@tlosm

Can you PLEASE buy another Amiga 4000 max it out and be
done with it and your angst over selling your last A4000 to buy the X5000?

Sorry but you don't provide much with your daily personal rationalisation
that everything else BUT the machine/emu you have for the day rules and
everything else sucks.

Buy an Amiga, come back and relax.

_________________

@geen_naam

I have have made several videos of my X5000 running!

They can be found on my youtube channel



Mad Skatemans Youtube Channel

This video shows the X5000 doing all sorts of stuff





Mad Skatemans X5000 Demo

_________________
AmigaOne X5000 -> 2GHz / 16GB RAM / Radeon R7 250 / M-Audio 5.1 -> AmigaOS 4.1 FE / Ubuntu Linux
Amiga 1200 -> Recapped / 68ec020 ACA 1221ec / CF HDD / RetroNET connected to the world

@Geen_naam

I’d still advocate getting an X5000. Yes you are not going to do everything on AmigaOS 4 but I still get a lot of use out of it for all sorts of things like classic and OS4 specific games (there are loads of games to play), some productivity, internet where Odyssey allows and watching stuff off YouTube and streaming sites.

Playing around with the disappointment that is a PlayStation Classic this Xmas also made me realise that you also have a pretty good set of emulation packages on AmigaOS 4 if you are into that too with the power and ability to upscale games at higher resolutions and high def and still run them at a decent speed. Final Burn Alpha is a personal favourite, once it’s all set up correctly it is a pretty nice piece of software to use and probably much better and more user friendly than say Emulation Station on a Rapsberry Pi .

You also have the back up of Linux too on a dual boot for anything missing. In reality even that doesn’t get switched on too much 😉

_________________
Amigaone X5000/20 4GB Radeon RX 550 Polaris 12 AmigaOS4.1 Final Edition Update 1
Amiga 1200 Workbench 3.1.4
Amiga CD32

@all
just for info

here my benchmark hardinfo of mine ppc vs mine x86 vs mine rasperry pi ... last rasperry pi 4 hardinfo was made using x86 emulation (exagear) ... take your conclusion

_________________
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

@tlosm

As much as I love my Made in Britain, Raspberry Pi 4, Retropie has yet to officially appear and is not officially supported, . It also doesn’t run AmigaOS natively too so any benchmarks are of course completely pointless 😉

_________________
Amigaone X5000/20 4GB Radeon RX 550 Polaris 12 AmigaOS4.1 Final Edition Update 1
Amiga 1200 Workbench 3.1.4
Amiga CD32

@tlosm
Interesting stuff. Thank you. Its both interesting and sad. Even The Pi4 pulverizes the X5000/40 ... (oh you got one .. ;) ?). Wondered that the X1000 with its altivitec doesnt compete better. Whats that PPC2080 stuff?

@LarsB

the 2080 is the cpu of the PPC notebook project.
X1000 have a good cpu but is too slow for today standard the P6T is an old isa like is the G5 .
Now mobile phones are more faster than our ppc

Last edited by tlosm on 07-Jan-2020 at 01:43 PM.

_________________
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

The X5000 is a nice machine, I love developing on mine.

_________________
Sent from my Quantum Computer.

@tlosm

"here my benchmark hardinfo of mine ... take your conclusion"

There are two results for "PowerMac 2006 4x970MP 2.5 GHz", and latter one is 4x better.

So you are comparing single core x5000 performance vs multicore ARM or what?
Bu then... single core test of 2080 can not be better than multicore RPi4, so, I do not understand those results.

Conclusion - bad testing or bad output of results.


(but it is true that our PPC options are behind mainstream, we dropped behind mainstream in CPU perf in 1994, so, that is not new)

btw. for fair comparison, we should have those test run on 060/50 to compare.
((Our CPU perf need (vs SW that we have) is currently closer to 060/50 than Ryzen.
Realistic way to get meaningful performance improvement is to go multicore.))

Last edited by KimmoK on 07-Jan-2020 at 02:14 PM.
Last edited by KimmoK on 07-Jan-2020 at 02:08 PM.
Last edited by KimmoK on 07-Jan-2020 at 02:07 PM.
Last edited by KimmoK on 07-Jan-2020 at 02:06 PM.
Last edited by KimmoK on 07-Jan-2020 at 02:04 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK
i was waiting one reply like this :)
two are make with hard info 06 ... debian 64 and Arm 64.
but the exager emulation is make with hard info 05 on Debian 8 i386 the same of all the top benchmarks :)

I did the same with blender and Pi4 kill my G5 quad there.

Last edited by tlosm on 07-Jan-2020 at 02:16 PM.
Last edited by tlosm on 07-Jan-2020 at 02:14 PM.

_________________
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

@LarsB

Whatr's sad for me is that I can't even view the results on my X1000.

Not on OS4 anyway.

@tlosm Maybe it could be useful to add an x5000/40 benchmark running Linux.

@LarsB

Is there, all the top bench are did with ubuntu mate 16.04.x with hardinfo 0.51 on second line the PI4 exagear is did with Pi 4 that emulate an x86 thru exagear on debian 8 i386 hardinfo 0.51

the other two are did with debian 10 PPC64 on G5 Quad and Arm64 with hardinfo 0.6
the debian PPC64 is full 64bit full multi thread like is on Pi with Arm 64.
You can see PI4 kill the Quad G5 .

I dont have any more an X5000 whre to run debian 10 if some one here have can do the same with hardinfo 06 and share the result.

If need later after work i will share the blender benchmark with PI4

_________________
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

@LarsB

Quote:

LarsB wrote: @tlosm Interesting stuff. Thank you. Its both interesting and sad. Even The Pi4 pulverizes the X5000/40 ... (oh you got one .. ;) ?). Wondered that the X1000 with its altivitec doesnt compete better. Whats that PPC2080 stuff?

That is because @tlosm is probably running a fits-all, does-all kernel on the X1000.
If he took the time to build a performance kernel things would be different.
Here are my results for the same tests on the X1000:
Cryptohash 79.83
N-Queens 12.24
FPU RAYTracing 14.02
Zlib 0.10
GPU Drawing 916.75 (on a very old low memory Radeon)

Click for Blowfish
Click for Fibinacci
Click for Raytracing

Benchmarks are only good for comparing CPUs of the same family, like 750, 750FX, 750GX etc..

_________________
Tinkering with computers.

@OneTimer1

Quote:

Floating Point Operations where done via built in libraries, developers could always relay on this libraries in AOS3. If an FPU was present in the system this libraries used them, so even software without FPU optimisation got the benefits of a FPU. On AOS4 they started with a PPC that had a built in FPU and they never made developers using some libraries, now they introduced a PPC without compatible FPU and they are running into troubles. That's when weak OS design meats weak hardware design.

The so called "weak OS design" where one assumes presence of a floating point unit is a design everyone follows nowadays. For example, on soft-float ARM architecture it was done exactly the way classic OS did ages ago - it was up to the library to provide correct FPU functions. And you know what - softfp ARM is nearly dead today. Everyone switched to hard-float, architecture which takes presence of FPU for granted and uses it.

You say it was so cool on classing AmigaOs. Let's have a look at adding two double numbers together. When using mathieee* libs, you just call IEEEDPAdd. This is how it looks behind the scenes:

1. Your code puts two double numbers into pairs of registers. D0:D1 and D2:D3.
2. mathieeedoubbas library base is put into A6 register
3. Your code executes JSR to an offset in LVO table
4. At the position in LVO there is a JMP to the IEEEDPAdd function, CPU jumps there
5. In the IEEEDPAdd implementation following happens:
5a. D0:D1 are pushed on the stack
5b. D2:D3 are pushed on the stack
5c. FP1 (double) is loaded from the stack
5d. FP0 (double) is loaded from the stack
5e. The actual FADD opcode (one cpu instruction) is executed
5f. The result from FP register is saved on the stack
5g. The result is loaded form the stack to D0:D1
5h. Stack pointer is adjusted
5i. RTS opcode - we go back to your code
6. Your code has result of adding two doubles in D0:D1 pair and does whatever it wants with it.

Compare this awesome approach with issuing one single FPU instruction in the code, and you will understand why such design is versatile maybe, but very far from optimal.

@Hypex
here for you
https://www.dropbox.com/s/cipympqgx55ibge/bench.pdf?dl=0

_________________
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

@Signal

Can you make a blender benchmark with all the thread enabled i will share it for the Pi and for G5 Quad

my X1000 bench was gave me by Christian some times ago.


Blendermark on PI 4

native
1 cpu 01:44.71
4 cpu 00:29.35

exagear i386 emulation
1 cpu 09:42.47
4 cpu 03:25.86

Last edited by tlosm on 08-Jan-2020 at 06:25 AM.
Last edited by tlosm on 07-Jan-2020 at 05:39 PM.
Last edited by tlosm on 07-Jan-2020 at 03:21 PM.

_________________
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

In other news, shall we start a bounty for a webdesigner for A-EON? It seems the last one is still on a hangover.

_________________

@Everblue

if they was kind ... i will ask they for free but ...

one of my last free gift http://www.nonnaiole.it/

_________________
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

@michalsc

Quote:

You say it was so cool on classing AmigaOs. Let's have a look at adding two double numbers together. When using mathieee* libs, you just call IEEEDPAdd. This is how it looks behind the scenes:

And this is for FPU optimised code. I can only imagine how bad it would be for 68000 code. But at least it shouldn't need the stack.

Does it really need to use stack? They didn't provide an instruction to move to/from data and FPU registers?

They made the same mistake on the PowerPC!