PPC970 at WWDC

The PowerPC began life as the single-chip version of the POWER chipset. POWER was and is 64-bit. With the PPC 970, the PowerPC is returning to its roots. It won't have to emulate anything. By your logic, the G1 through G4 emulated 32-bit.

Yes, emulation is not necessary... But without any specific recoding to take advantage of the new 64-bit chip, the OS/software won't benefit from the new improved CPU. In some cases, it might actually run a bit slower. Of course, if Apple makes use of the new Hypertransport bus as expected, this would probably offset any temporary speed deficits due to not having 64-bit clean OS/software.

You clearly don't understand how the original PPC emulated 680x0 code. Apple wrote a 680x0 emulator which it added to the Toolbox ROM. Until the advent of the NewWorld Macs, the 680x0 was emulated in firmware. With the advent of the NewWorld Macs and ROM in RAM, Apple replaced the Toolbox ROM with a ROM file. NewWorld Macs emulate the 680x0 in software.

You clearly don't understand the expression "sort-of". The comparison wasn't a technical one of specifics, but of generalities.

When Apple first moved to PPC, none of the software or the OS was optimized for it. All of the 68K code had to be processed through an emulator, which was transparently included in the OS. The first generation high end PPC Mac (8100), while a fast machine, initially was hobbled by these lack of optimizations. When compared to the Quadra840AV which the 8100 replaced, the 8100 usually came out slower in most tests (Photoshop especially). Once more of the OS and the applications dumped the 68K code and replaced it with PPC code, the situation was reversed - the 8100 ending up besting the 840AV in most tests. Unfortunately, it took Adobe over a year to update Photoshop to take advantage of PPC, and it took Apple even longer to remove all the 68K code from the OS (System 9 was the first version to be almost 68K code free).

If the 970 debuts before Panther does, you can expect it to be comparitively hobbled by an OS (10.2) that is not tuned for a 64-bit CPU, and even more so from applications that are not as well.

The situation is similar in that Apple will have released next generation hardware before it could get the next generation software out the door.
 
Originally posted by serpicolugnut
....

You clearly don't understand the expression "sort-of". The comparison wasn't a technical one of specifics, but of generalities.

When Apple first moved to PPC, none of the software or the OS was optimized for it. All of the 68K code had to be processed through an emulator, which was transparently included in the OS. The first generation high end PPC Mac (8100), while a fast machine, initially was hobbled by these lack of optimizations. When compared to the Quadra840AV which the 8100 replaced, the 8100 usually came out slower in most tests (Photoshop especially). Once more of the OS and the applications dumped the 68K code and replaced it with PPC code, the situation was reversed - the 8100 ending up besting the 840AV in most tests. Unfortunately, it took Adobe over a year to update Photoshop to take advantage of PPC, and it took Apple even longer to remove all the 68K code from the OS (System 9 was the first version to be almost 68K code free).

If the 970 debuts before Panther does, you can expect it to be comparitively hobbled by an OS (10.2) that is not tuned for a 64-bit CPU, and even more so from applications that are not as well.

The situation is similar in that Apple will have released next generation hardware before it could get the next generation software out the door.
You need to reread my post. The reason that Photoshop and other applications were slower when run on the PowerMac 8100 was that they were still running 680x0 code. Native PPC applications execute much faster than 680x0 applications executed by the 680x0 emulator. Emulation versus native code is a completely different issue than generic code versus optimized code. Optimization refers to maximizing the performance of an application for the particular processor being used. Just because code is native to a processor does not mean it is optimized for that processor.
 
I think apple would be stupid to release a new processor with out having at least a couple of high end apps (adobe possibly - and apps like shake, final cut pro) that support the new processor.

When they release it it will be the first 64bit processor for the 'consumer' market (if we can afford it) and it will need to have huge performance gains. Releasing a new chip - the latest and greatest - that appears to run 'only' as fast as the 'old' G4 would be a PR disaster. This beast needs to be impressive.


and on a completely random tangent:

I think apple should make a dual processor powerbook and find a way to make it run on one processor when under battery power (to conserve battery). I imagine this would be really useful to people doing things like mobile digital video (news media/doco makers etc...) who need a portable computer but one that can go as hard as a top end desktop when plugged in.

just my 2 cents
 
You need to reread my post. The reason that Photoshop and other applications were slower when run on the PowerMac 8100 was that they were still running 680x0 code. Native PPC applications execute much faster than 680x0 applications executed by the 680x0 emulator. Emulation versus native code is a completely different issue than generic code versus optimized code. Optimization refers to maximizing the performance of an application for the particular processor being used. Just because code is native to a processor does not mean it is optimized for that processor.

Yes, but the end result was that the newer 8100 was temporarily hobbled by the OS and applications that were not ready for the new architecture. This could also be the end result of the 970 being released before Panther.

While the 970 should handle OS 10.2 and the current releases of pro apps fine, neither will be specifically tuned for the chip. Since there are no prototypes to test, nobody really knows how well IBMs claim of full 32-bit support will pan out. Plus we still don't know at what speeds it will debut. Initial reports said up to 1.8ghz. Then there were reports that the chip had hit 2.5ghz in initial production. Then those reports we debunked, and the original 1.8ghz report stood as the best guess. So, outside of IBM and Apple, nobody really knows how fast this chip will be at launch.

Obviously, if we get the 1.8ghz chips at launch, those systems should be much faster than what we currently have.

But if Apple is only able to initially procure 1.4ghz 970's, I wouldn't expect a huge speed increase over the 1.43ghz G4, uless of course, the rest of system is upgraded too (hypertransport 900mhz FSB? GForce FX? USB 2? - hopefully all of the above...)
 
...a possible G970 will offer anyways :p a speed boost at the same speed of a G4 proc, even for current apps, simply because:
-It will offer a MUCH better data BUS solution
-It will use current AND faster DDR RAM solutions even better
-When running 32 bit apps actually is like running 64 bit apps or if you prefer a little bit of emulation will going on
-It will be able to scale even better with not only Dual configs but Quad and above as well
-It will have more L1&L2 Cache
-Its "Altivec" capabilities are enhanced

All in all, the new CPUs combined with the new OS X.3 (aka Panther) will make a formidable opponent even when running current 32 bits apps and "only" at 1.4GHz simply because they will be able to interact with all the other hardware a lot better than G4 currently does...

Still, I would like Apple to give us a similar PowerMac line which I think is logical too!?:
-Single G5/1.4 ~ Fast
-Single G5/1.8 ~ Faster
-Dual G5/1.4 ~ Fastest
-Dual G5/1.8 ~ Ultimate

:D ;) :)
 
Still, I would like Apple to give us a similar PowerMac line which I think is logical too!?:
-Single G5/1.4 ~ Fast
-Single G5/1.8 ~ Faster
-Dual G5/1.4 ~ Fastest
-Dual G5/1.8 ~ Ultimate

I wouldn't bank on Apple marketing the new IBM 970 Macs as the "G5". First of all, Motorola has produced a "G5" chip, and it's currently in mass production for embedded systems. The desktop version of the G5 was a total disaster for Moto, and the product has been scrapped.

Apple really needs to convey that this chip isn't just a minor evolution from the G4, but the start of a completely new architecture (if they do use Hypertransport, which they probably will).

How about "Power Mac 64"? Sounds a little too reminiscent of the "Commodore 64", though, doesn't it? :)

Personally, I like "Power Mac 970".
 
Originally posted by profx
I think apple would be stupid to release a new processor with out having at least a couple of high end apps (adobe possibly - and apps like shake, final cut pro) that support the new processor.

When they release it it will be the first 64bit processor for the 'consumer' market (if we can afford it) and it will need to have huge performance gains. Releasing a new chip - the latest and greatest - that appears to run 'only' as fast as the 'old' G4 would be a PR disaster. This beast needs to be impressive.

Apple has been lagging performance wise (on hardware) for quite some time. Imagine if we had waited for Quark Express before we got OS X. Most of us would still be hobling along in OS 9.

Apple needs to get their fastest out there as soon as possible. Once people see what the new chip will do they will come running to it.
 
When the first G4 came out (450 I think), I immediately bought one based on the Altivec hype. After running a battery of tests on multiple machines, I quickly realized that Altivec and the G4 wasn't all it was cracked up to be.

The G4 has scaled reasonably well and is now decent on a DP system.

My point is simply that Apple are masters of hype. I consider myself to be a high end Mac user. I need every ounce of performance. Apple, don't give me any more stupid Photoshop filter tests. Don't talk about BS Altivec and its miracle advantages. Just give me a fast processor with a modern architecture. Don't put DDS RAM in the thing if the system can't even take advantage of it!

As others have said. This cannot be a speed bump. This needs to be a whole new CPU experience. If you can, read page 45 in the April issue of Videosystems magazine. It's enough to make you want to give up and go crawl in a cave.

My last five Macs were purchased without any in-store testing or even waiting for reviews. I simply bought because it was the latest Mac. I have been mildly disappointed each time. Not because they weren't good machines, but just because the BS hype was a little over the top. Misleading, if you will.

I'm due for a new box come January 2004. This time I will be going to the local Apple store just to test. By then, plenty of reviews will have come and gone. There may even a small speed bump of the new chip by then. either way...Apple, impress me. You've got the best OS in the world, how about providing the best hardware?
 
Check out the benchmarks on MacBidouille. Totally awesome! The PPC 970 literally blows away every other processor on 32bit programs & OS. Performance seems to double with duals. If all this is true (and there is a fair chanche) the next PowerMacs are weel worth waiting for!
 
I would be (and I am) skeptical of MacBidouille...Their track record is pretty shoddy...

About the only rumor site I might even think of believing is ThinkSecret. Their track record, while not perfect, has been pretty good.

Either way - mindbend brings up another point about cart/horse with Altivec. Apple heavily hyped it, and although some applications made good use of it (Photoshop, Blast!, iDVD), others did not (AE, Lightwave, iMovie), and it wasn't the feature it was hyped to be (in fairness, though - that usually was due to the developer not implementing, no fault of Apples).
 
Originally posted by mindbend
When the first G4 came out (450 I think), I immediately bought one based on the Altivec hype. After running a battery of tests on multiple machines, I quickly realized that Altivec and the G4 wasn't all it was cracked up to be.

Just out of curiosity, what do you mean about Altivec not being everything it was hyped up to be? Did you find that apps that were built to take advantage of Altivec were not performing better than those that weren't? Did the speed increase not warrant the price differential?
 
My tests were done using a G3, a G4 and a PIII at similar clock speeds. When possible, I used identical applications.

I ran a bunch of Photoshop filters, transferred files, opened/saved files, ripped MP3, rendered video, rendered 3D objects and more.

I am here to tell you that, pound for pound, the G4 did not deliver in most cases across the board. There were/are a few exceptions where the performance really is way beyond the pure CPU expectation, but 99% of the time the determiner of speed was simply the CPU speed.

To answer binarydigit's question specifically...Altivec did NOT deliver on its speed promise, even in the case of supposedly optimized apps. My numbers show that Altivec typically gets you a measly 5-10% gain (again, there are a few exceptions, but they are rare). Even the Photoshoip tests were a joke, maybe pushing 30% gains on the optimized filters (from what I remember-I don't have the spreadsheet in front of me).
 
Originally posted by mindbend
My point is simply that Apple are masters of hype. I consider myself to be a high end Mac user. I need every ounce of performance. Apple, don't give me any more stupid Photoshop filter tests. Don't talk about BS Altivec and its miracle advantages. Just give me a fast processor with a modern architecture. Don't put DDS RAM in the thing if the system can't even take advantage of it!

As others have said. This cannot be a speed bump. This needs to be a whole new CPU experience. If you can, read page 45 in the April issue of Videosystems magazine. It's enough to make you want to give up and go crawl in a cave.

I'm due for a new box come January 2004. This time I will be going to the local Apple store just to test. By then, plenty of reviews will have come and gone. There may even a small speed bump of the new chip by then. either way...Apple, impress me. You've got the best OS in the world, how about providing the best hardware?

This is widely known as Jobs' "Reality Distortion Field" No matter what he says at a keynote, everybody eats it up and is drooling for more. If nothing else, he is an excelent public speaker.

As far as the 970 being an entirely new experience... I think apple has been listening to everybody complaining about the performance of their machines. Up to this point they just frankly couldn't do anything about it. They had to tout every little advancement as the next great thing because they didn't have the resources to create a truly great advancement.


This new proccesor really needs to go hard or go home. If it doesn't live up to expectations I'm not sure what Mac users will have to look forward to. The next generation chip is probably WAY down the line.
 
If MacBidouille's benchmarks are even vaguely accurate, the new processors are really going hard. I don't think the benches are outright fake and made up. First, they have no reason to do that, they would get a peak of traffic now, and next to none when proven wrong.
Second, you don't simply conjure benchmarks out of thin air. They know they are serving a techie crowd, and all the numbers except the 970 can be tested independently. So they will have made sure of not claiming something impossible, which makes the numbers credible and even probably next to true. Since there probably are going to be several changes in the hardware setup and the OS (they allegedly used and alpha build of Panther), they can be wrong, but not far off IMHO.
 
The PPC970 info from IBM has been impressive, I must admit. IBM has also been one to show it how it is, not to inflate benchmarks.
The only problem is that these benches are not of the production processor. these are pre production models that could undergo drastic changes before they're released.

Believe me, I'm pushing for these numbers to be true more than most people on this board. I've been disapointed in the past. I'm just trying to keep from getting my hopes up :D
 
If MacBidouille's benchmarks are even vaguely accurate, the new processors are really going hard. I don't think the benches are outright fake and made up. First, they have no reason to do that, they would get a peak of traffic now, and next to none when proven wrong.

They've made similiarly preposterous claims before and they still seem to get a good big of traffic, so I don't think they are too concerned about their track record...

I doubt these benchmarks, mostly due to MacBids geographics. The only place in the world right now that a fully operational IBM 970 Mac would be running is in Cupertino. The simple fact that MacBid is a French based rumor site gives me pause. Sure, they could have moles from within Apple, but I seriously doubt it judging from their past rumor history.
 
Originally posted by serpicolugnut
...
I doubt these benchmarks, mostly due to MacBids geographics. The only place in the world right now that a fully operational IBM 970 Mac would be running is in Cupertino. The simple fact that MacBid is a French based rumor site gives me pause.
...

If the rumours are correct and the systems are supposed to ship wwdc timeframe, there had better be systems out there other than in Cupertino. At a minimum there would have been developer systems out there for months. Any systems reaching peoples hands right now should be based on actually production runs.
 
I don't think that they'll be shipping anytime around WWDC. I think they'll be introduced then. THey could have as much as a 3 to 4 month lead in time before you see any of these machines in the hands of end users.
 
no way they are available at wwdc


the g4 powermacs for gosh sakes need 3 months lead time much less a new processor. it does seem like WWDC is a logical place to say something about a new processor though. developers need to know these things.
 
Back
Top