What is Steve Jobs Thinking? Log Out | Lost Password? | Topics | Search
Contact | Register | My Profile | SO home | MOL home

M-SO Message Board » Technology & The Internet » Archive through February 17, 2006 » What is Steve Jobs Thinking? « Previous Next »

  Thread Originator Last Poster Posts Pages Last Post
  ClosedClosed: New threads not accepted on this page          

Author Message
Top of pagePrevious messageNext messageBottom of page Link to this message

Duder
Citizen
Username: El_duderino

Post Number: 930
Registered: 2-2004


Posted on Thursday, January 19, 2006 - 3:25 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Jobs recently introduced the new dual core Intel chips on the new iMac lines. According to specs it is a waste of money if you are a professional and use Adobe applications.

Here is a direct quote from Steve Jobs from the Macworld keynote speech: "Rosetta is going to be a great bridge until all the apps are universal."

Then soon after: "The Performance of Photoshop in Rosetta isn't going to be powerful enough for a professional ... it's fast enough for those of us who use it occasionally."

What gives? Is he intentionally trying to take the Mac out of the professional market?


Top of pagePrevious messageNext messageBottom of page Link to this message

monster
Supporter
Username: Monster

Post Number: 1890
Registered: 7-2002


Posted on Thursday, January 19, 2006 - 3:31 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Well, a rumor going around is that Apple may be buying Adobe, then Apple can have the apps built on their schedule, not Adobe's.
Top of pagePrevious messageNext messageBottom of page Link to this message

Duder
Citizen
Username: El_duderino

Post Number: 931
Registered: 2-2004


Posted on Thursday, January 19, 2006 - 3:33 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Well, that would explain some things. Seems really odd for the CEO to admit his product, at his biggest speech of the year no less, couldn't be used professionally.
Top of pagePrevious messageNext messageBottom of page Link to this message

LazyDog
Citizen
Username: Lazydog

Post Number: 116
Registered: 6-2005


Posted on Thursday, January 19, 2006 - 5:44 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

is it possible Apple will introduce a pro-line of machines ? Maybe a high-end machine that could be used as either a smokin' workstation or as a server and the Mac-tel's for common folk ?
Top of pagePrevious messageNext messageBottom of page Link to this message

AlleyGater
Citizen
Username: Alleygater

Post Number: 1089
Registered: 10-2004
Posted on Thursday, January 19, 2006 - 8:26 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

I think you guys are missing the point. MOST software will run correctly on the new Macs. But some apps need to be RECODED to run at their fastest speeds. Adobe Apps need to be recoded. So Jobs is being truthful and letting people know that the new machine ISN'T for everyone quite yet. But the code revisions I HEAR are quite easy to impliment, so I suspect that Adobes apps will be working at their optimal speeds on the new hardware QUITE SOON!!! The new hardware IS TOTALLY PROFESSIONAL for some apps, and will be for others soon enough.

Make more sense?
Top of pagePrevious messageNext messageBottom of page Link to this message

Dave
Supporter
Username: Dave

Post Number: 8443
Registered: 4-1997


Posted on Thursday, January 19, 2006 - 9:36 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Helps Apple move more copies of Aperture, which is what pros should be using anyway.
Top of pagePrevious messageNext messageBottom of page Link to this message

monster
Supporter
Username: Monster

Post Number: 1893
Registered: 7-2002


Posted on Friday, January 20, 2006 - 11:34 am:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Point not missed, as far as the ease of recoding, I believe Mathematica was recoded to work on the new MacIntels in just a 24 hour period, with just a couple of tweaks here and there afterwards.
Pretty cool.....
Top of pagePrevious messageNext messageBottom of page Link to this message

TarPit Coder
Citizen
Username: Tarpitcoder

Post Number: 17
Registered: 12-2004
Posted on Tuesday, January 24, 2006 - 9:13 am:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

It's gotta just be the math kernels - I can't believe that they have two separate baselines for the PPC version and the x86 version.

Now it could be that they had nicely tuned math kernels for the PPC and the x86 - but the machines were constrained in such a way that adding threads slowed things down - so they were just a single threaded loop. That may have changed with the core duo - and they can easily saturate a single CPU but for best effect they need to use both.

I'd be pretty shocked if that was the case though.

I would believe that the tuning of the math kernel on the Core Duo may not have been done, it was done for the P4 and the Athlon and the performance sucks, but that's where Intel and Vtune could step in and speed that stuff up lots...

--Tarp

Top of pagePrevious messageNext messageBottom of page Link to this message

TarPit Coder
Citizen
Username: Tarpitcoder

Post Number: 18
Registered: 12-2004
Posted on Tuesday, January 24, 2006 - 9:20 am:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Yeah but the mathematica folks have their act together. I mean think about the range of processors:

Alpha
MIPS
x86
PPC
Power
HPPA
Itanium
Sparc

Big Endian, Little Endian, 32 bit, 64 bit. They tackled this stuff head on.

http://www.wolfram.com/products/mathematica/platforms/

Top of pagePrevious messageNext messageBottom of page Link to this message

LazyDog
Citizen
Username: Lazydog

Post Number: 123
Registered: 6-2005


Posted on Tuesday, January 24, 2006 - 5:20 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

What would a software vendor really need to change ? Definately low level code interfacing with the BIOS and the OS api's, anything else major ? These should, arguably, be remapping/reworking of functions.......now I come to think about it, is there a major data typing issue here.........
Top of pagePrevious messageNext messageBottom of page Link to this message

Tom Reingold
Supporter
Username: Noglider

Post Number: 12039
Registered: 1-2003


Posted on Tuesday, January 24, 2006 - 5:28 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

I think what tarpit coder is saying is that with math-intensive software such as mathematica, knowing the processor intimately can let the programmer program better, to get better performance. So it may not be a matter of just recompiling for the new processor.
Top of pagePrevious messageNext messageBottom of page Link to this message

LazyDog
Citizen
Username: Lazydog

Post Number: 124
Registered: 6-2005


Posted on Tuesday, January 24, 2006 - 5:36 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Tom, absolutley, recompiling without understanding the interface is non sensical. I guess what I'm driving at is not neccesarily the intimate knowledge of the interface (which can be obtained via specs) but the fact that Mathmatica must have a pretty cool design to be able to isolate OS-dependent code without, necessarily, redesigning their product. Koodo's to them
Top of pagePrevious messageNext messageBottom of page Link to this message

monster
Supporter
Username: Monster

Post Number: 1924
Registered: 7-2002


Posted on Tuesday, January 24, 2006 - 7:22 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)


Quote:

Architectural Differences

The PowerPC and the x86 architectures have some fundamental differences that can prevent code written for one architecture from running properly on the other architecture. The extent to which you need to change your PowerPC code so that it runs natively on an Intel-based Macintosh computer depends on how much of your code is processor specific. This chapter describes the major differences between architectures, organized alphabetically by topic. You can use the information to identify the parts of your code that are likely to be problematic.




Read all about here, http://developer.apple.com/documentation/MacOSX/Conceptual/universal_binary/inde x.html#//apple_ref/doc/uid/TP40002217
Top of pagePrevious messageNext messageBottom of page Link to this message

Politicalmon
Citizen
Username: Politicalmon

Post Number: 84
Registered: 9-2005


Posted on Monday, January 30, 2006 - 8:52 am:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Monster;

You hit it right on the head - processor specific coding is the key. Remember the PowerPC CPU was a more advanced RISC (Reduced Instruction Set Code) design where as the Intel is a combination of CISC (Complex Instruction Set Code) and RISC features reveresed enginnered from the DEC Alpha chip. I was perplexed as to why Jobs signed on exclusivley with Intel and didn't give the superior product developed my AMD the light of day?
Top of pagePrevious messageNext messageBottom of page Link to this message

monster
Supporter
Username: Monster

Post Number: 1971
Registered: 7-2002


Posted on Monday, January 30, 2006 - 11:24 am:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

here's another quote


Quote:


AMD was approached, but supply problems similar to those with IBM appeared likely. Also, Intel has been courting Apple for many years. Perhaps some of that effort has paid off. Further, this announcement was designed to be one of political expedience, convenience, exclusivity, and simplicity. The focus was to not scare the analysts and financial markets; those circles are happy with the "Intel" announcement. Further, you must look at Intel's roadmap one to two years into the future to see the types of processors Apple will be using; they're not Pentium 4, but will represent the advanced next generation of Intel's microprocessor designs. Finally, since Apple has shown it is willing to switch processor architectures entirely, nothing prevents Apple from availing itself of the best x86 and x86-64 processor technologies available from other x86 vendors once the x86 transition is complete or well underway - including those from AMD, as do many other PC vendors.




and more

Quote:

#


Will Mac OS X be more susceptible to Windows viruses and other malware, or other Windows-specific problems now that it will be running on Intel-based hardware?

No. It is the nature of the operating system, not the underlying hardware, that gives Mac OS X its security.

But if Windows (or another x86-based OS) is also run on an Intel-based Mac, it could be subject to the vulnerabilities of that operating system. However, when Windows or another OS is running in an emulation or a virtual machine environment, it is typically:

# completely isolated from the network, operating "behind" the host operating system
# not used for traditional tasks typically used as attack vectors, like email, web browsing, and downloading files
# usually not used as the primary operating system
# able to be completely restored relatively easily and quickly from a default image, if compromised

So, this becomes less of an issue. It certainly could be vulnerable to issues just as a full Windows machine would be, but there is less exposure for a variety of reasons. Another key factor further mitigating problems is that, generally speaking, the things that affect the Windows or other alternate environment still won't affect the Mac OS X environment.

If Windows is used as a primary OS, or the computer is directly booted into Windows in a hypothetical dual-boot configuration, the machine could certainly be subject to any and all Windows vulnerabilities, and Windows security best practices should be maintained. In this configuration, any other OS installations (such as a Mac OS X installation) on the same machine would still generallty be unaffected by any potential issues.




for more of this Apple/Intel FAQ, go to this site,
http://appleintelfaq.com
Top of pagePrevious messageNext messageBottom of page Link to this message

Dogbert
Citizen
Username: Dogbert

Post Number: 6
Registered: 1-2006


Posted on Monday, January 30, 2006 - 1:17 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

It's just like when the Mac moved from 68K to PPC. Even on the highest-end new PPC hardware, 68K programs weren't especially fast, but things were fine once the vendor recompiled for PPC.

Personally, I'm expecting nagging problems with some applications because the shift from PPC to Intel means a shift from big- to little-endian, and large programs often end up incorporating assumptions about endianness. The programs, like Mathematica for example, or anything with a Windows version and a lot of common code, will already have had to deal with this long ago, but programs that have always been Mac only could have some debugging ahead for them.

Endianness (I think the term is actually a play on something from Gulliver's Travels) is the order of byte storage for a word in memory. A word, typically 32-bits or 4 bytes on either the PPC or x86, is the standard unit of data on which the processor operates. In a big-endian processor the bytes in a word are stored big end-first, in other words from the most significant byte to the least significant byte. In a little-endian processor the bytes are stored least significant byte-first.

PowerPC could do both little- and big-endian, but Mac software was always big-endian. x86 is little-endian only so new Mac software will have to be work little-endian and Rosetta will have to deal with endianness in the conversion.
Top of pagePrevious messageNext messageBottom of page Link to this message

Eponymous
Citizen
Username: Eponymous

Post Number: 45
Registered: 6-2004
Posted on Monday, January 30, 2006 - 8:07 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Dogbert wrote:

"It's just like when the Mac moved from 68K to PPC. Even on the highest-end new PPC hardware, 68K programs weren't especially fast, but things were fine once the vendor recompiled for PPC."

My memory is that the first-gen PPCs, especially the 8100, ran even 68k software faster than the previous 68k machines. The Finder sure flew. I remember showing off how quickly the zoom rectangles opened up.

I think, with a bit of help from Wikipedia, that the faster 68k machines were the Quadras running at 33MHz, while the PPCs started at 60MHz. Not apples to apples (so to speak), of course.

Dating myself...
Top of pagePrevious messageNext messageBottom of page Link to this message

TarPit Coder
Citizen
Username: Tarpitcoder

Post Number: 19
Registered: 12-2004
Posted on Monday, January 30, 2006 - 8:46 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Probably worth a word about super linear speedup here too - which is something that many people find 'impossible' but fairly easily explained.

It happens when you take an existing algorithm and split it across N processors, and the resultant speedup is greater than what would be expected for the number of processors.

e.g. You have a working set that takes 8 hrs on 1 CPU, you throw in another CPU and suddenly it runs in 1 hr.


The basic concept is that sometimes when you parallelize an algorithm the new algorithm resides within cache for the two processors. So where as your single core processor with 512 KB of cache running a working set of 800K had quite a few cache misses, your dual core (say 512KB cache each ) processor running the same working set, but recoded to evenly split say the data fits inside both the processors D-Caches.

It's pretty cool when it happens. Chuck in a second CPU and suddenly something that took say 6 hrs runs in 1 hr.

Pretty rare but very cool

--Tarp
Top of pagePrevious messageNext messageBottom of page Link to this message

Dogbert
Citizen
Username: Dogbert

Post Number: 7
Registered: 1-2006


Posted on Monday, January 30, 2006 - 11:35 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Eponymous: you might have seen that with apps that spend a lot of time in the UI and generally using OS calls, which of course were all native code. A data/compute-intensive application like Photoshop would flounder in such an environment, even with a radically disproportionate clock speed increase.

Of course, the 68K was as complex an instruction set as you'll get and the PPC reasonably RISC (it's a relatively complex architecture for a RISC chip), so in the normal course of things you'd expect a high ratio of translated PPC instructions to 68K instructions.
Top of pagePrevious messageNext messageBottom of page Link to this message

Dogbert
Citizen
Username: Dogbert

Post Number: 8
Registered: 1-2006


Posted on Monday, January 30, 2006 - 11:37 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Tarpit Coder: Oh, if only most algorithms were so parallizable! SMP on the desktop would have been popular years ago. As you say, it's a rare thing.
Top of pagePrevious messageNext messageBottom of page Link to this message

TarPit Coder
Citizen
Username: Tarpitcoder

Post Number: 20
Registered: 12-2004
Posted on Tuesday, January 31, 2006 - 7:55 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Dogbert,

Yeap - Amdahls law. The truth I tend to believe is actually somewhere inbetween - lot's of good parallel stuff is out there - it's just that we were all corrupted by Von Neumann machine thinking.

It's like F.P. (Not floating point - The Functional Programming language that John Backus gave way back in the mid 70's).

I wish that Steve Jobs would resurect the transputer, switch to Occam and start selling boxes with say a hundred little processors. Nothing like 100 chickens plowing the field versus two strong oxen...


Top of pagePrevious messageNext messageBottom of page Link to this message

TarPit Coder
Citizen
Username: Tarpitcoder

Post Number: 21
Registered: 12-2004
Posted on Saturday, February 4, 2006 - 3:19 pm:   Edit PostDelete PostPrint Post   Move Post (Moderator/Admin Only)

Anandtech has a review of the new Intel Core Duo iMac. Have a gander http://www.anandtech.com/mac/showdoc.aspx?i=2685

Topics | Last Day | Last Week | Tree View | Search | User List | Help/Instructions | Credits Administration