PPC 970 Faster Than AMD 64bit Chip!

hvfsl

macrumors 68000
Original poster
Jul 9, 2001
1,794
100
London, UK
Tomshardware has some figures on the new AMD 64bit CPU
http://www.tomshardware.com/technews/20021016.html#093300 and macrumors as some figures on the new IBM PPC chip so I have compared then below.

AMD 64bit CPU 2Ghz
SPECint2000 score of 1202
SPECfp2000 score of 1170

NEW IBM PPC (Power4 Lite) 1.8Ghz
Specint2000 score of 937
Specfp2000 score of 1051

So the AMD chip is faster but only at a higher clock speed. But the AMD chip will be out first. One thing is sure the IBM chip has a better Floating point unit than the AMD, which means it is better at 3D graphics and massive number cruching.

Also the AMD chip is about the same speed as a 3.2Ghz P4 chip will be. So Mac fans do not need to worry that much.
 

The problem is...

...AMD is bringing the Hammer out early 2003, beating PowerPC 970 by several months, and they are highly likely to scale much further by the time PowerPC 970 finally comes out.

but:
AMD 64bit CPU 2Ghz
SPECint2000 score of 1202
SPECfp2000 score of 1170

NEW IBM PPC (Power4 Lite) 1.8Ghz
Specint2000 score of 937
Specfp2000 score of 1051
Since when was 1051 ever higher than 1170. Same is true for 937. 937 isn't higher than 1202. So technically, the AMD is beating the PowerPC 970. Who cares if it's at a higher clock. 200 MHz, yay! That doesn't mean the AMD is less efficient... from the looks of it, the AMD is just as efficient, just coming out earlier.
 

springscansing

macrumors 6502a
Oct 13, 2002
922
0
New York
Re: The problem is...

Originally posted by MacCoaster

Since when was 1051 ever higher than 1170. Same is true for 937. 937 isn't higher than 1202. So technically, the AMD is beating the PowerPC 970. [/B]
Yeah... I donno what this post is about exactly :confused:
 

4409723

Suspended
Jun 22, 2001
2,222
0
Re: The problem is...

Originally posted by MacCoaster
...AMD is bringing the Hammer out early 2003, beating PowerPC 970 by several months, and they are highly likely to scale much further by the time PowerPC 970 finally comes out.

but:

Since when was 1051 ever higher than 1170. Same is true for 937. 937 isn't higher than 1202. So technically, the AMD is beating the PowerPC 970. Who cares if it's at a higher clock. 200 MHz, yay! That doesn't mean the AMD is less efficient... from the looks of it, the AMD is just as efficient, just coming out earlier.
The AMD chip is delayed to the same time as the IBM.
 
Re: Re: The problem is...

Originally posted by W-_-W
The AMD chip is delayed to the same time as the IBM.
For production or mass-consumption? IBM's chip is starting production (as in, just about to manufacture them) late 2003 so we won't see final products based on 970 until a little bit later.

Edit: bingo, yes:

The AMD Opteron is set for early 2003 shipping, the AMD Athlon based on Hammer is set for late 2003 (but SHIPPING, not PRODUCTION).

"Shipments of the AMD Opteron processor for servers and workstations are planned to begin in the first half of 2003. Shipments of AMD’s eighth-generation AMD Athlon processor for desktops are planned to begin in the fourth quarter of 2002."

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_4699_5707^5728~23293,00.html

So yes, months before the PowerPC 970.
 

hacurio

macrumors newbie
Jan 17, 2002
6
0
Illinois
THINK ABOUT IT THIS WAY!!!

Originally posted by MacCoaster
...AMD is bringing the Hammer out early 2003, beating PowerPC 970 by several months, and they are highly likely to scale much further by the time PowerPC 970 finally comes out.

but:

Since when was 1051 ever higher than 1170. Same is true for 937. 937 isn't higher than 1202. So technically, the AMD is beating the PowerPC 970. Who cares if it's at a higher clock. 200 MHz, yay! That doesn't mean the AMD is less efficient... from the looks of it, the AMD is just as efficient, just coming out earlier.
I agree that the AMD and the P4 will be faster by the time the 970 comes out, but let’s not forget that the Power Mac will probably be DUAL!!!!! Also you should consider that by the time the 970 comes out, there will only be two 64bit processors in the desktop market: AMD Hammer and the PPC 970 (the Itanium is not a desktop processor.)
Apple knows they need to improve their power lines. Did you read Apple’s Q4 results? Their Power Mac sales are extremely low; they will definitely push performance with the 970. So think about it this way:

-Dual 970 1.8GHz
-900 MHz BUS, Hyper-Transport (Remember Apple is a member of this consortium as well as AMD. Intel isn’t) http://www.hypertransport.org/
-Up to 6 GB of memory DDR or RDRAM (this is possible with a 64 bit processor, and yes there are apps that will use that much memory, for example MAYA, AFTER EFFECTS, FINALCUT PRO, CAD, and most 3D apps)
-Raid ultra ATA 133 standard.
-2 HD, 100 GB each, 200 GB total.
-Fire Wire 2 (Gigawire)
-USB 2

Plus the 970will have two SIMD units, IMB calls it “VMX” because “Altivec” is a Moto Trademark but, they have the same purpose: Increase performance in multimedia applications.

This “assumed” PowerMac doesn’t sound bad at all, considering on how bad the G4 ranks against the P4 and Hammer in SPEC 2000 benchmarks.
 
Re: THINK ABOUT IT THIS WAY!!!

Originally posted by hacurio
I agree that the AMD and the P4 will be faster by the time the 970 comes out, but let’s not forget that the Power Mac will probably be DUAL!!!!! Also you should consider that by the time the 970 comes out, there will only be two 64bit processors in the desktop market: AMD Hammer and the PPC 970 (the Itanium is not a desktop processor.)
The Hammer architecture is also capable of symmetric multiprocessing as well. Power Macs will probably never be more than a dual PowerPC, whereas you can do quad, oct, etc. configurations of Itaniums, Hammers, etc.
 

nixd2001

macrumors regular
Aug 21, 2002
179
0
UK
Re: Re: THINK ABOUT IT THIS WAY!!!

Originally posted by MacCoaster

The Hammer architecture is also capable of symmetric multiprocessing as well. Power Macs will probably never be more than a dual PowerPC, whereas you can do quad, oct, etc. configurations of Itaniums, Hammers, etc.
Isn't the question for each camp here whether N-way (N>2) will actually be produced rather can be produced? 2-way has been part of Apple line up for a while now. 2-way has been possible in the PC space for quite a while, but still seems a bit esoteric when it comes down to the actual volume shipped.

For machines that ship in any quantity to be worth talking about for large mnfs, I don't think N>2 is going to figure in any x86 variants. They may exist as geek/specialist high end machines for prestige purposes, but I really don't see them clocking up the volumes...
 

hacurio

macrumors newbie
Jan 17, 2002
6
0
Illinois
Re: Re: THINK ABOUT IT THIS WAY!!!

Originally posted by MacCoaster

The Hammer architecture is also capable of symmetric multiprocessing as well. Power Macs will probably never be more than a dual PowerPC, whereas you can do quad, oct, etc. configurations of Itaniums, Hammers, etc.
YES.....but I insist we are talking about Desktops and low-end workstations. The Itanium is not a desktop processor. And just for your info the 970 can go up to 16-way SPM. We will probably never see a Mac with 16 processors but the 970 doesn’t have that technical limitation.
 

DaveGee

macrumors 6502a
Jul 25, 2001
677
2
Re: Re: Re: The problem is...

Originally posted by MacCoaster
For production or mass-consumption? IBM's chip is starting production (as in, just about to manufacture them) late 2003 so we won't see final products based on 970 until a little bit later.
You have no idea what you are talking about... If you wanna use numbers that were quoted before the intro of the 970 that's up to you but CURRENT info says. (From IBM's own slides no less)

Samples 2Q03
Production 2H03

2nd Qtr as in April-May-June
2nd Half as in July-Aug-Sept-Oct-Nov-Dec

Oh and if you need more PROOF that dates don't mean a thing? Have a look at this... I've put bold tags on the areas you need to pay attention to.

Quote from geek.com
====================================
posted 2:03pm EST Wed Feb 28 2001

NEWS - IBM is sampling 400, 500, and 600MHz versions of their PowerPC 750CXe chips. Samples are available immediately and volume production begins in April. Now, weren't these the chips being used in the new iMacs that are available right now? I'm guessing that Apple has a separate production deal with IBM to bring the 400-600MHz chips out right now in volume, while IBM samples the chips to other manufacturers in more limited quantities for the time being. Most of the 750CXes are expected to be used in embedded applications. The chips are going for about US$57, $94, and $144 respectively, in volume. The 400MHz 750CXe dissipates a mere 3.5 watts of power. All 750CXe chips are produced on a new .18 micron production process with an upgrade from the .18 micron process used to make the plain 750CX that topped out at 550MHz and fits in a tiny 27 square mm package, smaller than the 42 square mm package of the 750CX.
====================================

Dave
 
Re: Re: Re: Re: The problem is...

Originally posted by DaveGee
2nd Qtr as in April-May-June
2nd Half as in July-Aug-Sept-Oct-Nov-Dec
Right. They haven't announced when they're starting volume production. It could be October 2003. It could be December 2003. But that's besides my point. However, volume productions doesn't mean they're immediately available by that time. It takes time to come to the demand of this product for Apple. You might want to reread the Arstechnica article I posted.
NEWS - IBM is sampling 400, 500, and 600MHz versions of their PowerPC 750CXe chips. Samples are available immediately and volume production begins in April. Now, weren't these the chips being used in the new iMacs that are available right now? I'm guessing that Apple has a separate production deal with IBM to bring the 400-600MHz chips out right now in volume, while IBM samples the chips to other manufacturers in more limited quantities for the time being.
Great! Guessing! And I really doubt that IBM would do that. That's risky stuff. Again, volume production != shipping immediately. It varies company by company.

LATE 2003/Early 2004 at the earliest the 970 would be shipping.

And besides, this is a pretty much brand new chip. The PowerPC G3 is old but tried and true so manufacturing and engineering defects are much lower. Especially with IBM.
 

DaveGee

macrumors 6502a
Jul 25, 2001
677
2
Re: Re: Re: Re: Re: The problem is...

Originally posted by MacCoaster
Right. They haven't announced when they're starting volume production. It could be October 2003. It could be December 2003. But that's besides my point. However, volume productions doesn't mean they're immediately available by that time. It takes time to come to the demand of this product for Apple. You might want to reread the Arstechnica article I posted.
So I guess we could say YOU ARE GUESSING that it isn't going to be available until 2003 late or 2004... Well then we might also GUESS that it's gonna be JULY 03 or AUG 03 since after all they fits 2H03 just as well and DEC 03!

You are the one pulling numbers from the air (or wherever) and then you say I'm guessing? (below) Oh and BTW I wasn't guessing...

Originally posted by MacCoaster
Great! Guessing! And I really doubt that IBM would do that. That's risky stuff. Again, volume production != shipping immediately. It varies company by company.
Umm why would you REALLY DOUBT IT when infact that is EXACTLY what happened. Apple started shipping the iMac with the 750CXe in early Feb (announced @ MWSF I think) and then at the END of Feb IBM made the following statement. "IBM is sampling 400, 500, and 600MHz versions of their PowerPC 750CXe chips. Samples are available immediately and volume production begins in April." that clip was a news item posted to a Mac web site that I found via google. If you really don't trust me look it up yourself. That is EXACTLY what happened.

Originally posted by MacCoaster
LATE 2003/Early 2004 at the earliest the 970 would be shipping.
This is just a guess on your part... If you have a source that backs this up that's one thing but you don't even HINT at that... This is just your gut talking as far as I'm concerned.

Originally posted by MacCoaster
And besides, this is a pretty much brand new chip. The PowerPC G3 is old but tried and true so manufacturing and engineering defects are much lower. Especially with IBM.
Power4 code name Gigaprocessor
PPC 970 code name GPUL

The PPC 970 has roots...

Sorry but when people just throw stuff out from their gut but state is as fact it really bugs me. Read your quote again and tell me you don't make it sound like a FACT instead of a GUESS.

"IBM's chip is starting production (as in, just about to manufacture them) late 2003 so we won't see final products based on 970 until a little bit later."

Oh and since you quoted this too...

"Shipments of the AMD Opteron processor for servers and workstations are planned to begin in the first half of 2003. Shipments of AMD’s eighth-generation AMD Athlon processor for desktops are planned to begin in the fourth quarter of 2002."

Don't tell me... You think that (FOR AMD ANYWAY)

'1H03' is January 1st and '4Q02' is last day of September right? :rolleyes:

Dave
 
Re: Re: Re: THINK ABOUT IT THIS WAY!!!

Originally posted by hacurio
YES.....but I insist we are talking about Desktops and low-end workstations. The Itanium is not a desktop processor. And just for your info the 970 can go up to 16-way SPM. We will probably never see a Mac with 16 processors but the 970 doesn’t have that technical limitation.
I said we probably would never see any more than a dual PowerPC 970 Mac. Did I say that the 970 couldn't do SMP?

Oh, since you're insisting about low-end workstations--why couldn't the Itanium fit right in? Itaniums are far cheaper than a lot of other 64bit vendors.

Besides, the Opteron is for both servers and workstations.

Depends on how you define "workstations."
 

TheFink

macrumors member
Sep 26, 2002
39
0
The fact that you people are already starting this pissing contest before either product is even NEAR production is pathetic.
 

j763

macrumors 6502a
Nov 25, 2001
661
0
Champaign, IL, USA
Originally posted by TheFink
The fact that you people are already starting this pissing contest before either product is even NEAR production is pathetic.
true... let's wait until the product comes out before we start speculating on what the chip market will be like then and how the 970 will compare.

There's really no point arguing about something that we have such limited information about.
 

hacurio

macrumors newbie
Jan 17, 2002
6
0
Illinois
Re: Re: Re: Re: THINK ABOUT IT THIS WAY!!!

Originally posted by MacCoaster

I said we probably would never see any more than a dual PowerPC 970 Mac. Did I say that the 970 couldn't do SMP?

Oh, since you're insisting about low-end workstations--why couldn't the Itanium fit right in? Itaniums are far cheaper than a lot of other 64bit vendors.

Besides, the Opteron is for both servers and workstations.

Depends on how you define "workstations."
I never disagreed with you there (Power Macs being dual only) but I disagreed when you started to talk about Quad, Oct, etc. low-end “workstations.” So let’s put it simple for you………According with a dictionary, a workstation computer is a stand-alone computer, often with a high-resolution display, used for computer-aided design and other complex and specialized applications. A workstation is a machine oriented to professionals and ranges in price anywhere between $2,000 to up to $15,000 or sometimes even more. Ok, lets analyze what a low-end workstation is. If the prices range between $2,000 (like the dell workstation 530, the cheapest dual workstation dell has and it’s not capable of quad configurations) and $15,000 (like the HP Workstation zx6000 with dual Intel Itanium 2; to be precise it costs $14,160,) then a low-end workstation is one that ranges in price between $2,000 and $5,000. Non of the $5,000 range work stations I’ve seen are quad, nor oct; furthermore, there in no low en Itanium workstation. The cheapest Itanium workstation I’ve seen is the HP zx2000 which costs around $6,000+ and is not even dual.