Linked by Eugenia Loli on Wed 18th Jun 2003 15:46 UTC
AMD Hewlett-Packard apparently slipped up and revealed some of the specifications for AMD's upcoming Athlon64 chip, a processor that HP seems to be preparing to use in its PCs. Although the specifications fit within the expected performance range for the chip, the 1.8GHz speed and 3100+ model number may begin to raise questions about AMD's ability to keep up with Intel, C|Net says.
Order by: Score:
We dont need 64 bit !
by Nadav on Wed 18th Jun 2003 19:51 UTC

I dont want to have a 64bit cpu that will heat twice as much as the 32bit one. I dont want to have a 64bit monster in my laptop.

RE: We dont need 64 bit !
by Eugenia on Wed 18th Jun 2003 19:52 UTC

I guess that's what my uncle was saying in the days of 16-bit. ;)

Re: We don't need 64 bit !
by bytes256 on Wed 18th Jun 2003 19:55 UTC

There's one in every crowd :-P

But seriously, it will be perty cool having 64 bits to play with...even if 32 bits are more than enough for all current desktop applications. I guess just knowing that you COULD run an enterprise grade server on your desktop is neat for bragging rights

Eugenia: You cant run away from the fact that 64bit cpus
by Nadav on Wed 18th Jun 2003 19:55 UTC

You cant run away from the fact that 64bit cpus produce more heat and takeup more power.

Re: Nadav...
by bytes256 on Wed 18th Jun 2003 20:00 UTC

You cant run away from the fact that 64bit cpus produce more heat and takeup more power.
That's why I'm gonna go buy a 286 laptop from http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=3418994848&categ... right now!!! It should use half the power and takeup half the heat of today's processors

RE: Eugenia: You cant run away from the fact that 64bit cpus
by Andrew G on Wed 18th Jun 2003 20:02 UTC

Yes, but you can't run from the fact that progress will mean that in two years your 64 bit cpu will produce less and consume less power than the current 32 bit processors.

Also can you run from the fact that 32 bit cpu's produces more heat and consumes more power than a 16 bit cpu.

Sorry to tell you this, but this incorrect.
Die shrink / lower voltage will drop power usage and thus heat generated.
Only when you are trying to get the absoulte max clock out of a given generation fab will you have this heat 'problem'.
Do some research and you will learn.

On second thought...
by bytes256 on Wed 18th Jun 2003 20:07 UTC

I think I'm just gonna play it safe and stick with my Casio pocket calculator...It's solar powered, so it must use very little power consumption...and it doesn't feel very hot at all to me...so it must run really cool...must be 1 bit or something...maybe it uses 1/32 the power of a Pentium IV and generates 1/32 the heat

Don't focus on clock speed...It's very fast!
by TheOwl on Wed 18th Jun 2003 20:11 UTC

Clock for clock this cpu smokes the P4!
If you don't believe me search on the web.
This new 64 bit x86 is an excellent piece of engineering.
Questions still remain about manufacturing thou.
This is AMD Achilles heal - when they will be available in quantities and will it be too late.

Thankfully we still have choice
by eighttiesdude on Wed 18th Jun 2003 20:17 UTC

Thankfully we still have a choice with AMD around. Intel's cpus are expensive with AMD around imagine without them. When a person can get a 1.5 amd cpu for $50 makes life good for many people. I remember a time that I only wished for a 500mhz cpu. To get that much cpu power for about $50 is incredible. AMD new 64bit cpus are bringing something that intel has denied the public unless you had thousands of dollars. I suggest we support this new cpu to keep them going and our pocketbooks full. For me anything after 1ghz is overkill not to say I dont enjoy the thrill or rush but I would rather have a slightly slower cpu if they are to a more expensive cpu and even a more expensive price that they are now.

Re: Don't focus on clock speed...It's very fast!
by ? on Wed 18th Jun 2003 20:18 UTC

But AMD:Intel Clock speeds are never the same.

It would but great if AMD made one of the
64bit chips using the 9nm technology and running at
3.0 GHZ and not give out 200 watts of heat.

it is not all about the clock speed...
by grimlock3000 on Wed 18th Jun 2003 20:20 UTC

"Clock for clock this cpu smokes the P4!"

AMD's problem has _nothing_ to do with being compared to Intel clock for clock. What people care about is the end performance. Would someone buy a Ahtlon64 system that is faster per clock cycle over a P4 which is faster overall? (Assuming they are the same price). Also, AMD has been getting rather liberal with their "model number" ratings lately, so this chip might actually manage to underperform compared to a P4 3+ GHz. If AMD puts out a 64 bit capable chip that is roughly the same speed as a chip Intel has had on the market since November of last year, did they actually accomplish anything?

Light cpu
by ? on Wed 18th Jun 2003 20:32 UTC

I can remember whem people was talking about cpu's
using photons insted of electrons to boost cpu speed.

We will soon come to a point when well can not use
silicon and electron (about 0.001 mn).

In 2010 we will be looking at 128bit cpu because the
64bit cpu's had that annoying 64 Exobyte limit.

In 2020 We will looking at cpu's with only 64Mb of L2 cache
and try to make it to 128Mb L2 cache. Windows and Linux
will require 1GB of RAM just to boot and 10GB to do anything
useful.

Memory
by Harbinjer on Wed 18th Jun 2003 20:39 UTC

What you all aren't talking about is memory. The Athlon64 will be able to address much more memory than a P4. While you may not need 4GB+ of ram for most desktop applications, there are some that do. I for example, work with a program that can suck down 2GB of ram on a relatively small data set(we do medical imaging.)

I imagine that there are plenty of CAD/CAM programs that work similarly.

This will be nice because we can get 4+GB of ram in a <$10,000 workstation.

Re: Light cpu
by whaaa on Wed 18th Jun 2003 20:43 UTC

"In 2020 We will looking at cpu's with only 64Mb of L2 cache "

Well, in 2002 the Power4 had 32MB of L3 cache, so there is no need to wait that long ;)

ppc 970
by steve on Wed 18th Jun 2003 20:45 UTC

Nobody will want an amd64 or opteron when the ppc 970 comes out. <ducks>

"You cant run away from the fact that 64bit cpus produce more heat and takeup more power."

And the most uninformed comment award goes to....

Re: Re: Light cpu
by ? on Wed 18th Jun 2003 20:45 UTC

Not for the desktop though

opteron is already there
by grimlock3000 on Wed 18th Jun 2003 20:53 UTC

"What you all aren't talking about is memory. The Athlon64 will be able to address much more memory than a P4. While you may not need 4GB+ of ram for most desktop applications, there are some that do."

The Opteron is already availible to address this limitation at a reasonable price. The Athlon64's target market is not going to be workstations that may need more than 4GB of RAM, people will use Opteron's for that.

Confusion
by Jim on Wed 18th Jun 2003 20:58 UTC



The Opteron is not the Athlon 64, but Anandtech did a few god articles on the processor here.

http://anandtech.com/cpu/index.html

Desktop performance of the server chip here

http://anandtech.com/cpu/showdoc.html?i=1818

That should give a rough idea of how the architecture will perform. And I should mention that the CPU will not be available until September. I am neither an AMD nor Intel fan but AMD has been blowing hammer smoke for years now. If they would have released the chip when samples first became available a year ago it would have been a large step forward. Now it is nearly par and may have a difficult time keeping up with the Prescott. Lets hope for AMD's sake that the platform will scale.

Re: We dont need 64 bit
by Megol on Wed 18th Jun 2003 21:12 UTC

> I dont want to have a 64bit cpu that will heat twice as much as the 32bit one.

The power consumtion depends on a lot ot things but even if we speculate that Opteron/Athlon64 uses almost no static CMOS* that is clearly wrong.
The proof is simple: In todays processors a large part of the power budget is used for the clock drivers** and even if all logical units would double their power consumtion the total power usage would not double.
So you are a troll and/or completly clueless.

(* static CMOS design takes almost no power when the signals doesn't change)
(** the parts that makes sure that a synchronized global clock is transported over the whole chip)

Once, the pci express is wide use and I will buy a new desktop machine with 64bit CPU for myself. I am still happy with my 464Mhz machine. ;-)

pci express is a replace of agp.

RE: We dont need 64 bit !
by Marty McFly on Wed 18th Jun 2003 22:57 UTC

Doubling the bitness of a processor does not double the size of the processor, the size of the code, the number of execution units the processor has, the power requirements, the size of the cache or the amount of memory it can address.

CPU rating system needs to change
by HunterA3 on Wed 18th Jun 2003 23:13 UTC

They really need to change the rating system for CPUs to Instructions Per Clock cycle (IPC) rather than strictly on MHz clock speed. That's why 1.8 GHz Athlon64 can keep up with a P4 with more than 3 times the clock speed.

It's like basing your monitor purely on the size if the screen and ignoring the dpi, refresh rate, and resolution.

AMD Speed ratings too liberal
by Dial_tone on Wed 18th Jun 2003 23:18 UTC

I've hated these speed rating numbers ever since Cyrix did it. Just give me the actual speed and I'll make a decision.

re: CPU rating system needs to change
by hmmm on Wed 18th Jun 2003 23:41 UTC

No, more like they need to change the rating system to IPC * Mhz. We need to know how many instructions are executed per second for integer and floating point operations.

I think we need to make a law that requires a label similar to foods or this type of benchmark will never get done.. I guess if we took all the money out of the process people might start thinking about eachother and their customers more and providing this type of feedback instead of building a marketting department to do that for us through the use of their psycheology department.

RE: Clock Speed & Speed ratings
by Antarius on Wed 18th Jun 2003 23:49 UTC


It would but great if AMD made one of the
64bit chips using the 9nm technology and running at
3.0 GHZ and not give out 200 watts of heat.


Ah, but it would be cool if we could reclaim the heat and use it for some other purpose. A small space heater, making coffee, lighting cigarettes, power generation... ;-)


Dial tone:
I've hated these speed rating numbers ever since Cyrix did it. Just give me the actual speed and I'll make a decision.

Ahh... But you aren't the target market for that marketting practice. If you know the fact that you cannot compare two different platforms by clock-speed alone, then you're actually capable of making that decision.

If you're one of the masses that claims to be 'computer illiterate,' then you'd overlook the AMD and even the Macs because of its "slower speed."

<tongue_in_cheek>
Personally? I blame Intel for making a processor that was designed to purely increase clock speeds at the expense of actually doing anything useful!
</tongue_in_cheek>

RE: I will wait for pci express when it's wide use..
by Anonymous on Wed 18th Jun 2003 23:51 UTC

"pci express is a replace of agp."

Yup, and probably more eventually. I'd like being able to have more than one gfx card on a fast connect.

RE Nadav (IP: ---.red.bezeqint.net)
by CooCooCaChoo on Thu 19th Jun 2003 00:34 UTC

Obviously you know jack-squat about computers or otherwise you would know that the heat given off from the UltraSparc IIe is so low that it only requires a heat sink to keep it cool. Yes, the UltraSparc IIe is a 64bit CPU.

Re: AMD Speed ratings too liberal
by rajan r on Thu 19th Jun 2003 03:41 UTC

The actually speed? That would be rather difficult - relative to what? If there is anything proven in the past couple of years is that AMD has a nifty good record of being very close to their rating in relativity with Intel processors.

Clock speed - so what
by Anonymous on Thu 19th Jun 2003 12:24 UTC

A Via C3 933MHz is about as slow as a 500 MHz P111 and runs very cool. A 1 GHz Mac is about 3 times as fast as the Via processor, about 50% faster than a 1 GHz Athlon and much slower than a 1.4 GHz Opteron.

Re: AMD Speed ratings too liberal
by OoSync on Thu 19th Jun 2003 13:53 UTC

One thing to remember, and is often overlooked, is that the speed ratings are not Athlon versus P4. The speed ratings are current Athlon versus an Athlon Thunderbird.

So, indeed, the 3200+ may very well perform like an Athlon Thunderbird 3200 MHz, not a P4 3.2 GHz, but the number is probably stilly driven by marketing, not engineering.

16 bit versus 32 bit versus 64 bit
by slash on Thu 19th Jun 2003 15:29 UTC

Does anyone actually know anything about bits? Comparing the technological leap from 16 bit to 32 bit to the one made from 32 bit to 64 bit is completely unfair. The 16 bit barrier was actually a major barrier. 16 bit was limited to hardware manipulation of numbers that were limited to 2^16 power (65536) while 32 bit allowed manipulation of numbers that where 2^32 power (4294967296). A simple example of this limitation is that with 16 bit cpu's, you would be limited to 65 megs of ram. With 32 bit systems, you are limited to 4 gigs of ram. Yes 64 bit allows manipulation of even more ridiculously large numbers (2^64), but for anything but enterprise level applications, it is overkill and even comes with a slight penalty because you now have to deal with bigger registries. This is exactly why Intel was able to compete and succesfully sell their cpu's while Apple, Sun and everyone else was selling 64 bit cpu's. When the top of the line system was running at 300 mhz with 256 megs of ram and was dealing with numbers that were at most 24 bits, there was no performance hit.

Re: AMD Speed ratings too liberal
by Maynard on Thu 19th Jun 2003 16:09 UTC

I know, I have told people this before. They do not get it.

The model clock number was based on the thunderbird. The reason AMD seems to have been pushing things of late is because Intel meanwhile has done 2 things that pushed its performance up with the same clock speed. It has doubled the FSB clock and doubled the cache. So their own prodcts are now mistmatched. There was a time when the P4 3GHz 800FSB was faster performing than the P4 3.06GHz 533FSB one and no one seemed to find that crazy for processors whose peformance is represented by clock speed.

Cpu speed rating
by aaron on Thu 19th Jun 2003 17:21 UTC

Should be in

Intergers operations per seconed. IOPS
Or
Floating Point operations per seconed. FLOPS

Re: slash
by Bascule on Thu 19th Jun 2003 17:21 UTC

Does anyone actually know anything about bits?

Yes.

Comparing the technological leap from 16 bit to 32 bit to the one made from 32 bit to 64 bit is completely unfair. The 16 bit barrier was actually a major barrier. 16 bit was limited to hardware manipulation of numbers that were limited to 2^16 power (65536) while 32 bit allowed manipulation of numbers that where 2^32 power (4294967296). A simple example of this limitation is that with 16 bit cpu's, you would be limited to 65 megs of ram. With 32 bit systems, you are limited to 4 gigs of ram. Yes 64 bit allows manipulation of even more ridiculously large numbers (2^64), but for anything but enterprise level applications, it is overkill and even comes with a slight penalty because you now have to deal with bigger registries.

Several things to keep in mind with this:

* A program compiled for the AMD64 instruction set, even if it doesn't contain any 64-bit integers, has a performance advantage over IA32 because IA32 has essentially 4 general purpose registers (technically 8, only 4 of which are useful) whereas AMD64 will have 16. This alone will be a great performance boon to virtually all applications.

* 64-bit integers are already seeing widespread use in today's 32-bit systems, most notably within the filesystem. Using 64-bit integers on a 32-bit architecture comes with a performance penalty, and this will be remedied with the introduction of a 64-bit processor.

* 32-bit architectures are hurting for address space. There are many features which could use a larger address space, such as memory-mapped I/O (on large files) and prebinding (systems utilizing prebinding can quickly run out of addresses if all libraries on the system are preassigned addresses)

This is exactly why Intel was able to compete and succesfully sell their cpu's while Apple, Sun and everyone else was selling 64 bit cpu's.

Apple has never sold 64-bit systems. PPC970 will be the first. Intel can't really sell a commodity 64-bit processor at this point unless they plan to support AMD's instruction set (which has been rumored). So Intel simply claims that 32-bits are good enough for anyone.

Re: Re: slash
by slash on Thu 19th Jun 2003 21:07 UTC

I wasn't trying to insult anyones intelligence or suggest that 64 bit is useless. My post was just aimed at people who ridiculled the person who suggested that it possibly would be bad for laptops. For some purposes, a less powerful/simpler chip design is all that is needed.

re confusion
by patrick_darcy on Fri 20th Jun 2003 00:02 UTC

i agree that they are way late with their 64 bit processors.
i remember a few years back they (the amd pr people)
made the statement that amd was going to put the hammer
on the back burner. they did just that and now look, they
havent made any money in quite a while. if hammer
does not scale well, then i believe amd is over and they
have noone to blame other than themselves. u cannot put
your latest technology on the back burner unless u intend
to get burned there ;) i hope they do well. we will soon find
out. if they for some reason dont release the athlon 64on time
this time i believe it is over for them.