“Apple Computer is making the leap to Intel thanks in part to a software translation technology from a 65-person company in Los Gatos,” Dean Takahashi reports for The Mercury News. “Transitive Technologies confirmed Tuesday that it is providing Apple with technology that allows old Macintosh software programs to run on computers based on Intel rather than IBM chips. Transitive’s technology will be part of software called Rosetta, which will work for current Macintosh OS X programs that run on PowerPC systems but not for older programs that run on OS 8 and OS 9 software, according to Apple.”
Takahashi reports, “Apple Computer Chief Executive Steve Jobs demonstrated Rosetta on stage at a developer’s conference Monday, where he announced the alliance with Intel. ‘Steve was nice enough to recognize a relationship with us,’ said Bob Wiederhold, CEO of Transitive. ‘Our efforts involve integrating our technology into their system software.’ Rosetta will be important for Apple to hang onto its loyal Macintosh customers at a time when it is making a major switch to new hardware. If Rosetta lives up to its promise, consumers won’t have to throw away their old software when they buy a new computer from Apple with Intel chips.”
“Transitive uses only about 25 percent more memory than an older application otherwise might use. Wiederhold said Transitive’s team figured out how to do translation at roughly 70 percent to 80 percent of the speed at which it ran on the original computer,” Takahashi reports.
Full article here.
MacDailyNews Note: In an attempt to clear up confusion we see everywhere, including in some recent MDN reader feedback:
Rosetta is intended to translate PowerPC-only Mac software for “Macintel” (Macs with Intel processors) computers. PowerPC Macs don’t need Rosetta. Developers are already preparing “Universal Binary” versions of their software which, simplified, contain both “Macintel” and PowerPC Mac versions of their programs. All of the major software packages from the big players like Adobe, Microsoft and Apple will offer Universal Binaries. If you have a “Universal Binary,” you won’t need Rosetta to translate, so there will not be any performance hit. Rosetta is simply there to translate for “Macintel” users who wish to run older Mac PowerPC software.
Keep in mind that Rosetta is the last resort for Macintel users who do not have Universal Binary versions of their PowerPC software. Macintel users will want Universal Binaries. PowerPC Mac users already have the native software.
Related MacDailyNews articles:
Apple+Intel news is no reason not to buy a new PowerPC-based Mac today – June 07, 2005
RUMOR: Apple planning Mac OS X ‘Tiger’ release for x86 PCs? – February 25, 2005 (Transitive Technologies)
Startup claims ‘near-universal emulator’ allows any software to run all platforms with almost no performance hit – September 13, 2004 (Transitive Technologies)
iPod success opens door to Mac OS X on Intel – March 04, 2004
CNET on Transitive/Rosetta says,
“”History says that binary translation basically doesn’t work,” said Illuminata analyst Gordon Haff. “The day may come when someone can do a good enough job with it, but that concept has been thrown out there many times in the computer industry, and it’s always fallen flat on its face.”
But Los Gatos, Calif.-based Transitive is willing to set high expectations when comparing software compiled natively for the new processor to that compiled for the older processor and running on the new one.
In the case of Transitive’s first customer, Silicon Graphics Inc., software for the older processor generally reaches at least 80 percent of the speed of native software, Wiederhold said. But that high score stems partly from the fact that the SGI systems are used for graphics tasks, which have little or no translation penalty, he said.
With more computationally intense tasks, the performance of translated software is between 60 percent and 80 percent of native software, Wiederhold said.
Another skeptic is Nathan Brookwood of Insight 64. “Everybody always has said 50 (percent) or 60 percent and delivered 30 (percent) or 40 percent,” he said. Among those who have tried: Digital Equipment Corp.’s FX!32 to run x86 Windows programs on computers with Alpha chips; Hewlett-Packard’s Aries software to run HP-UX software for PA- RISC chips on Itanium; and Intel’s IA32-EL software to run software for x86 chips on Itanium.
Jobs was satisfied, though. During his demonstration, Jobs said translated software runs “pretty fast,” though his presentation’s slide said performance is “good (enough).” His demonstration computer had a 3.6GHz Pentium 4 and 2GB of memory.”
So we all waited with baited breath for WWDC this year to hear of new & improved hardware & software. The reality from the newly published XBench scores at ThinkSecret are that the new Macintels run about 1/2 as fast as a Mac Mini (!!) and get absolutely CREAMED by the G5. (Score: 305 to 63)
Skip the excuses and Microsoft-like promises of “wait till next year”. I’m outta’ Kool-Aid.
So….during SJ’s keynote, the program “Mathematica” was the only software presented w/o the use of Rosetta?
Daddy-Oh! –
I suspect the new podcasting-capable iTunes that was previewed also was compiled from a Universal Binary and did not require Rosetta.
John Gruber also points out on DaringFireball that Classic apps will not be supported by Rosetta. So if you have some old classic apps that you will be wanting to run for a few more years you should be considering buying a PPC Mac now.
http://daringfireball.net/2005/06/classic_not_supported
Anybody that puts one solitary ounce of credibility into the Xbench scores of the P4 development machines currently floating around on the web is both ignorant as hell and a complete tool. These are NOT production machines, they’re simply development boxes that are not anything like what will actually ship starting in 2006…
Now my question: Which apps use AltiVec? I know Apple’s apps do. I use FCP. If I want to keep using that software on a Mac notebook, I shouldn’t wait with the purchase of a new notebook until mid 2006, right? I should just buy me a PowerPC notebook. Or maybe I should wait and buy an Intel notebook next year and buy the FC upgrade. Oh, the dilemma.
Developers have a year at least to get Universal Binaries into their apps. Any app worth its salt would surely be receiving an upgrade of some sort plus any tweaks for leopard anyway so as long as the Universal Binaries are as easy to produce as stated then with any luck it shouldn’t be much of an issue.
Hopefully, the only stuff that will need rosetta will be those small apps that don’t require a load of processing – so wouldn’t be so badly hit by any rosetta induced drop in speed. Add to this the fact that we also have a year + of speed increases to the machines themselves and I’m not overly worried (as a user).
Classic wise it’s gonna be a pretty small share of the market by then and you’d probably find that Apple would have removed support for it in Leopard anyway.
Bye-bye….this rosetta and Universal Binaries discussion is going to go on for the next two-3 years at least.
The Apple apologists will cry “No Problem”…the other will cry “Yes, there is”.
In the end….
Microsoft will still have 98% of the market and Apple 2%…and then word will come that Intel chips are having problems with mac OS…then Steve will announce the jump to AMD….then….being an Apple owner is getting tiring…
What would be really cool is if Rosetta translated the program once and then saved the program in translated or universal binary format. That way, you would only get the performance hit the first time it ran a program, and from then on out, it would run the already-translated code. The translated code might be a little slower than on a PPC, but it wouldn’t be nearly as slow as translating on-the-fly every time a program is run, would it?
Then you could also set up Rosetta to translate programs overnight, in the background, or when the processor is idle.
Out of Koolaid,
So any benchmarks on the Xbox 360 using the PowerMac G5s should be considered as the shipping performance on the production boxes?…
Remember to mix water in with your Koolaid.
You may have noticed that Steve didn’t do much more than load a couple PPC programs in memory. I think he was concerned more complicated instructions would crash “rosetta”. I’m sure a great deal of improvement is yet to come.
MacIntel is going to kill Windows in a manner so ingenious I am more impressed the more I think about it.
Also Steve Jobs demonstrated several main stream applications including Photoshop CS that worked just fine with Rosetta. So CNET better do a little more research on the subject before declaring it doesn’t work. Several million people watched it work quite well in fact.
Those Xbench scores are really promising actually. XBench is not a univeral binary, so it was running through the Rosetta emulation layer. Look at the individual scores – in graphic rendering, it was right on par with PPC, even running through emulation. In floating point and integer operations it had no chance – native execution vs emulated execution is a total joke. Even so, the fact that the scores are as high as they are is a testament to the effectiveness of Rosetta. I can’t wait to see XBench scores using the universal binary version of XBench!
Right now all Apples pro apps use AlitVec. If I was debating the purchase of a comp now or later for pro use… I would buy soon. Wait for the PPC upgrades. The Pro apps (IMHO) will need a good year or two on Mac Intel to come up to speed and the new PPC comming are going to Rock. Don’t wait. Buy now and then again in two to three years. Best of both worlds. At least you will be VERY on top of the curve for a long time.
My 2 cents.
MDM MW fear. I’m not kidding.
I think I will be buying my last PowerPC desktops at the end of the year. And when those Intel Macs arrive, I will probably buy one of those new fast notebooks, an iBook probably. I think that will be my strategy.
XBench scores at ThinkSecret are that the new Macintels run about 1/2 as fast as a Mac Mini (!!)
UHM! Those are Rosetta speeds.. of course the whole point of releasing this information 1 year early is to give devs time to make their apps native.. so.. those scores are NOT native scores..
I mean.. everything I’m saying is obvious.. wait until the apps are native to whine about speed.. it says right there 25% increase in memory usage..
sooooooooo are you surprised it’s slower than native apps?
OBVIOUSLY photoshop isn’t x86 native for OS X.. that’s why it was using Rosetta!
I don’t care what happens as long as we don’t keep calling the new Mac’s
MACINTEL
What a f*@ked up name that is
Back from indigestion after eating my own words (I dared say Intel in Mac was ridiculous) …
I have a question :
Rosetta performance is 80% of what ?
Are we comparing Ghz ? 3.6 Ghz P4 x 80% –> 2.8 Ghz G4 (or G5)
Are we comparing throughput ? 3.6 Ghz P4 = 2.4 Ghz G5 –> 1.9 Ghz G5
Personally, I think Mactel is a lot better than Macintel.
Macintel BAD! Mactel GOOD!
//PowerMac
MW: Quality…as in “Apple will, and still makes quality machines.”
.. I prefer InMac. Together with Power, gives Power InMac.