Rosetta Supports Altivec?

[MacRumors]

OSX86Project.org posts unconfirmed reports that the latest version of Mac OS X 10.4.3 for Intel appears to contain some substantial upgrades.

First and foremost, they claim that Rosetta has been upgraded to fully support a PowerPC G4 processor including Altivec support.

Also, new ATI drives appear to offer more support for additional PC ATI graphic card chipsets.

[link92]

Altivec would be a major bonus, as it'd been quite big preformance upgrades for Pro apps, and some that wouldn't've worked otherwise now will.

[Lacero]

Anythings possible with Apple. If NASA can fly men to the moon, Apple can make Rosetta translate altivec code.

[AoWolf]

What is is altivec I thought it was some kind of 128bit thinggy that needed to be done on the CPU?

[Benjamin]

Originally Posted by AoWolf

What is is altivec I thought it was some kind of 128bit thinggy that needed to be done on the CPU?

http://en.wikipedia.org/wiki/Altivec
http://developer.apple.com/hardware/ve/summary.html

[AoWolf]

Originally Posted by Benjamin

http://en.wikipedia.org/wiki/Altivec
http://developer.apple.com/hardware/ve/summary.html

Thanks!

[lopresmb]

wow, thats good news for PPC App users in the future

[bigandy]

now i didn't see that one coming.

would be interested to see some hard evidence this were true...

[entropy1980]

If this is true I think this almost guarentees we will Intel Powerbooks soon. The biggest question mark was in regards to pro apps and if Rosetta now supports altivec then I think we are going to see new Powerbooks come January!

[Bear]

Originally Posted by entropy1980

If this is true I think this almost guarentees we will Intel Powerbooks soon. The biggest question mark was in regards to pro apps and if Rosetta now supports altivec then I think we are going to see new Powerbooks come January!

Of course, this also depends on the speed of the Altivec emulation in Rosetta.

[jane doe]

This is bogus information.

[qubex]

When PowerPC was first conceived, it was pretended as a RISC CPU so brutally powerful it could emulate x86 code without breaking out into a sweat. This was going to be a temporary measure to ensure legacy compatibility with existing applications while IBM transitioned its customers onto the new architecture. Maybe few of you will remember this, but PPC (and the "Mocassino" project) were so widely expected to replace the Wintel paradigm that Microsoft went as far as releasing Windows NT for PPC.

A decade later, here we are discussing the emulation of the PPC's fastest component on commodity x86 Intel chips. Things don't always go the way you expect them to.

It's sad that there's no evidence whatsoever of an "Anti-Rosetta" for running OSX86 code on PPC machines. In a couple of years my current machines (including my Quad G5) will be totally useless and obsolete, with no new software to run, and probably no new OS releases either. It's even sadder because PPC, especially in the G5 incarnation, is the better architecture.

[entropy1980]

Originally Posted by Bear

Of course, this also depends on the speed of the Altivec emulation in Rosetta.

Yes but lets assume we take a 30-40% performance hit. I would not be surprised if a dual core Yonah would be at least 30-40% faster than a 1.67 G4 (for starters the bus speed on Yonah is like 3 times faster alone not to mention we are talking essentially having 2 processors thanks to being dual core). That said I think it's totally plausible that we see decent speeds IF they actually did get altivec working in Rosseta.

[dr_lha]

Originally Posted by qubex

It's sad that there's no evidence whatsoever of an "Anti-Rosetta" for running OSX86 code on PPC machines. In a couple of years my current machines (including my Quad G5) will be totally useless and obsolete, with no new software to run, and probably no new OS releases either. It's even sadder because PPC, especially in the G5 incarnation, is the better architecture.

I'll guess we'll have to see how it goes, but I think for the near future Apple will insist on "Universal Binaries" for all applications. I.e. they will still be coded for PPC. Of course this will inevitably slowly go away, but as Mac owners tend to keep their hardware longer than PC users, it hopefully will go away slowly.

[Josh396]

Originally Posted by entropy1980

Yes but lets assume we take a 30-40% performance hit. I would not be surprised if a dual core Yonah would be at least 30-40% faster than a 1.67 G4 (for starters the bus speed on Yonah is like 3 times faster alone not to mention we are talking essentially having 2 processors thanks to being dual core). That said I think it's totally plausible that we see decent speeds IF they actually did get altivec working in Rosseta.

I was thinking the same thing. If they can get a 2 GHZ or better Pentium chip in a Powerbook they shouldn't be all that slow.

[kainjow]

This is big news, if it's true. I wonder if it does mean Intel Macs are coming in the next few months? 2006 will be a very exciting year

[BenRoethig]

I have heard rumors of something to emulate Altivec calls using SSE3. I wonder if this is it. Interesting if true.

[thequicksilver]

Presuming this is true, it's yet another reason to wait for the Intel based Macs if you can possibly afford to.

IMHO etc.

[PPC970FX]

Dudes the SSE3 is the closest we will ever et to Altivec on a x86 chip EVER. Emulating hardware is like dressing up a man as a women, he may look like a girl but if you shag him you will fast understand and feel that it is not a girl.

[Lacero]

Originally Posted by PPC970FX

Dudes the SSE3 is the closest we will ever et to Altivec on a x86 chip EVER. Emulating hardware is like dressing up a man as a women, he may look like a girl but if you shag him you will fast understand and feel that it is not a girl.

^

Great analogy! We should make that into a guide.

[nagromme]

If true, then I'd say the point is NOT to gain the SPEED benefits of AltiVec. There's no getting around the need to port to Intel if you want top speed.

Rather, the point is that AltiVec-dependent apps will RUN!

I'd rather have my apps during the transition at any speed than not at all.

Hope it's true.

PS: Many apps that SAY they need a G4 are actually not AltiVec dependent. Quake-3 engine games, for instance, say they need a G4, but run (and quite well) in Rosetta--even the Rosetta that demands G3-compatibility. That's because the G4 requirement is stated simply because no G3 was fast enough--not because of AltiVec. But Rosetta can run apps faster than those G3s.

[aegisdesign]

Originally Posted by entropy1980

Yes but lets assume we take a 30-40% performance hit.

That's quite an assumption bearing in mind Intel native vector code is way slower than AltiVec already.

Originally Posted by entropy1980

I would not be surprised if a dual core Yonah would be at least 30-40% faster than a 1.67 G4 (for starters the bus speed on Yonah is like 3 times faster alone not to mention we are talking essentially having 2 processors thanks to being dual core). That said I think it's totally plausible that we see decent speeds IF they actually did get altivec working in Rosseta.

The Yonah bus is a quad-pumped 667Mhz bus ie. 4 x 166Mhz. Current G4 is 166Mhz. Yonah gets about 6.4GB/s. G4 gets 1.3GB/s. That doesn't look good does it.

However, it's not that simple. Much of the time the CPU FSB isn't involved in transferring data and it's done entirely using direct memory access. With DDR sitting on a 167Mhz bus that gives you 2.6GB/s. Apple's Intrepid2 system controller supports DDR2 memory and talks to it on a 64bit 333Mhz bus. ie. 5.2GB/s theoretically - I'm not sure if the RAM they ship does that. I think that's actually faster than most of the Dothan based laptops which max out around 4.3GB/s. Still not good enough for some people I guess as they'll always fixate on the 166Mhz FSB with complete disregard for other components in the system.

In any case, as soon as you hit a hard drive the argument over which FSB is faster is kind of stupid.

[dr_lha]

Originally Posted by nagromme

If true, then I'd say the point is NOT to gain the SPEED benefits of AltiVec. There's no getting around the need to port to Intel if you want top speed.

Rather, the point is that AltiVec-dependent apps will RUN!

I'd rather have my apps during the transition at any speed than not at all.

Hope it's true.

PS: Many apps that SAY they need a G4 are actually not AltiVec dependent. Quake-3 engine games, for instance, say they need a G4, but run (and quite well) in Rosetta--even the Rosetta that demands G3-compatibility. That's because the G4 requirement is stated simply because no G3 was fast enough--not because of AltiVec. But Rosetta can run apps faster than those G3s.

Exactly. Clearly Altivec is going to be no speed demon emulated on x86, but it will increase the compatibility with PPC software.

[CmdrLaForge]

If true by far the best news in weeks!!!

[Photorun]

Bah to you naysayers, we're goin' Intel, DEAL WITH IT!!! If Jobs/Apple was right, Intel doesn't wear the Moto curse (that IBM got) and their road map pans out you wont remember all the griping here in these forums when we're rockin' out on cheaper dual-dual core chips or using fast competitve Powerbooks that get over 10 hours of battery life.

So it's emulation, every bit helps! This is great news, or would you whiners rather have NO altivec translation? Geez!