@wawa

Quote:

so you expect a major blender engine overhaul somewhere between 2.48 and 2.49 that might have significant influence on the benchmark?

No, I only say results aren´t directly comparable (see Blender benchmarks page for more results to compare).

Quote:

lets say 200% speedup?

Certainly not. However even 10 % is visible difference.

@pavlor

you really think, we need to care about some 10% here? me, id even multiply x1k results with those 10% from the start if it makes feel anybody any better.

@wawa

i don't think so
Blender 2.48a and 2.49 are not so different in rendering time, so, imho, this g5 test is interesting and good to compare

OT
very soon (hope this night) i upload on youtube my G4@2 ghz emulated by PearPC!!!!!
unfortunately i don't find PPC linux distro to test it..
/OT

Last edited by Seiya on 03-May-2012 at 11:34 PM.

_________________

About those blender tests...
Hard to believe that G4 at 1.67Ghz can beat 2.3Ghz G5 and 2x1.8Ghz PA6T cores.
Something fishy in those test binaries?
Surely same binaries should be used to be able to compare? (or maximally optimized for every chip)

(I wonder if the e600 core version is manually optimized, while there is no such optimizations for G5 or PA6T???)

Some older tests: http://www.barefeats.com/g5sum.html
http://www.barefeats.com/g5.html ((here the single 1.6Ghz G5 seems faster than dual core 1.4ghz G4 in Bryce rendering ... WTF?))
http://www.barefeats.com/pentium4.html

Last edited by KimmoK on 04-May-2012 at 08:40 AM.
Last edited by KimmoK on 04-May-2012 at 08:37 AM.
Last edited by KimmoK on 04-May-2012 at 08:37 AM.
Last edited by KimmoK on 04-May-2012 at 08:14 AM.
Last edited by KimmoK on 04-May-2012 at 08:07 AM.
Last edited by KimmoK on 04-May-2012 at 08:06 AM.
Last edited by KimmoK on 04-May-2012 at 08:05 AM.

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

http://www.youtube.com/watch?v=eD5x263GTuo

@Mufa

Funny video.
Next the SW should be updated to use the advantage.

btw. PA6T had 2x64bit 1066Mhz interface to RAM, while G4 uses up to 1x64bit 166Mhz. (IIRC)

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

@KimmoK

That Bryce render test was indeed strange.

All other tests: Very little difference between G4 and G5 performance, clock for clock. G4 seems to be slightly faster at work / MHz. And much better at power consumption / MHz

I assume these tests use Altivec, and the Altivec implementation on the G4 is the most efficient one.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Mufa

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

_________________
OpenWindows Initiative. Port PS3 hardware to bananas. For free. Join today and receive expired $50 cupon from AI!

@KimmoK

Well. An MPC8641 is a G4 based SoC, which basically means it has a direct memory interface (plus a direct PCIe, which also helps). Not 1066MHz, but still much faster than the 60x bus of the earlier 74x0 and 75xx chips.

I would have loved to see some actual benchmarks of that compared to these other chips. I'm beginning to have a sneaky feeling that it would be very good

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@opi

Well, to be fair that was one out of several points there

Lemme guess, you subscribe to the view that SMP won't happen for OS4 ever?

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Quote:

Lemme guess, you subscribe to the view that SMP won't happen for OS4 ever?

It may, but I think it will play out as USB2 drivers and XMOS support.

_________________
OpenWindows Initiative. Port PS3 hardware to bananas. For free. Join today and receive expired $50 cupon from AI!

@opi

"two cores equals double performance."
It seems to depend on application(s) and chip.

According to piru's graph PA6T delivered 2x speed with dualcore for blender, while G5 gained 1.8 speed.

(might be some testing clitch as well, though)

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

so.... based on this, an ARM port of MOS/OS4 doesn't look like a step back. A dual core A9 should be competitive with these chips at much lower cost and TDP.

_________________

@WolfToTheMoon

based on what? Did I miss something?

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@KimmoK

Quote:

It seems to depend on application(s) and chip.

Not every application can scale with number of cores. At least not in linear fashion.

_________________
OpenWindows Initiative. Port PS3 hardware to bananas. For free. Join today and receive expired $50 cupon from AI!

@WolfToTheMoon

According to some linpack tests Atom455 and ARM v7-A9 1.5Ghz might be roughly compareable with each other. And PA6T might be on the same ballpark.
Source: http://www.roylongbottom.org.uk/linpack%20results.htm


For OpenCL dreamers: http://www.luxrender.net/wiki/LuxMark_Results

Last edited by KimmoK on 04-May-2012 at 02:47 PM.

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

some test:

Scene: Room (complex)

i7 2600 + HD7870 OC 2 GB

OpenCL GPU = 763
OpenCL GPU + CPU = 869
OpenCL CPU = 292

E8500 + GTS250 1 GB

OpenCL GPU = 56

Last edited by Seiya on 04-May-2012 at 07:53 PM.

_________________

@Spectre660

Sam460ex 1150 MHz
(Debian Squeeze)

CPU Blowfish- 58.944
CPU CryptoHash -19.339
CPU Fibonacci -16.085
CPU N-Queens -42.252
FPU FFT - 60.214
FPU Raytracing -50.750

@KimmoK

Quote:

KimmoK wrote: @WolfToTheMoon According to some linpack tests Atom455 and ARM v7-A9 1.5Ghz might be roughly compareable with each other. And PA6T might be on the same ballpark. Source: http://www.roylongbottom.org.uk/linpack%20results.htm

Quote:

"This benchmark was produced by Jack Dongarra from the "LINPACK" package of linear algebra routines. It became the primary benchmark for scientific applications from the mid 1980's with a slant towards supercomputer performance."

So prolly best possible desktop benchmark....


For OpenCL dreamers: http://www.luxrender.net/wiki/LuxMark_Results[/quote]

Even linux doesn't have free OpenCL drivers so AmigaOS will have them in near future ?

Last edited by Rose on 05-May-2012 at 02:29 AM.
Last edited by Rose on 04-May-2012 at 11:24 PM.
Last edited by Rose on 04-May-2012 at 11:23 PM.
Last edited by Rose on 04-May-2012 at 11:22 PM.
Last edited by Rose on 04-May-2012 at 11:19 PM.

@virgolamobile

Thanks.

Sam460ex is about 1.06% faster per clock cycle than Sam440ep .

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