Looks like Apple is starting to shy away from PowerPC and has released a chip made by Intel.

Mac users.... is this going to cause problems with legacy apps? discuss!

---

Mac users.... is this going to cause problems with legacy apps? discuss!

Well, the only "news" is that they're actually shipping the first Intel-based iMacs six months ahead of schedule... BOINC Manager and the SETI application have been compiled already for these, but not released yet, in part because nobody actually has a "shipping" system to test on, only the developer boxes. I gave one of the volunteer developers David's email address a while back, as the two projects _he_ was interested in compiling for were SETI and Rosetta; it wouldn't surprise me to see a Rosetta app _fairly_ soon (and no, I have no knowledge of any of what's going on there, if he was able to get the source code, nothing). The Einstein guys are awfully good at the Mac side too, so I doubt they'll be far behind - _although_, all the Altivec code will have to be removed, so the Einstein app may actually run _slower_ on the new Macs.

Will it cause problems with legacy apps? Definitely. Apple's PPC emulation technology (confusingly named Rosetta...*) for these boxes emulates a G3, NOT a G4 or G5. So painfully, the newest/fastest legacy apps are the ones that are going to run poorly, or just not run at all. Unless of course the vendor has a new version, which I'm sure the "majors" will very soon. (Luckily, _IF_ you happen to use Apple's own development tools, and no Altivec, all you need is pretty much a straight recompile. Of course, if you use a 3rd-party compiler or toolset - as I always did when I was doing Mac development - you're out of luck. Rewrite time!)

Just for one example of the "minor" problems; if I followed the discussion correctly, the SETI and Einstein screensavers don't work on the new Macs. Will _I_ rush out and buy one? Not hardly, even if I had the $$; I have _way_ too many "legacy" apps that would cost too much (time _and_ money) to upgrade, including some that run under Classic (that's Mac OS 9, not SETI Classic...) which isn't supported on these boxes. For someone just converting from Windows, or someone who only uses Microsoft Office and the iTools suite, they would be great.

*Footnote: Until the Mac/Intel version is released, we can run the Rosetta app for the Rosetta project through the Rosetta translator on the Mac while running the Rosetta app for WCG as well. >_<

---

tHEY HAVE A PORTION OF MAC OS 10 THAT RUNS a PROGRAM CALLED ROSETTA THAT FIXES THE COMPATABILITY PROBLEM FOR ALL BUT THIER MOVIE AND SOUND SOFTWARE. THE MOVIE AND SOUND STUFF HAS TO BE UPGRADED TO a universal version.

sorry about the caps but my keyboard needs to be replaced.


cheers... have a nice day all......

---

The problem with a "translation" layer is that it typically will make everything dog slow. This is especially bad for intense number crunching which is what we are doing here.

For science applications and numeric intense applications this is a HUGE step backwards. Any application that used Altivec will run slower. None of them will ever run faster than a standard Dell type PC. So, a number of niches that Apple dominated because of the specifics of the hardware, well, they just blew off both feet.

My old dual G5 at 2.0 GHz does Einstein@Home work in 3 hours, the same work on a 3.4 GHz dual Xeon takes roughly 8-12 hours. Because of the high variability on the length of processing time it is hard to pin down, but the G5 is significantly faster. That advantage is now gone ...

*I* still think it was a bad move. Worse, if, according to Steve's presentation it was a speed/power question, AMD would have been a better choice.

The technical reasons that the newer machines are faster are simply that there are more CPU cycles available, significantly more cache, and a later generation of memory technology. And, if I remember the presentation correctly the performance was on speed/power basis not on speed alone. Why is this important? Because if raw speed is the goal, and for BOINC work that *IS* the goal, then they may not perform all that well at all ... The PowerPC was never about speed/power, just speed ...

Oh, and for BOINC work, Steve just marginalized all Mac owners. With two architectures to support, how many new projects are going to just support Windows/Linux ...

Sorry, OS-X alone is not compelling enough of a reason to buy Apple ...

---

Mac users.... is this going to cause problems with legacy apps? discuss!

It won't do Classic, it won't do AltiVec, and it won't do Windows. That's reeeeal convincing. Yep. Nothing like having to shell out a $hi+load for Inteicized versions of Final Cut, DVD Studio, Motion, Photoshop, and Illustrator. Joy oh joy. I'm gonna head over to my Apple Store right now and order one. You betcha. I just looooove Rosetta.

...as long as we're talking about the Rosetta handwriting recognition on the Newton MessagePads.

Seriously, this is 100% pure bad timing. The Quad G5 signaled that the Mactels were still a way away, and the PS3 (and therefore Cell) is getting closer and closer. Worse still is that Apple hadda call in what's-her-name from M$ in order to convince the masses at the expo that Mac x86 is a Good Thing.

As for BOINC... well, let's see how the numbers turn out. But if the top host list over at SETI is any indication, I'm willing to bet that G5s will still outpace their Intel counterparts clock for clock and watt for watt.

---

Well, they have some of the new computers for laptops and the lower end desktops showing with the new Intel cores. And if it is a heat thing, and in those models I saw, goinng away from the PowerPC chips was probably a smart thing to do.

But, for raw speed, bad move ... I totally agree ... look at Einstein@Home ... my G5 does work in less than 3 hours in almost all cases. My Intel and AMD systems are 4-12 hours ... it is a little harder to get good numbers with the variance in Albert work ... but I am certainly convinced ...

---

So, is the G5 the last of the real Apples?

Those of us who have not yet sampled Apple technology, should we rush out & get a G5 while we still can?

Discuss

I will as soon as the price tag on the Quad G5 comes down to reality.

I'd rather spend $3K on hardware that'll run the programs faster rather than spend $6K in completely new hardware and software.

For that I could build my own P4-64b EE with HT and put Christmas trees to shame with all the lights I'd stuff in it...

---

I will as soon as the price tag on the Quad G5 comes down to reality.

I'd rather spend $3K on hardware that'll run the programs faster rather than spend $6K in completely new hardware and software.

For that I could build my own P4-64b EE with HT and put Christmas trees to shame with all the lights I'd stuff in it...

Well, the dual 2.0 GHz G5 I have gives my 3.4 GHz Xeons a run for their money. Both systems cost roughly the same for the box (with the apple I also got a Cine display).

I doubt that the Quad G5 will come down in price much if any. More likely it will be replaced by an Intel at the same price point. What will be sad is that you will be able to go to Dell and get a similar PC equipped with the same CPU for a couple hundred less ...

---

What will be sad is that you will be able to go to Dell and get a similar PC equipped with the same CPU for a couple hundred less ...

As I said in my blog "The specs are rather decent, and I think everyone can agree that the price is reasonable (by Sony Vaio standards, that is)."

Here's to the glimmer of hope that the Quads will drop in price while I can take advantage of the student discount... otherwise I think I'll have to go and build myself an x86 box (I've got a number of issues with Dell, and Linovo doesn't seem to have their act together).

---

all the Altivec code will have to be removed

Apple's Rosetta translator supports the G3 and G4 processors, thus it does translate Ativec code. It does not support the G5. This is not to say that any individual computationally intensive application (like the BOINC Rosetta or the Einstien app) will work with the Rosetta translator, it's a "try it and see" situation.

--Nathan

The technical reasons that the newer machines are faster are simply that there are more CPU cycles available, significantly more cache, and a later generation of memory technology. And, if I remember the presentation correctly the performance was on speed/power basis not on speed alone. Why is this important? Because if raw speed is the goal, and for BOINC work that *IS* the goal, then they may not perform all that well at all ... The PowerPC was never about speed/power, just speed ...

The speed/power issue is all about the Powerbooks. Apple's Powerbooks are still running on the G4 because IBM never gave Apple a version of the G5 that had a low enough power requirement to actually use in a notebook. Apple has been losing market share in the notebook market for a while now because of this. A month after Steve announced that Apple was switching, IBM announced that a lower power version of the G5 was coming soon. :(

See Inside the big switch for information on some other possibilities on Apple's switch to Intel.

--Nathan

Nathan,

interesting article and matches my experiences. Heck when they did the Lisa the stores could not keep them on the shelves, to the point that Mac sales sufferred... the most approprite response ... kill the computer that sells because Jobs like the Mac ... go figure ...

And if it HAD been speed/power purely, Apple would have chosen AMD ...

Still a real shame ... I will likely buy a Quad this year, but, will also then likely sit on the sidelines for a yeear or two to see if they change back to PowerPC chips or not. As good as OS-X is, I buy computers for the speed in doing the models under BOINC. The G5 is the cat's pajamas at Einstein@Home, and very close at SETI@Home ... the new G5 with the much larger cache is very likely to be a real killer there ...

Ah well, next upgrades will be to put dual core AMD chips into my slowest box to see what I get there ... if that works ... :)

---