:36:50 PM

I doubt AMD themselves expect the 64-bit component of Hammer to benefit the consumer in any way. I'm guessing the reason they included it was because they needed to enter the Enterprise/Workstation market and didn't have enough resources to develope 2 core designs separately. Hence Hammer is more of a one-size-fits-all design. With an improved core that will perform 32-bit code very well and 64-bit extensions for the market that needs it.
As for will 128-bit computing ever be needed. I would hope that by the time that we actually reach that limitation, we would've gone beyond the point of fixed data sizes and come up with something much more revolutionary. Of course, that's wishful thinking.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 6, 2002 3:39:11 PM

very well and 64-bit extensions for the market that needs it

Does who really need 64bit will still go with pure 64 bit CPU.Sientific apps need FPU precision like IA-64 wich can go up to 86 bit.

At the end i have speak with a horny lady
September 6, 2002 4:18:13 PM

actually you are wrong ... an nth bit processor means it can process nth bit's per cycle. Along with being able to access more data as well which would increase performance as you don't need to keep a bigger lookup table for the virtual memory. Along with many many other goodies usually needed in large businesses.

<A HREF="http://www.anandtech.com/mysystemrig.html?id=9933" target="_new"> My Rig </A>
September 6, 2002 4:19:13 PM

64 bit addressing allows for the use of a lot more memory, something very useful in workstations and servers.

I like the Pentium IV, I really do! And it's so versatile. You simply won't find a more stylish or decorative key chain ornament or paperweight.
September 6, 2002 5:06:08 PM

Quote:
One of the key operations that the 64bit addressing allows is the larger page size. The current page size for 32 bit CPUs is 4k (12bits). The page size for 64bit CPUs is 2mb (21bits). This allows more efficient paging on large memory systems.


I am completely wrong on this one. Sorry. All processors after the 386 allow for variable page granularity.

Complicated proofs are proofs of confusion.<P ID="edit"><FONT SIZE=-1><EM>Edited by Schmide on 09/06/02 10:07 PM.</EM></FONT></P>
September 6, 2002 6:06:17 PM

And one thing that I think people are forgetting is just how many people are slapping together their own cheap little servers running open source software these days. I'm sure that the number of these things will grow dramatically once chips like the ClawHammer come out and allow people to not only toss together a quick web server, but even a 64-bit database server at the drop of a hat for dirt cheap. (Dirt cheap compared to the server market anyway.)

<i>That</i> is why AMD is releasing a 32/64 bit crossover chip. It has the potential to be a great cash cow by filling in an ever-growing niche for dirt-cheap servers run by enthusiasts and small businesses.

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 6, 2002 7:44:12 PM

Well, the workstation market has expressed a great need for 64-bit computing. Not neccessarily for the ability to handle larger integers but rather for the memory address capability. Currently, the workstation market is stuck with Xeons, which can only use virtual addressing. Hammer could be potentially very successful in this market. As for it competing with the likes of Itanium, I doubt it. It is an x86 processor after all.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 6, 2002 7:59:35 PM

Quote:
Well, the workstation market has expressed a great need for 64-bit computing. Not neccessarily for the ability to handle larger integers but rather for the memory address capability.

One thing that still bugs me is just how hard it really is to even configure a PC with 4GB of RAM! It's absurd to even try for a typical desktop. Even if you <i>can</i> afford the RAM, just finding a motherboard to run it on with stability takes a lot of effort.

Quote:
Currently, the workstation market is stuck with Xeons, which can only use virtual addressing. Hammer could be potentially very successful in this market.

That's also true. I just plain don't understand what Intel is trying to do with the Xeons lately. I think they've lost their marbles.

Quote:
As for it competing with the likes of Itanium, I doubt it. It is an x86 processor after all.

Darn straight! Heh heh. ;) 

Seriously though, I think maybe it could complete there, but purely because of price. Say you can build an 8-way Hammer for the price of a 4-way Itanium and in the end the 8-way Hammer performs better. It might be able to compete that way.

Plus, as I already said, it <i>should</i> open up a whole new market for dirt-cheap 64-bit servers (where minimal price is far more important than performance) where Itanium just can't compete.

For the most part though, Hammer (especially ClawHammer) just isn't going to make any major impact on the mid or high range server market.

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
September 6, 2002 8:26:37 PM

someone meantioned processors without a bit rating. Biological processors made of cells etc. Could in theory adapt to process different data lengths, cool huh?


I do agree the workstation market requires a cheap 64-bit multi processor option. 8 way 64 bit sledgehammer coming my way.
September 7, 2002 1:54:43 AM

With their powerful FP power and the added IPC for K8, those chips in 8-way configs in multiple racks, would be damn awesome for rendering and scientific stuff. A big plus for AMD CPUs on that.
Of course I heard the McKinley's FP power is truly a wonder, so we'll need to see benches of Sledge vs Itanium2 in FP only, to see if it's true. I am not implying Itanium 2 VS Sledge overall, both are not targetted at the same market, but just to see in FP.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
September 7, 2002 2:27:17 AM

?? i simply dont follow you

At the end i have speak with a horny lady
September 7, 2002 6:57:33 PM

Ya, Mckinley sure is a wonder, when it's running in it's native 64-bit environment. I envision many great things for IA-64 in the future. When sledge comes out, it'll have to contend with that monster-of-a-CPU, Madison. Sledge vs. Madison, Sledge probably won't stand a chance. Sledge vs. Mickinley, there's a small chance they <i>might</i> be close.

I mean, you must realize, that in FP in SPECint2000, Mckinley even beats an IBM Power 4, whcih is quite a feat all on it's own.

Imagine what a P4 could do if it had that kind of execution power!

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
September 7, 2002 7:13:46 PM

IA-32 dont that ILP

At the end i have speak with a horny lady
September 7, 2002 7:15:46 PM

This won't be possible, Sledge is not targetted against Itaniums, never was. They are both different markets. Sledge will undoubtedly be less performing than McKinley and surely Madison, because it is not designed like them.

FP power, not sure yet, we need that bench on Sledge to know. K7-K8 still is the most powerful FPU in x86.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
September 7, 2002 7:53:02 PM

K8, it's not certain for sure (FP power). I'm so sure that Madison (in FP power) will cream sledge in 64-bit, that Im willing to bet money on it. K7 may have powerful FP in x86, but that's specifically why Intel designed IA-64: to get rid of all the bottlenecks and limitations of x86. The amount of potential that Ia-64 has is just staggering. Actually, according to Intel (if you go over to their website and go to CPU's) the Itanium is a server & workstation CPU. That's the same <b>general</b> category that Sledge is heading to. So, in general, they will be in the same category, plus add to the fact that sledge is 64-bit, so it'll have to be compared with Itaniums.

Juin, sorry, but I didn't understand what you said.

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
<P ID="edit"><FONT SIZE=-1><EM>Edited by Dark_Archonis on 09/07/02 03:53 PM.</EM></FONT></P>
September 7, 2002 8:11:31 PM

Sledge's price is far from Itanium's, which is why Sledge is targetted in the lowest corporate server-workstation market. Itanium, IIRC was designed to go way up there against the big CPUs, so I have about 90% certainty that Sledge vs Itanium is a big no-no.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
September 7, 2002 11:09:05 PM

Ya i know, but something interesting in Deerfield, it's what Intel calls a "budget" IA-64 CPU. It could possibly compete with sledge.

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! :wink: </font color=green>
September 8, 2002 4:00:51 AM

I agree that Sledgehammer won't be competing with Itanic. I don't think Itanic really competes with anything based on it's dismal sales. If anything, Sledgehammer will hasten Itanic's demise. While McKinley or Madison may outperform Sledgehammer, maybe even to a large degree in some areas, they will also require a humongous investment, both in hardware and software. Neither will be true for Sledgehammer. And with Hypertransport it will be ridiculously easy and inexpensive to add as many processors and performance as you need. So I think you are very wrong in thinking that Sledgehammer won't be targetted against Itanic and other high-end platforms. It's price/performace ratio will be unbeatable.

I like the Pentium IV, I really do! And it's so versatile. You simply won't find a more stylish or decorative key chain ornament or paperweight.
September 8, 2002 6:45:24 AM

Performance outweighs price in that sector. It's dominated by IBM and Sun, that's why Intel is having a hard time squeezin their chips in there. Hammers cant compete w/ the Itanium2, Alphas, and UltraSparcs of the world. Theyre not geared that way. Theyre gonna be in desktops, and pitted against P4, and xeon based systems.

And whats up w/ your post? You're arguing w/ yourself in it.
Quote:
I agree that Sledgehammer won't be competing with Itanic.


and then....
Quote:
So I think you are very wrong in thinking that Sledgehammer won't be targetted against Itanic and other high-end platforms.

That was a pretty Anti-Itanium post. Itanium2 is a damn good chip......<A HREF="http://www.intel.com/ebusiness/products/server/index.ht..." target="_new">Here</A>, read up about it.

This sig runs too hot.
September 8, 2002 7:30:07 AM

Quote:
Performance outweighs price in that sector. It's dominated by IBM and Sun, that's why Intel is having a hard time squeezin their chips in there.

Intel is having a hard time giving the Itanic away, let alone selling them.
Quote:
Hammers cant compete w/ the Itanium2, Alphas, and UltraSparcs of the world. Theyre not geared that way.

Really? They're not geared that way? What does mean? Do you know anything about the architecture of Sledgehammer? Not just the processor but Hypertransport and the supporting architecture. AMD's next generation processors will wind up in everything from desktops to very high-end servers.
Quote:
And whats up w/ your post? You're arguing w/ yourself in it.

I was being facetious. My point is that nobody wants the Itanic so it doesn't compete with anything. And frankly, I don't think that's going to change.
Quote:
That was a pretty Anti-Itanium post. Itanium2 is a damn good chip......Here, read up about it.

Hehehe...you want me to read Intel's propaganda about the Itanic? You can't be serious!?

Take a look <A HREF="http://www.swox.com/gmp/gmp-speed.html" target="_new"> here </A>. The lowly Athlon putting the smackdown on some very expensive hardware. Imagine Hammer....

I like the Pentium IV, I really do! And it's so versatile. You simply won't find a more stylish or decorative key chain ornament or paperweight.<P ID="edit"><FONT SIZE=-1><EM>Edited by Nikko on 09/08/02 03:43 AM.</EM></FONT></P>
September 8, 2002 7:49:59 AM

You have to realize, in that market sector, performance is everything. A 20% performance delta could be the difference between a chip receiving huge sales or a chip not receiving sales at all. The first Itanium didn't perform that badly. In fact, it was able to keep up with most of the other high-end chips out there. The only problem was, it didn't offer too signficant a performance increase to justify people moving to it. If IT managers have been using Sun systems for years and nothing else will offer much of a difference, they'll stick with Sun systems. It really doesn't matter than Itanium is actually a lot cheaper compared to Sun systems or that it has a much better price/performance ratio. The few thousand that is saved for a mere 5% performance increase is well worth it in the millions of extra transactions that'll translate in the time that server is running (most large clusters are kept for at least a year). It's not like the desktop or even workstation market where you have a bunch of rabid fanboys that jump on a 4% performance increase and try to save $50. In order to enter such a market, you need a huge advantage in performance. Itanium 2 comes close to this although still not as grand as it should be. Looking at Sledgehammer, it is nowhere near enough to compete with Ultraspark or Power. I don't think AMD even wants it to compete there, they're limiting Sledges to 8-way processor designs. That just says it all right there. What it will kick ass in is when competing against Xeons and Xeon MP's. Which power clusters or workstations and/or low-end servers which is a more profitable market anyway, and easier to sink your teeth into if you've got the right partners.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 8, 2002 3:54:31 PM

Quote:
don't think AMD even wants it to compete there, they're limiting Sledges to 8-way processor designs.



In terms of the level of processor, 8-way is a massive scale for one board to support. I think the Ultrasparc maxed out at 14 and that was with a sub ghz processor. Remember each hammer processor has its own dedicated ram. A major break from the chipset controlled single memory system SMP. I think AMD has come to compete and might just raise the bar quite a bit.

Complicated proofs are proofs of confusion.
September 8, 2002 5:10:23 PM

I agree, Schmide. AMD WILL absolutely compete in the very high-end. Hypertransport and the architecture of AMD's 8th gerneration platform has more in common with exotic mainframe technology than anything Intel offers. If I'm not mistaken even Itanic uses the GTL+ bus topology which is ancient and ineffecient. Daisy-chainning 8-way Hammer boxes will be both easy and, considering the competition, relatively cheap.

I like the Pentium IV, I really do! And it's so versatile. You simply won't find a more stylish or decorative key chain ornament or paperweight.
September 8, 2002 5:57:34 PM

What about 16 and 32 processor servers? About 75 percent of the market is 1 and 2 way servers, where Opteron can dominate. Thats where you find xeons, and 32bit machines. Another 10 percent is 4-8 way......Then theres the high end where its above 8. AMD can't go there. "daisy chaining" servers.....8-way x2 to make a 16 proc server isnt how it works.

This sig runs too hot.
September 8, 2002 6:20:54 PM

If you're referring to the Unisys ES7000 type servers, they are clusters for 4 way servers. So a 16 processor server is a 4x4 and a 32 processor server is a 4x8. Sledges are designed to cluster in 8 processor increments. Please give other examples if you got um.

Complicated proofs are proofs of confusion.
September 8, 2002 6:21:37 PM

Quote:
AMD can't go there. "daisy chaining" servers.....8-way x2 to make a 16 proc server isnt how it works.

Really? Please educate me on how Hypertransport works.


I like the Pentium IV, I really do! And it's so versatile. You simply won't find a more stylish or decorative key chain ornament or paperweight.
September 8, 2002 9:17:00 PM

Hypertransport is limited by the number of links on each CPU. As they are point to point, you are limited to how many combinations of interconnects you can use to another processor. Sledgehammer has 3 links I think and in order to maintain low latency between chips, it's limited to 8-way. You can of course cluster but at the very high end, latency is everything and clustering together servers just doesn't work. Bank databases, data centers and transaction servers absolutely rely on ultra-low latency processing such as that found in a 32-way server.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 9, 2002 1:25:44 AM

Who makes a 32 way server?

Complicated proofs are proofs of confusion.
September 9, 2002 1:46:56 AM

IBM......can get a 640 processor cluster if ya want :) 

This sig runs too hot.
September 9, 2002 2:10:46 AM

The max I've ever heard a PowerPC go is 6 way. They may cluster out to 640+ processors, but their node segmentation I believe maxs out at 6. Do you have any links?

Complicated proofs are proofs of confusion.
September 9, 2002 2:40:23 AM

They use MCM's (MultiChipModules) 4 POWER4 chips on a MCM form an 8-way, up to 4 MCMs to a system. Search IBM dot com for 32-way, or Power4 processors....too much to look through to find that specific link.

This sig runs too hot.
September 9, 2002 2:42:59 AM

So were talking the same node granularity for the Sledge and the Power4.

Complicated proofs are proofs of confusion.
September 9, 2002 2:45:16 AM

Also, I havent read this about Opteron systems, but these high end servers have hotswapable components. You can even change processors on the fly if you want......w/o restarting the servers.

This sig runs too hot.
September 9, 2002 2:56:22 AM

The MCMs are symmetrical......it's not really a "node". Not everyone needs 32-way at first. You can pop another MCM in there. It's like throwing the 2nd Proc in a dual board that's been running w/ just one. That's common in small business that keep the door open for upgradability. And it's 4 chips that make an 8-way, instead of sledge's 8 for an 8-way. Also (someone correct me if Im wrong) the MCM's can go in while the system is "hot".
It's a hard 32-way system.....not 16x2, 8x4, or 4x4.

This sig runs too hot.
September 9, 2002 3:23:11 AM

Ironically, now that I think about it. Because the Opteron exists on a point-to-point protocol without a shared memory system, it has an actual node granularity of 1. So it actually is a 1-way processor. The use of n-way term is often abused. So we're really comparing apples to oranges. For someone to say the Opteron will not scale above clusters of 8 is just playing stupid. I guess we'll just have to wait and see.

Just because there exists some underlying system that keeps the L3 caches (main memory) synchronized, does not mean the processors exist on the same bus. Processors controlled by a single APIC (advanced programmable interrupt controller) are considered truly n-way.


Complicated proofs are proofs of confusion.
!