Poster | Thread |
WolfToTheMoon
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 17:05:46
| | [ #41 ] |
|
|
 |
Super Member  |
Joined: 2-Sep-2010 Posts: 1410
From: CRO | | |
|
| @pavlor
considering just how much newer the PA6T is compared to G4... hm, I'd say that results are more bad than good... I was expecting more _________________
|
|
Status: Offline |
|
|
pavlor
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 17:16:19
| | [ #42 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9660
From: Unknown | | |
|
| @WolfToTheMoon
Quote:
considering just how much newer the PA6T is compared to G4... |
Year 2007 product - compare it to Freescale e600 flagship of that time (8641D). PA6T has many advantages over G4: better performance per watt, better features on SoC, 64 bit etc.Last edited by pavlor on 02-May-2012 at 05:16 PM.
|
|
Status: Offline |
|
|
pavlor
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 18:58:13
| | [ #43 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9660
From: Unknown | | |
|
| Older "application benchmarks" in this poster confirm excellent FFT performance of PA6T (p. 8 and 10) and also G4 like performance in some other benchmarks (MM800, same pages).
Interesting to note that PA6T gives here better result with greater matrix (MM3000). It seems PA6T performance highly depends on application (see "strange" results in some HardInfo benchmarks) and workload (MM800 vs MM3000). |
|
Status: Offline |
|
|
Seiya
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 19:41:40
| | [ #44 ] |
|
|
 |
Super Member  |
Joined: 19-Aug-2006 Posts: 1479
From: Italia | | |
|
| @pavlor
but those benchmarks are made by them..so not much realistic.. _________________
|
|
Status: Offline |
|
|
pavlor
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 20:25:39
| | [ #45 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9660
From: Unknown | | |
|
| @Seiya
Quote:
but those benchmarks are made by them..so not much realistic.. |
If you mean poster I linked, then no, it isn´t made by PASemi.
Of course, I too believe only in benchmarks I forged myself.  |
|
Status: Offline |
|
|
sundown
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 20:27:40
| | [ #46 ] |
|
|
 |
Elite Member  |
Joined: 30-Aug-2003 Posts: 5120
From: Right here... | | |
|
| @Seiya
Run the OS4 blender version on a x86 system & tell us what you get with the benchmark test.  _________________ Hate tends to make you look stupid... |
|
Status: Offline |
|
|
tibby.dude
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 21:55:46
| | [ #47 ] |
|
|
 |
New Member |
Joined: 3-Jan-2012 Posts: 7
From: Cape Town | | |
|
| @Spectre660
Mac G5 2.3Ghz 970MP running Ubuntu 10.10
CPU Blowfish- 43.495 CPU CryptoHash - 36.826 CPU Fibonacci - 16.966 CPU N-Queens - 19.877 FPU FFT - 11.435 FPU Raytracing - 19.426 |
|
Status: Offline |
|
|
Spectre660
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 22:13:38
| | [ #48 ] |
|
|
 |
Elite Member  |
Joined: 4-Jun-2005 Posts: 3918
From: Unknown | | |
|
| @tibby.dude
now we are talking ! . _________________ Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card |
|
Status: Offline |
|
|
Seiya
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 2-May-2012 22:51:14
| | [ #49 ] |
|
|
 |
Super Member  |
Joined: 19-Aug-2006 Posts: 1479
From: Italia | | |
|
| @sundown
Blender 2.48a on intel core2Duo E8500 and Slax 6.1.2
02:24.18 (1 thread)
Blender 2.48a on Athlon64 3500+ on Salx 6.1.2
04:24.22 (of course 1 thread)
Last edited by Seiya on 02-May-2012 at 10:54 PM. Last edited by Seiya on 02-May-2012 at 10:53 PM.
_________________
|
|
Status: Offline |
|
|
sundown
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 3:05:41
| | [ #50 ] |
|
|
 |
Elite Member  |
Joined: 30-Aug-2003 Posts: 5120
From: Right here... | | |
|
| @Seiya
Debian Squeeze has no h/w acceleration, only s/w. I ran glxinfo in a terminal to test. Not much point in posting benchmarks, they would just be slow because of it. Blender times in linux vs os4.1 with 1 core are the same, neither has h/w acceleration yet.
That wasn't my point anyway, all benchmarks are optimized for x86, some versions of linux might support h/w acceleration, I'm not a linux expert/fan to know. There are no fair tests for the x1000, some ppl here understand that, others use the results to degrade a-eon's efforts.
The x1000 runs os4 faster then any x86 system build, thats all I care about.
_________________ Hate tends to make you look stupid... |
|
Status: Offline |
|
|
tommysammy
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 4:15:08
| | [ #51 ] |
|
|
 |
Cult Member  |
Joined: 20-Jan-2010 Posts: 664
From: Isselburg,Germany | | |
|
| @sundown
Quote:
sundown wrote: @Seiya
Debian Squeeze has no h/w acceleration, only s/w. I ran glxinfo in a terminal to test. Not much point in posting benchmarks, they would just be slow because of it. Blender times in linux vs os4.1 with 1 core are the same, neither has h/w acceleration yet.
That wasn't my point anyway, all benchmarks are optimized for x86, some versions of linux might support h/w acceleration, I'm not a linux expert/fan to know. There are no fair tests for the x1000, some ppl here understand that, others use the results to degrade a-eon's efforts.
The x1000 runs os4 faster then any x86 system build, thats all I care about.
|
Yes right ,at this moment we don`t have any hardware2D and 3D driver. We have only 2D and 3D softwaredriver,so we don`t need to make any benchmarktests_________________ Amiga600/Vampire2/PrismaMegaMix |
|
Status: Offline |
|
|
itix
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 4:32:25
| | [ #52 ] |
|
|
 |
Elite Member  |
Joined: 22-Dec-2004 Posts: 3398
From: Freedom world | | |
|
| @sundown
Quote:
That wasn't my point anyway, all benchmarks are optimized for x86, some versions of linux might support h/w acceleration, I'm not a linux expert/fan to know. There are no fair tests for the x1000, some ppl here understand that, others use the results to degrade a-eon's efforts.
|
Hmm... why you think the results degrade A-eon's efforts? Like pavlor said some results are surprisingly good. X1000 performing as good as Apple G4 is not bad at all since they are very fast._________________ Amiga Developer Amiga 500, Efika, Mac Mini and PowerBook |
|
Status: Offline |
|
|
itix
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 4:53:28
| | [ #53 ] |
|
|
 |
Elite Member  |
Joined: 22-Dec-2004 Posts: 3398
From: Freedom world | | |
|
| @pavlor
Wow, Honeywell, now that company has got bad reputation in Finland... anyway:
Quote:
One-Thread vs. Two-Thread PA Semi Results · Nearly 2x speedup provided for 3000x3000 matrix multiply - Smaller matrix multiply suffers due to dataset size · HSI application speedup limited to 1.63x by highly serialized Weight Computation stage - Autocorrelation Sample Matrix (ACSM) and Target Classification stages take advantage of both cores fairly efficiently · FFTW actually slowed down for multi-core implementations - Suspect likely due to inefficiencies in fine-grain parallelization of 1D FFT, expect much better performance for 2D FFTs with coarse-grain parallelization · Similar trends observed on 8641D (and 970FX SMP with 2 processors)
|
Quote:
· Largest speedups on large matrix multiply - Best exploits parallelism in multi-core architectures · FFTW does not efficiently exploit both processor cores, limiting speedup
|
_________________ Amiga Developer Amiga 500, Efika, Mac Mini and PowerBook |
|
Status: Offline |
|
|
sundown
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 6:07:27
| | [ #54 ] |
|
|
 |
Elite Member  |
Joined: 30-Aug-2003 Posts: 5120
From: Right here... | | |
|
| @itix
Quote:
Hmm... why you think the results degrade A-eon's efforts? Like pavlor said some results are surprisingly good. X1000 performing as good as Apple G4 is not bad at all since they are very fast. |
Guess because being a beta tester, I see the system getting faster in some areas now. I admit, benchmarks today will be interesting to look at after 4.2, but it will never beat the faster x86 systems, nor was it mean't to. Also current benchmarks are not optimize for PPC, so the x1000 will always seem slow. I really see no point doing these test because of that & I do see the MOS guys wanting to put the X down.
Now OS4 benchmarks on the A1's, Sam's, & x1000 to show the difference would be interesting, without the Mac or x86 systems, same OS, different h/w. _________________ Hate tends to make you look stupid... |
|
Status: Offline |
|
|
sundown
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 6:10:17
| | [ #55 ] |
|
|
 |
Elite Member  |
Joined: 30-Aug-2003 Posts: 5120
From: Right here... | | |
|
| @tommysammy
I was hoping to try Debian Mint soon, it might have a better gfx driver? _________________ Hate tends to make you look stupid... |
|
Status: Offline |
|
|
KimmoK
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 6:40:08
| | [ #56 ] |
|
|
 |
Elite Member  |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| from tibby.dude results... G5 2.3Ghz vs others: CPU Blowfish- G5 2.3Ghz is 1/4 faster than x1000, 1/10 slower than Piru's G4 CPU CryptoHash - G5 2.3Ghz is 1/10 faster than x1000 or G4 CPU Fibonacci - G5 2.3Ghz is 1/10 slower than x1000, 1/3 slower than G4 CPU N-Queens - G5 2.3Ghz is 1/5 faster than others FPU FFT - G5 2.3Ghz is almost 2x faster than x1000, almost 4x faster than G4 FPU Raytracing - G5 2.3Ghz is about 2x faster than others
(G5 has 1.3x the clock rate of PA6T & 1.4x the clock rate of G4) So the 25W PA6T (and G4) seems very nice against 100W G5.
Only surprice to me is the higher FPU performance of the G5. Why is that?
e600 is superb core, it will be nice to see some e6500 results in the future.
Last edited by KimmoK on 03-May-2012 at 06:44 AM. Last edited by KimmoK on 03-May-2012 at 06:44 AM. Last edited by KimmoK on 03-May-2012 at 06:42 AM.
_________________ - KimmoK // For freedom, for honor, for AMIGA // // Thing that I should find more time for: CC64 - 64bit Community Computer? |
|
Status: Offline |
|
|
itix
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 6:53:08
| | [ #57 ] |
|
|
 |
Elite Member  |
Joined: 22-Dec-2004 Posts: 3398
From: Freedom world | | |
|
| @sundown
Quote:
so the x1000 will always seem slow
|
Again, who said that? Was it you? Yes, it was you.
Quote:
I do see the MOS guys wanting to put the X down.
|
That is just how you want to see it. You find G4 class performance disappointing and wish it is just other people being bitter because they dont have X1000.
Quote:
Now OS4 benchmarks on the A1's, Sam's, & x1000 to show the difference would be interesting, without the Mac or x86 systems, same OS, different h/w.
|
Yes because you know X1000 is going to beat any existing OS4 hardware. To me it is not interesting how much faster PA6T is to existing OS4 hardware. To me interesting is "what if morphos was running on this hardware?" and these benchmarks answer my question very well.
Benchmarks are not always just for personal egoboost. They can be used to evaluate hardware options as well._________________ Amiga Developer Amiga 500, Efika, Mac Mini and PowerBook |
|
Status: Offline |
|
|
sundown
 |  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 7:09:14
| | [ #58 ] |
|
|
 |
Elite Member  |
Joined: 30-Aug-2003 Posts: 5120
From: Right here... | | |
|
| @itix
You & I can go around in circles forever & never agree. I have my thoughts & you have yours. Why is it ppl like to twist what others say into something negative all the time, are we all liars now? Well, some good news, x1000 benchmarks are becoming a lie.  _________________ Hate tends to make you look stupid... |
|
Status: Offline |
|
|
Jupp3
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 8:43:32
| | [ #59 ] |
|
|
 |
Super Member  |
Joined: 22-Feb-2007 Posts: 1225
From: Unknown | | |
|
| @sundown
Quote:
Debian Squeeze has no h/w acceleration, only s/w. I ran glxinfo in a terminal to test. Not much point in posting benchmarks, they would just be slow because of it. Blender times in linux vs os4.1 with 1 core are the same, neither has h/w acceleration yet. |
Does blender ever use h/w acceleration for rendering anyway?
I thought it was only for the user interface (and the game engine side)
The problem with OpenGL (related to rendering) is that it doesn't make any guarantees on how the end result will look. You can get (slightly) different results from different graphics cards (for rather extreme example, consider using both 16 and 24bit gfx cards) - potentially even different driver versions could cause different results.
And while with shaders OpenGL can be used for ray tracing, I'd assume blender to be "CPU-only". |
|
Status: Offline |
|
|
Seiya
|  |
Re: Debian Squeeze Hardinfo Benchmarks Posted on 3-May-2012 9:47:39
| | [ #60 ] |
|
|
 |
Super Member  |
Joined: 19-Aug-2006 Posts: 1479
From: Italia | | |
|
| i understand When benchmark are good for X1000, never mind if OS4 or inux lacks many features like 2D or 3D accerlation, but when these tests are not good, the problems is operating system...
however i tryed again on vmware disabling hardware acceleration
02:25.42
_________________
|
|
Status: Offline |
|
|