Clawhammer... to be waaahaaahaay cheaper than IA64

My dads brother has a friend who said his mother told him that too.

rumors...

The Inquirer > Vans

That writhing background of 1's and 0's on Van's site makes me nauseous. Someone needs to tell him that animated gif backgrounds do not make repeat visitors.

"Laziness is a talent to be cultivated like any other" - Walter Slovotsky

The sledge hammer will be priced around there. But still probably not even 60%, less than half the price of Itanium.

<font color=red><i>Poor is the pupil, who does not surpass his mentor</i> - Leonardo daVinci</font color=red>

Intel is going to keep on making IA64 better. Well family Desktops with IA64 in 6 or 7 is reasonable to say. IA64 has a very llllooooonnnngggggg life.

Nice Nvidia and ATi users get a Cookie.... Yummy

Just like before when the world switched to 32 bit. I predict suddennly MS will release a 64 bit OS. Sure it'll run on 32 bit machines, like molasses in January. They'll have to convert on the fly etc. An analogy: sure you can run Win95 on a 386...

I honestly do not know why home users would switch their system to 64-bit processors (with accompanying motherboard, etc.) at all currently. We have yet to reach the limit on 32-bit processors. Our current motherboards do not even support the maximum amount of memory that our 32-bit processors are capable of supporting. Have you seen any motherboards with 4GB of memory or more?

Everyone must realize that programming on a 64-bit CPU makes use of 64-bit memory address (pointers.) There are so many of these types of variables in any application that you could end up nearly doubling the size of the executable without adding any new functionality to it at all. All of a sudden your processor's caches hold half of the amount of functional code as it used to, simply because of the expanding of the size of your application. If all your 64-bit processor offers is access to a 64-bit address space instead of a 32-bit address space (as the Hammer does), and you do not really need more than 4GB of memory for your application, then building your application as 64-bit will actually greatly hinder performance. It will require more cache to fit less of the functional application. It will require more memory bandwidth.

In essence, targetting a 64-bit platform increases most of requirements of your application. It does not inherently increase performance as most think. Targetting a 64-bit platform that offers no further benefits over a 32-bit processor than the presence of a 64-bit address space, without having the need for the extra memory, will only hurt performance.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

I agree. I'm talking about down the road here. 3 years, 10 years, whatever.

I confess I'm ignorant about the construction of the Hammer. Does it give more cahce (I'd assume so), and morebandwidth (that one is harder to guess).

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>

"We have yet to reach the limit on 32-bit processors. Our current motherboards do not even support the maximum amount of memory that our 32-bit processors are capable of supporting."

I think you hit the nail on the head. The jump is probably being made for addressing reasons. 4 Gigs is not far off, but obviously the 32 bit resolution is good enough for most mortal tasks.

It seems to me since we hardly really need 64 bit we could use addressing tricks to nurse 32 bit systems along for a while longer. After all won't just moving to a 64 bit architecture double costs? I'm assuming the folks that really need 64 bit already have it, and pay healthily for it.

Except for one small issue. There will not be any 64-bit applications for the home market within the next 6-12 months. You are paying for something you will not be using. Have fun. By the time 64-bit applications come around, that processor you bought will be dated, and will run both 32-bit and 64-bit applications much more slowly than the other processors out there.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

Except ray, that they most likely will make it affordable, and It will be the fastest thing out at the time, and there will be no competition in the 32 bit arena for it from AMD themselves, the 64 bit support will just be icing on the cake!

~Matisaro~
"The Cash Left In My Pocket,The BEST Benchmark"
~Tbird1.3@1.5~

Which is why I said "MOST LIKELY!!!!!!!!!!

I did not state anything as fact! I try very hard not to make statements I present come off as being fact when they are not, and I try to disclaim anything I say which is my oppinion. I know me and you have been butting heads lately but I dont appreciate being called a troll unjustly ray.

~Matisaro~
"The Cash Left In My Pocket,The BEST Benchmark"
~Tbird1.3@1.5~

I think the move to 64bit computing is more than just the linear address space. You could say that the 32bit computers are actually 48 bit computers as you have a segment address and a linear address. 16:[32] Moreover, in a Virtual Memory Model, the information in the linear address is actually split into [Directory]:[Page]:[Offset] parts. These parts use bits [31..22]:[21..12]:[11..0] respectively. They are combined to form a physical address by use of a page directory and table. The actual physical address comprised of a [Page Frame]:[Offset] where bits [31..12][11..0] are allocated to each part respectively. This gives a Meg 2^20 of page frames and a page granularity of 4k 2^12. From what I’ve read on the x86-64 (was typed as IA-64 Intel's codename sorry) there is a new 2meg page granularity mode along with the legacy 4k mode. A fair amount of the reserved space in the page descriptor pointer has been allocated to discerning this difference. I don’t know what effect a 512x page granularity will have but I would guess less page table loads vs more memory loss due to the lower segmentation.

Schmide

My view on the world is often clouded by my own misconceptions.


<P ID="edit"><FONT SIZE=-1><EM>Edited by Schmide on 09/28/01 12:55 PM.</EM></FONT></P>

In case every 1 here forgot about APPLE computers..its motorola processors have a 256-bit arch.
so a move towards 64.bit is not that far out as memory is just an adressing issue the code string length that the processor is able to process in a cycle increases so optimised programs codes would run much faster in less time, thus more effective use of CPU time for servers and future APPZ for even the home market.

just look at you graphic card, the GF3 is using a 128-bit GPU..

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>

"Except for one small issue. There will not be any 64-bit applications for the home market within the next 6-12 months. You are paying for something you will not be using."

Wasnt that the same thing almost a year ago with the Pentium4 and SSE2. Nothing really took advantage of it until only recently?

Blame the newbies not the technology

Right...and another thing....performance on 32 bit apps won't be significantly decreased unlike P4. Yes, P4 will perform better once software is *optimized* for it, but truth be told, I think it should have been released with performance per clock at least equal to the P3...but it wasn't Hammer, otoh, according to AMD, will offer much better performance than Athlon on 32 bit apps at the same clock speeds as Athlons. So, it will be a logical upgrade path.

Mark-

When all else fails, throw your computer out the window!!!

Then wouldn't you consider it safe to say that if the hammer is greater optimized for 32bit, that it would infact be slower than the IA64 in 64 bit code? Considering the IA64 was more or less Designed for 64bit strictly in mind. Where AMD designed theirs for both

Blame the newbies not the technology

Still, this does not change the fact we we do not currently use the memory space available to us on 32-bit platforms. In truth the only component that actually is still down at 32-bits is the processor. Upgrading to a 64-bit processor will not change the other components in your system that have already been upgraded past 32-bits. Simply moving to a 64-bit instruction set only serves to increase your code size. Unless you need the extra memory, it is actually a performance hit.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

"Wasnt that the same thing almost a year ago with the Pentium4 and SSE2. Nothing really took advantage of it until only recently?"

There is a rather huge difference here. Simply recompiling your application to use SSE2 instructions offers a huge performance increase. On the other hand, simply recompiling your application to use x86-64's larger address mode offers a performance decrease due to the increase in the size of the generated code. The only reason you would want to do the latter is if your application requires gigabytes of memory. Performance actually suffers slightly compared to 32-bit code run on the same processor.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

i have to correct you there raystone, even if companies would recompile to a 64-bit legth per code string, they wouldnt have coz 32 bit processors would take twice the time to split the code cach it and process it later, right now 32 bit is mainstream and anyway this is not the issue.

we never did reach the limit of 16-bit platform in the subject of available adreses, thogh most PC CPU manu' did change the core to a 32-bit one, i dont know about you bot the new'er the stuff we got out today the new'er things they can start developing/interdusing tomorow.

and SSE2 is not alike 64-bit processor core, as SSE instructions are like software installed into the CPU, the 64-bit issue is much more then that, even the memory issue is good for servers, more expandability on a single server point plus costume server apps to run on it.

and the new 64-bit platform would allow new slots operating at 64-bit ,which means faster more robust hardware.

i say the sooner the better..

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>

A 64 bit cpu is not necessary to have 64 bit "slots" (assuming you refer to the PCI bus). They exist, and are fully functional.

"Than why couldn't we had gigabytes of memory in the current 32bit setup and forget about 64bit if you lose some performance, or is that not possible?"

We could. Our current 32-bit processors support up to 4GB of memory per selector. There are 65,536 selectors available. This allows up to 262,144GB of memory. I have yet to see a motherboard targetted for home users that even uses the basic 4GB of memory.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

"even if companies would recompile to a 64-bit legth per code string, they wouldnt have coz 32 bit processors would take twice the time to split the code cach it and process it later"

Huh? The point of a 64-bit processor is 64-bit registers and a 64-bit address space. The instructions (code) are not increased in size to 64-bits. We do not have that many instructions in our processors to require such huge instructions. The data with which we work becomes 64-bit. The pointers, the integers, etc. all become 64-bit rather than 32-bit. This causes a decrease in performance relative to the performance of 32-bit code on the same processor.


"we never did reach the limit of 16-bit platform in the subject of available adreses"

Yes, we did. Programmers were being forced to implement overlays. An overlay is where you have multiple sections of code, but only keep one in memory at a time. This is useful in situations where you just do not have enough memory for your application. The limit of 16-bit processors was about 1MB at first. With the introduction of the 80286, it expanded to 16MB. Neither of these were truly enough memory for the applications that were under development. Were you around during the era of messing with the tweaking of free conventional memory? Did you ever fiddle with settings in himem.sys, emm386.exe, etc? Did you rejoice when DOS came with support for DOS=HIGH,UMB in the config.sys file? We were very much at the end of the rope for 16-bit address spaces.


"the memory issue is good for servers, more expandability on a single server point plus costume server apps to run on it."

I fully agree that 64-bit processors are great for modern servers. They deal with vast quantities of data. However, this has nothing to do with the home user.


"the new 64-bit platform would allow new slots operating at 64-bit ,which means faster more robust hardware."

Sorry to kill your hopes, but the word size of your processor has no relevance on the various slots on your motherboard. We have had 64-bit bus slots for a while now. SDRAM operates on 64-bit memory quantities. None of these require a 64-bit processor.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

Yeah I rejoiced, but then I freaked out when I had to go through DPMI to handle raw input and output. (I.e. keyboard, com, and dma). It was very awkward until the virtual device driver came out.

Schmide

God dang... that guy has NO concept of web design. His website sucks. Just looking at it hurts my eyes.

-MP Jesse

"Signatures Still Suck"