A64 may release in 1st quarter 2004. They are having problems going past 2.0 Ghz.
do you really think its going to have an impact on the typical end user? Why would a user need the 64bit part of the processor?
I think the main benefit is more than 64bit math. Especially since the default operand size is 32bits. I think most of the benefits will be seen from the 16 extra registers (double the current amount) and the low latency memory controller.
It does need to be at a competitive speed and after looking at the <A HREF="http://www.xbitlabs.com/articles/cpu/display/athlon64.html" target="_new">xbitlabs</A> article 2ghz+ is necessary.
I think it needs a <A HREF="http://www.tech-report.com/etc/2003q2/athlon64/index.x?pg=1" target="_new">dual channel memory controller</A> to keep up with the P4. Then again, the Athlon has done well so far without a dual channel controller. It does seem that you need 5% here and there just to stay competitive.
I'm guessing <A HREF="http://www.theinquirer.net/?article=9216" target="_new">TheInquirer</A> article is the basis for this thread.
Dichromatic for your viewing plesure...
i dont think there are any probs with A64... i spoke to an amd techie a few weeks back and he was very confident about a september release date at ove 2GHz
The extra registers are all 64-bit, I don't understand why the performance boosts when enabling 64-bit, could come from these expansions than 64-bit computing itself, as imgod2u stated, if from what I've seen of these reg., they are all 64-bit and not 32-bit.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
Shots in the dark me thinks, AMD stated that 32bit code wount benfit from the 64bit ententions what so ever. So why everyone is saying "oh ya itll run 32bit code twice as fast" is beyond me.
-Jeremy
<A HREF="http://service.futuremark.com/compare?2k1=5341387" target="_new">Busting Sh@t Up!!!</A>
<A HREF="http://service.futuremark.com/compare?pcm=1060900" target="_new">Busting More Sh@t Up!!!</A>
The 64-bit instructions were a decent idea for a company such as AMD. They do not have the market share to move anyone to a new instruction set. Intel already has a 64-bit processor with a new instruction set and will not license it to them. Their only option is, indeed, what they are now doing. It's not so much vision as it is the only step they can take to enter the 64-bit market. You do what you must do.
Where I believe they made their mistake, and only time will tell, is to sell 64-bit silicon to the home market at this point in time. While others are using that silicon to speed up 32-bit operations, offer more cache, etc., AMD will be using a large portion of it for 64-bit operations that will go largely unused in most home systems for a very long time. That silicon could have been put to better use. As it stands currently you will be paying for that portion of the processor whether you use it or not. Perhaps they could not afford to maintain a separate 64-bit processor line. Who knows...
-Raystonn
= The views stated herein are my personal views, and not necessarily the views of my employer. =
When you load a 32-bit operand into a 64-bit register it is sign extended to 64-bits. When you further store a 64-bit register to a 32-bit operand it is truncated. You need not store extra bits in memory if they are not necessary to your task at hand.
Dichromatic for your viewing plesure...
They can use as a 32 register the same way I2 can use there 82 FPU reg for 32 ops on FPU
[-peep-] french
That come from Inquirer dont be take with caution.
Yes 2.6 GHZ was the intial target as benn drop to 2.4 maybe they wil drop it again to 2.2 ghz.
SOI process have never reach more that 1.8 ghz wich is opteron.No SOI from IBM have never reach over 50% yeild bad compare to FAB 11-X at 89% on 0.13 300mm.
If true presscott will be delay again maybe Nocona will be release in the 1 place.
[-peep-] french
64-bit extensions on AMD's silicon take no more than 10%, this is by no means a waste of our money Ray. It's in a fact a near-freebie. And they are not advertising 64-bit for the desktop, they will for server, but home users will likely not have that advertising. If it does, who cares, if you got Win 64-bit like 2003 Server will support, you will have support and be using some 64-bit despite not using the real amount that 64-bit can do.
It is already a powerful CPU, the Opteron at 32-bit is a powerhouse, and has sometimes IPC over 40% per clock. How in the world you see that as a silicon who has some wasted space is beyond me.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
Stating that the IA64 and the A64 have similar register models is quite a leap of reasoning.
In integer the IA64 has a base 32 register set and a framed 96 register set. The 128 floating point register are visible to all processes at all times, however, you are forced to run in a multiply accumulate mode. There is no integer multiply; and thusly you have to use the floating point unit for integer multiplies.
The A64 is like x86, with an extra 8 general purpose registers and an extra 8 xmm registers, an instruction prefix needed to load store 64-bit operands. The exception to this is explicit operations that use addressing, i.e. pointers and stack access.
The only thing IA64 and x86-64 have in common is the 64 in their name and the fact that there is an x86 compatibility mode in IA64.
Dichromatic for your viewing plesure...
WHO i nerver say that IA-64 register is the same model that A64 far from it.
I was just saying that you can use the extra reg without 64 data.You can store 32 data in the new reg as long the code is recompile.
[-peep-] french
I'm sorry.
What is the relevance of your analogy with the IA64?
Why are we talking about the IA64?
Dichromatic for your viewing plesure...
is all 4858 bit or any others wide register can allwayse be use as 1 32 bit reg like AMD64 or IA64.
Also FPU reg in Ia64 also have 32 static and 96 rotatif
[-peep-] french
Again, what is the relevance of that with the A64?
Every chip on the market can load and store 32 bit floats as well as bytes, words and dwords.
Dichromatic for your viewing plesure...
| Quote :
|
Wait a minute. The Clawhammer version of the core was named Athlon 64. So name in it self is a big advertising saying it's a 64 bit processor.
KG
"Artificial intelligence is no match for natural stupidity." - Sarah Chambers
| Quote :
|
Wait a minute. The Clawhammer version of the core was named Athlon 64. So name in it self is a big advertising saying it's a 64 bit processor.
KG
"Artificial intelligence is no match for natural stupidity." - Sarah Chambers
Last ligne nothing you were saying that on Int or GPR that there have 32/96 setting i just make sure that everyone know that also the case of FPU.
[-peep-] french
Yeah however they will not be lying once a 64-bit Windows is out. No one is being lied to, I'm sorry but no one is! They are not in any position to being sued over false advertising if a Windows will use it.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
There's a rumor out there that we'll never see A-64. I think Inquirer started it. Essentially, the rumor goes that AMD will just push the Opteron as a desktop processor since it performs so much better than A-64 will...
<font color=green>The Netherlands is where you go when you're too good for heaven.</font color=green>
That was not a rumor, that was my suggestion.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
A good idea if you were on 300 mn wafer but AMD capacity of production is 1/10 of intel or Samsung.They will need to outsource a lot of production to IBM or TCMS.The 1 become will be the 1 contender.
Also the full 128 bit mothers will cost a lot deleting any capacity to run on Mobile or low cost or small case.
Option nb 1 is to disable the last controleur and flexibility issue appear greately.As Opteron and A64 wont use the same memory as DDR-2 should be use in sever only very late compare to desktop so you wil need a controleur for DDR1 and DDR-2 by the way not put QBM and RDRAM and SRAM.
[-peep-] french
Thats when Hammer. No Clawhammer. No how about Nohammer. Oh well lets just call it Opteron and the heck with Hammer. Maybe we will see it next year.
Actually juin there is a rumor that the 0.9 version of Opteron and A64 in the first half of next year will include a ddr-II controller
It foolish to think AMD can have 90 nm only 1 Q after intel/Samsung/IBM/Toshiba.
DDR-2 and a DDR-1 controleur hmm lack of flex start to show
[-peep-] french
Well thats the rumor, and remember IBM is helping AMD with its manufacturing so it isnt too far fetched
IBM is late compare to Intel on 90nm Samsung and toshiba are faster also that IBM.
[-peep-] french
Maybe with IBM's help they will have 90nm by end of next year. Thats when A64 will be released finally!!!!
There are 1153 identified and unidentified users. To see the list of identified users, Click here.
You are about to answer a thread that has been inactive for more than 6 months.
If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.

