64bit, really all that?

[Vee]

Originally posted by: oralpain

Originally posted by: Vee
Well, that sounds a bit strange. But it's all up to judgement. Anyway, there's no technical similarities between the 16 to 32 bit migration and the 32 to 64 bit migration. Two completely different things.

How are they that much different? It's basically the same idea. Going from 16-bit to 32-bit registers and address space should be quite similar to going from 32-bit registers and address space to 64-bit (though the a64/opteron only has 40-bit addressing it should not be a limit that will be reached any time soon).

By the 16-bit to 32-bit migration, we moved from segmented addressing to flat linear address space. This was really the big thing. But also since we were actually introducing a new software format, the opportunity could be taken to also introduce a number of other modern features, like preemptive multitasking, multithreading, non serialized message qeues, real virtual memory, strong memory protection. All in all, a very nice migration that made the PC as pleasant to work with as other platforms.

With the 32-bit to 64-bit migration, we are doing none of those things. We already have linear process space and all the other stuff. This time the essential detail is that we are removing the 2GB limit on process space. Compare it to when the '286, and DPMI and Windows16 software removed the 8086/MSDOS 640KB barrier instead. Because that's where we are again.

It's important to understand that this is NOT about physical RAM addressing! The '386 addressing scheme as employed by our 32-bit cpus, allow, I think, addressing 64GB physical RAM! This is of no use for our current problem. Because it's the program's space to live in, with all things, including useless gaps between blocks of sequential addresses, that is limited to 2GB! The total amount of numbers it can use for every byte. And please note that this space becomes fragmented, so the problem is more severe than it may immediately sound.

The address space of X86-64 is not 40 bits or one Terabyte. It is 64 bits or 16 Exobytes! Currently announced implementations' (real cpus), memory managers allow mappings of 52-bits or 4 Petabytes to effective addresses. The 40 bits you're talking about are just the physical addressing. And 1 Terabyte seem quite sufficient for a while. But X86-64 is by no means limited to 40-bit physical addressing. That's just how current cpus are made.

Then, when we are going to have to change the ISA and binary format of software, we might just change a few other things, from the old '386, as well. That's why we also get more flexible registers, which should give performance a boost. How much remains to be seen, but I'm sure there are excellent opportunities for optimizations.

 

[glugglug]

Originally posted by: Vee

The address space of X86-64 is not 40 bits or one Terabyte. It is 64 bits or 16 Exobytes! Currently announced implementations' (real cpus), memory managers allow mappings of 52-bits or 4 Petabytes to effective addresses. The 40 bits you're talking about are just the physical addressing. And 1 Terabyte seem quite sufficient for a while. But X86-64 is by no means limited to 40-bit physical addressing. That's just how current cpus are made.

This is another incorrect common assumption. The MMU of current AMD64 chips can "only" handle 48-bit addresses, for 256 terabytes of virtual address space.

But you are right, my 40-bit number is for physical addressing in the current forms - only 40 address lines reach the outside of the chip.

 

[Vee]

Originally posted by: glugglug

Originally posted by: Vee

The address space of X86-64 is not 40 bits or one Terabyte. It is 64 bits or 16 Exobytes! Currently announced implementations' (real cpus), memory managers allow mappings of 52-bits or 4 Petabytes to effective addresses. The 40 bits you're talking about are just the physical addressing. And 1 Terabyte seem quite sufficient for a while. But X86-64 is by no means limited to 40-bit physical addressing. That's just how current cpus are made.

This is another incorrect common assumption. The MMU of current AMD64 chips can "only" handle 48-bit addresses, for 256 terabytes of virtual address space.

But you are right, my 40-bit number is for physical addressing in the current forms - only 40 address lines reach the outside of the chip.

Thank you, you're right. Sorry, let me rephrase that slightly:
The address space of X86-64 is not 40 bits or one Terabyte. It is 64 bits or 16 Exobytes. The scheme for memory mapping allows that to be mapped to 52-bit effective addresses, or 4 Petabytes. Current CPU implementations allows for mapping 48-bit virtual addresses (64-bit canonical form) to 40 bit effective addresses.

 

[Gamingphreek]

Originally posted by: Vee

Originally posted by: Gamingphreek
Also people think 64bit...WOW!!! Some ignore the fact that there are processors that are 256bit. (Lol cant remember the name). Computing still has a long way to advance beyond 64bit. Were only at the beginning.
-Kevin

No. You are totally misunderstanding the CPU bit issue! Todays "32-bit" cpus, like the P4, are already width-wise 128-bit wide. This is really so, we have 128-bit registers and instructions operating on 128 bit wide vector fields. Also in other ways, there are even paths inside the cpu that are 256 bit wide. And this is on a "32-bit" CPU. And we are probably going to see 256-bit, maybe 512-bit registers and vector operations on AMD's K9 and Intels Conroe.
The 16/32/64-bit cpu issue has mainly to do with the length of the address field used by instructions to refer to data or code. And since 64 bits are good for a program space of 16 ExoBytes, there is certainly no need for more than 64 bits in our lifetime.
You should not understand computer code- and cpu-bits in the game console bit paradigm. Computing width is already, as I said, 128 bits wide in SSE2. And we are probably going to see computing bit-widths hit four figure numbers on "64-bit" cpus. Computing width can be anything, in a "32-bit" or "64-bit" cpu. It's not the issue.

You misunderstanded me. There are CPU's that 128 and 256bit capable. The are the Crusoe and Efficeon. I think Efficeon is the 256bit one.

To sum up something AMD chose not to make a full 64bit memory address. The chose IIRC 40bit, because lets be honest, we are deffinately not going to need a terabyte of RAM. Heck we probably wont need much more the 4gb for a while. That is mainstream users not the enterprise class and all that good stuff.

-Kevin

 

[Nemesis2038]

Despite being different architectures Opteron and Itanium. Performance measurement can be relatively similar accross 64 bit platforms. Also with HP dropping Itanium can give us a clearer indication of how well Opteron and Intel's 64 bit processors must be performing in 64 bit modes for HP to drop Itanium2 systems as well as many other vendors dropping Itanium even when there is a Windows OS out for the Itanium2 processors. So why would they drop Itanium architecture?

If 64 bit is not all that then why does an Itanium2 significantly outperform a XEON 32 bit CPU at much lower Mhz? Simple 64 bit is kick arse.

The so called 64 bit 10-20% improvement is not so in certain applications. So lets use some Itanium2 Information to determine how good 64 bit is.

HP reckons that an Itanium 2 processor running at 1GHz is:
2.1 times as fast as a 1.67GHz AMD Athlon XP 2100+ processor
2.1 times as fast as the 1.6GHz Pentium 4 Xeon from Intel

Disclaimer.
Exactly where HP gets these comparisons is unclear, since Intel has yet to test a Xeon MP processor on the SPEC suite as well, and we couldn't get the same ratios that HP got using any of the limited numbers out on the SPEC web site.

So if Mhz were linear then a Itanium2 at 1Ghz = 3.36Ghz Xeon?
Hey thats what they are saying an Itanium2 = 2.1 x 1.6Ghz Xeon.

Since the Opteron Runs at 2.4Ghz in 64 bit mode can we expect 8.06Ghz Xeon Performance on an Opteron when the Opteron is in 64bit mode?

Or can we say the Opteron 2.4 could potentially be 2.4 times faster than an Itanium2 in 64 bit mode?
No but lets use the following statement anyway and see where we are.

OPTERON 2.4Ghz = 8.06Ghz Xeon?

This is what I see in 64 bit modes in my Opteron.

My Opteron is Nearly 3 times faster than my P4 processor and around 4 times faster than my Athlon XP 3300+ when compressing files. This makes my Opteron = ~9Ghz P4.

Video conversion the Opteron in 64 bit mode is Nearly 2 times faster than my P4 and Athlon XP.
The means my Opteron is around the performance of a my P4 if it were capable of around 6Ghz.

So is 64bit all that. YES YES YES. It certainly can be significantly faster.

I dont expect the same results on every application since games rely mostly on Video card performance and not CPU performance. For most games the Video card is the bottleneck so in there I expect the same as if you were CPU scaling. This is proably why alienware is reviving an SLI Config of video cards simply because the speed is reaching a limit. I expect to see video cards companies to announce or secretly work on Duals soon because of this limitation.

For other items where video is not the bottleneck and are CPU intensive then YES major speed increase coming soon.

I would love to see what a XEON 3.6Ghz with 3meg on die cache would do since intel is the king of Mhz. As well as Dual or Dual Core in 64 bit is going to be breathtaking.

If Itanium2 was going to be vastly superior at 64 bit computing than the Opteron or future Xeon 64 chips why are vendors dropping Itanium in support of Opteron64 and XEON64? I cant see HP dropping Itanium2, a Joint venture costing HP Millions, It doesnt make sense to drop a CPU with Windows OS support to favor an Opteron does it? If so then all vendors except Dell are idiots not to carry Itaniums over Opterons. Its not simply and easy upgrade path that makes the x86 64 bit architecture being accepted. I also just read that one vendor is dropping Itanium in favor of XEON64.

64 bit is the bomb dont let anyone tell you differently who doesnt own one of use one.

 

[Gamingphreek]

Ok you have no idea what your talking about. More or less your cofused. I dont want to sound like im outting you down or anything but you seriously have no real clue what your talking about

The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

In terms of performance there are 2 major things that you said wrong. The Itanium 2 is not meant for ANYTHING we use our P4 class CPU's today. They say it is 2.1 times faster they mean specifically one area of computing which is enterprise. AMD doesn't really have an enterprise class system. The opteron is as close as it comes. But the I2 is out of its league with a 128 way instruction set when an Opteron 840 (highest end opteron) has only an 8 way configuration.

Your A64/Opteron couldb't be able to touch a CPU running at 2.1 times it current speed!!

Also the environment the the I2 runs in is much different as they use Unix.

Yes the EPIC architecture combined with the IA-64 microarchitecture is better at Enterprise class computing then the A64 will every be.

As for your claim, people everywhere are not dropping the I2. The I2 like i said before is Enterprise class, and may not be as good as some other enterprise class processors it is still competitive.

As for you claim about the 2.1x faster, i mentioned that only at specific things, Gaming is not one of them. An A64, or P4 will massacre the I2 in gaming as it was not designed, nor are there any games that support it to run it, to play games.

Just next time before you post get your facts straight.

-Kevin

 

[LTC8K6]

Nemesis2038, what planet are you from? You sure post an awful lot of junk that you can't back up. :frown:

 

[Vee]

Originally posted by: Gamingphreek
You misunderstanded me. There are CPU's that 128 and 256bit capable. The are the Crusoe and Efficeon. I think Efficeon is the 256bit one.

OK. My point was you need to consider what those bits mean, in computing context, when making comparisions. I do not know terribly much about Transmeta's CPUs. I tried to read some of the early documents, but it was awfully unconventional and complicated. But my rough guess is still that in the comparable sense that the Crusoe is 128-bit, so is the P4 "128-bit". This in the meaning of what width of bits can be held in a register and operated upon. I'll check up on the Crusoe and maybe get back.

To sum up something AMD chose not to make a full 64bit memory address. The chose IIRC 40bit, because lets be honest, we are deffinately not going to need a terabyte of RAM. Heck we probably wont need much more the 4gb for a while. That is mainstream users not the enterprise class and all that good stuff.

There are two types of addresses. The virtual address is the number used by a program to refer to and identify each byte of code and data. It's very useful with a large virtual address space. The virtual space is the ultimate restriction of a program. Our current 32-bit processors virtual address space is only 4GB inside a flat memory model. However, a Windows32 process has only 2GB space for code and data. The 3'rd GB is used for mapping shared resources and the 4'th GB is used for mapping OS resources.
AMD'86-64 (the architecture) provides full 64 bits for virtual space. However, you cannot use all 16 Exobytes. You can use 4 Petabytes maximum. Since that is the number of addresses that can be mapped to an effective address. The mapping scheme provides for a mapping of 64bits -> 52bits.

That is not necessarily exactly the case with actual existing CPUs however. As Gluggplug pointed out, current AMD64 cpus provide only a 40-bit physical address, and a 48bits -> 40bits mapping scheme. So for a "start", we can "only" use the 48 lower bits in the address field. Bits 63 - 48 should be filled with either 0 or 1 depending upon the state of bit 47. But we have a massive virtual space, 256 Terabytes. Which should provide a nice process space in Windows64, instead of current 2GB.

 

[Vee]

Originally posted by: Gamingphreek
The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

X86-64 is only x86 in 32-bit mode. The long mode provides what is essentially a new ISA, that can use registers much better.

Yes the EPIC architecture combined with the IA-64 microarchitecture is better at Enterprise class computing then the A64 will every be.

The benchmarks I've seen lately, can't remember them all, but a lot of them comes from IBM, tell a different story. Not only is AMD64 a serious workstation threat against Itanium, on server benchmarks HP's latest 4-way Opteron system only trails 11% behind fastest ever 4-way Itanium system. That is utterly devestating considering the price difference. Frankly, Itanium is dead. Intel knows that too.

 

[Nemesis2038]

I will say it again since your missing the point. Reguardless if you want to compare Itanium2 to Opteron.
It will be funny to see how everyone will be changing their tune about 64 bit performance once the Xeon 64 is here.

FACTS ABOUT x86 64 BIT MODE
My Opteron is Nearly 3 times faster than my P4 processor and around 4 times faster than my Athlon XP 3300+ when compressing files using the Opteron in 64bit mode. This makes my Opteron = ~9Ghz P4.

Video conversion the Opteron in 64 bit mode is Nearly 2 times faster than my P4 and Athlon XP.
The means my Opteron is around the performance of a my P4 if it were capable of around 6Ghz.

These are CPU intensive operations not bottleneck video card gaming benchmarks.
64bit is much faster when using 64 bit code than 32bit.
Daily OS operation is NOT cpu intensive thus OS operation will not show much improvement.

Your Theory is wrong in what you think the performance increase will be when moving to 64 bit as being only 10%. This is not across all applications only a hand full will see a small increase because they arent CPU intensive to begin with.

Look at the performance increase of the Apple G5 compared to its previous 32 bit chip. There is a lot more than 10% in there for it going to 64 bit.

64bit mode games will perform around 10-30% faster since the video card is the bottleneck.

I can only assume you favor Intel chips thus the reason to debunk actual performance increases seen in 64 bit. I look forward to all of you changing your tune once Xeon64 is out. Then will you all admit x86 64 bit is all that? Yes you will begin to tout 64 bit as being as incredible as I am seeing.

Major Speed Improvements were seen with G5 and Itanium2 going to 64 bit. The same is seen in AMD 64. The same is expected in Xeon64. 64 bit is all that.

Now I expect some more comments from those who dont own a 64 bit CPU running some 64 bit OS with 64bit apps to tell me why my machine is not as fast as it is.

This is not a Intel vs AMD thread its X86 and 64 bit performance advantages on CPU intensive applications. If you dont like AMD then simply wait for Xeon64 I am not here to tell you who is better I am telling you 64 bit is great.

 

[jhu]

Originally posted by: Vee

Originally posted by: Gamingphreek
The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

X86-64 is only x86 in 32-bit mode. The long mode provides what is essentially a new ISA, that can use registers much better.

i'm not really sure you can call it a new isa. looking at the instruction encoding, most of the instructions look pretty much the same

 

[jhu]

Originally posted by: Nemesis2038
I will say it again since your missing the point. Reguardless if you want to compare Itanium2 to Opteron.
It will be funny to see how everyone will be changing their tune about 64 bit performance once the Xeon 64 is here.

FACTS ABOUT x86 64 BIT MODE
My Opteron is Nearly 3 times faster than my P4 processor and around 4 times faster than my Athlon XP 3300+ when compressing files using the Opteron in 64bit mode. This makes my Opteron = ~9Ghz P4.

Video conversion the Opteron in 64 bit mode is Nearly 2 times faster than my P4 and Athlon XP.
The means my Opteron is around the performance of a my P4 if it were capable of around 6Ghz.

These are CPU intensive operations not bottleneck video card gaming benchmarks.
64bit is much faster when using 64 bit code than 32bit.
Daily OS operation is NOT cpu intensive thus OS operation will not show much improvement.

Your Theory is wrong in what you think the performance increase will be when moving to 64 bit as being only 10%. This is not across all applications only a hand full will see a small increase because they arent CPU intensive to begin with.

Look at the performance increase of the Apple G5 compared to its previous 32 bit chip. There is a lot more than 10% in there for it going to 64 bit.

64bit mode games will perform around 10-30% faster since the video card is the bottleneck.

I can only assume you favor Intel chips thus the reason to debunk actual performance increases seen in 64 bit. I look forward to all of you changing your tune once Xeon64 is out. Then will you all admit x86 64 bit is all that? Yes you will begin to tout 64 bit as being as incredible as I am seeing.

Major Speed Improvements were seen with G5 and Itanium2 going to 64 bit. The same is seen in AMD 64. The same is expected in Xeon64. 64 bit is all that.

Now I expect some more comments from those who dont own a 64 bit CPU running some 64 bit OS with 64bit apps to tell me why my machine is not as fast as it is.

This is not a Intel vs AMD thread its X86 and 64 bit performance advantages on CPU intensive applications. If you dont like AMD then simply wait for Xeon64 I am not here to tell you who is better I am telling you 64 bit is great.

there's nothing inherent about 64-bit computing that makes it faster. these new 64-bit processors are faster due to architectural changes from previous generations. compare an athlon xp to a sun ultrasparc i. the athlon xp, despite being 32-bit, will trounce the ultrasparc i even though it's a 64-bit chip. on the otherhand, you could use the 386 as your explanation for why 32-bit computing is dog slow.

i beg to differ with your name dropping of itanium2. the itanium line was already 64-bit. however, the first generation itaniums just sucked compared to intel's best 32-bit x86 processor during that time.

 

[glugglug]

Originally posted by: jhu

Originally posted by: Vee

Originally posted by: Gamingphreek
The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

X86-64 is only x86 in 32-bit mode. The long mode provides what is essentially a new ISA, that can use registers much better.

i'm not really sure you can call it a new isa. looking at the instruction encoding, most of the instructions look pretty much the same

Yea, it really is pretty much the same ISA, just new registers added and existing registers widened.

In fact, one of the architectural improvements in the AMD64 chips is that it takes advantage of the fact that only esp and ebp x86 registers can (and generally are) used for complex indirect addressing modes and pre-fetches the data surrounding these pointer values when the memory bus is idle, even if instructions referring to that data haven't made it into the pipeline yet.

 

[LTC8K6]

How about some actual benchmarks Nemesis2038? Unless you can provide them, you are just talking out of your arse.

Which P4 vs which Opteron? What programs? What were the times?

I would think that 200%-300%-400% speed improvements would have been big very news for AMD. Was this news reported anywhere? Is there a page I can look at that verifies your claims?

Shouldn't Opterons be compared to Xeons anyway?

 

[Jeff7181]

While I am a "fan" of the Athlon-64 and the move toward mainstream 64-bit CPU's, I really wish some people would quit making these outrageous claims about three times the performance and crap like that. All that shows is that you don't really understand what 64-bit computing is and what the Athlon-64 is. The simple transition to 64-bit computing is not going to increase computing power three fold... it would be rediculous for someone to think that.

Listen REAL close... in the simplest definition of 64-bit computing, the ONLY benefit is being able to work with larger integers. The instruction set, the number of registers, the amount and speed of the cache, and all the other features on the Athlon 64 are specific to the Athlon-64 and the x86-64 specification AMD has created.

People are using "64-bit" and "Athlon 64" interchangeably, and that's incorrect. All Athlon 64's are 64-bit processors, but not all 64-bit processors are Athlon 64's. So when you're talking about the Athlon 64 or Opteron, you can't make a blanket statement and say "64-bit computing isn't necessary or beneficial yet" with the intent to persuade someone not to upgrade to an Athlon 64 because that's not all the Athlon 64 is. It has powerful 32-bit capabilities, support for new instructions like SSE2, technological advances like an on die memory controller and SOI.

There is no one best solution for everyone. The Athlon-64 IS a good processor... the fact that it's 64-bit capable and there's no 64-bit software shouldn't effect your decision at all. If it does, you're obviously holding onto some sort of bias. Look at how it performs in a 32-bit environment. Does it perform well? Absolutely, it's one of the fastest processors you can get your hands on. Does it have a good performance per dollar ratio? Not necessarily... it's comparable to the competition, but the Athlon XP is a MUCH better value.

I just can't even put into words how utterly stupid it sounds to me that someone would NOT buy an Athlon 64 simply because 64-bit computing isn't a requirement yet. I don't care if it's a 1024-bit processor... it's "64-bitness" is not why people are buying it now... they're buying it because it's a superior product that offers a very nice looking upgrade path. If you can't see that then well... buy a Prescott or a Celeron or a Athlon XP... or hell... just switch to a Mac so you don't have to think about which processor you want... you just take whatever they have to offer.

 

[Gamingphreek]

Originally posted by: Nemesis2038
I will say it again since your missing the point. Reguardless if you want to compare Itanium2 to Opteron.
It will be funny to see how everyone will be changing their tune about 64 bit performance once the Xeon 64 is here.

FACTS ABOUT x86 64 BIT MODE
My Opteron is Nearly 3 times faster than my P4 processor and around 4 times faster than my Athlon XP 3300+ when compressing files using the Opteron in 64bit mode. This makes my Opteron = ~9Ghz P4.

Video conversion the Opteron in 64 bit mode is Nearly 2 times faster than my P4 and Athlon XP.
The means my Opteron is around the performance of a my P4 if it were capable of around 6Ghz.

These are CPU intensive operations not bottleneck video card gaming benchmarks.
64bit is much faster when using 64 bit code than 32bit.
Daily OS operation is NOT cpu intensive thus OS operation will not show much improvement.

Your Theory is wrong in what you think the performance increase will be when moving to 64 bit as being only 10%. This is not across all applications only a hand full will see a small increase because they arent CPU intensive to begin with.

Look at the performance increase of the Apple G5 compared to its previous 32 bit chip. There is a lot more than 10% in there for it going to 64 bit.

64bit mode games will perform around 10-30% faster since the video card is the bottleneck.

I can only assume you favor Intel chips thus the reason to debunk actual performance increases seen in 64 bit. I look forward to all of you changing your tune once Xeon64 is out. Then will you all admit x86 64 bit is all that? Yes you will begin to tout 64 bit as being as incredible as I am seeing.

Major Speed Improvements were seen with G5 and Itanium2 going to 64 bit. The same is seen in AMD 64. The same is expected in Xeon64. 64 bit is all that.

Now I expect some more comments from those who dont own a 64 bit CPU running some 64 bit OS with 64bit apps to tell me why my machine is not as fast as it is.

This is not a Intel vs AMD thread its X86 and 64 bit performance advantages on CPU intensive applications. If you dont like AMD then simply wait for Xeon64 I am not here to tell you who is better I am telling you 64 bit is great.

I can not believe you keep saying this. Listen Carefully!!!!!!!

Your Opteron is no 3x as a P4. Your Opteron is no where near 2x as fast or even in most case not even .5x faster

PLZ do not post what you have no clue what youre talking about. Post only what you know how to do!!

Right now what your doing is setting a bad track record for yourself. Right now you have 2 Diamond Members 2 platinum members and 2 senior members telling you that you know nothing about what your talking about! Plz stop... post after you read a lot about microarchitecture!

Get it straight guy!

-Kevin

 

[Gamingphreek]

Originally posted by: Vee

Originally posted by: Gamingphreek
The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

X86-64 is only x86 in 32-bit mode. The long mode provides what is essentially a new ISA, that can use registers much better.

Yes the EPIC architecture combined with the IA-64 microarchitecture is better at Enterprise class computing then the A64 will every be.

The benchmarks I've seen lately, can't remember them all, but a lot of them comes from IBM, tell a different story. Not only is AMD64 a serious workstation threat against Itanium, on server benchmarks HP's latest 4-way Opteron system only trails 11% behind fastest ever 4-way Itanium system. That is utterly devestating considering the price difference. Frankly, Itanium is dead. Intel knows that too.

I disagree with your post. Good to know huh, lol.

The Itanium is in a COMPLETELY different class than the opteron. The Itanium has a 128 way instruction set, The max an opteron has is 8.

You cannot compare them!! Opteron is server, workstation class... Itanium II is enterprise class.

-Kevin

 

[Nemesis2038]

So what your saying is these members who do not currently own a 64 bit processor to verify for themselves know more than a person owning a 64 bit proc doing the encoding. Maybe in your 32 bit divx world that is not true in my HDTV conversion world. 3 TIMES FASTER THAN MY P4 at 3.3Ghz.

Im through arguing and explaining. You will see for yourself in 3 months hopefully when MS releases XP64 and this encoder is released. It all hinges on Billy G.

As well as File Compression is done on 64 bit level is also significantly faster.

This thread is not worth my time.

In the mean time please lock the thread The members of this forum not using 64 bit cpu's must be smarter than me somehow. I am back to smoking pixie dust apparently.

 

[Nemesis2038]

As to the benchmarks. Something called an NDA? You do know what they are? You do understand why a company doesnt want you releasing benchmarks on beta software on a beta OS right? Ask anand or get yourself on a beta list and find out for yourself why they dont want benchmarks released on beta products. If anything its only going to get faster.

 

[jhu]

Originally posted by: Gamingphreek

Originally posted by: Vee

Originally posted by: Gamingphreek
The Itanium 2 uses a completely different Micro-architecture, and a completely different 64 code. While the AMD, and Intel CPU's (P4, P3, K6, K7 etc...) run on the x86 architecture or an extension of it the Itanium 2 runs on the EPIC (Explicitly Parallel Instruction Computing).

X86-64 is only x86 in 32-bit mode. The long mode provides what is essentially a new ISA, that can use registers much better.

Yes the EPIC architecture combined with the IA-64 microarchitecture is better at Enterprise class computing then the A64 will every be.

The benchmarks I've seen lately, can't remember them all, but a lot of them comes from IBM, tell a different story. Not only is AMD64 a serious workstation threat against Itanium, on server benchmarks HP's latest 4-way Opteron system only trails 11% behind fastest ever 4-way Itanium system. That is utterly devestating considering the price difference. Frankly, Itanium is dead. Intel knows that too.

I disagree with your post. Good to know huh, lol.

The Itanium is in a COMPLETELY different class than the opteron. The Itanium has a 128 way instruction set, The max an opteron has is 8.

You cannot compare them!! Opteron is server, workstation class... Itanium II is enterprise class.

-Kevin

what do you mean the itanium has a 128 way instruction set? you might be referring to the number of registers, in which case, the itanium does have 128 registers whereas the opteron has 16.

also, you're not the only mentioning the words server, workstation, and enterprise without actually being able to differentiate between them all. as such, i would argue that such distinction are arbitrary and rather meaningless since most modern processors can handle "server", "workstation", and "enterprise" applications without any problems.

 

[Gamingphreek]

First to nemesis. DO NOT TALK ABOUT WHAT YOU DO NOT KNOW!!! Your processor is no where near 3x as fast. End of discussion you have no clue what your talking about!

Ok ,

Server- Just a standard Xeon class server, no special things, used in schools, buisnesses, etc.

Worstation- Again basically the same as server only its a single PC thats being run by the server, government, computers hooked up to a central server.

Enterprise- The top dollar system, run mission critical operations, scientific caliber, sometimes used in server farms. $$$$$

The Itanium has a 128way instruction set, where as the Opteron has a max of an 8 way. Hence the name 240 for a 2way 440 for a 4way and 840 for an 8 way...lol i might have missed one along the line somewhere. Ok, im getting a bit out of my legue here, i just remember reading it at the end of the AT review when the opteron came out. It went something like this. "we would be very interested to see how the Opteron competes against the Itanium 2. Intel however does not want this done, not because they are affraid but because the Itanium 2 is out of the opterons league with its 128way instruction set and the opterons measly 8". Again that quote may be butchered a bit, but if you search around for the review on the opteron, youll find it on the last page of the last part of the review .

-Kevin

 

[glugglug]

Vee: is "52-bit" addressing a clean linear mode or is that segmented crap like PAE to get the extra 4 bits?

 

[jhu]

Originally posted by: Gamingphreek
First to nemesis. DO NOT TALK ABOUT WHAT YOU DO NOT KNOW!!! Your processor is no where near 3x as fast. End of discussion you have no clue what your talking about!

Ok ,

Server- Just a standard Xeon class server, no special things, used in schools, buisnesses, etc.

Worstation- Again basically the same as server only its a single PC thats being run by the server, government, computers hooked up to a central server.

Enterprise- The top dollar system, run mission critical operations, scientific caliber, sometimes used in server farms. $$$$$

The Itanium has a 128way instruction set, where as the Opteron has a max of an 8 way. Hence the name 240 for a 2way 440 for a 4way and 840 for an 8 way...lol i might have missed one along the line somewhere. Ok, im getting a bit out of my legue here, i just remember reading it at the end of the AT review when the opteron came out. It went something like this. "we would be very interested to see how the Opteron competes against the Itanium 2. Intel however does not want this done, not because they are affraid but because the Itanium 2 is out of the opterons league with its 128way instruction set and the opterons measly 8". Again that quote may be butchered a bit, but if you search around for the review on the opteron, youll find it on the last page of the last part of the review .

-Kevin

ok, that's a good distinction. however, as i said before, practically all modern cpus today can handle all three of those. just look at www.top500.org. most of those supercomputers contain p4s, xeons, opterons, itaniums, ppcs, etc (i think there's even an athlon in there somewhere).

the second thing you're talking about is multiprocessor systems.

 

[Vee]

Originally posted by: Gamingphreek

Yes the EPIC architecture combined with the IA-64 microarchitecture is better at Enterprise class computing then the A64 will every be.

The benchmarks I've seen lately, can't remember them all, but a lot of them comes from IBM, tell a different story. Not only is AMD64 a serious workstation threat against Itanium, on server benchmarks HP's latest 4-way Opteron system only trails 11% behind fastest ever 4-way Itanium system. That is utterly devestating considering the price difference. Frankly, Itanium is dead. Intel knows that too.

I disagree with your post. Good to know huh, lol.
The Itanium is in a COMPLETELY different class than the opteron. The Itanium has a 128 way instruction set, The max an opteron has is 8.
You cannot compare them!! Opteron is server, workstation class... Itanium II is enterprise class.
[/quote]
Disagreeing is ok, up until the moment the fighting starts.
I argue you can compare them. The important property deciding what the CPU can do is 64-bit addressing. That's the ticket.
(I'm not sure I understand what you mean with "128 way instruction set", but I assume I understand what you're refering to, and my answer is based on that assumption. If I'm wrong, please explain in more detail.)
Do you happen to know how many 8-plus-way Itanium systems have been sold? I don't know, I'm just curious.

Somewhere, there's always some company with someone without all paddles in the water, tasked with buying computers, with a large budget. And who's going to criticize him for buying Intel, right? My guess is that those are the guys Intel keeps up this "enterprise class" for.

There are considerable differences between Itanium and Opteron in architecture, yes. There's a considerable difference in die size and cost. But there is no difference in capability, in favor of the Itanium, that is significant enough for most of the market to care about. I'm betting on the relentless march of lower end technologies, conquering higher echelons of computing. If you think Itanium2 has some important advantage, I have to ask you what that is. I believe dual core Opterons and corresponding Intel x86-64 dual core CPUs are going to be the final nails in the coffin.

Also, hyper transport seem to be very useful and flexible, - you might want to check out this document:
http://www.lanl.gov/orgs/ccn/salishan2003/pdf/camp.pdf
Is that "enterprice class" enough for you?

In the somewhat limited way I understand these things, I'm inclined to think that Explicit Parallel Instruction Computing is a very complicated, troublesome and horrendously expensive -still incomplete- solution to a problem Intel once were very afraid of, branch prediction in a superscalar architecture. But which happened to never materialize as the crucial bottleneck in reality. High bandwidth and Out of Order execution beats EPIC.

 

[jhu]

speaking of which, do you ever wonder why there aren't any official spec benchmarks for the original itaniums? because they were worse than alpha at the time. now that alpha isn't being developed anymore, out come the itanium2 benchmarks.