PowerPC 970 up to 2.5GHz?

[Frobozz]

Originally posted by bertagert
I want to see what one looks like. Anyone have a link to a picture of the 970? Or, get your photoshop out and lift that heatsink off so i can take a peak.

Edited for spelling

Just type "IBM PowerPC 970" into Google. I came up with this:

The page URL. And the image:

 

[backspinner]

it looks like this (using transparant heatsink mode)

 

[backspinner]

- on the left is the processor under the heatsink
- below it is a very large and expensive Xilinx programmale chip, which acts as the memory interface in this design (in later stages, these normally will be finalized in cheaper asic chips)
- above the heatsink the voltage regulators for the processor core
- under the harddrive a Lucent io chip
- in the middle two legacy PCI slots
- at the right two (serial) PCI-X slots with broadcom controllers

 

[Pedro Estarque]

32, 64, 128 bit

I don't know much about 32, 64, 128 bit but I thought that one of the differences between then was that because it processes information in 64-bit chunks, the 970 would process more information per cycle then a 32 bit ppc.
Isn't that the reason why altivec instructions are so much faster then normal ones?

 

[pgwalsh]

the 970 will process 32 bit and 64 bit. My understanding is that 32 bit apps will run throug 32 bit mode. It wont see any advantage of 64 bit. So most apps, unless altered, won't see any advantage. However, it' s good for server stuff since most - to my understanding - is done in 64

 

[Pedro Estarque]

yes I know, but there are people saying we will never need 128 bit addressing, because we will never need that much memory. But what about performance?I think a 128 bit chip with OS and applications optimized for it would be much faster then 32 bit just because it processes information in a bigger chunk per cycle.

 

[cr2sh]

Every thread on the 970 degenerates into this:

"32 vs 64, some apps will run...."

then it goes to

"Will osX be able to?"

and then the code-monkeys come out...

"Yes but it will need recompiled.. or blah."

But it's always fun to watch the progression.

 

[Frohickey]

Re: Re: Re: Re: Blade

Originally posted by beatle888
yes it will be in a server room, but the fact remains, its ugly, and so is their site.

Beauty is in the eye of the beholder.

 

[Frohickey]

Originally posted by backspinner
it looks like this (using transparant heatsink mode)

In transparent aluminum heatsink mode, you would be barely able to see the heatsink.

Geez... and I travelled millions of miles.... err thousands of miles to tell you this.

 

[Dont Hurt Me]

Originally posted by cr2sh
Every thread on the 970 degenerates into this:

"32 vs 64, some apps will run...."

then it goes to

"Will osX be able to?"

and then the code-monkeys come out...

"Yes but it will need recompiled.. or blah."

But it's always fun to watch the progression.

No kidding the 970 will run 32 as well as 64 no emulation needed . 32 Native. yes it will be faster beacause of cpu speed, more instructions in and out per clock cycle, wider front side bus. Just thought i would say that for all those 32 -64 huh? people.

 

[beowolf]

32 bit on 64 bit 970 issue

To quote this page, http://www-3.ibm.com/chips/products/powerpc/newsletter/dec2002/newproductfocus2.html , on the sixth paragraph, "In addition to its support of new 64-bit solutions, the 970 retains full native support for 32-bit applications. This not only protects 32-bit software investments, but provides these 32-bit applications with the same high-performance levels that it extends to 64-bit uses. This native, nonemulated, 32-bit support is not limited to application code, which runs unmodified. 32-bit operating systems with minor updates can also take advantage of the PowerPC 970's outstanding performance." This is what arn was saying when he said the 970 would give 100% compatability, and, from the look's of this, something more since 32 bit will get the same high performance that 64 bit gets. Nice

One other thing of note, on this page http://news.zdnet.co.uk/story/0,,t269-s2131244,00.html?rtag=zdnetukhompage
in the 9th paragraph they say "Sources familiar with IBM and Apple's plans said that Apple will be a customer for the PowerPC 970 next year, but Apple has declined to comment." Is this of any important note coming from zdnet?

Phew, that was a long first post ever. Hi.

 

[beowolf]

one other thing

On that zdnet page I linked to, it pointed out this, "A 1.8GHz PowerPC 970 dissipates 42 watts, according to IBM's figures, compared to 68.4 watts for a 2.8GHz Pentium 4 and 30 watts for a 1GHz G4e chip from Motorola. The more heat a chip dissipates, the more space and equipment is required for cooling." Does this mean a lower clocked 970 (1 GHz-1.3 or 1.4 GHz) could possibly run in a powerbook with a slightly updated design off the bat? Or maybe Stevo will try and go for a home run and update across the line to 970 (VERRRYY doubtful in my mind, but would be awesome none the less). Can anybody add to this thought (i.e. economic benifits of only selling puters with one chip to support?). Thanks

 

[Foxer]

Re: Re: Business

Originally posted by law guy
What I find discouraging is that I just called the Apple store in Tysons Corner, VA and they tell me at least four (4) weeks until they begin to get 1.42 inventory. So, we won't even be seeing the 1.42 G4 machines until April

According to the Apple store, mine will be built next week. Hopefully his info is better than Tysons.

 

[shadowfax]

Re: 32 bit on 64 bit 970 issue

Originally posted by beowolf
This is what arn was saying when he said the 970 would give 100% compatability, and, from the look's of this, something more since 32 bit will get the same high performance that 64 bit gets. Nice
...
Phew, that was a long first post ever. Hi.

I don't know just what they mean by taking advantage of the 64 bit performance--it could just be some propagandized sort of thing exaggerating the fact that it will still run 32 bit apps &c. i'd be surprised if it did something like putting 2 32 bit instructions in a single operation to enhance performance. i suspect these will just be running 32 bit apps about as well as a 1.8-2.5 GHz G4 would. i could be off though .

anyways, welcome to the forums; it's always good to start off relevant and backed by sources right up front!

 

[shadowfax]

Re: one other thing

Originally posted by beowolf
On that zdnet page I linked to, it pointed out this, "A 1.8GHz PowerPC 970 dissipates 42 watts, according to IBM's figures, compared to 68.4 watts for a 2.8GHz Pentium 4 and 30 watts for a 1GHz G4e chip from Motorola. The more heat a chip dissipates, the more space and equipment is required for cooling." Does this mean a lower clocked 970 (1 GHz-1.3 or 1.4 GHz) could possibly run in a powerbook with a slightly updated design off the bat? Or maybe Stevo will try and go for a home run and update across the line to 970 (VERRRYY doubtful in my mind, but would be awesome none the less). Can anybody add to this thought (i.e. economic benifits of only selling puters with one chip to support?). Thanks

i think when they pop the .09 micron design into the PPC 970 it will take considerably less power, which looks even better for powerbooks, provided the chips are cheap enough.

 

[Frobozz]

Originally posted by Pedro Estarque
I think a 128 bit chip with OS and applications optimized for it would be much faster then 32 bit just because it processes information in a bigger chunk per cycle.

Well, I'm no chip architect, but I believe that also means it'll waste more time and become less efficient if you can't fill up those chunks.

 

[beowolf]

....one more thing

I was letting my mind chew on this while feeding cows and had to wonder. A lot of people on here keep saying that they average joe isn't going to need a 64 bit computer. Well, from what I remember from senior economics in high school (just last year), a lot of getting somebody to buy something is both percieved need and want. Well, I'm pretty certain that the want is going to be there (hey, it's bigger, better, and faster! mindset) and I think that marketing should be able to make people PERCIEVE that they need it, whether they really do or not. Just my $.02

 

[shadowfax]

Re: ....one more thing

Originally posted by beowolf
I was letting my mind chew on this while feeding cows and had to wonder. A lot of people on here keep saying that they average joe isn't going to need a 64 bit computer. Well, from what I remember from senior economics in high school (just last year), a lot of getting somebody to buy something is both percieved need and want. Well, I'm pretty certain that the want is going to be there (hey, it's bigger, better, and faster! mindset) and I think that marketing should be able to make people PERCIEVE that they need it, whether they really do or not. Just my $.02

the want is going to be there, that's certainly given. whether or not it will be affordable enough to convince people to succumb to their desire is much more sketchy.

 

[eric_n_dfw]

Re: ....one more thing

Originally posted by beowolf
I was letting my mind chew on this while feeding cows and had to wonder. A lot of people on here keep saying that they average joe isn't going to need a 64 bit computer. Well, from what I remember from senior economics in high school (just last year), a lot of getting somebody to buy something is both percieved need and want. Well, I'm pretty certain that the want is going to be there (hey, it's bigger, better, and faster! mindset) and I think that marketing should be able to make people PERCIEVE that they need it, whether they really do or not. Just my $.02

What can kill the "want" is a lack of applications that show off the advantage of 64 bits. All it would take would be an x86 32Bit machine coming close to the same performance on video redering or something and the 64bit marketting bubble would pop.

 

[Catfish_Man]

Re: Re: 32 bit on 64 bit 970 issue

Originally posted by Shadowfax
I don't know just what they mean by taking advantage of the 64 bit performance--it could just be some propagandized sort of thing exaggerating the fact that it will still run 32 bit apps &c. i'd be surprised if it did something like putting 2 32 bit instructions in a single operation to enhance performance. i suspect these will just be running 32 bit apps about as well as a 1.8-2.5 GHz G4 would. i could be off though .

anyways, welcome to the forums; it's always good to start off relevant and backed by sources right up front!

Nope, it should be about twice as fast (maybe a bit more) per clock than a G4 (for fp operations anyway). This is entirely unrelated to 64 bitness, which actually is a slight performance loss, not a gain, for most tasks (64 bit pointers take up more cache space). The extra power comes from a fast bus, out of order execution, big caches, and a second floating point unit.

In reply to the 128 bit posts. It wouldn't allow 4 32 bit chunks (that's what Altivec does) it would allow one 128 bit chunk. AFAIK, no one needs numbers THAT big (2^128).

 

[freemidnight]

Going to the average rumors here about speed increase compared to the real facts, this is still Blah Blah Blah! and for months and months to come!

This is becoming pathetic!

 

[Catfish_Man]

Originally posted by freemidnight
Going to the average rumors here about speed increase compared to the real facts, this is still Blah Blah Blah! and for months and months to come!

This is becoming pathetic!

I'm afraid I can't quite figure out what you're saying. Could you clarify it?

 

[shadowfax]

Re: Re: Re: 32 bit on 64 bit 970 issue

Originally posted by Catfish_Man
Nope, it should be about twice as fast (maybe a bit more) per clock than a G4 (for fp operations anyway). This is entirely unrelated to 64 bitness, which actually is a slight performance loss, not a gain, for most tasks (64 bit pointers take up more cache space). The extra power comes from a fast bus, out of order execution, big caches, and a second floating point unit.

right. the processor itself shouldn't be faster. if and when the G4 gets a DDR bus, and if you could put that bus at 900MHz and clock the processor at 2.5 GHz, i doubt there would be much of a speed difference. the speed difference isn't intrinsic within the processor, it's a change because of the bus and cache so on... the only internal advantage you mentioned was the 2nd FP unit.

 

[shadowfax]

Re: Re: ....one more thing

Originally posted by eric_n_dfw
What can kill the "want" is a lack of applications that show off the advantage of 64 bits. All it would take would be an x86 32Bit machine coming close to the same performance on video redering or something and the 64bit marketting bubble would pop.

hehe, unless they cost the same (or so). if you could get a 64 bit processor that performed about as well as a 32 with 32 bit apps for about the same or slightly more money, wouldn't you? it's like insurance against the future. it there were to come out a 64 bit app you needed and you went for the 32 bit x86 just because you didn't think you needed the 64, wouldn't you feel like an idiot? that said, they really are going to need some good apps to run on the 64, as it's probably NOT going to be of comparable price to a cutthroat pentium.

 

[astrocity20]

well whenever we see new apple computers is when I'll believe. Hopefully they'll have something good for us whether that be these or soemthing else.