I can know how much any CPU compute XOR function for write performance of parity calculation on standard frequency of CPU.

 

I am tested on desktop/"workstation" motherboards (values are from syslog) these CPU:

 

desktop E7200 2532.862 MHz   9537.2 MB/s pIII_sse

 

desktop Q9530 2,493GHz   9408 MB/s pIII_sse

 

mobile Core2Duo T7200 2GHz   7175 MB/s pII_mmx

mobile CoreDuo T2500 1991,904Hz   6412,80 MB/s pII_mmx

mobile CoreDuo T2300 1,66GHz   5344 MB/s pII_mmx

Pentium M740 1,73GHz   5723 MB/s pII_mmx

Pentium M760 2,00GHz   6604 MB/s pII_mmx

 

 

All tested with UnRaid 4.3.3.

 

 

Thanks for contributions

 

Core 2 DUO 2.66 - xor using function: pIII_sse (5910.400 MB/sec) (I think I'm running a t 1.6ghz).

Celeron M 1ghz - xor using function: pII_mmx (3310.800 MB/sec)

P4 2.8 - xor using function: pIII_sse (4486.000 MB/sec)

AMD Athlon 64 X2 4600+ (2.4 GHz) - xor using function: p5_mmx (8309.600 MB/sec)

 

I just saw what I think may be the champ, in this thread.

You've got bragging rights with your XOR processing speed, fastest I personally have seen:  "xor using function: pIII_sse (10090.000 MB/sec)".  I was particularly surprised because you are using a "CPU: Intel® Core2 Duo CPU E6750 @ 2.66GHz stepping 0b" clocked at "2720.125 MHz[/color]" in single core mode.

AMD Athlon XP (2500 barton) running at:

 

1100 MHz (3971.600 MB/sec)

1470 MHz (5295.600 MB/sec)

1837 MHz (6619.600 MB/sec)

 

from syslog:

 

CPU: Intel® Core2 Duo CPU    E4500  @ 2.20GHz stepping 0d

xor using function: pIII_sse (6604.400 MB/sec)

Nice survey messieurs.  Messieurs go ahead...

unRAID 4.3.3

 

CPU: Intel® Celeron® D CPU        430  @ 1.80GHz stepping 01

xor using function: pIII_sse (6723.200 MB/sec)

 

My xor stat on an Intel E2140 Allendale 1.6GHz (using 1 of 2 cores).

    md: xor using function: pIII_sse (5920.000 MB/sec)

 

Interesting stats, but just don't want anyone to think that this really means anything relative to unRAID performance.

 

The only time the CPU horsepower is important is if you are running other apps on the unRAID machine that require fast CPU (transcoding video. MP3 encoding, computing pi to the billionth decimal place, etc.), and then it won't be just the "xor" instruction that you'd want to be fast.

 

To give an idea, if your parity checks are running at 50 MB/sec (and that's fast for most users), INFINITELY FAST xor speed would only improve performance by at most 1% (that's if the CPU was running CONTINUOUS xors for the full second).  Your improvement would more likely be 0.001% or even less!  (Depending on how unRAID is coded, it could be that the xor functions are actually running while waiting for disk I/O, which could mean that unless the xor speed dropped below some threshold, that the performance improvement would be 0%.)  Remember that unRAID is only doing massive xors while while doing disk writes (which are far slower anyway because of the need to do 4 I/Os - 2 reads and 2 writes - for each logical write operation) and parity checks/rebuilds.  Just not a big deal guys.  If you want to speed up parity checks, get disks and network traffic off the PCI bus.  If you want to speed up writes, use a cache disk.

Interesting stats, but just don't want anyone to think that this really means anything relative to unRAID performance.

 

Nah, I don't think anyone's thinking that. If anything, I think this just goes to show that the recommendation of 2.0GHz on the hardware compatibility page is overkill for unRAID. That is, unless you load it up with a bunch of add-ons (VMWare, etc), of course.

 

 

model name      : Intel® Xeon® CPU            5148  @ 2.33GHz

stepping        : 11

cpu MHz        : 2000.000

 

Actually running at 2000Mhz

 

Oct 18 05:38:20 Tower kernel: md: xor using function: pIII_sse (7420.800 MB/sec)

 

 

CPU: AMD Athlon X2 Dual Core Processor BE-2400 stepping 02

 

2294.318 MHz

 

md: xor using function: p5_mmx (7652.000 MB/sec)

 

 

CPU: AMD Athlon Processor LE-1620 stepping 03

Detected 2411.099 MHz processor.

md: xor using function: p5_mmx (8040.800 MB/sec)