| [News] OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? | ANN.lu |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 201 of 211 | ANN.lu |
| Posted by Lennart Fridén on 13-Apr-2002 07:18 GMT | In reply to Comment 198 (NeRP):
*Giggle* |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 202 of 211 | ANN.lu |
| Posted by Lennart Fridén on 13-Apr-2002 07:18 GMT | In reply to Comment 199 (NeRP):
LOL! |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 203 of 211 | ANN.lu |
| Posted by Alkemyst on 13-Apr-2002 07:34 GMT | In reply to Comment 200 (NeRP):
Nerp
i have to agree
that i some of them say i had to move to pc & windows becuase AmigaOs does not have the apps &
so i dont use amigaOs anymore.
so why do they want AmigaOsx86 as so they can have it sitting on the harddive & not use it.
cos it does not have the apps once again. |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 204 of 211 | ANN.lu |
| Posted by Wittgenstein on 13-Apr-2002 12:31 GMT | In reply to Comment 195 (aTmosh):
> Looks like time to step in and debunk some FUD again..
Well, it looks like you are the FUDer.
>> The Hammer is more vapor than the G5 and it has only been
>> shown once (correct me if I am wrong).
> You are wrong, it has been shown running Linux at CeBIT and engineering
> samples are in the hands of people concerned. How that constitutes more
> vapour than the mythical G5 anyone is yet to see is beyond me.
Yeah, just like I said? Shown once. As I might know MOT doesnt use to show unfinished products so this prooves nothing. Just look at Intel and the prototoypes of Itanium.
>> The price per performance of the the 64 bit x86 chip when running
>> 32 bits code will be very poor (because that the complexity of the wirering
>> increases with a factor 2) so only 64 bits code will be relevant.
>Perhaps you are confusing Hammer with Intel's failed Itanium, 32 bit code
>will be more efficient than current Athlons. AMD know what they are doing
>thankfully. But don't take my word for it, we'll see soon enough.
Uhh no? When you want to create a 64 bits CPU you end up with twice as many (in the range of twice as many) wires. This will make the die much larger and it will also increase the power consumption (this is the reason to Intels 'failure' of Itanium). This will also make the CPU much more difficult to produce in large quanteties so there is no way they can compete with the real 32 bits cpus in price per performance when runing 32 bits code.
>> The big advantage of the PPC8500 is that it will execute 32 bits ppc code
>> faster than the 7500 and it will reach clockspeed to atleast 3.5 GHz +- 0.5 GHz
>This is pure speculation, Motorola's own roadmap lists 800 MHz-2 GHz for
>85xx. Of course I could argue Intel will ramp P4 up to 10 GHz and IBM and
>some other manufacturers have shown purely lab transistor setups that go
>over 200 GHz in a bidding war, but that doesn't mean a flying fluff for us
>desktop users, or for anyone much in the next 5 years. Then again, people
>have already successfully overclocked Pentium 4s to over 4 GHz.. Personally
>I'll be more than happy with 2.5 GHz Athlon (overclocked of course) by summer >:)
Oh dear. You are compairing the f_T (transition value) of a transistor with a clockfrequence of a CPU. This comparision is totally irrelevant. The main problem isnt the speed of the transistors. I can easy get a transistor with f_T around 50 ghz and they have build 100 ghz mixers in many years now. This isnt the problem. The problem lower the heat generated. This heat comes from inpurities in the material. You can lower the heat by using smaller connectors or using a cleaner enviroment under production (less particles in the air).
But you are correct at one point. If you want speed today you should go x86. But dont expect that it will always be like that.
/ |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 205 of 211 | ANN.lu |
| Posted by aTmosh on 14-Apr-2002 08:38 GMT | In reply to Comment 204 (Wittgenstein):
>> You are wrong, it has been shown running Linux at CeBIT and engineering
>> samples are in the hands of people concerned. How that constitutes more
>> vapour than the mythical G5 anyone is yet to see is beyond me.
> Yeah, just like I said? Shown once. As I might know MOT doesnt use to
> show unfinished products so this prooves nothing. Just look at Intel and
> the prototoypes of Itanium.
The wrong part was meant to be about the Hammer being more vapourous than
G5, which it isn't, AMD's entire future hinges on it. But if you want to
nitpick, it was arguably shown twice at CeBIT, one motherboard ran Linux
(concurrently running one 32 and one 64 bit program, funnily enough they
had bouncing balls in two windows for a demo) and on another motherboard
Windows XP. Oh yeah, they had working motherboards, actually demonstrating
working programs and OSes, not just the chips in a glass cage or something.
This also means AMD has learnt from the past 3rd party chipset delays and
already have their own ready way before the launch, cutting time to market
significantly. Does G5 have working prototypes? Engineering samples? A
chipset? This takes a lot of time. It wouldn't surprise me if G5 didn't
appear this year at all.
What Itanium has got to do with all this is beyond me.
>> Perhaps you are confusing Hammer with Intel's failed Itanium, 32 bit code
>> will be more efficient than current Athlons. AMD know what they are doing
>> thankfully. But don't take my word for it, we'll see soon enough.
> Uhh no? When you want to create a 64 bits CPU you end up with twice as many
> (in the range of twice as many) wires.
No, that's where HyperTransport comes in. ClawHammer will not nearly have
twice as much pins as current Athlon/Duron, only 63% more in fact, and get
this, that includes integrating the memory controller into the CPU, as well
as the communication bus to the rest of the system (including other CPUs)!
This basically does away with the Front Side Bus as we know it and the
North Bridge (leaving only AGP as a discrete part). SledgeHammer will
double the number of pins to over 900, but then again it's meant for 2 to
8-way CPU servers.
The new serial packet driven protocol (HT) not only decreases the amount of
pins needed (it's scalable from 2 to 32 lines depending on how much you
need for what parts of the system), it greatly simplifies board layout,
therefore cost, and at the same time it increases bandwidth *throughout*
the system by a magnitude. We're talking anywhere from 5.33 GB/s to 25.6 GB/s
memory/IO bandwidths here, depending on how many CPUs are involved (since
they each have their own memory controller).
> This will make the die much larger
Nope. Current Athlon XP has a die size of 128 square mm, this will shrink
to 80 mm^2 with Thoroughbred (0.13 micron) and only increase to ~103 mm^2
for ClawHammer, i.e. still smaller than current XP, and much smaller than
the already 0.13 micron Pentium 4 Northwood (136 mm^2).
> and it will also increase the power consumption
Nope, AMD has been sticking to a maximum of around 70W for quite a while
now, and will remain under that level with process shrinks and improvements.
Even HyperTransport has power saving modes. For example Thunderbird went up
to 72W max (i.e. peak power, not sustained), then XP dropped to 45-50 and
as clock ramped up neared 70 again, this will once again drop with
Thoroughbred thoroughly (excuse the pun :) and once again slowly ramp up
until the end of the year when SOI is introduced in Barton (although they
might actually skip this and go straight to Hammer). SOI is supposed to
lower power draw by up to 3 times. After that they move to 90 nanometer
process, etc.
In short: power consumption will not increase, but fluctuate below a given
threshold.
> (this is the reason to Intels 'failure' of Itanium).
Not really, their main problem was that everyone thought in the early
nineties that either RISC or VLIW would be the future but due to the sudden
competition between AMD and Intel CISC got a huge boost and has surpassed
both (partly by employing RISC like techniques under the hood). This meant
that there was no pressing reason to switch codebases, therefore low volume
and no critical mass for IA-64. It now looks like McKinley, although better
than Merced, will be too little, too late. Intel is even working on an own
x86-64 backup plan in the background. My gut feeling says they'll need it.
> This will also make the CPU much more
> difficult to produce in large quanteties so there is no way they can
> compete with the real 32 bits cpus in price per performance when runing 32
> bits code.
Heh, I think you are missing the big picture here, AMD is replacing their
32 bit line (i.e. Athlon/Duron) entirely with (Claw)Hammer in 2003. It is a
very real 32 bit CPU, with 64 bit thrown in as a bonus. They now produce 32
million CPUs a year, and will ramp that up to 50 million by the end of this
year when their Dresden factory (built in cooperation with Motorola and
IBM) is going full tilt.
PowerPC simply has no economies of scale to even remotely be able to
participate in the desktop market. Why do you think Macs are (amongst
others) so damned expensive? Hoping for a desktop PPC comeback is an
exercise in futility.
Since AMD produces more chips per wafer (70% more for Thoroughbred and 32%
for Hammer) compared to Intel's P4, they have much lower costs. Coupled
with the performance boost of Hammer, and quite likely major clock boost
due to SOI this means Intel is in real trouble unless AMD mess up the
Hammer line/introduction somehow.
It also means by second half of 2003 AMD could not only lead the
IPC/performance league, but even the clock frequency league again, if not
sooner. They will *utterly* *crush* everything else. Guess what, Microsoft
is in on the party. And we have to grudgingly admit that they have done
very little wrong in the last couple of years when strategy is concerned,
except maybe for not taking Internet seriously enough for a while.
> Oh dear. You are compairing the f_T (transition value) of a transistor
> with a clockfrequence of a CPU. This comparision is totally irrelevant.
Didn't I say exactly that? A lab bidding war that doesn't mean a flying
fluff to us desktop users for the next 5 years or more.
> But you are correct at one point. If you want speed today you should go
> x86. But dont expect that it will always be like that.
Not just today, *especially* tomorrow. You know AMD engineers had the Intel
jingle in music script printed on the PCB of one of the Hammer mobos, with
the Hammer logo hovering over it, and on another the jingle was already gone? :)
I don't think people see the significance of the maddening race in the
current x86 market. Intel is doing something like a 200 MHz clock increase
every two months now and may even have to step that up, AMD is introducing
a potentially huge *architectural* leap by the end of the year. PPC for the
desktop is done for, period.
And that's why I think going with PPC for Amiga is a very sad, big mistake,
especially considering Apple practically controls the supply anyway.
For gods sake people, this isn't about elegance or idealism, it's about basic
survival; it's like porn just arrived on VHS and Amiga is betting on Betamax! |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 206 of 211 | ANN.lu |
| Posted by Samface on 14-Apr-2002 11:24 GMT | In reply to Comment 205 (aTmosh):
For me; it isn't about wether it's the best processor or not. I mean, Mr.Anykey couldn't care less about the difference. It's about the software market, the AmigaOS simply isn't suited for the x86 market right now, period. End of story.
Flames >NIL: |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 207 of 211 | ANN.lu |
| Posted by aTmosh on 15-Apr-2002 01:52 GMT | In reply to Comment 206 (Samface):
> For me; it isn't about wether it's the best processor or not. I mean,
> Mr.Anykey couldn't care less about the difference.
Mr. Anykey uses Windows and has never heard of an Ohmega, or possibly even
a Macintosh. Another proprietary PPC solution will not change that.
> It's about the software
> market, the AmigaOS simply isn't suited for the x86 market right now,
> period. End of story.
I don't see how there is a major difference between x86 and PPC in terms of
what needs to be done to get AmigaOS to run on them, both need adaptation
of the kernel and a JIT compiler for legacy software or parts of the OS
that have not been ported yet, and a gradual porting of the rest of the
system, plus (new) applications. If the API was kept the same, developers
would only have to do a recompile (or not even that if sufficiently
sophisticated HALs are brought into the picture, this is basically what
DE/AA aims at).
The most important question is which new system would draw more users and
developers, or whether having both would be worse than having only one. The
rest is a matter of business models and marketing.
That said, the balance of things is not easily defined and will depend a
lot on where you're standing, but it's important to keep in mind that one
of the most important variables is the CPU, the chipset(s) it needs and
speed/price/availability for these. The fact of the matter is that for this
particular variable the scales have been tipping towards x86 in an
unstoppable manner for a long time now.
Leaving x86 out of the roadmap of a desktop OS today that practically
starts anew is simply folly, PPC-only is a much worse idea now than it was
in 1995 or 1997 and it looks like it's only going to get worse.
I have nothing against a PPC AmigaOS, provided that it's not *PPC-only*. |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 208 of 211 | ANN.lu |
| Posted by Anonymous on 15-Apr-2002 05:09 GMT | In reply to Comment 207 (aTmosh):
Well, appearantly the "Flames >NIL:" commandline didn't work so...
>> It's about the software
>> market, the AmigaOS simply isn't suited for the x86 market right now,
>> period. End of story.
>
>I don't see how there is a major difference between x86 and PPC in terms of
>what needs to be done to get AmigaOS to run on them, both need adaptation
>of the kernel and a JIT compiler for legacy software or parts of the OS
>that have not been ported yet, and a gradual porting of the rest of the
>system, plus (new) applications. If the API was kept the same, developers
>would only have to do a recompile (or not even that if sufficiently
>sophisticated HALs are brought into the picture, this is basically what
>DE/AA aims at).
Look how much you've just written completely unnecessarily. Was I talking about wether it was possible to port the AmigaOS for the x86 or not? Nope. I was talking about the x86 market, the WinLin world, you know.
Please read my earlier post #130 in this thread. |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 209 of 211 | ANN.lu |
| Posted by Samface on 15-Apr-2002 05:10 GMT | In reply to Comment 208 (Anonymous):
That post was by me, btw. :-) |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 210 of 211 | ANN.lu |
| Posted by Wittgenstein on 15-Apr-2002 07:07 GMT | In reply to Comment 205 (aTmosh):
>> This will make the die much larger
>Nope.
[snip]
>> and it will also increase the power consumption
>Nope,
[snip]
uhh, ofcourse it will be larger with the same kind of manufacturing process. You cant compare apples and oranges. There are no free lunches. And ofcourse they will always come up with a better manufacturing processes otherwise it would be impossible to realize it.
>> This will also make the CPU much more
>> difficult to produce in large quanteties so there is no way they can
>> compete with the real 32 bits cpus in price per performance when runing 32
>> bits code.
[Lots of speculation and number]
>They will *utterly* *crush* everything else.
[More speculation and numbers]
Do you really know what you are talkning about? Do you have any clue about semiconductors at all or is this some kind of AMD-geek rambling? Do you really think that AMD have made all this progess themselfes? Ofcourse not. 95% of this are not AMD's results. It is results from the ppl from the research institutes and the research divisions from universities and this knowledge is certainly not only available for AMD. You are very naive when you think that AMD is in some kind of superior position compared to Intel, IBM and MOT. AMD doesnt have any new superduper technology that they discovered themself. I recommend a basic course in solid state physics.
I have seen this kind of hype in many years now. Always the same. 'AMD will *utterly* *crush* everything else'. Guess what. They have almost the same kind of technology and it will always be like that.
The hammer might be a success but it will not crush everything. You mentions that is will 'only' have 63% more pins, but the intressting isnt the number of pins. The intressting is the complexity of the wiring. This is the main problem when you increase from 32 bit to 64 bit and they can not predict how big this problem will be. They only have simulations which are very inaccurate. But lets stop this speculation right now. Only time will tell.
>And that's why I think going with PPC for Amiga is a very sad, big mistake,
>especially considering Apple practically controls the supply anyway.
This wasnt a choice. This was the only alternetive at that time. Now, when Amithlon has appeard you can speculate in a future for amiga runing 95% of the OS emulated. This might be a future, but why not focus in the path that was laid many years ago. Amiga cant change that and still have the main part of the community on there side.
/ |
|
| OSNews.com : Dear Amiga Inc., Could I Make a Few Suggestions? : Comment 211 of 211 | ANN.lu |
| Posted by aTmosh on 16-Apr-2002 02:51 GMT | In reply to Comment 210 (Wittgenstein):
> [More speculation and numbers]
Eh? You were speculating about G5, I responded with numbers. I'll disregard
the personal attacks, you are speculating in general terms while I respond
with facts and numbers. Do your homework instead.
> Do you really think that AMD have made all this progess themselfes?
Did I say they did? Go to http://164.195.100.11/netahtml/search-bool.html
(US Patent database), select Term 2: microprocessor, Field 2: abstract,
Field 1: assignee and rotate Term 1 through Intel, Advanced Micro Devices,
International Business Machines and Motorola.
The results: Motorola 60, IBM 158, Intel 191, AMD 299 patents. I think
these numbers seem to indicate AMD at least have done a lot of this
themselves and have an excellent cross licensing bargaining position if
nothing else.
> 95% of this are not AMD's results.
I trust you can substantiate that? Of course not. And would it matter? No.
They have access to the technology, while others will either license it
from them, cross license or whatever. The fact that Intel is working on
x86-64 as a backup plan after investing enourmous amounts and time in IA-64
says something about AMD's chances in this field.
> It is results from the ppl from the
> research institutes and the research divisions from universities and this
> knowledge is certainly not only available for AMD.
I never said it is only available to AMD. Of course there is a lot of cross
licensing, independent research and deals going on, but don't tell me AMD
have done nothing. The HyperTransport group includes Apple and HP (IA-64
partner) too for example, so these technologies will eventually be used by
them too. The point is that AMD have a major headstart as far as desktop
processor technology and its real world implementation is concerned, they
have working prototypes and chipsets already, while we have yet to hear
from the others.
> You are very naive when you think that AMD is in some kind of superior
> position compared to Intel, IBM and MOT.
As far as the desktop is concerned, they are in the best position right
now. Intel have extremely deep pockets, but they are at a major
disadvantage as far as costs are concerned (die size, plant upkeep), their
strategy is to command a price premium and throw money at it in a brute
force manner. They have a lot of fully owned plants that need to be
constantly modernized which costs a lot of money, while AMD is outsourcing
a lot of the production and cooperating with others. AMD have invested a
lot of time and research in process technology too in which they are
practically on the bleeding edge, and that will pay off.
Here's some interesting reading: http://www.upside.com/texis/mvm/story?id=3c98da541
> AMD doesnt have any new superduper technology that they
> discovered themself. I recommend a basic course in solid state physics.
I didn't say they have superior alien tech or whatever, but they have the
right strategy, implementation and a headstart to boot. With HyperTransport
they have a major advantage, until others start using it too.
> I have seen this kind of hype in many years now. Always the same.
Uh, yeah, like PPC will crush everything. Uhhuh. Don't take my word for it.
The fact of the matter is that IBM isn't competing in this field and
Motorola can't keep up.
> The hammer might be a success but it will not crush everything.
Fine, it will crush just PPC unless miracles happen.
> You mentions that is will 'only' have 63% more pins, but the intressting
> isnt the number of pins.
It seemed to be until you discovered it will not have over twice the amount
of pins. But I agree, the number of pins is not that relevant, rather the
amount of traces you need on the motherboard. And that's the forte of
HyperTransport, it cuts down on the amount of traces and complexity
drastically. Of course there are similar technologies coming by others (and
AMD) like Intel's 3GIO, IBM's RapidIO etc. but AMD have a headstart, and
that's significant. I don't know of any major desktop oriented 32 (or 64)
bit CPU integrating the memory controller or using HT for interprocessor
communication either.
> The intressting is the complexity of the wiring.
> This is the main problem when you increase from 32 bit to 64 bit and they
> can not predict how big this problem will be.
See above, wiring complexity will be drastically *lowered*. The real
challenge will be working out memory sharing strategies in multi CPU
environments, when you have 8 of them with each their own dedicated memory
controller, either on OS level or application software will possibly have
to be adapted to avoid extra latencies (i.e. a CPU accessing memory tied to
another). But even then, with the enourmous *system wide* bandwidth
increase compared even to intermediate advances like MuTiol, V-Link or
nVidia's channel bundling stuff (none of which are system wide), HT is
going to rule. And a single or even double CPU desktop (ClawHammer) isn't
affected, all you get is much more bandwidth, much less complexity, on
a cheap standard 4 layer motherboard.
>> And that's why I think going with PPC for Amiga is a very sad, big mistake,
>> especially considering Apple practically controls the supply anyway.
> This wasnt a choice. This was the only alternetive at that time.
At that time maybe. But at that time was years ago, and they should know
better by now. Honestly, there has not been a major investment in PPC that
needs to be upheld at all cost, this is a purely business/marketing
decision, PPC gives control over the hardware and therefore extra margins
if you lock the OS into it, see Apple. Also this is petty politics, the
people involved need this route for their own benefit, not for the love of
the community or whatever. The discussion should be about whether that
strategy is better, or even viable in itself (not in the long term IMHO),
not whether x86 is technologically superior or not, because it most
certainly is starting to be. As for the few hardware companies left like
Elbox, Eyetech etc., their survival does not depend on the Amiga market,
or they would have been gone a long time ago. Even Hyperion said the stuff
they do for Amiga is not commercially viable. But locking this into a niche
with few competitors could be. None of which is for the benefit of the user.
> Now, when Amithlon has appeard you can speculate in a future for amiga
> runing 95% of the OS emulated. This might be a future, but why not focus
> in the path that was laid many years ago.
Because that path has been obsoleted by events since then. Amiga have
changed direction so many times it isn't funny. Had the push for PPC been
serious back then, maybe by now it would have established enough support so
we wouldn't need this discussion. However *today* it is the wrong choice.
And the 95% emulated bit doesn't need to stay 95%, it could gradually be
lowered to a lot less (i.e. only legacy software support, OS and new apps
fully native). See AROS.
> Amiga cant change that and still have the main part of the community on there side.
The main part of the community is a few thousand people left, hell lets
exaggerate, maybe 20000, from a business point of view you have to know
this will not be enough. It remains to be seen whether enough new users can
be attracted to sustain the platform. I'd say a fivefold increase to 100k
is the least, but that still isn't enough for attracting serious software
development like office suites, databases etc.
Incidentally, Microsoft's 5 year contract with Apple has just ended, and in
spite of vague promises to continue to support the platform they are
already starting anti UNIX campaigns, continueing the anti Linux/GPL
campaign, while MacOS is heading for UNIX with MacOS X, and touting the
UNIX side. My guess is as the various cases against M$ falter, this support
will slowly fade. |
|
|
- User Menu
-
- About ANN archives
- The ANN archives is powered by #AmigaZeux. It was updated daily (news last: 22-Oct-2004; comments last: 18-May-2005).
ANN.lu was created, previously owned and maintained by Christian Kemp, www.ckemp.com.
- Contribute
- Not possible at this time!
- Search ANN archives
- Advanced search
- Hosting
- ANN.lu was hosted by Dreamhost. Sign up through this link, mention "ckemp" as referrer and he will get a 10% commission on any account you purchase.
Please show your appreciation for any past, present and future work on ANN.lu by making a contribution via PayPal.
|
