@K-L

Quote:

No, it is not. PA6-T at 1,8Ghz is slower than my previous G4 at 1,26 Ghz.

Please, re-read my post. G4 has much stronger AltiVec and so much better result.

@ferrels

And i dont understand the obsession with even older 68k.
Using 68k aps on os4 is a less hassle.

@OlafS25

Quote:

the archive has a problem

Extract it in Amiga enviroment (WinRAR reported problem, but Amiga lha extracted it correctly). Just tested version I linked (on Core i5-2500K 3.3 GHz, WinUAE 3.1):

OGRNG: 5,073,076 nodes/sec

@pavlor

Quote:

Please, re-read my post. G4 has much stronger AltiVec and so much better result.

.

Not really :

640x360 MP4 Video to 720x576 30FPS MPG Video :

*****************
AmigaOne XE G4
*****************

FFmpeg Altivec : 11mn56 (14FPS)
FFmpeg Non Altivec : 17mn12 (10 FPS)

*****************
AmigaONE X1000
*****************

FFmpeg Altivec : 6mn58 (25FPS)
FFmpeg Non Altivec : 14mn06 (12FPS)

AltiVec from the X1000 is way faster with FFmpeg.

_________________
PowerMac G5 2,7Ghz - 2GB - Radeon 9650 - MorphOS 3.14
AmigaONE X1000, 2GB, Sapphire Radeon HD 7700
FPGA Replay + DB 68060 at 85Mhz

@haywiresystem

I dont have the time setting up frankenamigas, so if time i money use NG amigas eg Os4. X1000 payed itself in som weeks.

@K-L

Quote:

AltiVec from the X1000 is way faster with FFmpeg.

Current dnetc is form 2010/2011, long before X1000 appeared. What version of ffmpeg you used for testing?


Edit: There could be another bottle-neck in FFMPEG test (eg. Pegasos II has much slower memory - how many MB/s are converted?).

Last edited by pavlor on 13-Jun-2015 at 09:59 AM.

@K-L

a) the X1000 is running at more than twice the clock (assuming 800MHz for the TeronOne)

b) memory bandwidth still plays a role

Whats really disapointing is the non-altivec result, something clearly wrong with the x1000 here (should have been atleast twice as fast).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

Quote:

TeronOne)

Quote:

assuming 800MHz

I see reading comprehension is limited today. Please, re-read K-L´s post 160 (it is 1.26 GHz G4).

Quote:

b) memory bandwidth still plays a role

That is what I think too.

Quote:

should have been atleast twice as fast

Reported DMIPS/MHz values for PA6T are lower than in case of G4 (2.2 vs 2.3).

@pavlor

1,972,597

And what does this now mean practical?

@pavlor

Quote:

pavlor wrote: I see reading comprehension is limited today. Please, re-read K-L´s post 160 (it is 1.26 GHz G4). .

Remove "comprehension" and your spot on

At 1.26GHz both systems are close to parity (clock by clock) despite the A1 being much weaker in the memory bandwith.

Something in line with the blender results I got when a I benched my my dual 1.8GHz Quicksilver against the X1000 (under linux).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@eder

please start not being polemic

otherwise I run the benchmark on my Win-System and compare it

@pavlor

I tested it on Aros 68k perhaps I also compare it it using 3.X

but general it should be on a similar level as yours i5 propably

@OlafS25

Quote:

And what does this now mean practical?

What x86 CPU you used for testing?

@Kronos

Quote:

a) the X1000 is running at more than twice the clock (assuming 800MHz for the TeronOne)

.

I've written G4@1,26 Ghz.

_________________
PowerMac G5 2,7Ghz - 2GB - Radeon 9650 - MorphOS 3.14
AmigaONE X1000, 2GB, Sapphire Radeon HD 7700
FPGA Replay + DB 68060 at 85Mhz

@OlafS25

That is not an amiga.

@OlafS25

4,389,276 with 3.X

@eder

I could call yours "frankenamiga" because it is "franken" too in 2015

but i do not do that because then I would be as polemic as you

@pavlor

AMD A6-4400M 2.70 GHz

@OlafS25

4,389,276 with 3.X
1,972,597 with AROS68k

3.x result is expected as AMD A6-4400M 2.70 GHz is not that much slower than my i5, result in AROS is strange.

For comparison:
WinUAE/OS4 scores cca 5,290,000 on same hardware (or 4,500,000 with slower debug OS4 kernel).

Your emulated hardware is as fast as 68060 860 MHz or 440EP 460 MHz in dnetc.

@K-L

Because PA6T is a G5 class CPU, it does not have all the 32bit instructions, so whenever it hits a missing instruction, it has to be emulated, so it becomes slow.

OGG / RC5 should have been recompiled for a G5 target with AltiVec enabled.
If OGG / RC5 was optimized for G4 CPU, it be really bad for performance.

Anyway we have old compilers that don't even support PA6T, so it's no wonder really.

Last edited by NutsAboutAmiga on 13-Jun-2015 at 03:42 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS