IBM 64-Bit PowerPC Details

[ddtlm]

barkmonster:

One simply cannot expect the P4 performance to scale linearly with increasing clock speed (as you showed in your theoretical numbers) because the memory system is not scaling that fast. Not to mention, I seriously doubt we'll see a 4.7ghz P4 available before the PPC-970 is available. Right now I don't think we'll see more than 4.0ghz before that time.

 

[beatle888]

Wow, I actually have something fairly significant
to point out

Ok this was taken from an article and it raises
the question....even if the P4 can go to 5ghz by
next year so what....the 970 will have better
architecture all around and will smack the A$$ of
the P4 5ghz (hypotheticaly 5ghz) because of all
it's advanced memory and front side bus capabilities.

"The 900MHz frontside bus

One of the most important and least-discussed features of the PowerPC 970 is its 900MHz DDR frontside bus. This bus physically runs at 450MHz, but it's double-pumped. Its architecture is interesting in that the bus actually consists of two, 32-bit unidirectional point-to-point links. David Wang described it in a post to comp.arch as follows:

It's two 32 bit links: one from CPU to "companion chip" [the northbridge], and one back from that chip to the CPU. Each link runs at 900 MHz (1.8 GHz CPU core. the interface link runs at integer fraction of the CPU core, in this case 1/2)

So 4 bytes to, 4 bytes from, at 900 MHz that's 3.6 GB/s raw BW each way. The link multiplexes command and address info over the same pins, so it's some sort of packet based protocol. The math gets you 7.2 GB/s of raw bandwidth, but after subtracting out command and address overhead, raw peak data bandwidth is supposed to be about 6.4 GB of that 7.2 GB/s.

This high-bandwidth link to the northbridge is one of the elements that's going to make this chip as a media machine; it's exactly what Apple's current bandwidth-starved G4 systems lack, and it's going to be a major selling point for systems based on the PPC 970. When coupled with the right memory in an SMP configuration, the 970 should do quite well in bandwidth-intensive applications."

Link picked up from MacCoaster's thread (he's a member here)

See? so it doesn't matter that the Putter4 (P4)
will be cranking so fast in mhz like a fiend on
cocain cause it still won't be able to fly with all
it's bottlenecks.

 

[nickgold]

Rock on!

This chip is going to go down in history as a Big Thing for us Mac-lovin' folks, I reckon. Seriously workstation-grade ****, if ya know what I mean. I can't believe people are even comparing this to the future P4... Gimme a break! This thing is a totally groundbreaking architecture... How can you compare that to pumping the MHz of the P4? Please.

Just imagine the power of a rack of Xserves sporting 970s a year and half or so from now. That would be a seriously bada$$ setup, and totally achievable for a pretty large segment of the computer-buying masses (small businesses, etc.) Even a desktop dualie or quad would be sick. From what I understand, the power sonsumption is pretty decent, right? Can it make it into a laptop? Now that would be something else...

All I know is, I am going to get a serious Apple Loan when these puppies come out...

 

[nickgold]

What will 64-bit mean?

I'm surprised that more people aren't talking about what moving to 64-bit will really mean for Macintosh users. Do any of us really know? Apple may very well be the first to market a 64-bit chip to the masses. What will people do with this power? What kind of totally new applications will be available, that to a degree were not previously possible (at least on the desktop level). What will individuals DO with pretty much literal super-computers at their disposal? I, for one, am psyched. But I don't quite know exactly why.

 

[beatle888]

well right off the top of my head i would expect
multi tasking to be one benefit. I know we can
already do this but i think with 64bit maybe the
processor can handle more tasks at once without
a major cpu hit.

 

[ddtlm]

nickgold:

Wowa, calm down. AMD's Hammer will be out first with 64-bit for the desktop, although M$ won't be supporting it in Windows for a while (maybe not till 2005?). People using Linux will have all 64 bits though, and that means servers and renderfarms.

64 bits is generally not important. It mainly helps with the max amount of RAM a single process can easily use, and it does not mean twice the performance. 64-bitness is not a very big deal, but it's still a good thing.

 

[ddtlm]

beatle888:

i would expect multi tasking to be one benefit. I know we can already do this but i think with 64bit maybe the processor can handle more tasks at once without a major cpu hit.

NO. Not to single you out, but for the benefit of everyone let me say very clearly that 64-bitness does not mean higher performance. For the vast mahority of the things we do, 32 bits is fine. End of story. 64-bitness is not important unless you are trying to give more than 4GB of RAM to a single task, or doing things with 64-bit integers (not a common thing).

Current 64-bits chips (well, those that are fast anyway) are fast per clock cycle for reasons other than the 64-bitness.

 

[Telomar]

Re: What will 64-bit mean?

Originally posted by nickgold
I'm surprised that more people aren't talking about what moving to 64-bit will really mean for Macintosh users. Do any of us really know? Apple may very well be the first to market a 64-bit chip to the masses. What will people do with this power? What kind of totally new applications will be available, that to a degree were not previously possible (at least on the desktop level). What will individuals DO with pretty much literal super-computers at their disposal? I, for one, am psyched. But I don't quite know exactly why.

The main benefactors will be the graphics community, anyone that requires over 4GB of RAM, some server/database apps and most importantly from my point of view the scientific/engineering community.

I used to do a lot of work using spaces that called on very high levels of precision. As a result a 64 bit workstation was a permanent figure on my desk. The ability to integrate that to my main working computer would have been a big plus.

 

[beatle888]

so what about the eight instructions per cycle?
or the front side bus...or the memory bottleneck
we're experencing now? wouldnt that increase
speeds of say mp3 ripping for an example of
something not graphics related.

ddtml you dont need to respond to this...i know
you've been over all this before.

 

[ddtlm]

beatle888:

Assuming that Apple puts some equally fast RAM (dual channel DDR or so) on their system controller, then the FSB will be absolutely great for encoding things like mp3s and movies. The PPC-970 should really dominate that. Certain things that are highly AltiVec optimized and deal with huge files should go like stink... at rates never before seen on any machine.

The 8 instructions per cycle is a peak number, and as such you will essentially never reach it. In fact, there is another peak number that gets in the way: the number of instructions that can be "retired" per cycle: 5. And one of those 5 must be a branch, plus some other restrictions. 8 is really not a number you are going to ever see. However, those are specs for the main scalar core... the AltiVec unit will probably be able to go at full speed along side without interference. Or so I am guessing.

The PPC-970 should be extra strong in all the places the G4 is now, and pretty good in the places the G4 is just lame. Overall I'd say a nice chip. I plan on getting a system with them ASAP.

 

[beatle888]

Originally posted by ddtlm
beatle888:
at rates never before seen on any machine.

wow that does sound super (pun intended)
sorta sounds like your describing super heroes.

well thank you for explaining it ddtml. im not a
very knowledgable guy when it comes to how
computers REALLY work. seems like all we have
to do now is be patient and see what this chip
can really do once it's released. i can't wait it
should be a blast from all that i've read here.
2003/2004 should be a wonderful time for mac
enthusiasts.

 

[ddtlm]

beatle888:

Just doing my part for the knowledge of the Mac community in general.

Not trying to start a hype-fest by saying "at rates never before seen" either... that is simply a logical conclusion based on the theoretical rate at which AltiVec can much masses of numbers, and based on the theoretical rate at which those numbers will be carried to and from the AltiVec units.

Seems to me that S Jobs will be able to greatly widen the number of filters he uses for the Photoshop bakeoffs when the PPC-970 ships.

The PPC-970 will not be the best chip at all things... but it will be a top-notch chip for things Macs tend to do.

 

[beatle888]

ddtml, i must agree, it sounds great, .
i don't really care if the mac is the fastest
computer, as long as they can be competive
which i think will benefit PC and Mac users.
im actually very happy with my 667 tibook and
512mb of ram. i mostly work in photoshop and
sometimes this system tends to slow me down.
However im using the internal 4200rpm hard
drive...so i have some room to improve this
system. I think that with a gig of ram and an
external firewire drive for a scratch disk my
photoshop work will be very efficient. AND if
i really needed the performance...hell i could
even get a RAID system...so i think where
all in pretty good shape however I know there
are some out there that need more power than
the current systems.

 

[nickgold]

Use your imagination

Originally posted by ddtlm
64 bits is generally not important. It mainly helps with the max amount of RAM a single process can easily use, and it does not mean twice the performance. 64-bitness is not a very big deal, but it's still a good thing.

Yes yes, the max amount of RAM a single process can easily use, yada yada. You didn't really address my question: What will it allow us to do? For instance, I imagine that the sort of apps creative professionals, which tend to push lots of data at once (3D and video rendering, multichannel digital audio, etc.) will benefit greatly. I know that now, it is a pain to manipulate files that are larger than 4GB or so. In video work this becomes a pain very quickly, when 5 minutes of DV footage is a gig. 64-bit should affect this dramatically, I'd imagine.

That's the kind of stuff I'm talking about. Heck, wouldn't even something as "banal" as a FPS benefit from this sort of leap? That is lots of data being processed at one time. Doesn't something akin to William Gibson's view of a VR-based "cyberspace" become more of a reality? That would require the processing of enormously large amounts of data at one time.

Use your imagination here, kids! There have been HUGE leaps in the computing paradigm between 8-bit, 16-bit, and 32-bit. Now, obviously the bit-depth of the processors over the years has been but one aspect of the shift in computing applications. So, let's think about what this chip represents on a more symbolic level. Between it being 64-bit, and having a 900MHz frontside bus (YOWZA), etc. Face it: THESE THINGS ARE GOING TO RAGE! Now what will we be doing with them that we can't be doing now? There has gotta be SOMETHIN', or else we wouldn't be sweatin' 'em so much!

 

[ddtlm]

nickgold,

No, you are still exaggerating the benefits of 64-bitness. Yes going from 16 bits to 32 was great, but thats because the biggest number you can store in 16 bits is 65536 or something, whereas with 32 bits its 4 billion, and in 64 bits its 4 billion squared. Yep, something like 16 with 18 zeros. Now tell me, who needs a number that big? Not many people. Who needs a number bigger than 32 bits? Not many people.

64 bit does not mean that data is processed twice as fast. It means that the same data takes up twice the space. It means that instead of storing the value "389723" in 32 bits, it is now in 64 bits, of which well over 32 are zero. As in blank. As in, useless.

Most programs do not need 64 bits, and so by not using 64 bits they will be more compact and therefore faster. Yes, you heard me right, 32 bits is generally faster. The only time 64 bits is faster is if 4 billion isn't a big enough number for you. (I should note that double precision floating point are already 64 bit, and the G4 supports exactly the same "doubles" as the PPC-970.) Litterally, there are two things where 64 bits can make you go faster: if you need mroe than 4 gigs of RAM in a single app, and if you need an integer bigger than 4 billion (or +/- 2 billion for signed integers).

Lets also draw a line between the 128-bit AltiVec registers and a 64-bit processor. AltiVec goes fast because it manipulates 4 32 bit chunks (or 8 16's, or 16 8's)... not because it manipulates 128 bits at once. A 64-bit processor cannot do 2 32-bit ops at once instead of one 64-bit op. It just does not work that way.

I know that now, it is a pain to manipulate files that are larger than 4GB or so. In video work this becomes a pain very quickly, when 5 minutes of DV footage is a gig. 64-bit should affect this dramatically, I'd imagine.

Sadly this is not the case. The 4gig files will still take forever to read off of the hard disk. 32 bit processors can handle massive hard disk files fine, just not more than 4 gigs of RAM dedicated to a single program.

 

[solvs]

Originally posted by 3G4N
Ummm, Maybe because... uuhhh..
THE PPC-970 DOESN'T FRIGGIN EXIST YET!! Jeezus.
Where *didn't* you read the chip won't go into production
until late 2003??? Does IQ need to become part of the
registration here?

End rant.
Boy, do I need a job!

Exactly, the Spec marks are estimates. We don't know what they will really be yet, or even if they will be in the next PowerMacs. I still say don't trust the Spec scores for much. AMD was just saying how the scores seemed to be biased towards certain types of CPUs. They scored well in Spec2000 and 2001, but were slaughtered in 2002. Benchmarks, shmenkmarks, I just want REAL WORLD PERFORMANCE.

IQ tests are the same way. Numbers don't say much. I can say that because I have a VERY high IQ. I just do a lot of really stupid things.

I need one of those job thingies too.

 

[ddtlm]

solvs:

The tests that AMD was bitching about were not SPEC... go to Anandtech sometime and witness the huge number of tests that PC sites run. In the case of the evil benchmark of doom, Intel essentially ran the damn thing and has changed it many times... each time the Athlon started beating Intel's best the rules changed. Note, that is NOT how SPEC operates... no one company controls SPEC.

 

[Shrek]

Originally posted by 3G4N


? is this an allusion to:


one of my fav lines.

Absolutely!

 

[solvs]

Originally posted by ddtlm
solvs:

The tests that AMD was bitching about were not SPEC... go to Anandtech sometime and witness the huge number of tests that PC sites run. In the case of the evil benchmark of doom, Intel essentially ran the damn thing and has changed it many times... each time the Athlon started beating Intel's best the rules changed. Note, that is NOT how SPEC operates... no one company controls SPEC.

My mistake, it was SYSMark, not SPEC. I still don't trust benchmarks as an "end-all-be-all" but I guess they can give you an idea, or at least a rough approximation, of raw speed.

Any speed test can be manipulated to make anything look faster or slower. And whenever a CPU looks slower, somebody has to cry foul. Sometimes they're right, sometimes it's just sour grapes.

If the new 970 is used in the next BIG Mac upgrade, and it performs fast (estimates and benchmarks be darned), I'll be happy.

 

[scem0]

Everyone here is talking about speed and benchmarks, etc. BUt what I am more concerned about is price. Will we be getting a computer that is actually a good deal? I really hope so.

 

[shadowfax0]

Originally posted by 3G4N

Pull up a chair, a cold beer, and a bucket, and start yer droolin.

I'll drink to that!

 

[MacCoaster]

Originally posted by ddtlm
nickgold:

Wowa, calm down. AMD's Hammer will be out first with 64-bit for the desktop, although M$ won't be supporting it in Windows for a while (maybe not till 2005?). People using Linux will have all 64 bits though, and that means servers and renderfarms.

64 bits is generally not important. It mainly helps with the max amount of RAM a single process can easily use, and it does not mean twice the performance. 64-bitness is not a very big deal, but it's still a good thing.

2005? Really? I doubt it will be that late because, *gasp*, Microsoft already has Windows XP 64-bit running on the Hammer architecture. I would think they'd release it as Hammer is released.

 

[shadowfax0]

So what makes these:
http://www.sgi.com/workstations/comparison.html
so great? I'm dead serious, WHY ARE THESE WORKSTATIONS $20,000??

 

[Catfish_Man]

I second the question...

Originally posted by shadowfax0
So what makes these:
http://www.sgi.com/workstations/comparison.html
so great? I'm dead serious, WHY ARE THESE WORKSTATIONS $20,000??

...I realize they're a ton faster than they look, but NONE of those numbers are any good. It seems like the sole redeeming feature of that machine is the support for 4 gigs of ram. Obviously I'm missing something.

 

[maxvamp]

Thoughts from a switcher..

I guess I am what you would call a switcher. I was an OS/2 to Windows NT to LINUX/UNIX to MacOSX ( better UNIX ) switcher. I now only use OSX and Windows 98 ( and then, only for 1 application )

Having said that, I have to say that this has been a fun discussion to watch. I want to put the memory concerns of this new processor to rest by simply saying that I would be shocked if Apple didn't user DDR 2 ( see http://www.commsdesign.com/design_center/broadband/news/OEG20021016S0038 ). This would satisfy the memory bandwidth needs of this new chip.

Second, as for the no performance gain, coupled with the faster P4 mantra, I have to say that there are some mis-informed people here. I suggest that we start a new thread and have everyone list the top 5 apps they use on the mac. This is what would clearly define any performance gain a user would or wouldn't see.

The reason I suggest this is that many high performance app developers on the window side ( and I am sure Mac too ) have been working with 64 bits for a while in their 32-bit applications. Even MS in several of their APIs will divide their values returned from objects in two 32bit chunks that must be reassembled into the original 64 bit values. I myself have seen this in file system operations in the Win32 writing I have done.

Since such operations described above require two ticks of the proc to process, I would take a slower 64 bit AMD or PPC processor in a heart beat over a much faster p4 for hard core applications. So a 5 GHz p4 at best on these types of apps still only keeps up with a 64 bit at half it's speed. An OS optimised for 64-bits make this more so.

Now, will every app see a difference? No.

If my average apps I use often are only Quicken, Mail, IExplorer, iChat, Word, etc. then a 2.4 GHz P4 will be snappier than a 1.8 PPC or AMD. If your average apps are DVD Studio, Final Cut Pro, Maya, Cinema4D, etc. then a 64 system will be an improvement. I won't even go into databases, except to say that in the future, more and more apps will start using them. Need proof ? iPhoto and iTunes both are primitive apps that use a crude database to track your data. Overtime, more and more apps with get more advanced in this idea, and you may see OSs come out with master Databases built into them that iApps, as well as other apps can use natively.

So, with all of this said, What are the apps you use on a daily basis?

Let the flames begin,
Max.