[24klogos]

Ok, I have been using Macs as my main machines since 2000. I am a multimedia designer and originally started back in 92 with an Amiga 1200 and in that 8 year in-between I used and abused Windows machines left and right. Even to this day I have to use Windows to test and proof a lot of my work and honestly i don't know whats wrong with me but i don't like the whole switching to Intel at all... I know its just hardware, but still something doesn't feel right to me.

In most forums where i've posted comments and such, using humor to soften up the topic, people seem to be very eager and thrilled about this switch, yet granted we're on a transitional mode where neither platform is the "way to go", I have been under the impression not many people will "miss" the current PPC line of products. It kinda pisses me off, I know i'll have a couple years, maybe less on my G5, but this machine is rock solid and i could not be happier with it.

Like any product that is no longer developed, the PPC machines are going to be forgotten in a couple of years once the transition has been at least 90+% complete. BUT, something still doesn't add up, it might be a more ambigous approach to the issue, but the personality of the Mac will be gone... Not from a PPC standpoint perse but the whole idea of being different, of having not only a different OS but a different MACHINE, that not only looks gorgeous but that sets you appart from the Win/PC users, even on the architecture. I don't know, if you ask me, putting aside the whole performance per watt jibba-jabba, the hardware, the other 50% that made Macs special and worshiped by few, wil be gone.

I don't expect people to agree, much less folks that don't really care what machine they use as long as it "does the job". But I believe Apple will lose an aspect important to few, non existant to many and vital for me (despite being a power user myself) and that is personality.

I really hope that the only thing we have left, besides gorgeous machines, OSX, will remain being developed and used by those who once understood the feeling of thinking different and being different in a world of clones.

Now back to my Zoloft.

[Big Mac]

Although my objections to Intel go far deeper than the point you're making, you should know you're certainly not a lone dissenter.

[mduell]

The personality of an computer will be lost? Please, stop before I vomit.

The cases will be the same. The software will be the same. A few chips on the logic board will change. Big deal.

"Oh noes, they'll be marginally less unique!"

[24klogos]

Originally Posted by mduell

The personality of an computer will be lost? Please, stop before I vomit.

The cases will be the same. The software will be the same. A few chips on the logic board will change. Big deal.

"Oh noes, they'll be marginally less unique!"

please vomit, that would rock.

[reader50]

I agree with the first two posters. Fringe bonus for the new Intel owners, a free TPM chip at no extra charge.

[Catfish_Man]

I'm gonna have to agree with mduell here. Instruction sets have nothing to do with individuality or personality. Neither do corporations. Design, art, engineering quality/attentiveness, and community... those provide the personality of a machine. All of those will still be here.

[new newton]

All the real personality of the Mac was lost during the switch to PPC. Gosh, it was horrible.

Some of you guys really don't have a clue as to what is important.

[Jason]

Personally I don't like the change from a consumers point of view, however I can certinaly see that if I was in Steve Jobs position, I would probably have a hard time telling all the other members of the company that we shouldn't make a switch to intel and make more money! Although I do agree that they have lost that 'special' feature about them, but in a few years when my G5 is no longer doing the jobs I want it to do, then I'll make the switch to intel... *shivers*

[Maflynn]

I agree the uniqueness that is mac will be gone. I think the switch to intel is a neccessary evil thanks to IBM's inability to produce a cool/fast running G5 and a low powered flavor or laptops

As for Mduell, perhaps he's a recent switcher and does not understand the whole Mac cult thing. I've been a user of Macs' since the mac plus days.

While I do hate to see this era ending, I am excited at what the future may bring, What I hope is apple to do something they have not been able to do for a few years - bring out fast machines that can compete with peecees.

Mike

[OreoCookie]

I do have some concerns about the switch, but they are not related to the points you are mentioning here. It's rather about competition. The fact that the consumer benefits most from competition implies that we need several architectures competing with each other. Now that Apple is using Intel chips, there is basically no competitor to the x86 architecture in the desktop space. Sure, you can get Power workstations or a Sun box with an UltraSPARC cpu, but they are hardly consumer machines.

The amazing and sad thing is that there are some very competitive PowerPC product available, be it the Power5(+), the Cell cpu, the PPC970MP or even (for special applications) the multi-core variant of the PPC440 (which is at the heart of the fastest super computers you can get for your money these days). I'm also kind of curious why IBM which is pushing Linux onto their desktops is not offering PowerPC linux computers `for the rest of us' since Linux is pretty agnostic about the underlying architecture.

While I see the advantages of the switch to Intel, I do have concerns that this might slow down competition. Intel does not seem to keen on replacing the x86 architecture with something new. I'm not sure about its latest chip, but even the Pentium 4 and the Intanic (I don't know about the latest re-incarnations, though) still include the infamous A20 gate which was conceived over 20 years ago to overcome the 1 MB limit (It still causes occasional problems with ACPI, for instance.) It's a veeery conservative company, the MS among the cpu manufacturers.

[OreoCookie]

Originally Posted by Maflynn

I agree the uniqueness that is mac will be gone. I think the switch to intel is a neccessary evil thanks to IBM's inability to produce a cool/fast running G5 and a low powered flavor or laptops

I don't think IBM is incapable, unwilling, maybe, but they are quite able.

[Todd Madson]

I believe it was a corporate inability to create what Apple required.

Apple wanted the fastest computer in the world with the least power
consumption and wanted it to be the most beautiful thing you've ever
seen.

IBM produced a behemoth of a chip with quite a lot of capability but
at the expense of power and heat. Apple had to scramble to make a
functional and attractive case for this thing.

The fact remains - it generates a lot of heat, it consumes a lot of power.
It's more of a workstation/minicomputer than a personal computer for
people.

Apple switched gears - they wanted a machine that was competitive in
terms of its functional ability and speed but generating large amounts
of heat and consuming immense amounts of power wasn't a good thing
in the eco-conscious 2000s.

I'll miss the G5 series, it was an interesting experiment. I love mine.
But I don't love the power bills and I dread the summer coming - it will
be 100 degrees in that room when it is under load. No wonder I'm
considering moving to the cool basement.

If you can have a machine with the performance of a G5 (or better)
with less than half the power consumption and with one third the heat
wouldn't you leap on it? I sure would.

Also, laptop battery life will be improved by this. I work in a fortune
100 company where we have a lot of new employees bringing in
computers to be set-up and 95% of them are laptops. I can count
the number of desktop setups I've done since last year on two hands.

People want convenient packages they can move around. If you had
the power of a desktop in the form of a laptop wouldn't you consider it?
I sure would.

Don't think of it as the personality of the computer changing since the
operating system is the same.

Yes, it will be sad to wave goodbye to the heady days of the G series
processors but there will be a new era to look forward to. And who
knows? Someday they might switch again to AMD or something else.

Evolution in action.

[Horsepoo!!!]

Well let's see...because *you* have the same type of brain that I do, your personality must be an exact replica of mine. Does that make sense? No? Exactly.

[Maflynn]

Originally Posted by OreoCookie

I don't think IBM is incapable, unwilling, maybe, but they are quite able.

I think they were incapable. Remember Steve's little speach that we'd see a 3GHz machine before too long [after the G5 debut] Well what did we get instead, a kludge of a fix to get the cpu up beyong 2.5 GHz but still no 3GHz. What about the heat issues that plagued and delayed the G5 imac and finally producing a cool running low powered G5.

I think they wanted to, because of the new relationship with Apple, but were unable to over come the obstacles. In time I'm sure they could have produced a 3GHz cpu that wasn't water cooled or a cpu that could be put in a laptop w/o giving the user third degree burns but in that time span Intel would have released even faster running CPUs. IBM (technically or business reasons) was unable to keep up with what they intially agreed to do with Apple.

[OreoCookie]

Originally Posted by Maflynn

I think they were incapable. Remember Steve's little speach that we'd see a 3GHz machine before too long [after the G5 debut] Well what did we get instead, a kludge of a fix to get the cpu up beyong 2.5 GHz but still no 3GHz. What about the heat issues that plagued and delayed the G5 imac and finally producing a cool running low powered G5.

Intel and AMD hit similar problems with their manufacturing techniques as far as clock speed is concerned.

Originally Posted by Maflynn

I think they wanted to, because of the new relationship with Apple, but were unable to over come the obstacles. In time I'm sure they could have produced a 3GHz cpu that wasn't water cooled or a cpu that could be put in a laptop w/o giving the user third degree burns but in that time span Intel would have released even faster running CPUs. IBM (technically or business reasons) was unable to keep up with what they intially agreed to do with Apple.

One manufacturer offers fanless PPC970FX blades. And until January of this year, Intel hasn't released revolutionary cpus, they got stuck with the Pentium 4 NetBurst architecture and pulled the emergency break. They got whipped performance and power-wise by their competitors (mainly AMD).

If you look at IBM's portfolio, they have a phletora of cpus, some of them custom designs for certain customers (e. g. Microsoft). Apple's relatively low volume made them less interesting. But if you look at the technology IBM has at their disposal, they can and have designed such cpus when they want to. They didn't for Apple for business reasons, but not for technical reasons. (The PPC440 which is used in the fastest super computers is a low-power chip.)

[tooki]

How many of you who claim the Mac's soul has been lost have actually used an Intel Mac?

You can't tell the difference. At all. Without opening the case, you'd never know. Except that they run fast. But everything that makes it a Mac is there. Try one and you'll see what I mean. It'll be very anticlimactic, actually: you'll just think you were sitting at a fast PPC Mac.

I do understand the pride in running something unique, but that has nothing to do with the Mac's soul, which is all about fit and finish -- and software.

tooki

[OreoCookie]

Originally Posted by tooki

How many of you who claim the Mac's soul has been lost have actually used an Intel Mac?

I wish I knew … I'm still waiting for my MacBook Pro

[24klogos]

Originally Posted by tooki

How many of you who claim the Mac's soul has been lost have actually used an Intel Mac?

You can't tell the difference. At all. Without opening the case, you'd never know. Except that they run fast. But everything that makes it a Mac is there. Try one and you'll see what I mean. It'll be very anticlimactic, actually: you'll just think you were sitting at a fast PPC Mac.

I do understand the pride in running something unique, but that has nothing to do with the Mac's soul, which is all about fit and finish -- and software.

tooki

I really hope so man, i dont eat that much cheesse to retire my G5 into a gouda grader yet.

[Maflynn]

Originally Posted by tooki

How many of you who claim the Mac's soul has been lost have actually used an Intel Mac?

...
I do understand the pride in running something unique, but that has nothing to do with the Mac's soul, which is all about fit and finish -- and software.

tooki

As you had put it I wouldn't say its soul but rather what made Mac unique is now lost. That is a little disheartening. I reveled in the fact I didn't use a peecee at home. Yes the mactel's run the same os, the same apps and are not a peecee but some of that uniqueness is gone.

I haven't lost all of my marbles, I understand its just a computer and a tool. One that does a better job then a windows machine, so to that end, I'm eagerly waiting to see what desktop apple releases but the doesn't change the fact I'm disapointed in the shift.

[goMac]

The only downside the Intel chips really have is the bass ackwards instruction set.

Which begs the question of how many of you guys actually do assembly programming and thus are affected by the switch?

[P]

I think that IBM could have made a great CPU, but they wanted Apple to pay for its development - like the Cell. The overlap of a desktop CPU with a server CPU (Power5) or embedded - (G3, G4) wasn't really all that great. The PPC advantage in that the ISA wasn't completely braindead has been going away with the recent move to x86-64. Intel could provide a chip that was as good and much cheaper, and Apple couldn't pass that up.

Yes, Macs are being commoditized to a large extent - but then that process began with the first Mac II. With the cost pressure of modern computing, a little better and twice the cost doesn't cut it anymore.

Apple still have one big diffrentiator - the OS. That's still what counts.

[OreoCookie]

Originally Posted by P

I think that IBM could have made a great CPU, but they wanted Apple to pay for its development - like the Cell. The overlap of a desktop CPU with a server CPU (Power5) or embedded - (G3, G4) wasn't really all that great. The PPC advantage in that the ISA wasn't completely braindead has been going away with the recent move to x86-64. Intel could provide a chip that was as good and much cheaper, and Apple couldn't pass that up.

Yes, Macs are being commoditized to a large extent - but then that process began with the first Mac II. With the cost pressure of modern computing, a little better and twice the cost doesn't cut it anymore.

Apple still have one big diffrentiator - the OS. That's still what counts.

I think you pretty much nail it here. (Except that IBM has nothing to do with the G4 … but there is a PowerPC sitting in both of the printers I work with, in our department's LaserJet 4250 and my Kyocera laser printer). Toshiba and Sony have invested quite a chunk of change in the Cell cpu and got a fully custom cpu which can be very powerful for some tasks. Also, the volume of Cell will be much larger than anything Apple could ever sell.

[Lateralus]

Originally Posted by Maflynn

As for Mduell, perhaps he's a recent switcher and does not understand the whole Mac cult thing.

He's a PC user, actually.

[mduell]

Originally Posted by OreoCookie

I don't think IBM is incapable, unwilling, maybe, but they are quite able.

Unable to produce 3Ghz G5s within a year... or just unwilling?

Originally Posted by OreoCookie

The amazing and sad thing is that there are some very competitive PowerPC product available, be it the Power5(+), the Cell cpu, the PPC970MP or even (for special applications) the multi-core variant of the PPC440 (which is at the heart of the fastest super computers you can get for your money these days).

Power5: Who wants the new $4000 Mac mini*? (*only 3 times the volume of the last one!)
Cell: Blazingly fast for some apps... generally considered to be poor for a general purpose desktop.
PPC440: How much of your software is optimized for 24-way SMP? How much can be optimized? Single threaded apps will run horribly.

Originally Posted by OreoCookie

One manufacturer offers fanless PPC970FX blades.

At, ta-da ta-da, 1.6Ghz! That's 20W for the power-optimized parts or 27W for the standard PPC970FX part. To get a 25% gain in clockspeed to 2Ghz, power consumption increases over 125%!
I can think of three things where a single 1.6Ghz G5 would be faster than a dual 2.16Ghz Core Duo.

Originally Posted by OreoCookie

And until January of this year, Intel hasn't released revolutionary cpus, they got If you look at IBM's portfolio, they have a phletora of cpus, some of them custom designs for certain customers (e. g. Microsoft). Apple's relatively low volume made them less interesting. But if you look at the technology IBM has at their disposal, they can and have designed such cpus when they want to. They didn't for Apple for business reasons, but not for technical reasons. (The PPC440 which is used in the fastest super computers is a low-power chip.)

IBM isn't going to design a custom chip for Apple, and none of their existing designs suit Apple's products very well.
Intel, OTOH, can spread the costs over a few hundred million chips.

Yes, as Lateralus says, I'm not a switcher yet.

[ReggieX]

Originally Posted by new newton

All the real personality of the Mac was lost during the switch to PPC. Gosh, it was horrible.

Now this is a laughable statement.

[svtcontour]

I'm not a mac user but I think I can understand where the poster is coming from. I'll use cars as an example and I know thats not a good thing to do but its the only way I can make this work


Lets say if Mazda decided to switch the engine out of the new RX8 from rotary to a 4 or 6 cylinder piston powered engine, there might be an uprising because the RX series (RX7 and older included) were famous because of their rotary engines. It made them unique. I understand that the engine could be replaced with an equal or even better engine but somehow, I think something would be lost that makes it a unique car. Its not just about the outward appearance.

Can this type of example apply to the mac/PPC arguement?

[OreoCookie]

First of all, chill. Then take a second and stop quoting others out of context.

Originally Posted by mduell

Unable to produce 3Ghz G5s within a year... or just unwilling?

You quote me out of context here. I was referring to the ability to design a mobile cpu, and that ability has nothing to do with willingness.

Originally Posted by mduell

Power5: Who wants the new $4000 Mac mini*? (*only 3 times the volume of the last one!)
Cell: Blazingly fast for some apps... generally considered to be poor for a general purpose desktop.
PPC440: How much of your software is optimized for 24-way SMP? How much can be optimized? Single threaded apps will run horribly.

Again, you quote me out of context and distort the meaning of my post completely. The point was that Intel covers everything from embedded PPC chips (e. g. the PPC405, PPC440 (well, it started as an embedded chip ) and the PPC750) to Power-ful servers. I've never said Apple should/could use a Power5(+) nor Cell nor the PPC440 for any of their products. But IBM has a cpu portfolio that encompasses almost everything from very small to very big. Hence my conclusion that if Apple were willing to pay for the cpu development, IBM would gladly do so.

Originally Posted by mduell

At, ta-da ta-da, 1.6Ghz! That's 20W for the power-optimized parts or 27W for the standard PPC970FX part. To get a 25% gain in clockspeed to 2Ghz, power consumption increases over 125%!
I can think of three things where a single 1.6Ghz G5 would be faster than a dual 2.16Ghz Core Duo.

The low-power PPC970FX is (at least according to IBM's documentation) available since it was first offered to customers. Which wasn't this January (when the Core Duo was released), but a bit earlier

Originally Posted by mduell

IBM isn't going to design a custom chip for Apple, and none of their existing designs suit Apple's products very well.
Intel, OTOH, can spread the costs over a few hundred million chips.

That's exactly what I said.

Originally Posted by mduell

Yes, as Lateralus says, I'm not a switcher yet.

So?

[24klogos]

Originally Posted by svtcontour

I'm not a mac user but I think I can understand where the poster is coming from. I'll use cars as an example and I know thats not a good thing to do but its the only way I can make this work


Lets say if Mazda decided to switch the engine out of the new RX8 from rotary to a 4 or 6 cylinder piston powered engine, there might be an uprising because the RX series (RX7 and older included) were famous because of their rotary engines. It made them unique. I understand that the engine could be replaced with an equal or even better engine but somehow, I think something would be lost that makes it a unique car. Its not just about the outward appearance.

Can this type of example apply to the mac/PPC arguement?

I think theres always a car analogy that can match any situation!

I would agree with you. Its more of a cult thing, that sense out non-belonging to the mainstream or not following the leader, that fires some of us in different aspects. I think your analogy is correct, if they switched the rotary engine to say a cheaper, more efficient, piston-casket based engine, only hardcore RX8 fans would be pissed, for that matter, to my mother it will be the same car, hell if it even saves a buck or two in gas it would be "better"... but i dig what you say, and by the way, mazda holds all the patents for that engine and nobody else can manufacture it.

Now, if you let me push it a little further, and i have used this analogy before... i'd like to think about it as BMW getting their engines from HYUNDAI, because their most popular model-customer's (probably the 3 series) dont want to pay extra gas and maybe, think anything beyond 3000cc is an overkill. if you ask me, and this is my oppinion. I could care less if my laptop resembles an X-large-chigago style pizza box, as long as it runs some mean, super hot, extra pepperoni PowerPC processors.

[P]

Originally Posted by ReggieX

Now this is a laughable statement.

You don't think it was irony?

(BTW: I know IBM didn't do antyhing with the G4 except help produce them when they were new. I made a general point and then edited myself because it got too long, and it seems I cut too much.)

[mduell]

Originally Posted by OreoCookie

You quote me out of context here. I was referring to the ability to design a mobile cpu, and that ability has nothing to do with willingness.

I was commenting that you were creating a distinction ("unwilling" v. "unable") without a difference. To the end user and Apple, it's the same situation

Originally Posted by OreoCookie

Again, you quote me out of context and distort the meaning of my post completely. The point was that Intel covers everything from embedded PPC chips (e. g. the PPC405, PPC440 (well, it started as an embedded chip ) and the PPC750) to Power-ful servers. I've never said Apple should/could use a Power5(+) nor Cell nor the PPC440 for any of their products. But IBM has a cpu portfolio that encompasses almost everything from very small to very big. Hence my conclusion that if Apple were willing to pay for the cpu development, IBM would gladly do so.

Mea culpa, thanks for clearing that up.

Originally Posted by OreoCookie

The low-power PPC970FX is (at least according to IBM's documentation) available since it was first offered to customers. Which wasn't this January (when the Core Duo was released), but a bit earlier

It has been out for a while, but it's still the best thing available today; 970MP is too power hungry for a laptop at a reasonable clockrate.

[rhashem]

Originally Posted by Maflynn

Iperhaps he's a recent switcher and does not understand the whole Mac cult thing. I've been a user of Macs' since the mac plus days.

The "Mac cult" strikes me as a bunch of folks more interested in "being different" than using a great computer. At the heart and soul of it, the x86 switch is about allowing Apple to do what it does best --- make a great computing platform, while ditching the baggage of single-handedly championing PowerPC on the desktop. The people who actually like Macs are thrilled about this. The people who see Macs as just a status symbol are not. It's as simple as that.

[BigBadBiologist]

Originally Posted by mduell

The personality of an computer will be lost? Please, stop before I vomit.

The cases will be the same. The software will be the same. A few chips on the logic board will change. Big deal.

"Oh noes, they'll be marginally less unique!"

But they might even start calling them motherboards instead of logic boards!

[rhashem]

Originally Posted by OreoCookie

Again, you quote me out of context and distort the meaning of my post completely. The point was that Intel covers everything from embedded PPC chips (e. g. the PPC405, PPC440 (well, it started as an embedded chip ) and the PPC750) to Power-ful servers. I've never said Apple should/could use a Power5(+) nor Cell nor the PPC440 for any of their products. But IBM has a cpu portfolio that encompasses almost everything from very small to very big. Hence my conclusion that if Apple were willing to pay for the cpu development, IBM would gladly do so.

Yes, IBM has a wide CPU portfolio. Conspicuously absent in it, however, is a good desktop chip. Desktop chips are interestingly enough one of the more complicated type of chip to design. It's hard getting high single-threaded IPC on a tight power and cost budget. With something like POWER, you can get good performance by throwing cache and memory bandwidth at the problem. With something like Cell, you can depend on parallelism and make lots of very simple, but very fast units. That is not to say that designing Power5 or Cell was easy. Making a Power5-like memory bus or a 4.5 GHz vector unit is not easy. However, at least the path to good performance in these domains is both clear and approachable.

A desktop chip is a different kettle of fish. The software you're expected to run consists of poorly scheduled, poorly optimized code originally compiled before your processor even existed. The software has little to no inherent parallelism, and is full of branchy integer code. You've got to make this software run fast, without using lots of cache (expensive), a fast memory bus (expensive), without making the chip too big (expensive), or clocking it too high (too hot). If you hope to get good desktop performance out of a chip, you've got to tune it to these requirements from the beginning. Taking a fast server chip like Power6 and neutering it by putting it on a PC-class memory bus with 2MB of cache just won't cut it.

Now, Apple could pay IBM to make them a desktop chip. However, the economics wouldn't make sense. Power5 was designed by a team of 300+ engineers over 3 years. That makes for a development cost of a few hundred million dollars, and that's considering that Power5 is a moderate update of the Power4 core. Such huge investments make sense for Intel and AMD --- they sell tens to hundreds of millions of chips a year. Apple would be lucky to ship five million chips a year. I don't know about you, but I'd rather that Apple spend that money improving OS X, and leave the expensive chip design to Intel!

[aristotles]

I was an Amiga user from 1989-96 and PC user from 1987-1989, 1996-2002 (at home). I've been a mac user since Oct 2002. As a cross platform user (Wintel at work), I will say that I was initially hesitant about the switch and in total disbelief of rumours about it.

Looking back, I can say that what makes a mac a mac is the unique set of hardware feature offered and software with the software being the most important part.

Look at the larger pbooks and macbook pro. They have unique features like a backlit keyboard with ambient light sensors, magsafe and the best bundled software and OS out there.

OSX is what makes the mac what it is and it is what attracted me to the mac platform in the first place.

PS. I speak as a windows software developer. I can tell you that based on my experience, the OS X software out there is usually of higher quality than windows counterparts. This is because of the wonderful tools Apple provides to developers via the core APIs and Cocoa.

[24klogos]

Originally Posted by rhashem

The "Mac cult" strikes me as a bunch of folks more interested in "being different" than using a great computer.

At the heart and soul of it, the x86 switch is about allowing Apple to do what it does best --- make a great computing platform, while ditching the baggage of single-handedly championing PowerPC on the desktop. The people who actually like Macs are thrilled about this. The people who see Macs as just a status symbol are not. It's as simple as that.

It was a nessesary evil and Apple knows it, business is business and this alliances between big fish will happen no matter what, We're all making comments here because we "actually like Macs"... c'mon we know Jobs ain't goinna read this post and go back to IBM cause we're all whinning about the good ol' PPC days. Trust me,

If you think some people (cult users youre talking about) use Macs as a status symbol, think again... that is exactly what Macs are going to become, instead of the a great, solid computers they are already (if it aint broken dont fix it) Macs will be nothing more than a luxury PCs that by the way... can run OSX as a bunus... there's your status change....

Businesswise makes a lot of sense, its all about the money, machines will be fast, and cute u can be sure of that, are they real MACs? lets say they're half, the good half.

[rhashem]

Originally Posted by 24klogos

It was a nessesary evil and Apple knows it, business is business and this alliances between big fish will happen no matter what

What's evil about it? The fact that one big fish (IBM) was replaced by another (Intel)?

If you think some people (cult users youre talking about) use Macs are a status symbol, think again... that is exactly what Macs are going to become, instead of the a great, solid computers they are already (if it aint broken dont fix it)

If anything, the new Macs will be more solid, not less. It seems pretty much every one of Apple's chipsets has some sort of quirk (eg: the memory controller on the G5 chipset). At least Intel's chipsets have a history of being utterly reliable and bug-free.

As for not being broken: who says Macs aren't broken? Isn't everyone tired of buying last-generation hardware at current-generation prices? Nobody is tired of the G4 and its late 1990's era performance? In almost everything I've run, my dual-core 2.3 G5 is the equivalent of a 1.8 GHz Opteron. That's a chip in a $2500 machine compared with a chip you'd find in a $1500 machine. Now, the new Intel chips (the ones that will be in the new PowerMac) are beating up on 2.8 GHz Opterons. If Apple stayed with the G5, it'd be the G4 fiasco all over again. I for one will be perfectlly happy when Apple starts using real desktop chips in their machines --- not neutered server chips or gussied up embedded chips...

Macs will be nothing more than a luxury PCs that by the way... can run OSX as a bunus... there's your status change....

Macs are already luxury PCs. I've got a dual-core PowerMac and a dual-core AMD Linux box. The only real difference between them is the CPU. The rest of standard stuff --- OEM hard drive, RAM, and graphics card, industry standard SATA, PCI-E, and USB busses, etc. There was a time when a Mac was a different sort of machine, with SCSI and Nubus, etc, but today, its a PC with a different CPU. The only thing that has changed here is that Apple has switched from a CPU that converts PowerPC into a undocumented native instruction set to one that converts x86 into an undocumented, native instruction set. If its the ISA that only even exists in the first four pipeline stages of the CPU that makes a Mac a Mac, well, I don't know what to say to that...

[24klogos]

Originally Posted by rhashem

What's evil about it? The fact that one big fish (IBM) was replaced by another (Intel)?



If anything, the new Macs will be more solid, not less. It seems pretty much every one of Apple's chipsets has some sort of quirk (eg: the memory controller on the G5 chipset). At least Intel's chipsets have a history of being utterly reliable and bug-free.

As for not being broken: who says Macs aren't broken? Isn't everyone tired of buying last-generation hardware at current-generation prices? Nobody is tired of the G4 and its late 1990's era performance? In almost everything I've run, my dual-core 2.3 G5 is the equivalent of a 1.8 GHz Opteron. That's a chip in a $2500 machine compared with a chip you'd find in a $1500 machine. Now, the new Intel chips (the ones that will be in the new PowerMac) are beating up on 2.8 GHz Opterons. If Apple stayed with the G5, it'd be the G4 fiasco all over again. I for one will be perfectlly happy when Apple starts using real desktop chips in their machines --- not neutered server chips or gussied up embedded chips...



Macs are already luxury PCs. I've got a dual-core PowerMac and a dual-core AMD Linux box. The only real difference between them is the CPU. The rest of standard stuff --- OEM hard drive, RAM, and graphics card, industry standard SATA, PCI-E, and USB busses, etc. There was a time when a Mac was a different sort of machine, with SCSI and Nubus, etc, but today, its a PC with a different CPU. The only thing that has changed here is that Apple has switched from a CPU that converts PowerPC into a undocumented native instruction set to one that converts x86 into an undocumented, native instruction set. If its the ISA that only even exists in the first four pipeline stages of the CPU that makes a Mac a Mac, well, I don't know what to say to that...

i'm sold.

next time I reply to your posts, it will be on an Alienware.

[rhashem]

Originally Posted by 24klogos

next time I reply to your posts, it will be on an Alienware.

What does Alienware have to do with anything?

I should point out that while my AMD machine is faster in pretty much every way (and quiet enough so I can sleep with it next to my head), I'm typing this from my Mac. However, it'd be great if I didn't have to compromise just to use OS X. I'd rather have the best, then live with something less just for the sake of being different.

[OreoCookie]

Originally Posted by rhashem

Yes, IBM has a wide CPU portfolio. Conspicuously absent in it, however, is a good desktop chip. Desktop chips are interestingly enough one of the more complicated type of chip to design. It's hard getting high single-threaded IPC on a tight power and cost budget.

The PPC970MP is a good desktop chip, it has a similar thermal envelope than its predecessor and achieves competitive SPECmarks with IBM's compiler. I'm not saying it's the fastest chip out there or whatever, but it's competitive.

So, honestly, the only chip that I think is absent in IBM's portfolio is a real low-power `desktop' cpu for notebooks and small computers (in the sense that I'm not talking about an embedded cpu).

Originally Posted by rhashem

A desktop chip is a different kettle of fish. The software you're expected to run consists of poorly scheduled, poorly optimized code originally compiled before your processor even existed. The software has little to no inherent parallelism, and is full of branchy integer code. You've got to make this software run fast, without using lots of cache (expensive), a fast memory bus (expensive), without making the chip too big (expensive), or clocking it too high (too hot). If you hope to get good desktop performance out of a chip, you've got to tune it to these requirements from the beginning. Taking a fast server chip like Power6 and neutering it by putting it on a PC-class memory bus with 2MB of cache just won't cut it.

That's beginning to change since every major cpu vendor switched to multi-core cpus. But for most things, you are correct.

That's why the UltraSPARC T1 is an ingenious design for servers, beating servers which are far more expensive, because of the different kind of work load. I'll be interesting to see how the T2 performs as Sun wants to give each core a powerful FPU

[Catfish_Man]

Originally Posted by OreoCookie

That's why the UltraSPARC T1 is an ingenious design for servers, beating servers which are far more expensive, because of the different kind of work load. I'll be interesting to see how the T2 performs as Sun wants to give each core a powerful FPU

I think the end result of the desktop processor wars is going to look a lot like a more general purpose CELL. A big core or two, and then a half dozen smaller cores. The parallel software (which should be the bulk of the really cpu intensive apps... fp tends to parallelize better) can run across all 7-8, while the serial/branchy/crappy code can run on the 'main core(s)'. Some sort of dynamic power management system like Intel's Foxton tech would be necessary as well... you'd have the main core slow way down while the mini-cores were in use and just act as a scheduler, thereby keeping the power usage under control.

So:
branchy int code:
[70+ watts from the main core] [0-10 from the mini-cores, which are turned off or nearly so]

parallel floating point code:
[20 watts from the main core] [10 watts per mini-core for 60 watts total]

Other fun tricks you could do would include stuff like doing runahead execution on idle cores (see, for example, http://www.csl.cornell.edu/~ilya/papers/pact05.pdf )

[OreoCookie]

Originally Posted by Catfish_Man

I think the end result of the desktop processor wars is going to look a lot like a more general purpose CELL. A big core or two, and then a half dozen smaller cores. The parallel software (which should be the bulk of the really cpu intensive apps... fp tends to parallelize better) can run across all 7-8, while the serial/branchy/crappy code can run on the 'main core(s)'. Some sort of dynamic power management system like Intel's Foxton tech would be necessary as well... you'd have the main core slow way down while the mini-cores were in use and just act as a scheduler, thereby keeping the power usage under control.

So:
branchy int code:
[70+ watts from the main core] [0-10 from the mini-cores, which are turned off or nearly so]

parallel floating point code:
[20 watts from the main core] [10 watts per mini-core for 60 watts total]

Other fun tricks you could do would include stuff like doing runahead execution on idle cores (see, for example, http://www.csl.cornell.edu/~ilya/papers/pact05.pdf )

I tend to agree with you. For most non-computing intensive things, optimization is `not necessary'. In the sense that if you need to drive a user interface or so, you won't need as much cpu horsepower. For specialized tasks (e. g. rendering, media encoding, etc.) you can rely on the full power of the specialized cores.

Personally, I like the design of both, Cell and the UltraSPARC T1, both are truly innovative designs. Rather than just gluing two cores together like Intel, AMD and (to a certain degree) IBM have done, they have put in new ideas.

However, both requires efforts from software vendors as well and a change in paradigm to more parallelism for even standard apps.

[rhashem]

Originally Posted by OreoCookie

The PPC970MP is a good desktop chip, it has a similar thermal envelope than its predecessor and achieves competitive SPECmarks with IBM's compiler. I'm not saying it's the fastest chip out there or whatever, but it's competitive.

It's maybe 25% slower than the top Opterons. Competitve, maybe, but not marketable. It's hard to sell people on a top-end (and priced that way) CPU that performs like AMD's low-end dual-cores. Either way, the question here isn't G5 versus Opteron, but G5 versus Intel Core. Now, PowerPC folks would say that if Apple had stuck with PPC, it'd have been G6 versus Intel Core, but I don't think that'd be any better. Power6 has reduced OOOE based on the assumption that a fast memory bus, huge caches, and SMT will offset cache miss costs. That's a great assumption for a big-iron server, and a pretty bad one for a 1 or 2-way desktop machine. Meanwhile, Intel is pushing OOOE even further with the Core architecture, because that's the only thing that makes sense on the desktop.

Regarding the IBM SPEC marks: XLC 8.0 isn't available on the Mac, so those SPEC marks are pretty much meaningless.

That's beginning to change since every major cpu vendor switched to multi-core cpus. But for most things, you are correct.

There are two distinct multi-core transitions going on. Cell and Niagra are seeing the movement towards lots of simple cores, each with low performance. That's an architecture that makes no sense for the desktop, where single-threaded performance still dominates and a good app will have 4-way parallelism tops. On the Core* and Opteron side, you're seeing the movement to a few high-performance cores. Conroe will have much better single-threaded performance than its predecessors, dual-core or not. On the desktop, this architecture makes a whole lot of sense.

[rhashem]

Originally Posted by OreoCookie

Personally, I like the design of both, Cell and the UltraSPARC T1, both are truly innovative designs. Rather than just gluing two cores together like Intel, AMD and (to a certain degree) IBM have done, they have put in new ideas.

Both are designs that make sense for a specific market, but not in general. The reality of the software world is three-fold:

1) Code has a tremendous shelf-life. There is 20+ year old code in OS X, for example. Expecting massive rewrites is foolish.

2) Most existing software is not parallel, and it is difficult to make new software parallel. This is true both in theory and in practice.

3) Developer productivity is king. Making a big piece of software (eg: Sun's Solaris), costs more than making a CPU, and costs way more than merely buying bigger CPUs. The movement is towards higher-level languages (which generate lots of integer-branchy-crappy code, as one poster put it), to reduce development time.

Hardware merely exists to run software. Apple knew that the software world wasn't suddenly going to embrace Niagra/Cell-style designs, and acted accordingly.

NOTE: We've been doing the single-threaded thing for decades now. All our theory, all our code, huge amounts of intellectual capital are all designed this way. Those expecting a massive paradigm shift might as well expect everyone to give up cars and start taking public transportation --- it's an interesting idea, but it isn't going to happen.

[spork]

i got my first mac about 3 years ago, a 12' G3 500 ibook and fell in love with a clean, understandable and easy to use desktop and interface. this of course in addition to it's recognition of simply anything i plugged into it, target mode and amazing program instalation, imagine, put it into the aplications folder.
I have since gone through a few different macs and made my way up to G4 towers, of which i love, for all the above reasons and a whole bunch more which I am forgetting right now. In fact, i have converte 3-4 people to macs and urge them on anybody and everybody, citing a huge list of benefits. does that make me a candidate for the apple tattoo or part of a cult?
I want blazing fast processor speed and in the long run, it doesn't really matter to me where it comes from, as long as i still have Final Cut, DVD Studio Pro, toast, and a host of other programs i have come to trust along with a beautiful and easy to use computer which gets the job done for me, day in and day out. What more do we really need?
Actually i want my tower to go faster so if anybody has a good cheap dual G4 upgrade in the closet, let me know.
Mac on

[Catfish_Man]

Originally Posted by rhashem

Both are designs that make sense for a specific market, but not in general. The reality of the software world is three-fold:

1) Code has a tremendous shelf-life. There is 20+ year old code in OS X, for example. Expecting massive rewrites is foolish.

2) Most existing software is not parallel, and it is difficult to make new software parallel. This is true both in theory and in practice.

3) Developer productivity is king. Making a big piece of software (eg: Sun's Solaris), costs more than making a CPU, and costs way more than merely buying bigger CPUs. The movement is towards higher-level languages (which generate lots of integer-branchy-crappy code, as one poster put it), to reduce development time.

Hardware merely exists to run software. Apple knew that the software world wasn't suddenly going to embrace Niagra/Cell-style designs, and acted accordingly.

NOTE: We've been doing the single-threaded thing for decades now. All our theory, all our code, huge amounts of intellectual capital are all designed this way. Those expecting a massive paradigm shift might as well expect everyone to give up cars and start taking public transportation --- it's an interesting idea, but it isn't going to happen.

At a certain point, though, it stops being possible to improve. As you (or mduell, I don't recall which) mentioned earlier, single threaded integer performance is the hardest kind to get. Unpredictable branches so you can't speculate, poor locality of reference so you can't prefetch, little parallelism... the only 'friendly' trait of the code is that it loves big caches.

Once you hit the ceiling on this sort of performance, you can still improve other sorts though. The 'main core' in my hypothetical design is in there to get the best possible integer performance. By the time we're at 35nm there'll be lots of room for both a near-optimal integer core and a number of smaller fp cores. The big limitation is power; you obviously can't run both at once without massive cooling, so it needs to be very smart about changing clock frequencies on the fly and redistributing which pieces of the chip are active.

[rhashem]

Originally Posted by Catfish_Man

At a certain point, though, it stops being possible to improve. As you (or mduell, I don't recall which) mentioned earlier, single threaded integer performance is the hardest kind to get. Unpredictable branches so you can't speculate, poor locality of reference so you can't prefetch, little parallelism... the only 'friendly' trait of the code is that it loves big caches.

I don't think its a given that single-threaded performance will become impossible to improve. The easiest way to improve single-threaded performance is to increase clockspeed. Clockspeeds hit a wall for awhile there, but they seem to have picked up again. Woodcrest will be capable of 3.3 GHz operation, while having better IPC than any core besides Itanium. There is also promising new technology on the horizon that can push single-threaded performance still further. Mechanisms like checkpoint-recovery, trace caches, and clustered register files can take complicated OOO structures out of the critical path, enabling high IPC and high clockspeed. So I think it's safe to say that there is enough good stuff on the horizon to keep pushing single-threaded performance over at least the next decade.

Your point about hetrogenous architectures is well-taken, but I don't think the "complex core, simple coprocessors" model will take off outside of specialized domains (eg: graphics). Even in floating-point intensive code, the movement has been towards more high-level code, not less. Some big engineering software these days (eg: FEMLab) is being written in Java of all things. How fast is a JVM going to run on a Cell SPE? I think it is far more likely that specialized processors like GPUs will grow upwards instead of general-purpose processors growing downwards. That's even a performance win --- a GPU running GLSL code can make a lot of performance-improving assumptions that a general-purpose processor like a Cell SPE cannot.

[Catfish_Man]

Originally Posted by rhashem

Some big engineering software these days (eg: FEMLab) is being written in Java of all things. How fast is a JVM going to run on a Cell SPE?

You're thinking of the PPE, but this is an excellent point; I had forgotten that a lot of "floating point" code is being run in JITC'd languages these days. That raises another question though: how good can JITCs get? Clearly any decent one can optimize a tight simple loop down to as good or better than a static compiler would do, but I don't really know how common those sorts of tasks are.

[rhashem]

Originally Posted by Catfish_Man

Clearly any decent one can optimize a tight simple loop down to as good or better than a static compiler would do, but I don't really know how common those sorts of tasks are.

JITC's can get pretty good, but they have some caveats:

1) The further you get from the "tight, simple loop" model, the worse performance gets. Some code is built on such loops. Other code (eg: FEMLAB, a finite element analysis package), is built on lots of more complex loops (for things like mesh generation, matrix partitoning, etc).

2) The more code you have, the bigger your code cache becomes, which makes your cache footprint bigger. On modern processors, this can become a big problem.

3) The JIT only has a few hundred milliseconds to generate code. A lot of the most powerful optimization algorithms just take a lot of time to perform, and thus cannot be used in a JIT.

(3) hits simple processors like the SPE or PPE particularly hard, because optimization techniques you need to get good performance on them (alias analysis, code scheduling, software pipelining, loop transformations, basic block reordering) can get very expensive. Now, one could imagine a Cell-style design with a more complex PPE, but then you hit clockspeed limits. The whole reason the PPE is narrow and in-order is because IBM wanted to get the SPEs to high clockspeeds, and a highly OOO chip wouldn't be able to scale along with the SPEs. Now, you can introduce seperate clock domains for the PPE and SPE, but then life just becomes very complicated.

[Catfish_Man]

Originally Posted by rhashem

Now, you can introduce seperate clock domains for the PPE and SPE, but then life just becomes very complicated.

Well, this is why I kept referencing Intel's Foxton. Each core (and the L3 cache) in Montecito can independently vary in clock speed. I was thinking of a "PPE"/main core more equivalent to a Conroe core.

[mduell]

Originally Posted by OreoCookie

The PPC970MP is a good desktop chip, it has a similar thermal envelope than its predecessor and achieves competitive SPECmarks with IBM's compiler. I'm not saying it's the fastest chip out there or whatever, but it's competitive.

So, honestly, the only chip that I think is absent in IBM's portfolio is a real low-power `desktop' cpu for notebooks and small computers (in the sense that I'm not talking about an embedded cpu).

I guess we need to be even more specific, and say a good desktop chip "for Apple" or "running OSX." All the SPECmarks in the world don't help when the compilier isn't available for your platform (XLC for OSX is still on version 6; those SPEC figures come from version 8).

IBM is missing a ~20-30W laptop/thin desktop chip today, however I haven't seen much on the roadmap for a ~50-80W desktop chip tomorrow (to compete with Conroe and Kensfield); are there rumors of a cut-down POWER5?

Originally Posted by rhashem

It's maybe 25% slower than the top Opterons. Competitve, maybe, but not marketable. It's hard to sell people on a top-end (and priced that way) CPU that performs like AMD's low-end dual-cores.

AFAIK the PPC970 chips are relatively cheap compared to AMD/Intel's 1Ku prices.