AOA Forums AOA Forums AOA Forums Folding For Team 45 AOA Files Home Front Page Become an AOA Subscriber! UserCP Calendar Memberlist FAQ Search Forum Home


Go Back   AOA Forums > Hardware > AMD Motherboards & CPUs

AMD Motherboards & CPUs Questions or comments on AMD products?


Reply
 
LinkBack Thread Tools Rate Thread
  #1 (permalink)  
Old 27th March, 2002, 05:45 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

What is the big deal about Hammer and X86-64 anyway?

What is the big deal about Hammer and X86-64 anyway? You might be surprised to learn that that 64 number doesn’t mean much to you and I! Then again some of you might not!

So then what is the big deal? The big deal is extra registers baby!

WTF is a register? They’re small parts of the CPU were values are stored. For example: X=1. The advantage to registers over system memory and cache is that registers can both be read(loaded) and written to(stored) in a single clock cycle. Even the P4’s super low latency 16K L1 cache is slow by comparison, it takes two clock cycles just to read the cache, but it also takes two clock cycles to read it! That makes registers 8 times faster! What’s even scarier is that registers are 32 times faster than the Athlon’s 64K L1 cache!

How many registers do X86 CPUs have? Technically all X86 CPUs have 8 general-purpose registers, and AFAIK all X87 FPUs have 8 registers of their own. This is a limitation of the X86 instruction set.

Let’s apply this information to the Athlon core. The Athlon has the capability to decode and execute three instructions per clock cycle. Let’s throw three virtual instructions at a virtual Athlon.

Code:
1. A+B=C
2. D+E=F
3. G+H= . . . . wait, that is already 8 pieces of data, whatever am I supposed to do with this?
So as you can see from this gross over-simplification, the low number of registers limits the real-world IPC(instructions per clock-cycle) rate, even with the most superscalar CPU design(the more IPC the core can decode and execute the more “superscalar” it is.)

However, fortunately for us there is more to it than that. Back when Intel introduced the P6 core(Pentium Pro, P2, Celeron, P3) they started using something called “register renaming”, AMD did the same with there K6/K7 cores. So what is register renaming? It basically allows for the addition of more registers by shuffling around registers between reads/writes, this happens transparently to the program code(the Athlon has something to the tune of 24 “hidden” registers). So with register renaming, there was hope yet for the superscalar designs used by the P6, P7(Pentium 4), and K7. But when all is said and done the K7 still only manages a real-world IPC rate of about 2.1! Without register renaming the Athlon would be lucky to manage a 1.5 IPC rate.

So how many registers will the Hammer(K8) have? First off, the 16 existing 32-bit registers will be extended to 64-bit, these registers will however still be able to be used as 32-bit registers. X86-64 calls for the inclusion of 16 new registers. 8 will be the general-purpose kind we so dearly need. 8 will be SIMD(Single Instruction Multiple Data, these are specifically for SSE, SSE2, and 3dNow!) only registers for the FPU. As for the “hidden registers” used for register renaming, your guess is as good as mine is! If I had to guess, I would say there will be 32 hidden registers(but that is pure speculation, there could be half that many).

So will doubling the number of registers(from the compilers point of view) double performance? Unfortunately, the answer is no. But, we can expect X86-64 applications to run 10-15% faster than there bog standard X86 counterparts, largely because of the extra registers.

Well that is it on Hammer for now. I’m going to look more in to Hammer’s decoding and executing capabilities later, I’ll also look more in to it’s instruction buffer size, and I’ll try and write something up for you guys(if I find anything worthy of a write-up).
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #2 (permalink)  
Old 27th March, 2002, 07:17 AM
AmbientFiction's Avatar
Member
 
Join Date: September 2001
Posts: 311

Very nice Dimm. I did not know that but now I do. I guess it is true you learn something new everyday. I see that as a large selling point of the x86-64 lines. I am not sure how well people are going to take to them seeing as how we just bought our XPs. I think it is rather messed up to time things the way that they have. I see it as a flaw in the market and they could be making much more money if they had have just put out the x86-64s rather than the XP and MP lines. I am not sure but I don't think AMD is good when it comes to timeing things.

BTW is DDR going to be supported(IE by new boards or by the controller) or are we going to have to go to DDRII. If we are being forced to buy DDRII I think a lot of people are going to be mad at AMD.
__________________

AMD XP 1700+ @ 1639mhz (Bad vid card can't hold high FSB)
Epox 8hka+
Aska Silver Mtn 40ºC full load folding 24/7
60gb Seagate 7200rpm
60gb Seagate 7200rpm
GF2MX400 w/ updated bios modded with 23cfm fan and AS2 under the heatsink on the GPU
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #3 (permalink)  
Old 27th March, 2002, 07:43 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

Quote:
Originally posted by AmbientFiction
BTW is DDR going to be supported(IE by new boards or by the controller) or are we going to have to go to DDRII. If we are being forced to buy DDRII I think a lot of people are going to be mad at AMD.
Yes, DDR will be the only memory for the K8. The DDR memory controller will be integrated in to the CPU die to drastically reduce memory latency. Many believe the memory controller will be of the dual-channel variety, this would be nice. So far AMD has only indicated support for PC2700 memories(and the older, slower, less attractive DDR options), but with VIA and other chipset makers bent on making PC3200(DDR400) supporting chipsets for the Athlon and P4, AMD may still elect to bung in support for it at the last minute.
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #4 (permalink)  
Old 27th March, 2002, 07:50 AM
AmbientFiction's Avatar
Member
 
Join Date: September 2001
Posts: 311

Very nice cause I can just see some guy with about 2 gigs of DDR cussing out AMD for not allowing DDR and just going right to DDRII. H3ll even I would be mad.
__________________

AMD XP 1700+ @ 1639mhz (Bad vid card can't hold high FSB)
Epox 8hka+
Aska Silver Mtn 40ºC full load folding 24/7
60gb Seagate 7200rpm
60gb Seagate 7200rpm
GF2MX400 w/ updated bios modded with 23cfm fan and AS2 under the heatsink on the GPU
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #5 (permalink)  
Old 27th March, 2002, 08:19 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

DDRII is good technology, and as happy as AMD will make a handful of consumers, they will piss off memory manufactures a great deal more.

DDRII just won't be ready for the Hammer launch date(Q4 of this year). Expect AMD to implement it(or something faster?) in 2005/6 with the K9 though.

Here is an interesting tid bit:
Today's X86 CPUs can support only 4GB of memory. An 8-way Hammer system will be able to support 128GB of memory!
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah

Last edited by dimmreaper; 27th March, 2002 at 08:22 AM.
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #6 (permalink)  
Old 27th March, 2002, 09:42 AM
Daniel ~'s Avatar
Chief BBS Administrator
 
Join Date: September 2001
Location: Seattle Wa.
Posts: 45,606

You just got AOAed to our Main page!":O}
__________________
"Though all men live in ignorance before mystery,
they need not live in darkness...
Justice is foundation and Mercy ETERNAL
."
DKE

"All that we do is touched by Ocean
Yet we remain on the shore of what we know."
Richard Wilbur

[img]/forum/attachments/random-nonsense/16515-sigs-dan_drag.jpg[/img]
Subscribers! Ask Pitch about a Custom Sig Graphic

Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #7 (permalink)  
Old 27th March, 2002, 12:54 PM
Superman's Avatar
Systems Administrator
 
Join Date: September 2001
Location: WI
Posts: 4,406
Send a message via ICQ to Superman

Quote:
Originally posted by Daniel ~
You just got AOAed to our Main page!":O}
Hey, a new verb, LOL!

EDIT: Good read, also, Jeff.
__________________
"And, most of all, remember this descendant of David who beat the hell out of death."
-from the book "Six Hours One Friday" by Max Lucado

"You have to go outside the sequence of engines, into the world of men, to find the real originator of the rocket. Is it not equally reasonable to look outside nature for the real Originator of the natural order?
-C.S. Lewis


Director of JavaScript section of the Allied Sites Support Team, web designer and programmer for DaOCPlace, and co-web designer and programmer for AOA Files

Avatar by Epox Tech



<>< I Believe-Do You?
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #8 (permalink)  
Old 29th March, 2002, 05:41 AM
martin's Avatar
Member
 
Join Date: October 2001
Location: Oshawa, just east of Toronto
Posts: 86

nice but....

You had posted a msg a few weeks ago on the clawhammer and I went to AMD's site to get some background on it. Yes, 64 bit is the eventual progression of the cpu and a start will definitely have to be more registers. I remember back in 82' when we studied the 8085 microprocessor in DeVry, the register's importance was harped on quite a bit but the data lines were equally important because if you can't get that data from A to B quickly with each clock pulse...well you're beat. Until an extra 8 data lines can be implimented and taken advantage of...then it's just a big friggin bottleneck for sure right?
__________________
If you want to feel- then you have to do....


"Why do a single thing today, when tomorrow's surely out there. But the days they turn into years and still no tomorrow appears....
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #9 (permalink)  
Old 29th March, 2002, 05:43 AM
Superman's Avatar
Systems Administrator
 
Join Date: September 2001
Location: WI
Posts: 4,406
Send a message via ICQ to Superman

Re: nice but....

Quote:
Originally posted by martin
You had posted a msg a few weeks ago on the clawhammer and I went to AMD's site to get some background on it. Yes, 64 bit is the eventual progression of the cpu and a start will definitely have to be more registers. I remember back in 82' when we studied the 8085 microprocessor in DeVry, the register's importance was harped on quite a bit but the data lines were equally important because if you can't get that data from A to B quickly with each clock pulse...well you're beat. Until an extra 8 data lines can be implimented and taken advantage of...then it's just a big friggin bottleneck for sure right?
What DeVry did you go to?
__________________
"And, most of all, remember this descendant of David who beat the hell out of death."
-from the book "Six Hours One Friday" by Max Lucado

"You have to go outside the sequence of engines, into the world of men, to find the real originator of the rocket. Is it not equally reasonable to look outside nature for the real Originator of the natural order?
-C.S. Lewis


Director of JavaScript section of the Allied Sites Support Team, web designer and programmer for DaOCPlace, and co-web designer and programmer for AOA Files

Avatar by Epox Tech



<>< I Believe-Do You?
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #10 (permalink)  
Old 29th March, 2002, 07:07 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

Re: nice but....

Quote:
Originally posted by martin
I remember back in 82' when we studied the 8085 microprocessor in DeVry
Microprocessors have come a long way since the 8085

The 8085 must not have been pipelined. I know it certainly wasn't superscalar. And I'm quite sure they didn't use decoders to break the X86 instructions down in to simpler RISC-like instructions for the execution units. :-D

8085 vs. K8 . . . apples and oranges, hell kiwis and pineapples . . .
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #11 (permalink)  
Old 29th March, 2002, 07:19 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

Re: nice but....

Quote:
Originally posted by martin
if you can't get that data from A to B quickly with each clock pulse
Pipelines eliminate that problem, they allow the clock speed to scale(MHz) despite however long(pico seconds) it takes a certain portion of the CPU to complete it's job. The down side is that the execution latency(in clock-cycles) is higher.

A non-pipelined non-superscalar CPU takes in one IPC, and spits out the result of that instruction in a single clock-cycle. A non-superscalar 6-stage pipelined CPU takes in one IPC, and spits out one result every clock-cycle, but it takes 6 clock-cycles for a particular instruction's result to be spit out, but the pipeline lets the CPU run roughly 6Xs faster. See how pipelines work?
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #12 (permalink)  
Old 29th March, 2002, 07:49 AM
martin's Avatar
Member
 
Join Date: October 2001
Location: Oshawa, just east of Toronto
Posts: 86

Sure thing, I was too obvious with my description but the issues still remain from those old days. Today there's pipelines, is there a newer model ready to go if and when the 0.09u technology goes into production?
__________________
If you want to feel- then you have to do....


"Why do a single thing today, when tomorrow's surely out there. But the days they turn into years and still no tomorrow appears....
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #13 (permalink)  
Old 29th March, 2002, 08:25 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

I'm afraid pipelines are the way for the foreseeable future. Smaller transistors will of course make everything faster, but design wise pipelines are it.

I've been reading up on molecular computing, and even there they seem to be leaning toward pipelines. Various research groups have had luck creating simple 2 and 3 stage pipelines from what I hear, but nobody has a clue how to go about making superscalar molecular-microprocessor.

Intel's IA-64 instruction set handles conditional branches in an interesting way that eliminates the branch prediction problem associated with super deep pipelines(beyond 30 stages). Instead of trying to predict the branch like X86 CPUs have to do, IA-64 CPUs simply assume the conditional branch is both true and false. It fills the pipeline with the instructions that would later be needed if it were true, and the instructions that it would need were it false. Then at the end of the pipeline it sorts it out, then knowing whether the conditional was true or false. Sort of a "kill 'em all and let God sort 'em out" approach, which will work well with superscalar designs with ultra-super-mega-deep pipelines(no I haven’t been watching the Power Rangers again). I think this approach may be adaptable to X86, maybe there will be an X86-64-2 someday.
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #14 (permalink)  
Old 1st April, 2002, 07:43 AM
Member
 
Join Date: April 2002
Posts: 3

molecular computing

dimmreaper,

You mentioned molecular computer. I've read that this is just one branch of quantum computing. It does have amazing potential. However, basic research still has to be done in the area of quantum physics. This is necessary before they can go anywhere with it anytime soon. I think when they finally do make it a reality, they will be using something similar to the magnetic resonance imaging (MRI) to control the molecules.

A new technology that may be develop sooner uses something called "photonic crystals." I first read about it in the December 2001 issue of Scientific American. I thought you might be interested in this. Here's a link to MIT's research regarding this exciting new field that may have applications in computer science.
http://ab-initio.mit.edu/photons/
__________________
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #15 (permalink)  
Old 1st April, 2002, 09:17 AM
dimmreaper's Avatar
Member
 
Join Date: September 2001
Location: road to insanity
Posts: 5,347
Send a message via ICQ to dimmreaper Send a message via AIM to dimmreaper

Cool, thanks for the link!
__________________
¤ Jeffery ¤

Wishing you were someone else is a waste of the person you are. - Kurt Cobain

AOA Team fah
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #16 (permalink)  
Old 1st April, 2002, 06:39 PM
Member
 
Join Date: April 2002
Posts: 3

No problem.

There is a solid background in basic research for this field. They know just about everything they need to about the properties of light. There has been a book published on photonic crystals already, written by those MIT people I believe. I've not read it, I think its a fairly new book.
http://www.amazon.com/exec/obidos/AS...650909-0059322
__________________
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #17 (permalink)  
Old 2nd April, 2002, 06:40 AM
martin's Avatar
Member
 
Join Date: October 2001
Location: Oshawa, just east of Toronto
Posts: 86

yes!

This is what I was thinking but didn't state it intelligently. That article was a very interesting read, waveguides and fiber optics fascinated me in college(my god, it's been 20 years this year), after reading up on that, i got to thinking that photons could be a great way to move obscene amounts of data.

Reading the article, I was especially interested in the descriptions of one-dimensional and 3D light movement. What a picture in my head, resonant cavities capturing light waves and redirecting them to all the points on a compass! All those little one's and zero's riding specific frequencies would be moving everywhere at the speed of...well, light.

Too wild! Our multimedia would move a hell of a lot faster then.
__________________
If you want to feel- then you have to do....


"Why do a single thing today, when tomorrow's surely out there. But the days they turn into years and still no tomorrow appears....
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #18 (permalink)  
Old 2nd April, 2002, 08:27 AM
Member
 
Join Date: April 2002
Posts: 3

It is very exciting! Its much more than the speed of the photon that makes it impressive. There are quantum properties of light that allow it to do amazing calculations. Photons can be split into mirror images and somehow this increases the number of possible calculations per cycle. It doesn't have the same level of potential as quantum computing with molecules, because molecules add rotation/spin as another factor. However, scientists and engineers have a much better knowledge base of how to control photons. Therefore, this concept is more likely to used in the near future. The article in Scientific American says, "photonic crystals, the semiconductors of light." With the right amount of investment from corporations we could see this technology soon. Of course the amount they invest depends on how long it takes for current semiconductors to reach their speed limit. Some estimations are at 15 to 20 years from what I've read. I might be optimistic, but with nanotechnology being the hot topic it is, I don't see why I shouldn't be.
__________________
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
Reply



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
New Castle or Claw Hammer 64Bitchin AMD Motherboards & CPUs 10 28th October, 2004 01:07 PM
24182 3Dmarks with AMD Hammer marcofra2000 Benchmarking 2 17th September, 2003 06:44 PM
Hammer Delay Reason Lionfish AMD Motherboards & CPUs 6 26th September, 2002 09:40 PM
Hammer already being fabbed? dimmreaper AMD Motherboards & CPUs 1 8th August, 2002 10:01 AM
Cray super-computers to use AMD Hammer? dimmreaper Random Nonsense! 19 14th July, 2002 04:01 AM


All times are GMT +1. The time now is 10:04 AM.


Copyright ©2001 - 2010, AOA Forums
Don't Click Here Don't Click Here Either

Search Engine Friendly URLs by vBSEO 3.3.0