Poster | Thread |
cdimauro
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 22:04:49
| | [ #41 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3621
From: Germany | | |
|
| @pavlor
Quote:
pavlor wrote: @cdimauro
Quote:
because only a brain-less would have chosen to begin the PowerPC port on late 2003, when the x86 supremacy was overwhelming. |
Port began in 2001.
|
Effective PowerPC port seems to be started on late 2003, according to itix:
"Think it was as late as 2003 when the PPC port happened."
Is it true or not? Quote:
However, G5 had in 2003 better performance ratio towards x86 than G4 in 2001.
|
I've already reported a link, and I do it again: Athlon 64 vs. Apple G5 Systems: Not Even Close
As you can see, there were some tests were the top x86s shown DOUBLE the performance of the G5, and better than the DUAL G5. Even the Pentium 4, which wasn't certainly well know for its performance, drawn circles around the G5s.
BTW, it was early October 2003. Quote:
Quote:
Vision seems to be not your one: |
True, 7 dioptre lens on both eyes. |
Time to get a better pair.
@matthey
Quote:
matthey wrote: Quote:
cdimauro wrote: We already discussed it: go for a more powerful architecture, and with a much brighter future. |
RISC processors like the Alpha, PA-RISC, MIPS, SPARC, m88k and PPC which were supposed to replace CISC never lived up to their hype. The 68k was dropped by Motorola due to their own hype
|
That was part of the reason, but history shown that Motorola wasn't able to keep-up the contenders, RISC included, with its 68K family.
Take a look a what she released, when, and the drastic decisions that she made to try to sustain it. Quote:
and marketing while the inferior CISC evil sister x86 became the dominant high performance CPU for desktops and laptops.
|
An architecture doesn't have to be only "beauty", but it has to perform too. Take a look at numbers of such "inferior" CISC. Quote:
So are you saying Amiga should abandon the superior 68k CISC processor and Amiga classic compatibility and to adopt an inferior CISC design on the dark side?
|
No. See below. Quote:
Is that now or back when Motorola was shoving PPC down everyone's throat including C=?
|
The context was 2001, Mat. By that time, x86s already shown their muscles, and the situation was inverted: RISCs had to keep-up the "inferior" CISC design.
@pavlor
Quote:
pavlor wrote: @matthey
680x0 lagged at least one year behind x86 since late 80s.
1989: 80486 (1990: 68040) 1992: 80486DX/2 (no comparable 680x0 CPU) 1993: Pentium (1994: 68060 on much lower clock speed) |
Exactly. And Motorola had to cut instructions and change the MMU design in order to reduce the amount of transistors used, killing the backward compatibility with the existing software. Whereas Intel, the "inferior", never did it.
Last but not really least, the 68060 only allowed to pair simpler and the shortest instructions, whereas the "inferior" Pentium allowed to pair much more complex ones, with an certain advantage with the executed code.
Again, that's because Motorola was constrained by the transistor (and power too: its chips, 68040 included, generated a lot of heat) budget, cutting features of its "beauty" but too much complex architecture. |
|
Status: Offline |
|
|
cdimauro
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 22:11:07
| | [ #42 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3621
From: Germany | | |
|
| @matthey
Quote:
matthey wrote: Quote:
pavlor wrote: 680x0 lagged at least one year behind x86 since late 80s.
1989: 80486 (1990: 68040) 1992: 80486DX/2 (no comparable 680x0 CPU) 1993: Pentium (1994: 68060 on much lower clock speed) |
The 68k was usually a little later and didn't clock as high but was more powerful for the clock frequency than the last x86 iteration.
|
Final, total numbers matter: not the intrinsic efficiency. Quote:
I believe new 68k designs would not clock quite as high
|
Why not? Quote:
as x86 designs while potentially giving better performance at a given clock rate (lower frequencies have advantages).
|
See above: what counts is the final number that you get. Quote:
The 68060 was a very good superscalar base design. It would have been interesting to see where it would have gone after being clocked up
|
It was limited by its complexity: that's why Motorola wasn't able to reach higher clocks. And the same already happened with the 68040. Quote:
and with a few bottlenecks removed.
|
Memory double indirection modes, I suppose. But it's too late: they are part of the ISA, and they were used.
Like Olaf already stated, the 68K software base is huge. But mostly closed source, so you have to deal with what was released. Pros and cons of legacy... |
|
Status: Offline |
|
|
pavlor
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 22:33:35
| | [ #43 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @cdimauro
Quote:
Effective PowerPC port seems to be started on late 2003, according to itix: |
Itix as source for history of OS4 developement? Priceless.
Quote:
Of course not. Simply trust me in matters of computer history...
Quote:
I've already reported a link, and I do it again: |
I wrote: G5 had in 2003 better performance ratio towards x86 than G4 in 2001
Your link shows G5 in some tests faster, in some slower (eg. word/search tests highly depend on harddisk speed more than CPU), in most on par (Render, Photoshop, Quake). Nothink surprising. Try similar test with G4 form 2001 and and then high-end x86 CPUs...
Quote:
And Motorola had to cut instructions and change the MMU design in order to reduce the amount of transistors used, killing the backward compatibility with the existing software. |
Killing is too strong word, limiting maybe - it wasn´t that great problem on Amiga after all. |
|
Status: Offline |
|
|
cdimauro
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 22:47:42
| | [ #44 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3621
From: Germany | | |
|
| @pavlor
Quote:
pavlor wrote: @cdimauro
Quote:
Effective PowerPC port seems to be started on late 2003, according to itix: |
Itix as source for history of OS4 developement? Priceless.
Quote:
Of course not. Simply trust me in matters of computer history...
|
YOU also reported that the preliminary OS4 version was mostly 68K on 2004... Quote:
Quote:
I've already reported a link, and I do it again: |
I wrote: G5 had in 2003 better performance ratio towards x86 than G4 in 2001
Your link shows G5 in some tests faster, in some slower (eg. word/search tests highly depend on harddisk speed more than CPU), in most on par (Render, Photoshop, Quake). Nothink surprising.
|
pavlor, I think that you seriously need a new pair of glasses.
The link shows that the top x86 only lost, slightly, on 2 tests, but against the DUAL G5.
If you look at the SINGLE G5, the situation is simply embarrassing: it was obliterated! Quote:
Try similar test with G4 form 2001 and and then high-end x86 CPUs...
|
I never saw a dual G4 on 2001 losing almost all tests against a single x86.
But if you have some data, I appreciate if you can give it, so I can take a look. Quote:
Quote:
And Motorola had to cut instructions and change the MMU design in order to reduce the amount of transistors used, killing the backward compatibility with the existing software. |
Killing is too strong word, limiting maybe - it wasn´t that great problem on Amiga after all. |
It was, with the 68060, because Motorola removed even some useful instructions in user mode (so, no MMU involved). |
|
Status: Offline |
|
|
pavlor
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 22:59:09
| | [ #45 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @cdimauro
Quote:
YOU also reported that the preliminary OS4 version was mostly 68K on 2004... |
I wrote: non-kernel components
ExecSG was of course PowerPC with many OS parts still runing under 68k emulation.
Quote:
I think that you seriously need a new pair of glasses. |
Quote:
If you look at the SINGLE G5, the situation is simply embarrassing: it was obliterated! |
I fear I don´t share your passion for overblown words: 30 % is not obliterated in my point of view.
Quote:
I never saw a dual G4 on 2001 losing almost all tests against a single x86. |
What of benchmarks you linked use both CPUs in G5? Premiere? judging by result not. Photoshop? Seems so, but there the difference is not big. Word? No. Quake 3? No.
Quote:
It was, with the 68060, because Motorola removed even some useful instructions in user mode (so, no MMU involved). |
Sure, but was this real problem on Amiga? |
|
Status: Offline |
|
|
matthey
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 10-Dec-2015 23:01:14
| | [ #46 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| Quote:
cdimauro wrote: That was part of the reason, but history shown that Motorola wasn't able to keep-up the contenders, RISC included, with its 68K family.
Take a look a what she released, when, and the drastic decisions that she made to try to sustain it.
|
R&D was going to more promising but flawed RISC research when Motorola saw greener pastures on the other side of the fence and was drawn into the AIM Consortium. Motorola/Freescale has always tried to reduce and simplify to the point of performance degradation. Look no further than ColdFire and MCore for example.
Some of the Intel advantage was as much luck that FPS games worked better on PCs and lack of companies like C= to quickly add chunky modes. Management probably also played a role. Motorola designs were generally more professionally done with less bugs while Intel designers seemed to be given more room to be creative but they made more mistakes. Some of the mistakes came from pressure from the marketing department to increase clock speeds which sold processors. I have heard rumors that is how the Pentium 4 space heater design came about. Both companies made mistakes along the way but Intel won the cash flow battle.
Quote:
Exactly. And Motorola had to cut instructions and change the MMU design in order to reduce the amount of transistors used, killing the backward compatibility with the existing software. Whereas Intel, the "inferior", never did it.
|
Some of the decisions were good and others less so. Motorola should have worried more about performance than being logic misers though.
Quote:
Last but not really least, the 68060 only allowed to pair simpler and the shortest instructions, whereas the "inferior" Pentium allowed to pair much more complex ones, with an certain advantage with the executed code.
|
The 68060 can handle very powerful instructions in each integer pipe but it doesn't have enough fetch to feed it. This should not have been a problem to fix. I don't believe instruction folding/fusing would have been a problem for the 68060 design either. It may already do this for decrement+branch into DBcc. I expect there was just not time to finish optimizing for the 68060 and it was already stronger per clock than the Pentium so they needed to get it out the door.
Quote:
Again, that's because Motorola was constrained by the transistor (and power too: its chips, 68040 included, generated a lot of heat) budget, cutting features of its "beauty" but too much complex architecture. |
The 68040 was not a good design because it was not the direction they needed to go for the future. The 68060 design had to be largely created from scratch. The 68060 was a good base design which later designs could have improved on. It should have been possible to clock it up but the EA before ALU designs are not high clockers (but efficient) and a new OoO design for the highest performance would have been needed at some point. The complex 68k addressing modes should be easier to handle with OoO although the 68060 and Apollo core figured out how to keep them from being much of a bottleneck. There are some complexities of the 68020 ISA like this which limit 68k clock speeds but the powerful addressing modes potentially make it more powerful per clock than x86 (68k's shorter average instruction length and simpler encoding are also small advantages for parallel processing). Each processor ISA gives some innate traits as we know ;).
Last edited by matthey on 10-Dec-2015 at 11:14 PM. Last edited by matthey on 10-Dec-2015 at 11:12 PM.
|
|
Status: Offline |
|
|
agami
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 6:16:50
| | [ #47 ] |
|
|
|
Super Member |
Joined: 30-Jun-2008 Posts: 1637
From: Melbourne, Australia | | |
|
| @Sky7 It seems you weren't around for the Wintel hate years; the computing community darn nearly tore itself apart. "Use an Intel compatible CPU? Ptew, I spit on the person who suggests that!" "Why don't you just marry Bill Gates?"
Moving to x86 in the post-Commodore '90s would have been admitting defeat. And back then we all beloved we were better than that. We still had some fight in us. It didn't matter that we were unreasonable. Also, we were "keeping up with the Joneses"; If it was good enough for Apple and Be (that Jean-Louis Gassée kept flattering us), then it was good for us.
(Continuing with multiple replies without referencing the original poster)
1. QNX is stil superior. 2. In all my real-world use cases of ripping CDs to MP3, ripping DVDs to MP4, and using software which was available on both Windows and Mac OS X e.g. Handbrake, the PowerPC G5 was on par, and in some cases faster than an equally clocked Intel processor. 2.1 PowerMac G5 single core 1.8GHz vs. Pentium 4 1.8GHz. Heck, even the Pentium M 1.8GHz outperformed the P4. 2.2 Later a quad core PowerMac 2.5 GHz easily beat the the early Intel Core2 Quad in the above tasks. It wasn't until about 2010-11 that a quad core i5 at 2.5GHz outperformed the 2005 G5. NOTE: Intel did wonders in terms of lowering cost, moving to smaller silicon and improving power efficiency. Credit where credit is due. 3. So if phase 5 decided to slap a Cyrix or AMD CPU on a 040/060 card and call it the Cyberstorm686 and Blizzard586, and H&P WarpOS came with an x86.library, MorphOS and AmigaOS 4.x would be x86 operating systems? Fricken domino effect.
_________________ All the way, with 68k |
|
Status: Offline |
|
|
CodeSmith
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 6:30:54
| | [ #48 ] |
|
|
|
Elite Member |
Joined: 8-Mar-2003 Posts: 3045
From: USA | | |
|
| One important thing to remember when looking at G5 vs x86 is that most of the PPC data we have from the time comes from Apple marketing, and at the time Apple was in a full PR offensive against x86 (ironically, considering the timing when OSX first came out, they must have had a full team working on it by then). It was pretty common to see really dubious benchmarking going on, so one needs to be careful when looking at sources from that time period.
Last edited by CodeSmith on 11-Dec-2015 at 06:38 AM. Last edited by CodeSmith on 11-Dec-2015 at 06:34 AM. Last edited by CodeSmith on 11-Dec-2015 at 06:33 AM.
|
|
Status: Offline |
|
|
ne_one
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 7:52:37
| | [ #49 ] |
|
|
|
Cult Member |
Joined: 13-Jun-2005 Posts: 905
From: Unknown | | |
|
| @billt
Quote:
While everyone seems to think that x86 is the path to success, if that were true, then how do we have both OS4 and MOS on PowerPC, and no Amithlon anymore? The x86 path imploded and vanished... |
That reasoning makes absolutely no sense.
It has nothing to do with the technology and everything to do with the ineptitude of the forces that have been guiding this platform for the last 15 years.
Dig in to some of the older mailing list archives and you will see many of us being vilified not only for challenging the decision to commit to yet another dying hardware architecture but for even posting to the lists using other platforms.
Surely by now it's clear that coupling the OS with *any* hardware is a bad idea?
|
|
Status: Offline |
|
|
tlosm
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 8:16:43
| | [ #50 ] |
|
|
|
Elite Member |
Joined: 28-Jul-2012 Posts: 2746
From: Amiga land | | |
|
| @pavlor
some one dont wanna understand the g5 start be optimized in late 2007/2009 in the apple compilers... before was not . only in xcode 3.1.4 for example the gcc was doing right it job on g5 all the stupid benchmarks did on programs optimized for g4 and make it running on g5 was making all worst results compared the low classes cpus and compared with p4 . this is the reality of the facts ...
ps: the quad have only 2 programs who are using the 4 cores and this was made by apple in 2009 logic studio pro and final cut pro.... all and im writing all are using only 1 or max 2 cores ... Last edited by tlosm on 11-Dec-2015 at 08:19 AM. Last edited by tlosm on 11-Dec-2015 at 08:17 AM.
_________________ I love Amiga and new hope by AmigaNG A 500 + ; CDTV; CD32; PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB; MacBook Pro Retina I7 2.3ghz; #nomorea-eoninmyhome |
|
Status: Offline |
|
|
matthey
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 8:58:19
| | [ #51 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| Quote:
ne_one wrote: Surely by now it's clear that coupling the OS with *any* hardware is a bad idea?
|
It's not clear to me. A tightly coupled OS to a narrow range of standardized hardware is more efficient. This is part of the reason why inferior specked consoles have respectable performance. The trick is controlling the hardware so that a manufacturer or intellectual property holder can't pull the hardware out from under your feet. This is much easier today with cheap FPGAs.
Quote:
tlosm wrote: some one dont wanna understand the g5 start be optimized in late 2007/2009 in the apple compilers... before was not . only in xcode 3.1.4 for example the gcc was doing right it job on g5 |
The PPC (and generally RISC) philosophy was to move complexity out of the hardware and into the compiler but there are assumptions which can't be made by the compiler. Apple, IBM and Motorola all worked toward the holy grail of PPC compilers but failed. Meanwhile, Intel did a good job of optimizing their processors internally to work well with less than optimal existing code. More is known and possible with hardware optimization. CISC is more tolerant of crap code also. The big advantages of RISC were supposed to be simpler hardware and higher clock speeds but they contributed to their downfall. How many failed RISC designs does it take before processor designers get it?
|
|
Status: Offline |
|
|
ExiE
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 9:22:59
| | [ #52 ] |
|
|
|
Regular Member |
Joined: 18-May-2004 Posts: 450
From: Czech Amiga News | | |
|
| @pavlor Quote:
No, it was otherwise: QNX first (1998), then Linux |
My bad, memory corruption error... |
|
Status: Offline |
|
|
OlafS25
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:07:37
| | [ #53 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6321
From: Unknown | | |
|
| @agami
Wintel hate still in 2002/2003?
I can remember early 90s, in that period most amiga users saw them as rebels against the evil empire (Microsoft). At that time a change to X86 (how economic reasonable it would have been) would have propably not been accepted by most users. PowerPC was there easier to "sell" to the customer base. But 2002/2003 when AmigaOS and MorphOS became reality situation had changed.
68k indeed seemed dead back then but at least at some point it should have become obvious that most users are not willing to change. So updating Amiga OS (I do not know if there are or at least were legal limitations) would have brought them much more money than the PPC line in my view. Customers finally decide success or economic failure and if customers decide that they prefer to stick with 68k it would have been better to support it. Today the situation has changed (at least for me) because Aros has a 68k branch and thanks to Toni Wilen there are replacements for the copyrighted roms but back then they would have sold a lot, propably much more 68k than PPC. But for Hyperion the 68k community only were a potential resource for new customers of PPC hardware and not a target market for products.
@thread
if I understand the original question right it was not about why using PowerPC vs. 68k but dropping 68k support at all. I think the responsible persons had a wrong impression and thought everyone would change to PPC. And it is certainly also a kind of arrogance that is still feelable today, propably even more today because you cannot really do business anymore. Last edited by OlafS25 on 11-Dec-2015 at 10:15 AM.
|
|
Status: Offline |
|
|
pavlor
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:10:11
| | [ #54 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @OlafS25
Quote:
but dropping 68k support at all. |
You mean they should develop OS4 for both 68k and PowerPC? |
|
Status: Offline |
|
|
OlafS25
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:16:12
| | [ #55 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6321
From: Unknown | | |
|
| @pavlor
at least they should have offered updates for the 3.X line |
|
Status: Offline |
|
|
pavlor
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:28:56
| | [ #56 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @OlafS25
Quote:
at least they should have offered updates for the 3.X line |
And slow down 4.x developement? |
|
Status: Offline |
|
|
OlafS25
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:29:38
| | [ #57 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6321
From: Unknown | | |
|
| @pavlor
slow down? |
|
Status: Offline |
|
|
pavlor
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:32:00
| | [ #58 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @OlafS25
Commiting more resource to 3.x branch would mean slowing down developement of 4.x branch (which was slow in itself - first public release in 2004). |
|
Status: Offline |
|
|
OlafS25
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:32:31
| | [ #59 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6321
From: Unknown | | |
|
| @pavlor
when they had earned more money that also would helped the 4.X line. Ignoring customer wishes is never a good idea for a business. |
|
Status: Offline |
|
|
OlafS25
| |
Re: Why was AmigaOS 4.X developed only for PowerPC? Posted on 11-Dec-2015 10:33:53
| | [ #60 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6321
From: Unknown | | |
|
| @pavlor
but earning money is what business is for. And as far as I know they represent themselve as business and not a hobby venture. |
|
Status: Offline |
|
|