Nah -- I'm one of those people, and it's great to be on a platform with a future.
Believe it or not, the existence of faster computers does not make yours slower!
Tom
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
PowerMac G5 Revision in Jan 2004?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Nah -- I'm one of those people, and it's great to be on a platform with a future.
Believe it or not, the existence of faster computers does not make yours slower!
Tom
Re: Faster!
Hold your breath, heck hold your tounge.
Have you seen what motorola has done to us?
How long did it take up to hit 1 GHZ... and then finally break it?
Are you disappointed with apple's prediction for a 1 GHZ increase?
Dealing with apple's past record... that is pretty damn good.
Originally posted by yamabushi
If the max speed is only 2.5GHz in 1Q 2004 I will be disappointed.
Hold your breath, heck hold your tounge.
Have you seen what motorola has done to us?
How long did it take up to hit 1 GHZ... and then finally break it?
Are you disappointed with apple's prediction for a 1 GHZ increase?
Dealing with apple's past record... that is pretty damn good.
I know that the prevailing wisdom has been to wait for a rev. B G5. So now we don't have long to wait. By then the compatability issues should be ironed out. The old rule works best, to purchase a machine when you need one. The 20th anniversary should even make it a better time to purchase.
This is incorrect on several accounts. First off, the AMD64 ISA differs in substantial ways from the ia32 ISA. Athlon XP is ia32 compatible, and has no mode that resembles amd64. The Opteron very cleverly retains opcode compatibility with ia32, but is essentially a 64-bit RISC-y chip in its native instruction set. Registers on amd64 are all GP (unlike ia32), and there are a lot more of them, which is very important for code optimization. (Ironically, on modern CPUs, CISC opcodes are easier to optimize on silicon for performance.)
Second, nothing is optimized for the Opteron. All the benchmarks you are seeing is the Opteron running in ia32 emulation mode, not amd64 mode. While there is GCC compiler support, there is still a lot of room left for fully exploiting the new amd64 ISA, which behaves like RISC but uses CISC-like opcodes. That the Opteron can school high-end ia32 native CPUs speaks volumes about its capabilities.
If you actually follow some of the early benchmarks of the Opteron that are being done in amd64 mode, it is significantly faster in amd64 mode than in ia32 mode. But there is still substantial work to be done before any Opteron compiler is vaguely optimal, especially since AMD doesn't produce a compiler (unlike Intel, whose compiler schools GCC in terms of the code it produces for ia32).
I'll clue you in that the difference between GCC and IBM in the typical case is about the same as the difference between GCC and Intel in the typical case. The difference is big enough to make it worth while for production code compiles, but not so big that it HAS to be part of the tool chain. We are talking <50% improvement in all non-toy cases. But at least the PPC970 has a highly optimized compiler for it; the Opteron does not. That said, the Opterons are still a touch faster than the PPC970 or Power4 on our codes.
You should also note that the early Opterons had slower memory than the G5s; something that is changing now though. This, Altivec (which I rarely use), and the lack of optimized amd64 compilers, are the only true advantages the G5 has in benchmarks. Nobody has really properly benchmarked the Opteron yet. Native mode support is still very sketchy and limited. That it is competitive in ia32 mode is impressive.
It sounds to me like you've read a lot of marketing whatnot, but haven't really considered it much beyond this. For a narrow range of floating point apps, the PPC/Power architecture is better, but for other types of apps that use limited FP (like ours), the Opteron is faster. And for most codes, they are close enough that it doesn't really matter. And yes, this is using the IBM compilers. I like PPC970 systems, but the reality is that it is neck and neck with Opteron and nothing more even with the IBM compilers. I'm curious to see what an Opteron optimized compiler can do.
Only tangentially related, the one thing that really blows goats about OS X (there is always SOMETHING -- I love OS X) is that there is still no true 64-bit option for it; it is a 32-bit OS even on G5s. I have large memory applications which have to run Linux on Opterons because we need a true LP64 operating system and LP64 LinuxPPC is still sketchy on G5 hardware (never mind that it is the fastest machine for our code, including Power and PPC options). FreeBSD is 64-bit clean, so it must be for backward compatibility purposes that they keep OS X 10.3 32-bit.
Oh noooooooooo... here we go again with overly long posts about optimization and compilers etc etc...
Yeah, I agree. They'll probably make some tweaks. The easiest things to update have to be optical drives, hard drives, and graphics cards... assuming economics don't factor in (profit margin.) It would REALLY be nice to burn a DVD at 8 speed. My 1x DVD burner, although incredibly useful, is dog-ass slow.
As someone mentioned before I'm sure, as soon as they're announced, be prepared to wait months to get one...
I wish Apple wouldn't announce things until they're ready to ship.
I wish Apple wouldn't announce things until they're ready to ship.
About the DVD burrners, read this:
Sony claims a startling 24x write speed for the PSX DVD burner, which is fast enough to write a one-hour TV program to DVD in only two and a half minutes. If the drive lives up to its promise, it will leapfrog two or three generations of DVD drives, as the fastest DVD writers available now run at a comparatively anaemic 4x speed.
More here:
http://www.zdnet.com.au/reviews/coolgear/av/story/0,2000023510,20279416,00.htm
Sony claims a startling 24x write speed for the PSX DVD burner, which is fast enough to write a one-hour TV program to DVD in only two and a half minutes. If the drive lives up to its promise, it will leapfrog two or three generations of DVD drives, as the fastest DVD writers available now run at a comparatively anaemic 4x speed.
More here:
http://www.zdnet.com.au/reviews/coolgear/av/story/0,2000023510,20279416,00.htm
You know I have a dvd 8x burner on my pc and it supposed to work with the mac too, but I have a G3... anyway I can't find any 8X media for it.
hahahaOriginally posted by mvc
Yeah, the ubergeeks need to optimise their postings
The DP 2 GHz G5 PowerMac is still the best choice of the three models. This is the only dual processor model and is thus likely to benefit the most from future updates to the operating system and applications. It also comes with the high-end PCI slots and 8GB capacity for RAM which are not guaranteed in the rumoured 'low-end' single 2 GHz G5 Powermac.
My prediction for the next PowerMac updates is that there will only be one single processor model and the other two will be duals. The low-end PowerMac would not be a dual because that would take sales away from a mid-range dual model... unless the high-end comes with four processors, which would be over-kill.
hahaha
excellent point. people complaining that they wont have the fastest computer on the block are ridiculous. if apple is continually making revisions and creating a better product then its good for everbody.
Re: It is great that Apple may be starting to catchup
I am currently running a 1.6G5 and the longest application I have to wait for would take about 2.5 minutes to do its stuff. The same task took 4 hours on a SE 30 10+ years ago Still I will happily lay out a 30-50% premium to resell my 1.6 and update to a single 2.4? a year from now.
Well for me my 450mhz cube 3? years ago was the first computer that felt fast enough for most applications. Ie office,mail pretty much anything that mostly involves typing. But for now their will allways be certain applications that demand more,mainly games
I am currently running a 1.6G5 and the longest application I have to wait for would take about 2.5 minutes to do its stuff. The same task took 4 hours on a SE 30 10+ years ago
Still I will happily lay out a 30-50% premium to resell my 1.6 and update to a single 2.4? a year from now.So when will the 64bit OS coming out?
I'm just wondering when a 64bit OS11 will be coming out?? I could see them releasing a 64bit OS in maybe another yr or so. I'm sure they are trying to get everyone on the G5 platform before going full 64bit. I wonder if they will still make updates for the older OS's atleast for few yrs so folks won't have buy new hardware. By then I will have two G5's but I will still have my G4PB and I would atleast like to keep it updated for few more yrs after the 64bit OS is introduced.
I'm just wondering when a 64bit OS11 will be coming out?? I could see them releasing a 64bit OS in maybe another yr or so. I'm sure they are trying to get everyone on the G5 platform before going full 64bit. I wonder if they will still make updates for the older OS's atleast for few yrs so folks won't have buy new hardware. By then I will have two G5's but I will still have my G4PB and I would atleast like to keep it updated for few more yrs after the 64bit OS is introduced.
In a Homeric fashion... "mmmm....3 GHz G5...gahhhhh...."
Unfortunately, I did not have the money this year to get a new computer, and I'd like to since my current computer is over 3 years old now. But it looks that if I keep waiting a little while longer, it might end up being a very good thing if I have to save up for another year and end up with an even more powerful computer.
About the only addition I'd like to see to the PM G5 is more space for hard drives. In my current PM G4, it can normally only hold 2 hard drives, and I'd like to be able to put in another drive so I can back up my drives instead of having to go and buy an external HD (however, I do use my iPod to back up my most important files).
If the MacWorld Expo is held in January, I wouldn't be surprised if there is no G5 announcement. There have been years where there was no new PM announcement, and by the end of January there were newer models out. But considering the wealth of Apple goodies which have come out in the past several months, I don't feel bad at all. Lots of great stuff coming out lately.
Now, an update to AppleWorks, now THAT would be an announcement.
Unfortunately, I did not have the money this year to get a new computer, and I'd like to since my current computer is over 3 years old now. But it looks that if I keep waiting a little while longer, it might end up being a very good thing if I have to save up for another year and end up with an even more powerful computer.
About the only addition I'd like to see to the PM G5 is more space for hard drives. In my current PM G4, it can normally only hold 2 hard drives, and I'd like to be able to put in another drive so I can back up my drives instead of having to go and buy an external HD (however, I do use my iPod to back up my most important files).
If the MacWorld Expo is held in January, I wouldn't be surprised if there is no G5 announcement. There have been years where there was no new PM announcement, and by the end of January there were newer models out. But considering the wealth of Apple goodies which have come out in the past several months, I don't feel bad at all. Lots of great stuff coming out lately.
Now, an update to AppleWorks, now THAT would be an announcement.
would this help spped up the 970 ?
Ok i am a bit from the old school of personal computers ( IBM XT , Apple II , Commodore 64 , ect. in other words when some of you were in diapers or a glint in your father's eye ) Now in the 80's there was talk of making faster processors , ram , chips , ect by making them out of Gallium Arsonide . Would such a process help speedup today's processors and chips ? I know it is made by the ten thousands of tanks of it cheaply. Has any development continued with it today or is it an abandoned idea because silicon is literally " dirt cheap " ?
Ok i am a bit from the old school of personal computers ( IBM XT , Apple II , Commodore 64 , ect. in other words when some of you were in diapers or a glint in your father's eye ) Now in the 80's there was talk of making faster processors , ram , chips , ect by making them out of Gallium Arsonide . Would such a process help speedup today's processors and chips ? I know it is made by the ten thousands of tanks of it cheaply. Has any development continued with it today or is it an abandoned idea because silicon is literally " dirt cheap " ?
Re: Re: It is great that Apple may be starting to catchup
Sorry, quantum mechanics rules that out. Not even Apple can circumvent the laws of QM. Or can they? Naahh...
Sorry, quantum mechanics rules that out. Not even Apple can circumvent the laws of QM. Or can they? Naahh...
It won't. All of the major updates of OS X made my PM G4 400 faster. Applications got faster. Maybe the speed improvements of OS X hit a ceiling now but I'm sure most applications will still run fine in the foreseeable future. You might be out of luck with gaming, though.
This is incorrect on several significant points.
First of all like it or not, simple Fact, Opteron is heavily based on AMD Athlon XP architecture. Thats not necessarily a bad thing since the Athlon XP is a great chip, but trying to argue otherwise is a bit pointless.
One only has to look at the exhaustive reviews that have been posted EVERYWHERE, to acknowledge that the Opteron is essentially based on an Athlon XP. Its fpu unit has remained completed unchanged as has its integer unit. Most of the internal pathway, the number of instructions it will handle in flight, hell even the internal pathways are pretty much identical. Adding more General purpose registers is correct however essentially to take proper advantage the chip must be in 64 bit mode. Like the G5 there is not 64 bit os currently.
I hope idon't sound condescending but why not head over to arstechnica and read the dissection of the AMD 64 architecture or anandtech etc....
Anyway, basing a processor on an older design and proven design like the athlon means that the athlon 64 is able to take advantage of optimized code. Incidentally there is extremely advanced Athlon optimization available in GCC..... 3d now, extended 3d now, SSE, SSE2 & x86-64 (if athlon 64) etc... before we say anything about the code being optimized for its internal pathways which again it does do. Funnily enough the Athlon does better with the Intel compiler than with the GCC.... but again this compiler does optimize code for it, generates SSE and SSE2 code automatically where possible, and does a great job of unrolling code to prevent branch mis predictions etc...
Simple fact is in 32 bit mode, the Opteron IS as optimized as it will get presently. The very fact that the Athlon has to remain ISA x86 compatible will always be a limiting factor. Also the Opteron can take advantage of years of optimizing for the Athlon arch in compilers.
Regardless the results speak for themselves. The G5 is far more heavily removed from its "predecessor" G4 than the Athlon 64 is from the Athlon XP.
Any in depth review i have seen has said that the G5/970 is so far removed from previous PPC architecture that running older uncompiled legacy software it will only be running at 50% best case scenario of its peak. Even when recompiled with GCC 3.3 most case scenario the chip is only running 50% its potential. A recompile with XLC and XLF definately substantiates this.
Also the G5 was designed to have lots of instructions flying around in flight (ignore apple marketing speak), and to have its units running in parallel. That was the design decission and this pretty much leaves the older PPC design decisions behind. However as arstechnica pointed out in their analysis of the chip, this is brilliant, great for performance , when software and compilers catch up. The present GCC 3.3 does NOT optimize the code properly for the internal pathways etc... How else can you explain a 300% increase in performace when moving to the XLC and XLF compilers...... there has to be something more to it than minimising branch misprediction. Indeed this would certainly point out and highlight arstechnica's theory that initial compilers would only be using 50% efficiency of the chip in best case scenario and that code would have to be compiled with a compatible compiler before the performance of the chip could be unlocked...... I would reckon from my own testing that this is very much the case. In heavy integer and floating point mathematical work i was seeing practically 200% improvement if not more by moving to XLC and XLF compiler. Others in the thread that i linked to have found identical results.
Can you honestly tell me that presently the Athlon 64 is running at only 50% of its potential....no it runs pretty much close to its peak 4.5 GFlops. When all instruction units of 970 are filled (which requires a compatible compiler.... not GCC!) it runs at peak 8GFlops, but presently with most GCC compiled code it only runs peak 4GFlops, again this information has been sourced from ars forums.
I use a dual Opteron 2 Ghz in work thats about nearly 1 month old. We actually compared results with the 970 G5 and were blown away. while merely competitive and neck and neck with GCC compiled code it blew the Opteron away when we re ran with XLF compiled code.
why dont you test for yourself and see if it does anything for you like maybe doubling your results!
. remember this doesn't contain any SIMD autovectoring ...yet! But as you already mentioned you don't need Altivec for your code etc....Im too lazy to go look for your post and find the remark about Altivec being of little use for hte majority of code... but i just wanted to say that i completely disagree.
A huge amount of people who buy macs are creative professionals who work with graphics, music, film etc... for these apps Altivec makes a massive difference where used.
If you are working with science based code like BLAST alitvec makes a huge difference again. Unfortunately for Apple they never managed to get an autovectorising compiler out like intel and amd had. but again this is changing. An autovectoring compiler would use Alitvec where ever possible ... so even an app that you wouldn't expect to make use of alitvec at all, might show a slight 3 -5% performance increase.
Kind Regards
i_wolf
P.S. when i have been comparing processors so far i have been comparing in 32 bit mode for both. I do agree that there is some work to be done optimizing for 64 bit mode athlon 64, the same as there would be for 64 bit mode G5. But thats another analysis thread! and presently there is no pure 64 bit os for either, bar linux for x86-64. ..... rumors of a G5 64 bit gentoo have been heard mentioned though
Originally posted by i_wolf
This is incorrect on several significant points.
First of all......... is no pure 64 bit os for either, bar linux for x86-64. ..... rumors of a G5 64 bit gentoo have been heard mentioned though
Wow. A lot of very long posts here
But anyway. All the comparisons I read so far in any Mac or PC magazine stated, that when compared in real speed - meaning real Apps like Photoshop, Quake ... - the G5 is not slower and not faster than the fastest PC machines. And thats what counts. At least for me.
Re: Re: Re: It is great that Apple may be starting to catchup
Ahh I was wondering if someone would drag Schroedinger's poor confused cat into this.
But, observe, since Steve Jobs reality distortion field is clearly a singularity, the ordinary laws of the universe no longer apply within his personal event horizon, and therefore nothing is impossible for Apple as long as he is around!
And that is the sole reason why I still expect a G5 powerbook in January.
Originally posted by tduality
Sorry, quantum mechanics rules that out. Not even Apple can circumvent the laws of QM. Or can they? Naahh...
Ahh I was wondering if someone would drag Schroedinger's poor confused cat into this.
But, observe, since Steve Jobs reality distortion field is clearly a singularity, the ordinary laws of the universe no longer apply within his personal event horizon, and therefore nothing is impossible for Apple as long as he is around!
And that is the sole reason why I still expect a G5 powerbook in January.
mvc: dude there are many aspects of my life and interests that i have. computing happens to be one of them. i come here because its a tech/computer forum where i would imagine like minded people reside. when im out with mates we take footie, films, women etc... If i want to do that il talk to my mates. if i want to talk comp il come in here. whats the problem.. what have i said that has offended you?? All i have done is mention a few facts.... which would explain why in some apps we are seeing double the perf of nearest competetive pc, while in other apps we are neck and neck. Is this forum the wrong place to discuss this, if so can you recommend a more appropriate place.
Commander laforve : yeh you are right a lot of mags bench's show neck and neck with the fastest. but they are legacy apps, and seamingly not even tapping the full potential of chip. seeing as a recompile with compatible compiler is showing double the results with the same code, i would imagine that future apps and even current apps if they were recompiled with compatible compiler to take advantage of the chip they are running on would show significant perf boost to leave a large gap between the 970 and the nearest competing x86.
But if anything its a testiment to IBM how fast their 970 is that even now on crippled GCC3.3 code that their 970 is neck and neck with the fastest optimized x86 out there. It just means that perf wise we got a LOT to look forward to from the same chip! Wouldn't it be interesting to see comparisons between the same processors and systems in a year from now on future editions of current software!
Falleron: thanks for standing up for me there. much appreciated.
Commander laforve : yeh you are right a lot of mags bench's show neck and neck with the fastest. but they are legacy apps, and seamingly not even tapping the full potential of chip. seeing as a recompile with compatible compiler is showing double the results with the same code, i would imagine that future apps and even current apps if they were recompiled with compatible compiler to take advantage of the chip they are running on would show significant perf boost to leave a large gap between the 970 and the nearest competing x86.
But if anything its a testiment to IBM how fast their 970 is that even now on crippled GCC3.3 code that their 970 is neck and neck with the fastest optimized x86 out there. It just means that perf wise we got a LOT to look forward to from the same chip! Wouldn't it be interesting to see comparisons between the same processors and systems in a year from now on future editions of current software!
Falleron: thanks for standing up for me there. much appreciated.