PDA

View Full Version : off topic (amd vs ppc)


Amdfan
Dec 8, 2003, 05:59 AM
All I have to say is wow!
Ok I have been reading the forums because my roommate wants a Mac, and I have been telling him that he should get an x86. Well I came to this posting and decided to read it. And well, WOW.
I would like to say that if you don’t know jack about computers don’t spew the lies from your mouth based on your stupidity.
I realize that I will get a lot of flames for this but I don't care because I feel that the truth must be said.

First off, DDR-2 533 is in no way shape or form pc4200.
Now to 64 bit, I love how the Apple people (you know who you are) were saying that the g4 was a 64 bit proc, based of something on the proc that was 64 bits wide. I know you will all deny it now but whatever.
---- This is off the Mac site from a year ago, made me laugh so hard!
"The fewer the steps, the shorter — and more efficient — the pipeline. Thanks to its efficient 7-stage design (versus 20 stages for the Pentium 4 processor) the G4 processor can accomplish a task with 13 fewer steps than the PC. You do the math."

Don’t get it? Hah Apple doesn’t seem to understand how a processor works and that all 20 stages are doing work at any given time, and that means MORE work is getting done with the P4. When will Apple learn not to LIE to people about there computers? I will agree with you guys that OS X is a good operating system, but I don’t think ill ever change from my Linux/windows system.
Speaking about OS X, wouldn’t it be nice if it was a 64 bit OS X? LMAO, note "64 bit extensions" does not mean that it is a 64 bit OS. Hence whey it can only utilize 8GB of ram, where as in an Opteron system you can use as much as you can physically put into it.
---- More Misinformation
"What makes a supercomputer “super” is its ability to execute at least one billion floating-point operations per second, a staggering measure of speed."

"The PowerPC G4 won renown as the first supercomputer on a chip, with a sustained performance of over one gigaflop. And it only gets better. The new Power Mac G4 desktop system with the dual 1.25GHz PowerPC G4 processor configuration hits speeds of over 18 gigaflops."
Yeah and well an AMD k6-2 did 3 gigaflops, and an INTEL 3.06 does 18, what’s your point?
---- Oh yeah how about this one.
“The G5 is the world’s fastest personal computer and the first with 64-bit processor”
Wonder why you wont find that on there site anymore.
http://spl.haxial.com/apple-powermac-G5/
Go there if you haven’t already.
Now onto the topic at hand. G5 revisions.
I fully understand how Apple will update on even numbers and going by that the fastest g5 to be released in January could be a 2.6 GHz model
Posted From sunnrunner
“Is it just me or does it seem like the IBM chip development pipeline is lighyears ahead of Motorola when it comes to speed of development and manufacture, (not to mention chip technology). Moto almost killed the Mac, it just wasn't a priority for them. IBM, on the other hand, is set to make Macs the fastest thing on a desktop once again”
Why would IBM make Apple the fastest thing on the desktop? Think for a second for me. IBM sells Intel computers and AMD computers, right? Ok, if IBM was to offer apple as the fastest desktop IBM would cut its profits. Nay, IBM doesn’t want Apple to be the fastest; they only want Apple to hang in the game.
As far as the speed bumps and Apple to make it to 3.0 GHz. I don’t believe that the current chip design will scale as fast as apple wants it to. The 16 stage pipeline of the G5 at 90 nanometers with SOI sauce and IBM’s processor technology means that they will need to do maybe 2 more revisions to the chip to get there. Assuming that they have no problems with those revisions they will make it. However I don’t think they will, because IBM has NEVER manufactured a chip at over 2 GHz and the pure heat of the chip will be a furnace! Assuming that the calculations of 70 W @ 2.5 GHz. At 3 GHz that would mean about 95 W !@!@!@!@! If you think that that computer will be quiet you must live in a vacuum. Of course I understand that with future revisions they WILL be able to get that number down, but by how much? I understand that Intel is also having a problem with there 90 nm tech, but you have to understand that they don’t not use SOI. And they are at 3.2 – 3.6 GHz.
For a more near clock for clock basis for comparing IBM’s process technology, compare AMD. AMD expects around 30 – 40 W when they move to 90 nm @ around 2.6 – 2.8 GHz. AMD also clock for clock is faster than the G5. I’m sorry guys but I hardly think Steve’s dreams of 3 GHz by the end of summer are reasonable, but they are possible.
Please if you respond to my post, post intelligent information and not just Apple Propaganda. I tried as had as I could to see both sides of the equation equally and I hope that you can as well.



--- sorry I, belive that the formating got messed up.

rickag
Dec 8, 2003, 08:21 AM
Originally posted by Amdfan

The 16 stage pipeline of the G5

I believe the G5 has between 9 and 20 stages. Has only about 52 million transistors when compared to the bazillion in the intel and amd top end processors, yet, hangs with them in real world applications and with those apps only partially optimized for the G5(ie. Final Cut Pro) actually blows away anything either intel or amd has to offer.

To quote you,"What's your point?"

the_mole1314
Dec 8, 2003, 08:51 AM
I love how many things you quote are many years old. Get recent information. Frankly, all of what you said has been debated here before and what you have said isn't even correct. They said it was the first 64-bit processor in a PERSONAL COMPUTER, not just the FIRST.

Amdfan
Dec 8, 2003, 09:42 AM
Well, since a cpu can only have a set number of piplines and that 16 piplines I got off of IBM's site I belive my information.
The G5 doesent really "hangs with them" as much as it gets beat.
Oh by the way, have you ever heard of optimizeing for and AMD proc? No, thats 'cause they dont need optomizations. I know of another company that uses "optomizations", their name is Nvidia.
I have yet to see a benchmark that shows the G5 beating the AMD opteron 248s
the AMD opteron only has around 100million transistors. half are in the l1 and l2 cache. "Whats your point"
They said it was the first 64-bit processor in a PERSONAL COMPUTER, not just the FIRST.
So your saying that a dual 246 system is not a personal computer? Well then what am I typing on? A server that serves noone?
------
If you would be so kind as to provide me with a link to show how well optomized applications run on an apple, that would be much apreciated.

Sunrunner
Dec 8, 2003, 09:48 AM
I love it when Mac-hating trolls show up and act like donkeys, it reminds me why the PC world is such a heartless barren wasteland of mindless lemmings.:rolleyes:

the_mole1314
Dec 8, 2003, 09:57 AM
Originally posted by Amdfan

So your saying that a dual 246 system is not a personal computer? Well then what am I typing on? A server that serves noone?

Average Joe Schmoe won't be using a server for their computing experence. You, my friend, are one of the lucky ones who know. Simply put, you can't walk into Best Buy and get a 64bit proc before the G5.

Amdfan
Dec 8, 2003, 11:06 AM
I can see your point the_mole and agree with it. However, why is that statment no longer on Apples site?

-------- responses to the personal insults directed twoards me
I love it when Mac-hating trolls show up and act like donkeys, it reminds me why the PC world is such a heartless barren wasteland of mindless lemmings.
WOW, I love how you automaticaly assume that I hate Mac's and call me a mac-hating troll. Sorry my friend but it seems that you are the one who sould be in your "PC world."
Personal insults or spam is grounds for instant banning. You called me a troll, I am sorry for trying to have a resonable disscusion. How old are we? Rather than refute my information you call me a troll?

ffakr
Dec 8, 2003, 11:41 AM
Originally posted by Amdfan


AMDFAN, you are a moron.

The G4 was the first personal supercomputer, by definition of the US Government munitions statutes. Yes this was mis-leading (as supercomputers were doing Teraflops in 1999), but it was accurate. Sure it was marketing, but it was legally correct. BTW.. a 500 MHz K6-2 could do 3 GFlops? You're saying that a K6-2 can retire 6 floating point ops per clock? what are you smoking? That's enough to show you for an idiot or a liar.

Nearly all the stuff you posted is ancient and irrelevent. The haxial site was a pure troll. A lot of the stuff he posted was inaccurate and debunked within days of his publication. Even the developers of GCC pointed out some of his wrong assumptions about Apple's use of GCC flags. Some corrections to Wrong statements on Haxial (off memory): SPEC DOES run faster with HT off, The tested G5 was preproduction and the registers on the CPU were modified to reflect the shipping product, The flags used with GCC DID take advantage of the SSE2...
It shows the TRUE colors of the Haxial author that he never went back and fixed his inaccurate statements after they were more than adequately disproven.

The G5 DOES have different numbers of pipeline stages dumbass. Altivec has a different number of stages than the rest of the processor.
Here's a 970 pipeline diagram:
http://www.cs.umass.edu/~weems/CmpSci635/Lecture8/L8.33.html

AMD NOT Apple insisted that Opteron is NOT a personal computer cpu. AMD has always insisted that Opteron was a workstation/server cpu (like Xeon). It's all marketing but that's every bit as much AMD's fault as Apple's. According to AMD, yes, AMD.. there wasn't a 64 bit x86 PERSONAL computer processor until Athlon64. Go piss on some AMD forum if you don't like it.
And yes, normal people don't have Opterons. They have been marketed from the start as workstations and servers. I've got an SGI Irix in the garage.. just cuz I've got one at home that doesn't make it a 'personal' computer either.

You're guessed numbers on Opteron heat disipation are a joke too. AMD licensed SOI from the AIM partners.. I'm very impressed that they'll be able to make a much larger processor run so much cooler than a technological lightweight like IBM on the same process.

IBM doesn't currently make processors over 2GHz because the 970 is new, and their other processors weren't designed for all out speed, they are designed with thicker, slower gates for more reliability. If you think clock is everything you're a bigger dufus than I first thought. Please explain to us why the 1.5 GHz Power4+ peforms so well, and so reliably if it only runs under 2 GHz.

Your argument about the Opteron running well without extra optimisation is a hollow one. The Opteron is just an extension to an existing architecure. It's the next revision of the K7 in some ways.
The PPC970 is based off the Power4, not the G4. It does some things differently than previous PPCs. You are only pointing out that the G5 Macs will ONLY GET FASTER as compilers and developer code catch up with the architecture. IBM's Beta of xlc generates code that is up to 70% faster than GCC 3.3.

So AMDfan, I suppose you never really bothered to actually look into the PPC 970 did you? Did you know it can retire more instructions per clock than the opteron? I guess you base all your detailed knowledge from lusers like AMDZone, haxial, and those incredibly relevent MS Word benchmarks.

OS X can address more than 8GB of ram natively with the 64bit extensions to OS X. You'de have found this out if you spent about 2 seconds checking. The G5 system controller can also address more than 8GB of RAM (16 actually) but Apple has ALWAYS historically rated their machines at whatever level of ram they could certify at release. Even now, http://www.crucial.com isn't selling 1GB DIMMS, only 512MB for the G5 and buying 4GB of RAM from Crucial would run nearly $800.
Now, if you advocate buying 8 or 16GB of generic RAM, your idiocicy knows no bounds. RAM problems are a pain in the ass, it's more than worth while to buy quality ram for a little more.

why don't you try this on for size.
Fast Fourier Tranforms (http://www.fftw.org/)
It's a site put together by some people from MIT.. benchmarking various FFTs in C and Fortran on various architectures. The 970 crushed the Opteron.
You'll always be able to find benchmarks where the Athlon64 beats the G5, and I'll always be able to find benchmarks where the G5 beats the Athlon64. It's the nature of benchmarking.
lies, damn lies, and benchmarks.

geez, get a life (though I have to say I really appreciate the fact that you spent all that time pulling together all those irrelevent links for the benefit of no one).

pgwalsh
Dec 8, 2003, 11:54 AM
Originally posted by Amdfan
WOW, I love how you automaticaly assume that I hate Mac's and call me a mac-hating troll. Sorry my friend but it seems that you are the one who sould be in your "PC world."
Personal insults or spam is grounds for instant banning. You called me a troll, I am sorry for trying to have a resonable disscusion. How old are we? Rather than refute my information you call me a troll? Well your original post has indirect attacks and direct attacks on Mac users as a whole. So you're sort of asking for people to call you names as you are mixing old marketing with new and old facts/opinions. You punch the G4 that is a very old processor with the newer Intel and AMD processors. I like AMD processors too, but the G5 does kick arse and I'm sure you could argue to no end for both processors. You do know that IBM and AMD are working together on these technologies...

So why come to a Mac centrist forum antagonizing people unless your looking for an argument or a bashing? Is it because you just wanted to heat people? Bored? WTF is your deal?

Tim Flynn
Dec 8, 2003, 11:57 AM
Deeper pipelines is not neccesarily better.
Deeper pipes allows higher clock rates, that is good. The problem occurs when a branch occurs in software ( the "if" or "while" etc). When a branch occurs, the pipelines need to be flushed and thats the speed penalty.
So for performance, you need to keep the speed up (MHz) and the pipelines down, always trade offs.
Some people call this efficency, but efficacy is perhaps a better term.
How much work gets done in a given unit of time. Then there are other factors...

My latest machine is a G5, because I like OS X and quiet.
My previous machine is a P4, I wanted an AMD because I like AMD, but I like quiet. The P4 was a quieter machine.

I'm not too worried about absolute clock speed, because how often is my machine "pinned".

rickag
Dec 8, 2003, 02:28 PM
Originally posted by Amdfan
Well, since a cpu can only have a set number of piplines and that 16 piplines I got off of IBM's site I belive my information.

I have yet to see a benchmark that shows the G5 beating the AMD opteron 248s

Thank you ffakr for providing the link to the pipelines used in the G5. Like I said, the G5 uses between 9 and 20 stages, your mileage may vary by 1 or 2 depending what you include as stages. This was discussed in depth on ArseTechnica's boards.

As far as no finding a G5 beating an AMD opteron, you haven't looked.

AMD's opteron and athlon 64 are excellent processors and their engineers have done a great job, but so has IBM's. The 970 will scale, compilers for it will improve and it and its' decendents will be running 64 bit in due time. Remember that the PPC ISA is 64 bit, has always been 64 bit. The PPC 32 bit instructions are a subset. Not an after thought cludge, that Microsoft's operating system is forcing intel and amd to use.

I believe you weren't trolling, but some of your facts are misplaced, and attacking the G5 on a Mac forum may not be the best place to do so.

Rincewind42
Dec 8, 2003, 03:42 PM
Originally posted by ffakr
OS X can address more than 8GB of ram natively with the 64bit extensions to OS X.

Actually, OS X has always been able to address more than 4GB of RAM - this capability it itself is required for the virtual memory system to work. What is new with the G5 system is a system controller that can actually address more than 2GB of RAM.

Sunrunner
Dec 8, 2003, 04:15 PM
Originally posted by Amdfan
WOW, I love how you automaticaly assume that I hate Mac's and call me a mac-hating troll. Sorry my friend but it seems that you are the one who sould be in your "PC world."
Personal insults or spam is grounds for instant banning. You called me a troll, I am sorry for trying to have a resonable disscusion. How old are we? Rather than refute my information you call me a troll?

It seems that I must clarify a few things...
FIRST: I never said "Amdfan, you are a Mac-hating troll, blah, blah".
SECOND: I merely pointed out my dislike for some of those PC people who show up in a Mac forum with no other apparent purpose than starting a fight.
THIRD: If I hurt your feelings then I will do the same thing I tell my 5 year old to do when one of her friends gets upset... sorry.
FOURTH: Its interesting that you seem to have identified yourself as a Mac-hating troll...

:rolleyes:

Amdfan
Dec 8, 2003, 04:55 PM
http://arxiv.org/PS_cache/hep-ex/pdf/0105/0105082.pdf
The first pc to be able to do over 1 gigaflop was the AMD k6-2 300 MHz NOT the g4. Sorry guys. Also you may say that the g4 in that PDF does 3 gigaflops, but the g4 could only do 64 bit calculations whereas the k6-2 was 80 bit and well the g4 wasn’t out yet :P. I was mistaken on the k6-2 gigaflops figure, but it still serves my point.
Here is another article: http://www.beowulf.org/listarchives/beowulf/1998/06/0190.html
Oh I don't think you know how to read that picture, so here is some help:
http://www.arstechnica.com/cpu/03q1/ppc970/ppc970-12.html
"You're guessed numbers on Opteron heat dissipation are a joke too. AMD licensed SOI from the AIM partners.. I'm very impressed that they'll be able to make a much larger processor run so much cooler than a technological lightweight like IBM on the same process."
What does this mean? Are you calling IBM a technological lightweight? ROFL. Well AMD does around 20 % of all processors sold in the world. Something tells me that they have a BIT more expertise than IBM, but to call IBM a lightweight is unfounded.
“CALGARY, Alberta, Dec. 8 -- VoodooPC, a world leader in the design and manufacturing of high-performance personal computer systems, today launched the world's first fan-free PC. The new Voodoo F-50 Powerful Personal Computer features an AMD Athlon 64 Processor 3200+ and supports high-end hardware, such as the ATI Radeon 9800XT video card. The F-50, which is configured with no fans, runs at an ambient noise level of under 20 decibels. The F-50 delivers outstanding performance in all applications, including gaming, and it provides the most realistic cinematic computing experience available today.”
Also if you don’t believe me that the Opteron runs cooler than the g5, then why is it in laptops and the g5 is not?
“Some corrections to Wrong statements on Haxial (off memory): SPEC DOES run faster with HT off”
So that’s why the p4’s with HT on get higher scores? www.spec.org
Also I get it; the Opteron is not a PC.
“ If you think clock is everything you're a bigger dufus than I first thought. Please explain to us why the 1.5 GHz Power4+ peforms so well, and so reliably if it only runs under 2 GHz.”
Well maybe because they have up to 128 Megs of l3 cache? And for around $130K for an 8 point system I would hope that they are fast.
As far as that site you linked me to, I can disregarded it on the fact that they don’t know how to properly label a graph, and if you can label a graph how can you run benchmarks? I must stat that I do nut understand what they are trying actually benchmarking. Seriously, the G5 was beaten on that site by the P4 Zeon. Also that benchmark program does not allow for the “Intel compiler” which gives Intel a significant performance boost.

“It seems that I must clarify a few things...
FIRST: I never said "Amdfan, you are a Mac-hating troll, blah, blah".
SECOND: I merely pointed out my dislike for some of those PC people who show up in a Mac forum with no other apparent purpose than starting a fight.
THIRD: If I hurt your feelings then I will do the same thing I tell my 5 year old to do when one of her friends gets upset... sorry.
FOURTH: Its interesting that you seem to have identified yourself as a Mac-hating troll...”
First: I did not call my self a Mac-hating troll.
Second: You didn’t flame me, until now of course.
Third: all I said in response to your post was that I don’t believe that IBM wants to make apple the fastest thing on the block. And then I gave my support for that statement.

AMDFAN, you are a moron.

Telomar
Dec 8, 2003, 07:15 PM
Originally posted by Amdfan
The first pc to be able to do over 1 gigaflop was the AMD k6-2 300 MHz NOT the g4. As it happens the G4 was the first to be able to sustain over 1 gigaflop, which was the test. The numbers given in that pdf aren't sustained performance.

ffakr
Dec 8, 2003, 08:16 PM
Originally posted by Amdfan
http://arxiv.org/PS_cache/hep-ex/pdf/0105/0105082.pdf
The first pc to be able to do over 1 gigaflop was the AMD k6-2 300 MHz NOT the g4. Sorry guys. Also you may say that the g4 in that PDF does 3 gigaflops, but the g4 could only do 64 bit calculations whereas the k6-2 was 80 bit and well the g4 wasn’t out yet :P.
all those Flops stats are quoted with SIMD single precision FP, Theoretical Max of course. I don't think you'd argue that Altivec is more powerful (better implemented, more expansive in scope..) than 3DNow. You are correct that the K6-2 could do a max of 4 single precision FPs in SIMD per clock, but I have to wonder what the max real performance was. The G4 was supposed to be the first personal PC Processor to have more than 1 Sustained GFlop of performance, not theoretical max.
"You're guessed numbers on Opteron heat dissipation are a joke too....run so much cooler than a technological lightweight like IBM on the same process."
What does this mean? Are you calling IBM a technological lightweight? ROFL. Sorry, it's called sarcasam.

Well AMD does around 20 % of all processors sold in the world. Something tells me that they have a BIT more expertise than IBM well, actually it's closer to 17% http://news.com.com/2100-1006_3-1001084.html IBM isn't big in the x86 market right now, but they are a HUGE chip producer and their process technology has long been top notch.
“CALGARY, Alberta, Dec. 8 -- VoodooPC, a world leader in the design and manufacturing of high-performance personal computer systems, today launched the world's first fan-free PC. The new Voodoo F-50 Powerful Personal Computer features an AMD Athlon 64 Processor 3200+ and supports high-end hardware, such as the ATI Radeon 9800XT video card. The F-50, which is configured with no fans, runs at an ambient noise level of under 20 decibels. The F-50 delivers outstanding performance in all applications, including gaming, and it provides the most realistic cinematic computing experience available today.”
Also if you don’t believe me that the Opteron runs cooler than the g5, then why is it in laptops and the g5 is not?

haha... I looked at the voodoo PC before I saw your post. It costs around $3400 in the base config (USD), it can run an Athlon fanless because the entire MASSIVE case is a heat sink and there are thermal pipes linking the cpu to the case. The thermal pipes are designed to dissapate up to 150 watts of heat! I didn't see a shipping weight, but the options are 'pick up' or pay $150 in shipping.
Also, It certainly is NOT the first fanless PC. The cube (running OS X server quite nicely) in my office was entirely fanless and it was a heck of a lot smaller than the Voodoo PC.
As for Opteron's Heat Output.. I got into a big argument about this before and I think I defended my position pretty well with facts and figures. Opteron's max wattage is much higher than the 970's max wattage. It isn't easy to come up with exact figures to compare since AMD likes to publish Max and IBM likes to publish typical. Athlon64s are hotter than 970s though, I won't go through the details though since they filled several posts.
AMD is licensing IBM's .09 micron SOI process. Your claim that .09 micron Athlon4s will be significantly cooler than PPCs is silly because both chips will be made on the SAME PROCESS. The 970 is MUCH smaller than the Athlon64, and it will remain smaller even if it moves to a 1MB of L2 cache. You are saying that AMD, by some magic, will make a bigger chip on the SAME PROCESS that turns out to be cooler than the smaller 970. Hmn.

“Some corrections to Wrong statements on Haxial (off memory): SPEC DOES run faster with HT off”
So that’s why the p4’s with HT on get higher scores? www.spec.org

Dell, the company whose computers Apple benchmarked against, was the one who stated in documentation that their machines ran SPEC faster with HT turned off. Apple was following Dell's guidelines while performing the benchmarks. Some of the SPEC code DOES run faster with HT turned off. Code that is not multithreaded often runs slower with HT on.
Check out this Dell paper on High performance computing and HT...
http://www1.us.dell.com/content/topics/global.aspx/power/en/ps3q02_khalid?c=us&l=en&s=esg


Well maybe because they have up to 128 Megs of l3 cache? And for around $130K for an 8 point system I would hope that they are fast.
The point is that they ARE fast, and that they are designed for 99.9999% uptime. If IBM wanted to, they could produce a Power4+ that clocked over 1.5GHz, but that wasn't what they were after.

As far as that site you linked me to, I can disregarded it on the fact that they don’t know how to properly label a graph, and if you can label a graph how can you run benchmarks?
Gee, that's odd, You'd think that a couple of MIT PhDs would know how to label a graph. I don't really see what's wrong with them, perhaps you can enlighten us?

I must stat that I do nut understand what they are trying actually benchmarking. Seriously, the G5 was beaten on that site by the P4 Zeon.

The site benchmarking some C libraries used to perform Fast Fourier transformations.
I'm not sure what you are looking at, but it looks like the 2GHz G5 bested the Xeon (not Zeon) too. I haven't overlayed all the graphs, but if you look at the scales for various benchmarks, the G5 performs better than the Opteron or the Xeon 2.8.
FFTs are math operations that are fairly pervasive in some fields of research.. let me allow someone smarter to me to explain a bit...
Linear transforms, especially Fourier and Laplace transforms, are widely used in solving problems in science and engineering. The Fourier transform is used in linear systems analysis, antenna studies, optics, random process modeling, probability theory, quantum physics, and boundary-value problems (Brigham, 2-3) and has been very successfully applied to restoration of astronomical data (Brault and White). The Fourier transform, a pervasive and versatile tool, is used in many fields of science as a mathematical or physical tool to alter a problem into one that can be more easily solved. Some scientists understand Fourier theory as a physical phenomenon, not simply as a mathematical tool. In some branches of science, the Fourier transform of one function may yield another physical function (Bracewell, 1-2).

Also that benchmark program does not allow for the “Intel compiler” which gives Intel a significant performance boost.
The Intel Compiler is well known for giving a boost to Intel processors, Especially on Benchmark Code. It's no great secret that Intel makes SPEC specific optimisations to their compiler, especially internal copies. In fact, SPEC claims that SPEC benchmarks CPU, System AND compiler performance.


Now let me clarify some things.
I'm typing on my iBook from work. MY machine (the only one not in parts) is upstairs, an Athlon. I use and support PCs and Macs.
I don't mind people pointing out when one system is better than another. I recommend PCs to people all the time.
I do mind when someone comes in with a lot of FUD, a lot of irrelevent info, a lot of outdated info and tries to tell a Mac list why all macs suck and all x86 PCs (or all Athlons, or Opterons, or Xeons) Rulez.

As I stated before (in this thread and others)... Opteron/Athlon64 is an impressive chip. It has advantages and dis-advantages over other architectures. That doesn't diminish the fact that the 970 is also an impressive processor with advantages and dis-advantages. The G5 is also a wonderful machine. I've got a dual 2GHz in the office and it screams. It's also the nicest Desktop machine I've ever seen from a design fit/finish perspective.

Amdfan
Dec 8, 2003, 09:07 PM
On the hyper transport item. HT has been revised so that in all programs there is no performance degradation, and in most cases there is an improvement. Also your explanation does not prove anything to me because on the spec website the scores with HT are faster than the scores with out HT.

Market share. I don’t think you heard to Q3 conference call, which Hector stated around a 20 % market share.

Floating point. The AMD Athlon which came out before the g4, did 2.4 gigaflops and 1gigaflop double precision. That makes the Athlon the first pc to do a gigaflop sustained.
http://ulita.ms.mff.cuni.cz/pub/techdoc/amd/21016.pdf

G4 cube. This little devil would crash all the time due to overheating. I could never get it to reboot, so I had to unplug it. OS X.0 was a nightmare compared to 2000, X.0 for clarification. Let’s just say that that was my last Mac.

The 970 does beat out the AMD system and tied with the Intel system(within the margin of error). Also on there site they only state the OS specs of the systems, but not the actual included software or other hardware, i.e. Chipsets, ram, pci cards.

manitoubalck
Dec 8, 2003, 11:02 PM
Originally posted by willmg
Anyone know anything about DDR2/3 availibilty by Febuary? I thought it was just recently that standards were even set for both of these formats for ram.

Both Ati and Nvidia scrapped DDRII, due to high costs, and uninspiring preformance gains. Currently no DDRII DIMM's are avaliable to the consumer to my knowlege. It would seem a strange move to move away from the proven standard DDR.

Go the RAMBUS @ 1600MHz (:( It won't happen I know, but I can dream:rolleyes: )

Amdfan
Dec 9, 2003, 02:34 AM
I belive that Micron will be ramping up production of ddr-2, quite posible just for apple.

Telomar
Dec 9, 2003, 02:49 AM
Originally posted by Amdfan
I belive that Micron will be ramping up production of ddr-2, quite posible just for apple. Or how about Intel who release their chipsets early next year...

manitoubalck
Dec 9, 2003, 03:24 AM
Originally posted by Amdfan
I belive that Micron will be ramping up production of ddr-2, quite posible just for apple.

And the cost of this would be:confused: not good to say the least.:mad:
Also the preformance gains are negligable at this time, using the Graphics cards that have been released with DDRII thus far as a guide.

If this is the case, it might be a case of Intel and RAMBUS all over again.

mvc
Dec 9, 2003, 04:31 AM
You shall not pass!

manitoubalck
Dec 9, 2003, 04:38 AM
Ha ha, cough:rolleyes:, while this is only mildly amusing, what on earth doest it have to do with this topic?

Back on topic: Ony time will tell whether or not DDRII will make its first system appearence in the REV B G5.

ffakr
Dec 9, 2003, 10:50 AM
Originally posted by Amdfan
On the hyper transport item. HT has been revised so that in all programs there is no performance degradation, and in most cases there is an improvement. Also your explanation does not prove anything to me because on the spec website the scores with HT are faster than the scores with out HT.

Here are two entries, FROM INTEL, on the SPEC site. They are back to back. The first is a 3.06 WITH HT on, the second is a 3.06 WITHOUT HT on...
Intel Corporation
Intel D850EMVR motherboard (3.06 GHz, Pentium 4 processor wi
1
1088
1098


Intel Corporation
Intel D850EMVR motherboard (3.06 GHz, Pentium 4 processor)
1
1167
1210

You may notice.. the 3.06 on the SAME motherboard ran faster with HT OFF.
Note, you'll find SPEC scores all over the place since A LOT of the performance has to do with system tuning the hardware/software/compiler to specifically run SPEC. It isn't fair to compare on system with HT off and a totally different system with HT on. You need to look a the same setup with HT off and on. Dell posted scores that showed SPEC running faster with HT off. Intel posted scores showing SPEC running faster with HT off. Somehow, Apple is lying when they say they turned HT off on SPEC for the benefit of the Dell?
THIS IS WHAT DELL TOLD THEM TO DO.

.. just a side jab at the validity of SPEC as a CPU performance meter.. Check out Intel's scores for the 3.0 P4 and the 2.8 P4. The 2.8 beat the 3.0. Also look at Dell's entry for the P4EE. It beat the 3.2 GHz P4 system by 250 in Int and FP. Where has the P4EE system ever performed that much better over a 3.2 P4 in any real world benchmark?


Floating point. The AMD Athlon which came out before the g4, did 2.4 gigaflops and 1gigaflop double precision. That makes the Athlon the first pc to do a gigaflop sustained.
http://ulita.ms.mff.cuni.cz/pub/techdoc/amd/21016.pdf
Page 1, right? You do realize that those figures are still max theoretical, right?
I'm fairly certain that Apple's claims regarding sustained FP performance were for Altivec single precision performance. I'm pretty sure neither the first Athlons nor the first G4s could sustain 1GFlop of double precison FP. Again, I think just about anyone familiar with both technologies would admit that Altivec is a better (more depth, more function) SIMD technology than 3DNow!. 3DNow! was impressive.. compared to MMX, but it was only the first modern desktop SIMD technology that had useful FP functions... it isn't the end all be all of SIMD.
The Athlon did come out before the G4, by one or two months and it was a very nice processor. I built one for my last office as soon as I could get the parts ordered.
I think this is all a bit irrelevent though. Somehow the argument of 'did Apple lie when they called a G4 a supercomputer?' turned into a pissing match over who got to 1GFlops first. I merely pointed out that it was the US Government (right or wrong) who supported this assertion and allowed Apple to design this marketing campaign. I think it was a pretty fun campaign, I'm sorry it got your panties in such a bunch.

G4 cube. This little devil would crash all the time due to overheating. I could never get it to reboot, so I had to unplug it. OS X.0 was a nightmare compared to 2000, X.0 for clarification. Let’s just say that that was my last Mac.
Did you ever consider that your cube had problems? Maybe Apple actually tested the cube for heat issues BEFORE they shipped it and the one you have is simply defective? Even though Apple has had one of the lowest DOA numbers in industry for years, it does occasionally have a faulty product. ;-)
I run mine as a server. It's got the stock GF2MX, and nearly a gig of ram in it. It's currently sitting on top of our aluminum G5 case in a small office that occasionally gets quite toasty from the 6 other computers.
It has NEVER crashed (that I can remember) and I have never seen any heat related issues with that cube.

Considering the the Cube still commands anywhere from $1000-$2000 (depending on features/upgrades), I'd say there are a LOT of people who don't have heat issues with their cubes.

Oh, BTW... Your cube problems aside, where is the condemnation for Voodoo PC for claiming they have the first totally passively cooled PC? Isn't the Mac a personal computer? Imagine a passively cooled PC that DOESN'T require a 150watt heat pipe and a giant aluminum slug for a case... Oh, that's so 2001.

the_mole1314
Dec 9, 2003, 11:00 AM
So you compare the beta of OSX to a final version of Windows? Frankly, that's kinda stupid.

ffakr
Dec 9, 2003, 11:29 AM
Originally posted by manitoubalck
And the cost of this would be:confused: not good to say the least.:mad:
Also the preformance gains are negligable at this time, using the Graphics cards that have been released with DDRII thus far as a guide.

If this is the case, it might be a case of Intel and RAMBUS all over again.

I think this would be a case of early adoption of the next big thing rather than some great performance leap. DDR2 isn't especially compelling right now but it's just in its infancy.
It is, however, a technology that is designed to scale farther than DDR. It would be a good move for Apple to adopt DDR2 in the near future, but it doesn't have to be in REV 2. Dual Channel DDR 400 still has life in it and the G5 would probably do fine by sticking with it till next summer. Apple could always simply rev the system to use a faster DDR variant. If they release an rev2 early next year, they could probably ship DDR 466 or DDR 500. Though limited to marginal 'OC' markets now, they'll likely be more popular in the near future.... especially if Apple orders a quarter million DDR 500 DIMMs.

Tim Flynn
Dec 9, 2003, 12:24 PM
Originally posted by Amdfan Well AMD does around 20 % of all processors sold in the world.

Actually PC processors only account for 2% of all processors sold in the world. So AMD's PC processors account for only 0.4% of world wide processor market.

World wide processor market includes DSPs, 4,8,16,32,64 bit processors.

Check out Jim Turley

P.S. Last year I heard a quote about if you rounded to the nearest whole percent MS Windows market realtive to all processors, it rounded to zero (0).

mvc
Dec 9, 2003, 02:28 PM
Originally posted by manitoubalck
Ha ha, cough:rolleyes:, while this is only mildly amusing, what on earth doest it have to do with this topic?

Even mild amusement is probably preferable to pointless argument. ;)

Amdfan
Dec 9, 2003, 03:37 PM
So you compare the beta of OSX to a final version of Windows? Frankly, that's kinda stupid.
Funny I didn't know that apple shiped ther computers with beta operating systems.

- about my cube, I got it for cheap cause they would not sell at my work. For 300$ it was a nice computer. I don't imagin that i would ever spend 1000$ on any outdated computer (yes outdated). any single g4 sytem sold for 1000$ is a ripoff. Just like any sub 2 GHz p4 system for 1000 is a rip off.
--- World wide PC market.
the dsp's 4 and 8 bit procs alnoe make up the vast majority of the market
and if you look on the front page of his site he says that the pentium acounts for 2% of the microprocesor market. AMD wiht all of its chips has out sold the Origional Pentium many many times over

the_mole1314
Dec 9, 2003, 06:40 PM
If I am not mistaken, they booted into OS9 and you had to change the start-up disk to OSX, if I'm not mistaken. That's what you had to do with all the new white iBooks back in 2001.

ffakr
Dec 9, 2003, 06:42 PM
Originally posted by manitoubalck
Didn't the very first G4's come with a beta version of OS-X?
Well, you can make a strong argument that 10.0 was a Beta OS. It didn't even have a DVD player (though XP didn't have the codec for playing DVDs either).

Amdfan
Dec 10, 2003, 02:45 AM
Or how about Intel who release their chipsets early next year...
Intel will not have a ddr-2 chipset till latter this year. And yes Intel is getting DDR-2 from micron but they have no plans to use it until later, when socket t comes out. I may be wrong, if you can find any information on future Intel chipsets let me know.

I remember buying a stick of 128 pc 2100 DDR the month it came out for $278. So I would bet that DDR-2 would be almost as expensive.
AMD licensed SOI from the AIM partners.. I'm very impressed that they'll be able to make a much larger processor run so much cooler than a technological lightweight like IBM on the same process.
The reason for AMD to be able to make there chips run coolers is because AMD's transistors have a lot less leakage when the transistor is off, compared to IBM and Intel. The less leakage you have the less power it requires to run the chip and thus, with less power, equals less heat.

ffakr
Dec 10, 2003, 01:20 PM
Originally posted by Amdfan
The reason for AMD to be able to make there chips run coolers is because AMD's transistors have a lot less leakage when the transistor is off, compared to IBM and Intel. The less leakage you have the less power it requires to run the chip and thus, with less power, equals less heat.

Correct me if I'm wrong, but this a 'feature' of a particular process. From an architectural standpoint (call it the 10 micron view as opposed to the '10,000 ft. view'), these different processors with different ISAs look essentially the same. Intel, Motorola, IBM, and AMD all have their own process though. They don't all make a .13 micron wafer in the same manner, they don't even use identical materials.
Your assertion regarding current AMD processor being more efficient may or may not be true, but it doesn't seem to remain likely when AMD moves to .09 micron since they will be licensing the .09 micron SOI process from IBM. I would expect that the resultant chips will perform very similarly regardless of whether they are designed to run the x86-64 ISA or the PPC ISA.

The real difference would be, what core voltage to they ship at, what clock speed, and how big they are. These factors will all be related though me thinks. I have no reason to believe that the .09 micron PPC 970 will be larger than the .09 micron Athlon64 even if the 970 doubles the L2 to 1MB. The Opteron is currently quite a bit bigger than the 970, more so than could be explained by the extra L2 and large L1 caches.

Side note, I've never personally heard this about AMD gates. I'm really surprised since AMD processors have never historically performed much better than P4s as far as thermal output. My 1.3 with the stock AMD heat shink (which I put on instead of a very loud Thermaltake dual fan orb, to no ill effect) has always run at idle at around 60° Celsius! PowerPCs, though smaller, have always run significantly cooler than PC chips on the same size process.

Amdfan
Dec 11, 2003, 02:26 AM
but it doesn't seem to remain likely when AMD moves to .09 micron since they will be licensing the .09 micron SOI process from IBM. I would expect that the resultant chips will perform very similarly regardless of whether they are designed to run the x86-64 ISA or the PPC ISA.
AMD Is co developing the 90nm w/SOI with IBM. Also, where do you think IBM got the PPC architecture? That can handle 64bits AND 32bits natively
Right now AMD's transitors have 2 gates insted of just 1 compared to IBM and Intel. That is a big big diffrence in the designes. Soon, 65nm, AMD's transistors will have 3 gates.
The only thing IBM is helping AMD with is SOI.

Sun Baked
Dec 11, 2003, 02:41 AM
Originally posted by Amdfan
AMD Is co developing the 90nm w/SOI with IBM. Also, where do you think IBM got the PPC architecture? That can handle 64bits AND 32bits natively
Right now AMD's transitors have 2 gates insted of just 1 compared to IBM and Intel. That is a big big diffrence in the designes. Soon, 65nm, AMD's transistors will have 3 gates.
The only thing IBM is helping AMD with is SOI. [/B]IBM and Motorola co-developed the PPC, with some input from Apple since they needed a new chip to replace the 68k series.

Before AMD decided to have IBM Fab processors, they were working with Motorola on process tech. And the Motorola-AMD partnership was buying process fab tech from IBM.

Motorola is still partners with IBM on the basic PPC ISA (32- and 64-bit) and the new Book-E subset of the PPC, but Motorola is using somebody else for Fab tech.

So yes I can see where the 64-bit and/or the PPC wouldn't exist without AMD.

Amdfan
Dec 11, 2003, 07:30 AM
Alderwood is an 800MHz dual DDR2 chipset which runs in so called "turbo" mode and is similar to Grantsdale P, which launches at the same time.(second quarter)
--- thats from the inq today
that chipset would equal 6.4 + 6.4 Gb/s = 12.8GB/s bandwidth for Intel or if its DDR 533 it would be more.
g5 bandwith @ 1.5ghz fsb 12GB/s. 2 chips @ 24GB/s
Of cource both these architectures pale in comparison of AMD's bandwidth due to the onchip memory controler. :P
If the g5 was to add an onboard memory controler as say dual channel @ 6.4GB/s per proc. That would be the most improvement that they would see next to going to scsi again. Think if you could remove half of the trafic on the FSB, and inclued really fast hard drives. Then add PCI express. All I can say is if they did that and it wasn't 10K a lot of people would buy it.
Now that what they should do for there 20th birthday :P

Telomar
Dec 11, 2003, 08:23 AM
Originally posted by Amdfan
Intel will not have a ddr-2 chipset till latter this year. And yes Intel is getting DDR-2 from micron but they have no plans to use it until later, when socket t comes out. I may be wrong, if you can find any information on future Intel chipsets let me know. Just do a search for Grantsdale. It's widely acknowledged it'll appear Q2 of next year. Nobody will have a chipset supporting DDR-2 prior to then contrary to what rumours claim.

Originally posted by Amdfan
The reason for AMD to be able to make there chips run coolers is because AMD's transistors have a lot less leakage when the transistor is off, compared to IBM and Intel. The less leakage you have the less power it requires to run the chip and thus, with less power, equals less heat.

AMD Is co developing the 90nm w/SOI with IBM. Also, where do you think IBM got the PPC architecture? That can handle 64bits AND 32bits natively
Right now AMD's transitors have 2 gates insted of just 1 compared to IBM and Intel. That is a big big diffrence in the designes. Soon, 65nm, AMD's transistors will have 3 gates.
The only thing IBM is helping AMD with is SOI. PPC was designed with 64 bit in mind and has a history of at least 5 years in the 64 bit arena, since the 620. That's not considering the POWER series or some of IBMs other designs though.

As for power leakage through transistors although it's true that becomes a far more significant problem as transistor sizes shrink it's hardly the only way to minimise power usage.

Finally that's just nonsense. IBM has dual gate transistors patents and Intel has its own tri-gate transistor that will appear around 2007, about the same time AMD and IBM and most other manufacturers will introduce their own funnily enough or about when the 45 nm manufacturing processes come online.

rickag
Dec 11, 2003, 08:58 AM
Amdfan
"AMD Is co developing the 90nm w/SOI with IBM. Also, where do you think IBM got the PPC architecture? That can handle 64bits AND 32bits natively"

I'm confused, totally. How is AMD at all involved in PPC, especially the ISA?

The PPC ISA was written during the AIM Alliance to be 64 bit and the 32 bit instructions as a subset of these instructions. This was thoroughly hashed out over @ Arse Technica with plenty of supporting links to IBM documentation. Not once in any of the discussions, did I read AMD was responsible for the PPC in any fashion, architectually or process technology.

Sun Baked
Dec 11, 2003, 05:22 PM
Originally posted by rickag
I'm confused, totally. How is AMD at all involved in PPC, especially the ISA? They were working with Motorola on the die shrinks, and advancing their process technology -- but that was also the brick wall that everybody splatted into at the same time, resulting in poor yields. And when AMD-Motorola needed help they turned to IBM for technology.

Now AMD is working with IBM, and Motorola went off to another technology source.

Can't you see the connection? If AMD was helping Motorola make PPCs, and now that they're using IBM as a technology partner/source, and IBM makes PPCs -- it means AMD invented the PPC. Sort of like Papa John invented the pizza. ;)

pgwalsh
Dec 11, 2003, 05:55 PM
Originally posted by Sun Baked
They were working with Motorola on the die shrinks, and advancing their process technology -- but that was also the brick wall that everybody splatted into at the same time, resulting in poor yields. And when AMD-Motorola needed help they turned to IBM for technology.

Now AMD is working with IBM, and Motorola went off to another technology source.

Can't you see the connection? If AMD was helping Motorola make PPCs, and now that they're using IBM as a technology partner/source, and IBM makes PPCs -- it means AMD invented the PPC. Sort of like Papa John invented the pizza. ;) Or AMD is going to do the same thing to IBM as Motorola.

:eek:

ffakr
Dec 11, 2003, 06:02 PM
Originally posted by Amdfan
AMD Is co developing the 90nm w/SOI with IBM. Also, where do you think IBM got the PPC architecture? That can handle 64bits AND 32bits natively
Right now AMD's transitors have 2 gates insted of just 1 compared to IBM and Intel. That is a big big diffrence in the designes. Soon, 65nm, AMD's transistors will have 3 gates.
The only thing IBM is helping AMD with is SOI.

hahaha... You are a dumbass.

The PPC ISA was designed from the start to be 64bit compatible. Though the product never came to market, IBM was working on a 64bit PowerPC back in 1995! It was called the PPC 620, and companies like Bull [supercomputers] actually marketed products on the chip, though IBM pulled the plug before release.

sheesh.

the contention that IBM learned how to make 64 bit (or even 64bit / 32bit) processors from AMD is by far the stupidest thing you've said yet.
PowerPC is a subset of the 64bit Power ISA. 64bit Power processors can run PowerPC code, even 32bit PPC code.

Amdfan
Dec 11, 2003, 10:22 PM
its more like the ceo of the proc devision at motorola was working at AMD and Motorola at the time. ehhh?
--- edit
Also AMD was working on x86-64 at the same time. Like in 1995


ffker, do you go to any 2600 meetings?

Sun Baked
Dec 11, 2003, 10:32 PM
Originally posted by Amdfan
its more like the ceo of the proc devision at motorola was working at AMD and Motorola at the time. ehhh?
--- edit
Also AMD was working on x86-64 at the same time. Like in 1995


ffker, do you go to any 2600 meetings? So by that logic, I guess that means Apple invented, scripted, and produced the movie Toy Story, Finding Nemo, Monsters, Inc. and/or any other Pixar movie?

[edit - oops wrong movie]

Somebody, PLEASE stop this silly fight...

Telomar
Dec 11, 2003, 11:09 PM
Originally posted by Amdfan
its more like the ceo of the proc devision at motorola was working at AMD and Motorola at the time. ehhh?
--- edit
Also AMD was working on x86-64 at the same time. Like in 1995


ffker, do you go to any 2600 meetings? x86-64 was under development for around 4 years I believe it was. As for Hector Ruiz he was at Motorola from 1997, well after PPC was developed. Anybody with any familiarity with x86-64 and PPC can inform you they are not similar or related though. x86-64 is similar to current x86 really in that it is just an extension of the old architecture, even AMD themself will tell you this. There is no connection between x86-64 and PPC and to claim otherwise is silly.

ffakr
Dec 12, 2003, 01:06 AM
I think it's pretty clear that every argument amdfan has posted in this thread has been trashed.

-AMD did not teach IBM how to make processors capable of running 32bit and 64bit code.
-I've found NO indication that AMD is currently using a dual-gate process, though I've seen references to them looking into it for future processes (and triple-gate by 2007!). I've found more info about Other companies work into dual/triple-gates
-x86-64 was NOT in development 10 years ago
-AMD will NOT SOON have .65nm triple gate transistors (see above.. 2007!)
-Cubes did not have rampant overheating problems
-A Voodoo PC with a 150 watt heat pipe (and a heatsink/case) does not prove that Athlon64s run cool
-SPEC Does run slower with Hyperthreading on (and spec isn't a great isolator of Processor performance)
-Apple Did NOT cripple the Dell in benchmarks (as reported on the soapbox)
-Opterons do NOT always beat G5s in benchmarks
-G5s Do have different pipeline lengths in different functional units (it is possible to do this dispite amdfan's claims)

I'm done. Does anyone here think he knows what he is talking about? He came in here telling us that Mac forums were full of people spouting crap and fud. Unfortunately, pretty much everthing he has said doesn't stand up to scrutiny. His arguments fall by the wayside, one by one, as they are all debunked.

On to the next topic. I'm going to bed, and I'm going to appologize to my Athlon for AMDFan. ;-)

ffakr.

P.S. No, I don't go to 2600 meetings. I've read a few issues but I don't find phone freaking interesting and most of the other topics don't stimulate my curiosity either. I've got a few old copies, but for the most part I flip through them at the rack and put them back.

Amdfan
Dec 12, 2003, 05:06 AM
your so silly ffaker.
you have disproved 3 at most. and those were not even close to the most important one I posted.
BTW you should be banned because you directly insulted me.

-Cubes did not have rampant overheating problems
-A Voodoo PC with a 150 watt heat pipe (and a heatsink/case) does not prove that Athlon64s run cool
-SPEC Does run slower with Hyperthreading on (and spec isn't a great isolator of Processor performance)
-Apple Did NOT cripple the Dell in benchmarks (as reported on the soapbox)
-Opterons do NOT always beat G5s in benchmarks
-G5s Do have different pipeline lengths in different functional units (it is possible to do this dispite amdfan's claims)

-my cube had overheating issues.
-athlon 64's use less watts then the g5
-why did apple use SPEC if it isnt a good judge of performance? And why if the p4's with HT on beat the apple benchmarks(even though HT APARENTLY slows performance)?
- opteron was only beat on one benchmark program hardly a victory.
-g5's DO have a set pipeline lenth, AltiVec can so a differing number of instructions

the_mole1314
Dec 12, 2003, 07:56 AM
I'm sick of this ****. Just nobody reply and let this thing die.