How much faster is Pro vs Pink?

[kmihelich]

Is Scimark using both cores on the Pro?

[Socaltom]

This is the only description I could find.
$this->bbcode_second_pass_quote('', 'T')his test runs the ANSI C version of SciMark 2.0, which is a benchmark for scientific and numerical computing developed by programmers at the National Institute of Standards and Technology. This test is made up of Fast Foruier Transform, Jacobi Successive Over-relaxation, Monte Carlo, Sparse Matrix Multiply, and dense LU matrix factorization benchmarks

[slycat]

top has a screwy way of showing cpu usage, too.

During the udevd/syslog-ng power hog fiasco I had syslog-ng at 99%, udevd at 34% and transmissiond (during download) at 26%

So I'm assuming Arch Linux ARM allows the separation of processes but does not let them cross-over both cores. Which is probably why the pro's scores were kinda low (one core pro < one core pink), mirite?

I'm sorry I don't have time to be configuring Scimarks on my pro, but if there is a need for further testing later on in the week I can offer my plug.

[kmihelich]

The kernel handles tasking processes out to separate cores based on availability. For scimark to use it, it needs to have a multi-processor or SMP function to fully test it. A single-threaded application can only use one processor. But the results show what we generally expected, that despite being clocked nearly half of the Kirkwood chips, the new OxNAS chips are faster. This again goes to comparing across architectures, which is impossible just in terms of clock speed. For example, AMD cores are typically clocked lower than Intel, yet have the same performance characteristics; hence their old "3000+" marketing technique to show that a processor clocked around 2GHz was capable of going as fast as 3GHz Intel processors.

Anyway, it does show that used to its full capacity the Pro will be faster than a Kirkwood system in multi-threaded scenarios. If you can find a way to get the benchmarks to test multi-threaded (I'm sure there must exist an option), we would see a significant crushing of Kirkwood.

[Socaltom]

I've tried to install several other tsts without sucess. Compress-7pbzip is suppsed to suppot multithreads.
Tom

[Socaltom]

So it turns out that P7zip has a multi thread capable bench mark built in. 7za b starts it with defaults. you can adjust the number of threads with -mmt (x) -mmt1 calls for single thread testt.
you can adjust the dictionary size ( impact memory use) with -md2 -md21 calls a 2 mb dictionary ( uses 35Mb)
Default uses about 113Mb

On the Pink vs I got the following results $this->bbcode_second_pass_code('', '[root@Arch Linux ARM install ~]# 7za b

7-Zip (A) 9.13 beta Copyright (c) 1999-2010 Igor Pavlov 2010-04-15
p7zip Version 9.13 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,1 CPU)

RAM size: 247 MB, # CPU hardware threads: 1
RAM usage: 211 MB, # Benchmark threads: 1

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 352 99 347 343 | 7324 99 670 661
23: 348 98 361 355 | 7203 99 666 659
24: 324 92 378 348 | 7057 98 665 654
----------------------------------------------------------------
Avr: 96 362 349 99 667 658
Tot: 98 514 503
')

On the Pro I got the following results using defaults
$this->bbcode_second_pass_code('', '[root@Videobox ~]# 7za b

7-Zip (A) 9.13 beta Copyright (c) 1999-2010 Igor Pavlov 2010-04-15
p7zip Version 9.13 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 122 MB, # CPU hardware threads: 2
RAM usage: 113 MB, # Benchmark threads: 2

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 329 147 217 320 | 9447 199 429 853
23: 160 81 201 163 | 9051 192 430 828
----------------------------------------------------------------
Avr: 114 209 242 196 429 840
Tot: 155 319 541
')
The Pro seems to hit a bottle neck, due to the lack of memory, which seriously reduces the CPU % in both tests, especially the second test. If you mathematically adjust for the % CPU time, then the Pro should be capable of 698 MIPS vs 513 for the Pink v2.

I also ran the test on the Pro with a single core to see how the result compared.$this->bbcode_second_pass_code('', '[root@Videobox ~]# 7za b -mmt1

7-Zip (A) 9.13 beta Copyright (c) 1999-2010 Igor Pavlov 2010-04-15
p7zip Version 9.13 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 122 MB, # CPU hardware threads: 2
RAM usage: 107 MB, # Benchmark threads: 1

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 233 100 227 226 | 4804 100 433 433
23: 193 89 223 197 | 4694 99 433 429
----------------------------------------------------------------
Avr: 94 225 212 100 433 431
Tot: 97 329 322
')
It calculates to about 331 MIPS @100%. So as we believed, the Pro is capable of more calculations than the Pink, but it is hampered in some operations by the limited memory. In the end, the balance of processing power and memory looks to provide similar performance between the v2 and the Pro as far as I'm concerned.
Tom

[slycat]

KILLER RECONNAISSANCE! These were both freshly booted and up-to-date? No connected peripherals and default rc.conf setups?
According to openbenchmarking, the 7za b process is an acceptable benchmarking tool. Awesome find, Tom. My PogoPro's e-penor just got bigger

[Socaltom]

The Pro only had the one eSATA drive connected, and no other non-stock deamons/programs running. The Pink was running samba in the background, but it was idle. The Pink has 3 drives, 2 printers and a USB hub connected.
Tom

[WarheadsSE]

just touching base, overclocked to ~850 i got this
$this->bbcode_second_pass_code('', '

7-Zip (A) 9.13 beta Copyright (c) 1999-2010 Igor Pavlov 2010-04-15
p7zip Version 9.13 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 122 MB, # CPU hardware threads: 2
RAM usage: 113 MB, # Benchmark threads: 2

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 396 147 262 385 | 11589 300 349 1046
23: 139 60 237 141 | 11014 192 525 1008
----------------------------------------------------------------
Avr: 103 249 263 246 437 1027
Tot: 175 343 645
')

[WarheadsSE]

Dredging up: @ 925mhz
$this->bbcode_second_pass_code('', '

7-Zip (A) 9.13 beta Copyright (c) 1999-2010 Igor Pavlov 2010-04-15
p7zip Version 9.13 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 122 MB, # CPU hardware threads: 2
RAM usage: 113 MB, # Benãhmark threads: 2

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 425 147 281 413 | 12581 199 569 1135
23: 152 58 270 155 | 11577 187 568 1060
----------------------------------------------------------------
Avr: 102 275 284 193 568 1098
Tot: 148 422 691
')