Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: billyfish on February 03, 2012, 10:56:32 AM
-
Hi all
Since we appear to be going benchmark crazy, even though I think that real world usage is more appropriate, for the sake of completeness, here are the Disk I/O benchmark conducted by Mufa at http://forum.amigaone.pl/topic65.html#p503 and the RageMem http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35053&forum=33&start=40&viewmode=flat&order=0#650254 benchmark conducted by Sam.
I don't (yet!) own either of the machines involved but I think we should show them all regardless of which systems we favour.
Also, big congrats to Trevor Dickinson and all that have made the X1000 a reality, I really can't wait to get one!
(http://amigaone.pl/robocze/kopiowanie.jpg)
@sam
RAGEMEM v0.37 - compiled 11/06/2010
CPU: P.A. Semi PWRficient PA6T-1682M B1 @ 1800 Mhz
Caches Sizes: L1: 64 KB - L2: 2048 KB - L3: none
Cache Line: 64
CPU
MAX MIPS: 3084
L1
READ32: 6851 MB/Sec
READ64: 13682 MB/Sec
WRITE32: 6851 MB/Sec
WRITE64: 13681 MB/Sec
L2
READ32: 3276 MB/Sec
READ64: 4784 MB/Sec
WRITE32: 2531 MB/Sec
WRITE64: 4090 MB/Sec
RAM
READ32: 2857 MB/Sec
READ64: 4000 MB/Sec
WRITE32: 2732 MB/Sec
WRITE64: 3383 MB/Sec
WRITE: 352 MB/Sec (Tricky)
VIDEO BUS
READ: 15 MB/Sec
WRITE: 161 MB/Sec
-
Considering how skewed (http://www.amiga.org/forums/showthread.php?t=60604) the initial Mufa lame bench was I would take any results from him with a grain of salt. The least there should be a clear specification how the test was performed in order to be able to repeat it.
In memory bandwidth X1000 clearly is superior though.
Here's a MPlayer decoding benchmark where X1000 could really show its power:
http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=35053&forum=33&start=240&viewmode=flat&order=0#650877
-
I don't see why we are revisiting the lame benchmarks as they are lame in every sense of the word. Surely we can get real world tests now that it's a product not a prototype with NDAs.
-
I don't see why we are revisiting the lame benchmarks as they are lame in every sense of the word. Surely we can get real world tests now that it's a product not a prototype with NDAs.
I'm waiting for Linux benchmarks coming from the A1X1K. Say running BF2 via WINE should do a nice stress test. Or would that be too much for the PA6T to even attempt?
-
I'm waiting for Linux benchmarks coming from the A1X1K. Say running BF2 via WINE should do a nice stress test. Or would that be too much for the PA6T to even attempt?
I'd like to see some stop-watch tests for normal tasks. Also for any benchmark it should be noted whether or not HIGHER is better or LOWER is better, otherwise they are useless.
-
Also for any benchmark it should be noted whether or not HIGHER is better or LOWER is better
best case they get head in head.
up till now, real life benchs show x1k=mac g4 on slightly lower clock.
-
I don't see why we are revisiting the lame benchmarks as they are lame in every sense of the word. Surely we can get real world tests now that it's a product not a prototype with NDAs.
Even though I started this thread, I agree 100%.
best case they get head in head.
up till now, real life benchs show x1k=mac g4 on slightly lower clock.
Sigh, not according to those my first email in this very thread.
Anyhooo, given that Hyperion have stated that the OS for the X1000 is currently unoptimised, all of these benchmarks are not really worth the virtual paper that they're written on. I'm much more interested in how people are finding them in every day use.
-
@billyfish:
but these are calculation benchmarks im talking about, what can be done about the os to make them run faster on a given hardware?? i mean except it is seriously busy-looping in the background at all times.
I'm much more interested in how people are finding them in every day use.
and then you post a highly artificial benchmark result to start the thread? learn to live with objective results you get. i hope you will find real life benchmarks to your liking even if it might be difficult considering the current performance to the gfx card and lack of 3d support?
-
I'd like to see some audio latency times.
-
i'm waiting for linux benchmarks coming from the a1x1k. Say running bf2 via wine should do a nice stress test. Or would that be too much for the pa6t to even attempt?
wine?
-
I'm waiting for Linux benchmarks coming from the A1X1K. Say running BF2 via WINE should do a nice stress test. Or would that be too much for the PA6T to even attempt?
and theres me thinking wine is not an emulator ... ;)
-
Hi all
Since we appear to be going benchmark crazy, even though I think that real world usage is more appropriate, for the sake of completeness, here are the Disk I/O benchmark conducted by Mufa at http://forum.amigaone.pl/topic65.html#p503 and the RageMem http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=35053&forum=33&start=40&viewmode=flat&order=0#650254 benchmark conducted by Sam.
I don't (yet!) own either of the machines involved but I think we should show them all regardless of which systems we favour.
Also, big congrats to Trevor Dickinson and all that have made the X1000 a reality, I really can't wait to get one!
(http://amigaone.pl/robocze/kopiowanie.jpg)
I've always found those old results weird. I don't know what is wrong with the said system (maybe highly fragmented SFS partition?), but these are the results I get:
(http://sintonen.fi/pics/copy_benchmark.png)
X1000 is of course still faster (as would likely be Sam 440/460 as well).
It should be noted that many things can affect such benchmark, such as the filesystem being used (copy may return immediately if delayed writing is applied by the filesystem). Another significant factor is the physical location of the partitions, are they located on the same or different HDDs? And finally, rotating HDDs are clearly slower than SSD.
Notes: The PowerBook has a newer Western Digital Scorpio Blue WD2500BEVE 250GB 5400 RPM 8MB Cache 2.5" HDD, while the Mac mini has the original Apple branded Seagate Momentus 5400.2 ST9808211A 80GB HDD).
-
I'm waiting for Linux benchmarks coming from the A1X1K. Say running BF2 via WINE should do a nice stress test. Or would that be too much for the PA6T to even attempt?
:roflmao: Wine? On linux PPC? QEmu, perhaps. Wine Is Not an Emulator.
-
:roflmao: Wine? On linux PPC? QEmu, perhaps. Wine Is Not an Emulator.
I knew it was only a matter of time before someone mentioned DosBox or "virtualisation". You people make me sick! :p
In all seriousness you don't really think people would spend all that money just to emulate Windoes do you? What a waste. Please, get that Windoes **** off here :rant:
-
I can't decide...
Should I buy 10 mac mini G4 computers with MorphOS licences (assuming I can get a mini with 512K and Morphos license for $300) or should I purchase one X1000?
What would be faster?
-
In all seriousness you don't really think people would spend all that money just to emulate Windoes do you? What a waste. Please, get that Windoes **** off here :rant:
Well, the benchmark discussions are pretty interesting. On the one hand, you have people pointing out that G4 class macs are at least as powerful (on the basis of tests so far) and others concluding that due to this, the X1000 is clearly a failure.
Label me a red-camp fanboy for saying this if you must, but I'm not sure I see it that way. Sure, it's an expensive system, no argument. If you consider it to be an over-priced yesterday's news machine, then don't buy it. You can pick up cheap PC kit and pay nothing for a high-quality, robust OS, like linux. Simple, really.
However, it's also a new system, designed and built from scratch that was (by those slamming it now) dismissed as pure vapour since the day it was announced. Now it's here and based on some quick benchmarks we can see that it's on par with old apple hardware and apparently that's really bad. Personally, I disagree. For such a project, going from "sheer vapourware that'll never appear" to "at least as good as Apple's G4-class PPC" is a pretty decent result. Something about saying that seems like deja vu. Weird.
I'm also not convinced we've really seen what the hardware is capable of yet. We know that OS4 certainly isn't making the most of it right now, being restricted to 32-bit operation on a single core. Even 64-bit PPC linux will have limitations if you use a modern graphics card with it as you are unlikely to get a vendor-supplied driver for PPC (unlike OSX) and will have to rely on whatever open source alternatives there are.
-
Well, the benchmark discussions are pretty interesting. On the one hand, you have people pointing out that G4 class macs are at least as powerful (on the basis of tests so far) and others concluding that due to this, the X1000 is clearly a failure.
Label me a red-camp fanboy for saying this if you must, but I'm not sure I see it that way. Sure, it's an expensive system, no argument. If you consider it to be an over-priced yesterday's news machine, then don't buy it. You can pick up cheap PC kit and pay nothing for a high-quality, robust OS, like linux. Simple, really.
However, it's also a new system, designed and built from scratch that was (by those slamming it now) dismissed as pure vapour since the day it was announced. Now it's here and based on some quick benchmarks we can see that it's on par with old apple hardware and apparently that's really bad. Personally, I disagree. For such a project, going from "sheer vapourware that'll never appear" to "at least as good as Apple's G4-class PPC" is a pretty decent result. Something about saying that seems like deja vu. Weird.
I'm also not convinced we've really seen what the hardware is capable of yet. We know that OS4 certainly isn't making the most of it right now, being restricted to 32-bit operation on a single core. Even 64-bit PPC linux will have limitations if you use a modern graphics card with it as you are unlikely to get a vendor-supplied driver for PPC (unlike OSX) and will have to rely on whatever open source alternatives there are.
You red-camp fanboy! I hope that curry does the flaming sword dance in your lower intestine! ;)
-
You red-camp fanboy! I hope that curry does the flaming sword dance in your lower intestine! ;)
How did you know I had curry today? :lol:
-edit-
Sometimes I wonder what the deal with the ongoing camp arguments is. IMHO, people genuinely interested in getting an X1000 aren't going to be swayed by any of these benchmarks. If price/performance was your main motivator for your choice of "next gen" Amiga, you'd opt for AROS, end of story. There's absolutely no way you can compete with "costs nothing" running on commodity hardware that's faster than anything the combined purples can come up with. The traditional "blue" argument against it used to be that AROS is woefully far behind, but that's simply not true any more. In fact, it's OS4 and MOS that are lagging behind AROS in some areas now (3D springs to mind immediately).
I struggle to see that there are really any "would-be NG" Amiga users left these days that aren't fully aware of the main options on the table and what they each offer. We're not arguing to convince them. We're just doing it because we can :)
-
How did you know I had curry today? :lol:
The owls are not what they seem!
Nah, you mentioned it in another post. Man, now I want some curry!
-
@Karlos. I agree with everything you just said :)
My stance on running DosBox is an ethical one and think it has no meaningful place on Amiga's. But it may be useful here in making meaningful benchmarks possible. I think it is funny how people are putting the x1000 down because it has only just been released and not even optimized yet. It is still in beta as the software that has been made for it has yet to be released. With this in mind it could be a while before it reaches it's full potential.
There's is nothing wrong with being a OS4.x fanboy. Although i have never used OS4.x i can see the appeal as it seems to recreate the classic 3.x workbenches whereas Morphos, as much as i like it, still feels like an Apple somehow and I'm not talking about the hardware. I have a love hate relationship with both the blue and red camps :)
It sounds to me you are thinking of getting an x1000?
-
The owls are not what they seem!
:lol:
There's a fake plastic one on the roof of a building opposite the office where I work. I thought it was to scare off feral pigeons. You can imagine my amusement when I noticed today that it's rather caked in bird crap. As if feral Manchester pigeons have ever seen an owl.
Nah, you mentioned it in another post. Man, now I want some curry!
Oh yeah, forgot about that. I had lamb too, albeit not the kind I mentioned (which was actually a clue to the thread tag thing). It was good stuff though :D
-
I think it would be interesting to run some sort of 68K based benchmark.
I could provide results for
UAE on a i7 740QM which is a 1.7 (i think)
MorphOS 2.6 on 1.42 mini
A1200/060/FastATA IV
if so, what benchmark application?
-
@Karlos
You have good points there.
However, it has to be pointed out that the obviously incorrect initial benchmark results were readily accepted as facts by many. This I believe shows the amount of unrealistic expectations some might have had of X1000.
-
I'm also not convinced we've really seen what the hardware is capable of yet. We know that OS4 certainly isn't making the most of it right now, being restricted to 32-bit operation on a single core. Even 64-bit PPC linux will have limitations if you use a modern graphics card with it as you are unlikely to get a vendor-supplied driver for PPC (unlike OSX) and will have to rely on whatever open source alternatives there are.
apart from multicore support i dont see how 64bit operation would speed up anything. in fact my 64bit version of lightwave9 was marhinally slower than a 32bit on my then 64bit capable system. 64bit operation enabled amounts of memory are anyway marginally useful under amiga-like applications. so thats imho not a point.
-
apart from multicore support i dont see how 64bit operation would speed up anything. in fact my 64bit version of lightwave9 was marhinally slower than a 32bit on my then 64bit capable system. 64bit operation enabled amounts of memory are anyway marginally useful under amiga-like applications. so thats imho not a point.
I think it depends a great deal on the CPU architecture.
For example, the original netburst P4 and the Itanic took a huge performance hit running legacy code.
-
@Karlos
You have good points there.
However, it has to be pointed out that the obviously incorrect initial benchmark results were readily accepted as facts by many. This I believe shows the amount of unrealistic expectations some might have had of X1000.
Indeed he does. I never had any of these 'unrealistic expectations' about the X1000.
What I did know, is that this 'custom built' computer was led and funded by an enthusiast (and a very decent guy) and was everything I wanted to buy, regardless of 'value for money'. I'm sick of the money 'and what it could buy' argument.
I was a HUGE fan of MorphOS back in the day, and no 'camp' meant anything. Alas, things changed... and so did my opinion.
The X1000 is here through dedication and hard work, as is the OS that runs on it. The users that choose to use it, are here to stay. I just wish the ambience between 'choices' could be so much more amiable.
I admit to being totally put off by some of the attitudes shown around here. Nevermind. :) 'Let's keep this party going' :D
Cheers.
Steve
-
apart from multicore support i dont see how 64bit operation would speed up anything. in fact my 64bit version of lightwave9 was marhinally slower than a 32bit on my then 64bit capable system. 64bit operation enabled amounts of memory are anyway marginally useful under amiga-like applications. so thats imho not a point.
That depends. I do have 32-bit compiled code for my Core2 that's faster than the same code compiled 64-bit native (as well as the sort of contrary examples I was expecting). However, as mentioned above by bbond007, whether any given system is faster or not in 64-bit mode ultimately comes down to how that chip is implemented and what features may or may not be available in a given operation mode. On thing 32-bit has going for it over 64-it generally is that pointers take up half the memory and thus half the amount of cache. That could make a difference to the performance of various inner loops of code where lots of pointer operations are used.
-
@Karlos
You have good points there.
However, it has to be pointed out that the obviously incorrect initial benchmark results were readily accepted as facts by many. This I believe shows the amount of unrealistic expectations some might have had of X1000.
The PA6T is a OOO PPC with a 2MB cache and a memory interface much better than any G5. It was designed by some of the best people in the industry so it should be competitive with a G5 at the same clock rate, if not faster.
There are any number of reasons the tests are coming out as they are. One pretty big possibility is it's the first release of the HW/OS and it's not been properly set up yet. The memory benchmarks might indicate this - they are good but they should be far higher.
Another problem is it's not always clear what people are testing against what. Tests should at least be repeatable. OTOH running different binaries on different OSs is a pretty good way of *not* testing a processor.
It'd certainly be interesting to see some Linux benchmarks. These should do:
http://www.phoronix-test-suite.com/
http://www.primatelabs.ca/geekbench/
BTW I've recently seen some some benchmarks of a Pandaboard which came out pretty badly. A couple of kernel changes later and the results are wildly different. All but one of the test was over 50% faster.
-
Can the PandaBoard now use all its memory w/o crashing?
As I've had conversations with the folks at Varisys in the past (before the connection to Aeon was announced), it hardly surprises me that the hardware works (well).
PA6T faster then the G5? Not really.
The original intent was to have a processor that would draw less power while being more powerful then the G4 (As G5s were not practical for laptops).
Camps? That's fiction. Some of us hold strong opinions, others may retain some hard feelings, and still others continue to strive to keep an open mind.
We're not two camps, but individuals with a broad spectrum of opinions.
As a MorphOS user, I've never felt any ill will toward this project.
And I like Treavor.
Ben has made some dicky statements though.
And, no, it doesn't surprise me that at comparable speeds the PA6T and the G4 are not that far apart in benchmarks.
That was predicted quite a long time ago when the specs for the X1000 were first announced.
-
There's absolutely no way you can compete with "costs nothing" running on commodity hardware that's faster than anything the combined purples can come up with.
Is X1000 sold at 3000$ your strategy to compete against AROS? :-)
This X1000 project has big "dead end" written in the wall. The computer built in 2012 and sold at $3000 is only marginally better than used Macs from yesteryears. The platform has stagnated. This is the final stop. Last passengers will be served and then it is over.
It is time to admit PPC is finished. And when you have done that start enjoying old Macs, old Amigas, old/new AmigaOnes, SAMs, or whatever, again. The train is still here. There just aren't tracks anymore.
The traditional "blue" argument against it used to be that AROS is woefully far behind, but that's simply not true any more. In fact, it's OS4 and MOS that are lagging behind AROS in some areas now (3D springs to mind immediately).
Hmm... I think you are confused. MorphOS never had problem with AROS and they have shared their work in the past. It is just that AROS not having binary compatibility and not so developed it has been dismissed by both MorphOS and OS4 users.
AROS is a research operating system interesting in its own way.
-
I'm also not convinced we've really seen what the hardware is capable of yet. We know that OS4 certainly isn't making the most of it right now, being restricted to 32-bit operation on a single core. Even 64-bit PPC linux will have limitations if you use a modern graphics card with it as you are unlikely to get a vendor-supplied driver for PPC (unlike OSX) and will have to rely on whatever open source alternatives there are.
Luckily OS4 don't have to rely on whatever open source alternatives there are ;)
-
It is just that AROS not having binary compatibility and not so developed it has been dismissed by both MorphOS and OS4 users.
AROS is a research operating system interesting in its own way.
I'm glad you think AROS is interesting in its own way, but this 'binary compatibility' issue that is a common complaint should really be examined more deeply.
All three modern Amiga-like platforms rely on emulation for 68k Amiga software, when they don't run on 68k CPUs (i.e. excluding 68k AROS). The only thing is that the emulation in AROS is different from the emulation in OS4 and MorphOS. The main differences are:
1. OS4 and MorphOS have a JIT for 68k software, whereas AROS doesn't yet.
2. OS4 and MorphOS don't have to swap the byte order of 68k code, whereas AROS does (not including AROS PPC and AROS 68k).
That's it. A JIT is a feature I concede is very nice to have, but it's the second issue I mention that I think is the cause of the confusion. Not swapping byte order != binary compatibility. What OS4 and MorphOS have is faster emulation, not no emulation.
Hope we can move beyond this 'binary compatibility' fallacy, and see the platforms as they really are. Thanks for reading.
-
All three modern Amiga-like platforms rely on emulation for 68k Amiga software, when they don't run on 68k CPUs (i.e. excluding 68k AROS). The only thing is that the emulation in AROS is different from the emulation in OS4 and MorphOS. The main differences are:
Sorry, I didn't know AROS was binary compatible with 68k software, I havent looked at aros-exec nor developer mailing list for a while. Is there a distro I could try in Virtual PC? I'd love to try some 68k software for benchmarking purposes. SysSpeed, P96Speed, AIBB...
There is also some 68k software I can't run on MorphOS but perhaps it works better in AROS.
1. OS4 and MorphOS have a JIT for 68k software, whereas AROS doesn't yet.
2. OS4 and MorphOS don't have to swap the byte order of 68k code, whereas AROS does (not including AROS PPC and AROS 68k).
I don't think swapping byte order is relevant (see Amithlon). Just like JIT is not so relevant anymore when you have got 1GHz+ speeds.
(Though, admittely, not having 68k binary compatibility was greater disadvantage in last decade than it is now...)
-
Sorry, I didn't know AROS was binary compatible with 68k software, I havent looked at aros-exec nor developer mailing list for a while. Is there a distro I could try in Virtual PC? I'd love to try some 68k software for benchmarking purposes. SysSpeed, P96Speed, AIBB...
Fair enough.
In that case, you've got a few options. If you'd like to try an x86 AROS distro, then Icaros is probably the best place to start. Of course running it in a virtual machine will always incur a speed penalty compared to running it natively, but I appreciate virtual machines are useful when you want to give a new OS a try with minimal fuss. Worth noting that you can run AROS from the CD/DVD without needing to install it (though again this offers slower performance than native). You can download Icaros from here:
http://www.icarosdesktop.org/
There are some nice features coming up for AROS that help with using it as a day to day system, for example this week we got the first public release of the AROS NTFS driver, it's buggy at the moment, but it's worth noting now as it'll enhance using AROS on the same computer as other OS, as you'll be able to share files with other OS you have installed.
Also worth mentioning that you can run AROS 68k in UAE (WinUAE probably works best, as Toni works on both). It's not perfect, but its development has been stunningly fast (all done in about 18 months) and improves pretty much daily. This is a useful thread to read to bring you up to speed with the state of play for AROS 68k:
http://eab.abime.net/showthread.php?t=56211
HTH. Please feel free to ask any further questions you have.
EDIT: Oh and if you'd like to keep up to date with progress on AROS, but don't have time to visit AROS-Exec regularly, the next best thing would be to visit Planet AROS from time to time, which is a blog aggregator, mostly focused on AROS:
http://planet-aros.heimstetten.net/
Not many people seem to know about it, probably because it hasn't got an easy to remember URL, but it is handy to know about. If you forget the URL, just search for Planet AROS in Google, it should be the top result.
-
Is X1000 sold at 3000$ your strategy to compete against AROS? :-)
It's not my strategy for anything. The whole reason this point has come up is the reiteration of the point that MOS runs on much cheaper hardware that so far seems to turn in equally good performance and that this is some sort of decisive factor in the X1000 being a failure; "I could buy 10 macs for that price and they'd each be as fast" may be a valid observation, but *if* performance versus cost was your only motivation you'd use AROS.
Neither OS4 or MOS can hold a candle to AROS on that front. It's free and it runs on commodity hardware as cheap or as expensive as you care to obtain. That's always been the case, but as you point out above, "red camp" trolls were always quick to point out that it lacked official sanction and that PPC was teh 1 true successor and so on. Likewise, despite your pleas to the contrary, "blue camp" trolls were always keen to point out that AROS was basically a hobby project not-quite-going-anywhere and certainly light years behind the level of maturity and compatibility of MOS and is thus not a serious option for the discerning Amiga users. I find it really hard to believe you've never come across the latter sentiment in your forum travels.
This X1000 project has big "dead end" written in the wall. The computer built in 2012 and sold at $3000 is only marginally better than used Macs from yesteryears.
As I said, considering many were insistent it would never arrive at any price, the fact it exists at all and is at least as good as a commercial machine that basically help put apple back on the map is quite an achievement, IMHO. It may very well be the last PPC desktop machine ever built, but that doesn't detract from the fact that it was realized.
It is time to admit PPC is finished. And when you have done that start enjoying old Macs, old Amigas, old/new AmigaOnes, SAMs, or whatever, again. The train is still here. There just aren't tracks anymore.
Erm, I currently have 2 PPC machines. One is a BlizzardPPC the other an A1. I'm not under any illusions about where they sit ;)
Hmm... I think you are confused. MorphOS never had problem with AROS and they have shared their work in the past.
Not at all. The exchange of code between developers of the two has never had any bearing on the opinions of camp trolls. This just about summed it up perfectly: http://www.amiga.org/gallery/index.php?n=3377
It is just that AROS not having binary compatibility and not so developed it has been dismissed by both MorphOS and OS4 users.
AROS has 68K binary compatibility. Janus UAE running in "coherence" mode has been around for a little while now. You may have to set it up initially (I had to, but I haven't tried more recent releases), but what do you want for free?
-
Janus UAE running in "coherence" mode has been around for a little while now. You may have to set it up initially (I had to, but I haven't tried more recent releases), but what do you want for free?
Actually, since late last year, Janus-UAE doesn't require any initial setup, it comes working out of the box, as it is shipped with AROS 68k. See here:
http://o1i.blogspot.com/
Also, Icaros has AmiBridge, which lets you launch 68k programs by double clicking on the program icon, and with Janus-UAE coherence mode the 68k programs can interact with x86 programs (sharing clipboard contents, for example).
-
1. OS4 and MorphOS have a JIT for 68k software, whereas AROS doesn't yet.
2. OS4 and MorphOS don't have to swap the byte order of 68k code, whereas AROS does (not including AROS PPC and AROS 68k).
You've forgot about:
3. OS4 and MorphOS emulation allows 68k applications to use native system components and vice versa.
-
You've forgot about:
3. OS4 and MorphOS emulation allows 68k applications to use native system components and vice versa.
Granted. Tell me, are there any 68k libraries in regular use on OS4 or MorphOS that don't have native equivalents? The Emumiga project should allow AROS to mix 68k and native libraries, and it'll be good to have some idea of the benefits this could bring.
-
@Karlos
You have good points there.
Thanks. It can be hard work trying to remain impartial, the old "red v blue" stuff is so tiring after all these years.
However, it has to be pointed out that the obviously incorrect initial benchmark results were readily accepted as facts by many. This I believe shows the amount of unrealistic expectations some might have had of X1000.
That's true too. I've mostly adopted a "wait and see" approach to this whole thing. Some people have assumed the PA6T is necessarily a G5 slayer and thus ready to stomp all other PPC machines, without considering that the PA6T was designed for a different market. Where it probably will smash the G5 is in performance per watt, but all considered, that's not saying much.
One thing that has come out of this, at least for me, is the realisation that the old PPC machines perhaps aren't quite as slothful for everyday use as many assume. At least if the lame benchmarks were any indication of single core performance.
-
It's not my strategy for anything.
But you work for Hyperion to improve OS4, don't you?
The whole reason this point has come up is the reiteration of the point that MOS runs on much cheaper hardware that so far seems to turn in equally good performance and that this is some sort of decisive factor in the X1000 being a failure; "I could buy 10 macs for that price and they'd each be as fast" may be a valid observation, but *if* performance versus cost was your only motivation you'd use AROS.
Indeed. MorphOS can stay alive only by improving its software. The operating system must go forward and key software must advance respectively. Everyone who have used MorphOS 2 recognizes it again when it is discussed or spotted somewhere.
Likewise, despite your pleas to the contrary, "blue camp" trolls were always keen to point out that AROS was basically a hobby project not-quite-going-anywhere and certainly light years behind the level of maturity and compatibility of MOS and is thus not a serious option for the discerning Amiga users. I find it really hard to believe you've never come across the latter sentiment in your forum travels.
AROS is a hobby project not going anywhere and is light years behind. There is no doubt about it. AROS is completely different cup of tea. It provides source compatible platform for Amiga users who want Amiga on their x86. Its strength is not in polished UI or software, its strength is in open source and x86.
As I said, considering many were insistent it would never arrive at any price, the fact it exists at all and is at least as good as a commercial machine that basically help put apple back on the map is quite an achievement. It may very well be the last PPC desktop machine ever built, but that doesn't detract from the fact that it was realized.
Nice, they did it, we can move forward.
Not at all. The exchange of code between developers of the two has never had any bearing on the opinions of camp trolls. This just about summed it up perfectly: http://www.amiga.org/gallery/index.php?n=3377
I have only met AROS developers so far :)
AROS has 68K binary compatibility. Janus UAE running in "coherence" mode has been around for a little while now. You may have to set it up initially (I had to, but I haven't tried more recent releases), but what do you want for free?
Janus UAE again? Ok, let's see if my old 68k MUI applications run under AROS Zune.
-
Granted. Tell me, are there any 68k libraries in regular use on OS4 or MorphOS that don't have native equivalents? The Emumiga project should allow AROS to mix 68k and native libraries, and it'll be good to have some idea of the benefits this could bring.
I don't know why it should matter because 68k MUI class can be embedded in applications.
-
AROS is a hobby project not going anywhere and is light years behind. There is no doubt about it.
For someone who a couple of posts ago admitted they'd barely paid any attention to AROS for a while, it seems odd for you to be so sure about this. Perhaps it would be better for you to take a fresh look at AROS to see where it's at now.
I don't know why it should matter because 68k MUI class can be embedded in applications.
Krashan thought it enough of a benefit to mention it, so I'd like to know how much this feature gets used.
-
But you work for Hyperion to improve OS4, don't you?
That depends on how you define work. Work usually implies some sort of formal working contract and regular/pre-arranged payment. On that basis, no, I don't work for Hyperion*. I have, however, volunteered time to work on areas of OS4, principally for the classic version. And yes, that requires me to delve into the OS code here and there and so yes, I have an NDA.
I'll continue to work on it too, at least as long as I have a working machine. I'd like to do more, but my day job and other responsibilities keep me occupied.
*I suppose obtaining the latest classic release as well as a fully working BVision to replace my malfunctioning one could be considered payment, but I see those as the minimum necessary tools for the task at hand.
-
For someone who a couple of posts ago admitted they'd barely paid any attention to AROS for a while, it seems odd for you to be so sure about this. Perhaps it would be better for you to take a fresh look at AROS to see where it's at now.
I just did and I am not sure if you wish to have my comments. First of all browsing files with Wanderer is pain and it is much better browse files using a dock. Running SysMon results in crash. I can't change theme without restarting computer. Even changing screenmode can't be done without closing windows. Is Restart AROS in Shutdown menu (which doesnt have Shutdown btw) supposed to work? It would be also nice if Intuition settings had click to front option. The default Zune (MUI) theme is stuck in year 1993.
It is not bad. I mean stuff is there and Icaros comes with lot of software. Initial appearance of Icaros was very nice (nice background, preconfigured toolbar) but I find it is difficult to use.
Anyway, you forgot to mention that I have to buy Amiga Forever to run 68k software in AROS natively. Are there instructions how to use AROS/68k in AmiBridge?
-
Tell me, are there any 68k libraries in regular use on OS4 or MorphOS that don't have native equivalents?
As a programmer of DigiBooster 3 I often deal with music modules packed with XPK SQSH or SMPL compressor. There are no MorphOS native (or AmigaOS 4 native AFAIK) versions of them. M68k versions work without any problems.
Another example is ARexx. AmigaOS 4 uses (and MorphOS can use) M68k interpreter.
In fact the opposite possibility is much more interesting. Legacy M68k apps can use native PowerPC libraries, which make them faster and enhance functionality. For example M68k apps use native ASL requesters, which are much more powerful and comfortable compared to AmigaOS 3.x ones.
-
That's true too. I've mostly adopted a "wait and see" approach to this whole thing. Some people have assumed the PA6T is necessarily a G5 slayer and thus ready to stomp all other PPC machines, without considering that the PA6T was designed for a different market. Where it probably will smash the G5 is in performance per watt, but all considered, that's not saying much.
I've never done any testing of this myself, but it looks from benchmarks I've seen like the G5 isn't actually much of an improvement in terms of performance-per-clock - more just that it broke out of the 2GHz barrier Motorola couldn't get the G4 past.
One thing that has come out of this, at least for me, is the realisation that the old PPC machines perhaps aren't quite as slothful for everyday use as many assume. At least if the lame benchmarks were any indication of single core performance.
They really aren't. Memory throughput and disk access can be a limiting factor, depending on the machine, but overall (Flash/fullscreen HD video excepted) they're surprisingly capable. TenFourFox on a 1.25GHz G4 runs noticeably better than the equivalent Firefox version on my 1.6GHz Eee, f'rexample.
The industry just keeps us buying the you-need-more-horsepower myth, that's what I say...
-
@itix
Well, at least you gave it a go. Thank you for that.
@Krashan
Thanks for the info.
-
The industry just keeps us buying the you-need-more-horsepower myth, that's what I say...
Indeed, even though you do need a lot of power for a lot of the crapware out there. I'll keep using my 667 mhz Pentium 3.
-
I realised that the computer I have now (dual core 2.x ghz) will pretty much be fine until it finally breaks down, and have to replaced through physical wear.
Given Im on linux, there's not much in the way of new games (lots of indies though), so keeping up with the race is basically pointless. Will 2 more cores help libreoffice open 0.3 seconds faster?
-
JIT is not so relevant anymore when you have got 1GHz+ speeds.
I can't say that I agree with that.
Anything that helps performance is useful.
-
Speed of the x1000 was a real dissapointment! Im happy i bought 60Euro Mac mini yesterday.
-
Speed of the x1000 was a real dissapointment! Im happy i bought 60Euro Mac mini yesterday.
60 Euros? Wow!
Most European Mac hardware is priced higher then it is in the US.
You got a bargain.
-
I can't say that I agree with that.
Anything that helps performance is useful.
I think what he meant was that even interpretive emulation on 1+ GHz class machines is usually more than capable of running 68K code faster than any real 68K used in an Amiga. However, JIT is always nice to have regardless and becomes essential on slower host systems. I don't think 68K emulation on OS4.1 classic would be much fun on my 603 sans JIT.
-
I think what he meant was that even interpretive emulation on 1+ GHz class machines is usually more than capable of running 68K code faster than any real 68K used in an Amiga. However, JIT is always nice to have regardless and becomes essential on slower host systems. I don't think 68K emulation on OS4.1 classic would be much fun on my 603 sans JIT.
Thanks Karlos,
I guess I misinterpreted that statement.
-
@itix
Thanks for trying AROS again and for reporting problems, however, as I've already told you on AROS-EXEC, you may have missed some important points:
1. patch 01, 03 and 04 for Icaros Desktop 1.3.3 fix most Wanderer problems and make it a completely different beast than 'bare' 1.3.3. You'd install them using LiveUpdater.
2. patch 03 adds AROS M68K Kickstart support to AmiBridge, so it's now possible to play Amiga games and demos on adf files without any Amiga ROM from Commodore, Cloanto, Hyperion or whatever
3. sysmon doesn't work on virtual machines. It shouldn't crash however, but just return weird results. This might be fixed with one of the patches. sysmon works well on physical processors
4. bring to front is a commodity you can enable by default just moving it to WBStartup folder, like on AOS 3.1
5. Please use the Icaros manual to see what you can actually do with Icaros Desktop. Like any other operating system, it hides many little big secrets you might learn to love.
...and, above all...
6. look for supported hardware, and make a Icaros PC with sound, network and 3D acceleration. You'd be surprised how Gallium3D works nice with modern game ports (shaders let you see 3D games like Cube 2, MegaGlest and so, like no other Amigoid system can show you).
-
I just ran into this statement on aw.net:
I won't post any more x1000 benchmarks, OS4.1.5 is not optimized for speed yet on the x1000. Its very stable & thats whats important for the moment. Benchmarks are interesting, but some ppl abuse the results by comparing it to other systems. I won't add fuel to the red/blue war."
- sundown (http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=31896&forum=33&start=260&viewmode=flat&order=0#651426)
Well, okay... (Though the same argument has been used about AmigaOS 4 on Pegasos2 for years now... without much change)
How about someone installs linux, disables the 2nd core (as root: echo 0 > /sys/devices/system/cpu/cpu1/online) and runs the the same benchmarks (lame, blender, mplayer). Then at least we could see if the problem really is in AmigaOS 4.1.5 as is claimed.
-
How about someone installs linux, disables the 2nd core (as root: echo 0 > /sys/devices/system/cpu/cpu1/online) and runs the the same benchmarks (lame, blender, mplayer). Then at least we could see if the problem really is in AmigaOS 4.1.5 as is claimed.
Now there is a dandy idea. :lol:
-
I just ran into this statement on aw.net:
- sundown (http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=31896&forum=33&start=260&viewmode=flat&order=0#651426)
Well, okay... (Though the same argument has been used about AmigaOS 4 on Pegasos2 for years now... without much change)
How about someone installs linux, disables the 2nd core (as root: echo 0 > /sys/devices/system/cpu/cpu1/online) and runs the the same benchmarks (lame, blender, mplayer). Then at least we could see if the problem really is in AmigaOS 4.1.5 as is claimed.
or, as MOS supporters (and at least one very anxious dev) are so eager to test the X1000 here, why not make a MOS port for the X1000 with duo core support please and thank you ;)
-
or, as MOS supporters (and at least one very anxious dev) are so eager to test the X1000 here, why not make a MOS port for the X1000 with duo core support please and thank you ;)
Great idea.
-
or, as MOS supporters (and at least one very anxious dev) are so eager to test the X1000 here, why not make a MOS port for the X1000 with duo core support please and thank you ;)
The Quark kernel supposedly could work on a multi-core system, but Abox doesn't support SMP.
Besides, I'm pretty sure if you ask Piru he'll tell you that a port for the X1000 ain't happening.
-
or, as MOS supporters (and at least one very anxious dev) are so eager to test the X1000 here, why not make a MOS port for the X1000 with duo core support please and thank you ;)
I have no interest in it myself, other than trying to figure out where the supposed performance failure is. Currently the official story appears to be that AmigaOS 4 is at fault. It would still be interesting to verify this claim.
-
I've always found those old results weird. I don't know what is wrong with the said system (maybe highly fragmented SFS partition?), but these are the results I get:
(http://sintonen.fi/pics/copy_benchmark.png)
X1000 is of course still faster (as would likely be Sam 440/460 as well).
It should be noted that many things can affect such benchmark, such as the filesystem being used (copy may return immediately if delayed writing is applied by the filesystem). Another significant factor is the physical location of the partitions, are they located on the same or different HDDs? And finally, rotating HDDs are clearly slower than SSD.
Notes: The PowerBook has a newer Western Digital Scorpio Blue WD2500BEVE 250GB 5400 RPM 8MB Cache 2.5" HDD, while the Mac mini has the original Apple branded Seagate Momentus 5400.2 ST9808211A 80GB HDD).
I dont know wy this values are so slow.I test with i5 760 with winuae and disable JIT.so sysspeed show 84 Mips and readfastl and writefastl both transfer 296.44 megabytes.
so my system get really slow then.
But the copy test need 0.5 sec on my solid State disk OCZ Vertex 2 120 gb (i boot windows so nothing in cache )
I do the test on a slow 2.5 drive with 5400 rpm WDC3200.here it need 0.6 sec.(after reboot windows too)
When i do a second copy, then time go less 0.3 sec.
when i enable the JIT i get 1184 mips and readfastl writefastl 1092 -1300 megabytes.i use min and max values i get.with jit the tests run so fast, that timer accuracy have some influence.
How do you measure the time ?
maybe the time is not exact measure.I think every system is fast enough to copy files at same speed.
The X1000 mem benchmark show for 2. level cache near same values as for mem transfer.
i think this is problem the testcode fit in the 2 megabyte 2. Level Cache of pa6.TO get real values, there need a test that copy 20 megabyte or more mem
-
I've always found those old results weird. I don't know what is wrong with the said system (maybe highly fragmented SFS partition?), but these are the results I get:
(http://sintonen.fi/pics/copy_benchmark.png)
X1000 is of course still faster (as would likely be Sam 440/460 as well).
It should be noted that many things can affect such benchmark, such as the filesystem being used (copy may return immediately if delayed writing is applied by the filesystem). Another significant factor is the physical location of the partitions, are they located on the same or different HDDs? And finally, rotating HDDs are clearly slower than SSD.
Notes: The PowerBook has a newer Western Digital Scorpio Blue WD2500BEVE 250GB 5400 RPM 8MB Cache 2.5" HDD, while the Mac mini has the original Apple branded Seagate Momentus 5400.2 ST9808211A 80GB HDD).
I dont know wy this values are so slow.I test with i5 760 with winuae and disable JIT.so sysspeed show 84 Mips and readfastl and writefastl both transfer 296.44 megabytes.
so my system get really slow then.
But the copy test (part to part) need 0.6 sec on my solid State disk OCZ Vertex 2 120 gb (i boot windows so nothing in cache )
I do the test on a slow 2.5 drive with 5400 rpm WDC3200.here it need 0.9 sec.(after reboot windows too)
When i do a second copy, then time go less 0.2 sec. on all systems
when i enable the JIT i get 1184 mips and readfastl writefastl 1092 -1300 megabytes.i use min and max values i get.with jit the sysspeed tests run so fast, that timer accuracy have some influence.
I think every system is fast enough to copy files at same speed.if not, its problem of harddrive, or file system handle read or write cache diffrent.enable the JIT do not speedup anything only second copy in windows cache is then faster.
The X1000 mem benchmark show for 2. level cache near same values as for mem transfer.
i think this is problem the testcode fit in the 2 megabyte 2. Level Cache of pa6.TO get real values, there need a test that copy 20 megabyte or more mem