@opi

Quote:

Wow, AmigaOS4 has SMP?! Man, that's awesome. You know what else is awesome? You thinking that two cores equals double performance.

Well blender in linux shows a 2.09x speedup, 430s single core, 206s both cores.
Performance remains to be seen WHEN os4 gets SMP.

_________________
Hate tends to make you look stupid...

@wawa

Good example of differences between Blender versions is this thread on a.org

Piru shows his PowerBook is faster than dual-core X1000 (and even single core 970MP 2.3 GHz ), only forgot to add which version of Blender he use... certainly not 2.4x.

Edit - exact numbers:
PowerBook G4 1667 MHz, Blender 2.49: 325 s
PowerBook G4 1667 MHz, Blender Piru: 205 s

Edit2 (Blender 2.48 - same as used on OS4):
PowerBook G4 1500 MHz, Blender 2.48a: 425 s
AmigaOne XE G4 1000 MHz, Blender 2.48: 591 s

Last edited by pavlor on 04-May-2012 at 11:13 PM.
Last edited by pavlor on 04-May-2012 at 11:00 PM.

@pavlor

Piru shows alot of things alright LOL...see this interesting benchmark thread where he cunningly forgets to mention that the X1000 is only running on 1 core and not to its full potential.

surprised he finds time for this stuff since he's so busy posting on every single OS4.x thread over at the .org

_________________
____________________________
c64-2sids, A1000, A1200T-060@50(finally working!),A4000-CSMKIII
! My Master Miggies- Amiga 1000 & AmigaOne X1000 !
mancave-ramblings
X1000 I BELIEVE

@klx300r

Quote:

see this interesting benchmark thread where he cunningly forgets to mention that the X1000 is only running on 1 core and not to its full potential.

HardInfo are single core benchmarks. So two cores are pointless here.

@pavlor

We need to get those optimizations ported to AOS4 version of Blender.
Piru is a code optimization wizard and it shows!!!!

Bounty to get Piru to speed up our Blender 2x ? !!!!!!!

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

@pavlor

yeah, you might have some point, g4 being 2times faster than higher clocked x1k on one thread seems a little strange. where do your numbers of powerbook come from? yourself?

@sundown

Quote:

Well blender in linux shows a 2.09x speedup, 430s single core, 206s both cores. Performance remains to be seen WHEN os4 gets SMP.

Interesting, blender is designed to be multithreaded so getting nearly 2x wouldn't suprise me slightly more is suprising, but it could be within error margins for the test I suppose.

The other posibilty is blender is doing purely rendering on the second thread / CPU and rendering and GUI updating on the other, thus for a purely rendering orientated benchmark if gives more speed up.

Last edited by broadblues on 05-May-2012 at 10:22 AM.

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@pavlor

Quote:

Edit2 (Blender 2.48 - same as used on OS4):

I doubt it OS4.1 blender is n't exactly any version, it's based on a snapshot between 2.48a and 2.49 (a in blender case refers to an update version a b c etc not alpha / beta stsuts)

It's not optimised for G4 altivec etc in any way (except for the libffmpeg part) .

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

do you guys suggest piru has posted morphos blender results of his powerbook?

Last edited by wawa on 05-May-2012 at 10:43 AM.

@wawa

Quote:

where do your numbers of powerbook come from? yourself?

Blender benchmark page:
http://www.eofw.org/bench/

@wawa

Piru posted: "PowerBook G4 1.67GHz: my own test run"

I assume Piru ran the test on MOS.
And as it seems it performs about 50% faster than a Linux version on the same CPU (and about 100% faster than AOS version (per Mhz)), he has some superb optimizations in it. (we need those for AOS4 blender as well)

And Apple G5 tests have indicated that for G5 or PA6T optimizations can have even bigger boost compared to traditional PPC/G4 compile.

Last edited by KimmoK on 05-May-2012 at 12:13 PM.
Last edited by KimmoK on 05-May-2012 at 12:11 PM.
Last edited by KimmoK on 05-May-2012 at 12:10 PM.
Last edited by KimmoK on 05-May-2012 at 12:10 PM.
Last edited by KimmoK on 05-May-2012 at 12:09 PM.
Last edited by KimmoK on 05-May-2012 at 12:08 PM.

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

@sundown

You are in the fast lane .
Blender test
Sam440ep-Flex 800 MHZ
Test.blend
Debian Linux. Blender 2.49b Radeon 9250 - 17:33:27
Debian Linux. Blender 2.49b Radeon HD 4850 - 18:10:57
Amiga OS4.1 Update 4 OS4.1 Blender 2.48a Radeon 9250- 19:11.64
Amiga OS4.1 Update 4 Wazp3D Blender 2.48a Radeon HD 4850- 25:36.45

Last edited by Spectre660 on 05-May-2012 at 03:05 PM.
Last edited by Spectre660 on 05-May-2012 at 02:47 PM.
Last edited by Spectre660 on 05-May-2012 at 01:47 PM.
Last edited by Spectre660 on 05-May-2012 at 01:46 PM.
Last edited by Spectre660 on 05-May-2012 at 01:17 PM.

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

Quote:

pavlor wrote: @klx300r Quote: see this interesting benchmark thread where he cunningly forgets to mention that the X1000 is only running on 1 core and not to its full potential. HardInfo are single core benchmarks. So two cores are pointless here.

Really? 90% of people have no idea those benchmarks do not support multi cores so it's misleading at best unless it's noted along with the test results !

So let's do a Blender test comparing machines and post results but just leave out that one of the machines is only running on one core!

Yup truly pointless indeed

_________________
____________________________
c64-2sids, A1000, A1200T-060@50(finally working!),A4000-CSMKIII
! My Master Miggies- Amiga 1000 & AmigaOne X1000 !
mancave-ramblings
X1000 I BELIEVE

@KimmoK

Piru used the standard debian unstable version

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

@Spectre660

looking again seems that pirus results are about in accordance with here:
http://www.eofw.org/bench/
now, this is really x1k (and g5) numbers that are surprisingly low..

@wawa

As you see 2.5x (and some 2.43 etc.) numbers are higher than 2.48/2.49 results.

Quote:

now, this is really x1k (and g5) numbers that are surprisingly low..

PA6T 1.8 GHz is as fast as G4 1.42 GHz with the same Blender version (OS4), result confirmed also by other benchmarks.

@pavlor

Is there SAM460 blender results somewhere?

Some of my observations:

With blender 2.49 or 2.48 & dualcore:
120s blender 2.49 G5 2.3Ghz Dualcore
121s blender 2.49 G5 2Ghz Dualcore (erroneously reported singlecore?)
127s blender 2.49 G5 3Ghz Dualcore (erroneously reported singlecore?)
206s blender 2.48? x1000 1800Mhz Dualcore

Then with blender 2.49 or 2.48 & singlecore:
223s blender 2.49 G5 2.3Ghz single core
234s blender 2.49 G5 2.1Ghz single core
235s blender 2.49 G5 3Ghz SingleCore
325s blender 2.49 PowerBook 1667Mhz
331s blender 2.49 PowerBook 1667Mhz
420s blender 2.48? x1000 1800Mhz Single core
425s blender 2.48a PowerBook 1500Mhz
540s blender 2.48 MacMini 1420Mhz
591s blender 2.48 A1XE 1000Mhz
1180s blender 2.48 SAM440 800Mhz (AOS4)

With same blender version G4 is a slower than G5, but faster than PA6T.

Performance scaling etc.:
G5-2300single vs G4-1667 - 1.3 x Mhz - 1.4 x result - G5 better

G5-2300single vs PA6T-1800single - 1.28 x Mhz - 1.88 x result - G5 better

G4-1667 vs PA6T-1800single - 1.08 x Mhz - 0.77 x result - G4 better
G4-1667 vs PA6T-1800dual - 1.08 x Mhz - 1.6 x result - PA6T is better

The last part shows PA6T advantage, one day, after AOS is updated to use both cores.

Simple "graph" about AOS4 HW: (imaginary 400Mhz Notebook/Netbook as a baseline)

A1notebX
SAM440_XX
SAM460_XXX
A1_G3__XXX
A1_G4__XXXXX
A1X1000XXXXXXXDDDDDDD
(wannabe AmigaNG desktops next)
P1022__XXXDDD+ T4240__XXXXXXXDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD++

D indicates the potential of multicore speedup in blender.
+ indicates the possible speed up from SPE unit
++ indicates the potential additional speedup of hyperthreading (can be high, depending on application, Freescale estimates about 2x MIPS per MHz for T4240 vs previous generation)
(I think that in integer performance the graphs would look similar)

Last edited by KimmoK on 07-May-2012 at 11:08 AM.
Last edited by KimmoK on 07-May-2012 at 10:44 AM.
Last edited by KimmoK on 07-May-2012 at 10:41 AM.
Last edited by KimmoK on 07-May-2012 at 10:39 AM.
Last edited by KimmoK on 07-May-2012 at 09:32 AM.

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

@broadblues

Quote:

The other posibilty is blender is doing purely rendering on the second thread / CPU and rendering and GUI updating on the other, thus for a purely rendering orientated benchmark if gives more speed up.

Then again, you don't need GUI to do rendering in blender (I'd say it's preferable to not use it, as it can increase rendering times slightly)

blender -b scene.blend -f 1

-f specifies the frame number. See documentation for other options.

Updating my summary:
With blender 2.49 or 2.48 & dualcore:
120s blender 2.49 G5 2.3Ghz Dualcore
121s blender 2.49 G5 2Ghz Dualcore (erroneously reported singlecore?)
127s blender 2.49 G5 3Ghz Dualcore (erroneously reported singlecore?)
206s blender 2.48? x1000 1800Mhz Dualcore

Then with blender 2.49 or 2.48 & singlecore:
223s blender 2.49 G5 2.3Ghz single core
234s blender 2.49 G5 2.1Ghz single core
235s blender 2.49 G5 3Ghz SingleCore
325s blender 2.49 PowerBook 1667Mhz
331s blender 2.49 PowerBook 1667Mhz
414s blender 2.49 x1000 1800Mhz Single core (AOS4)
420s blender 2.48? x1000 1800Mhz Single core
425s blender 2.48a PowerBook 1500Mhz
538s blender 2.49 MCP8610 1066Mhz
540s blender 2.48 MacMini 1420Mhz
591s blender 2.48 A1XE 1000Mhz
1180s blender 2.48 SAM440 800Mhz (AOS4)

Added MPC8610 and x1000+AOS4 information.
(6810 seems much faster per Mhz than macmini because of faster RAM)
Can anyone point me to SAM460 results?

Last edited by KimmoK on 11-May-2012 at 06:00 AM.
Last edited by KimmoK on 10-May-2012 at 07:36 PM.

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

@virgolamobile

could you please assist us with the Sam460ex Blender test.blend benchmark results.

http://www.eofw.org/bench/test.blend?PHPSESSID=d5406d28673851951267403174b77191

Thanks

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card