Efficiency - PowerPC G4 vs. G5

Apple hardware/software and related topics.
Forum rules
Any posts concerning pirated software or offering to buy/sell/trade commercial software are subject to removal.
Shiunbird
Donor
Donor
Posts: 497
Joined: Fri May 06, 2016 1:43 pm
Location: Czech Republic

Efficiency - PowerPC G4 vs. G5

Unread postby Shiunbird » Sat Mar 24, 2018 3:24 am

Hello gentlemen,

Yesterday eve I was super bored, and decided to force my G5 Quad on two cores only (both cores on the same CPU) and reduced performance mode and benchmark it against the G4 dual 1.25. Even though we all know the G4s are more efficient, some numbers still surprised me.

One of the numbers is against what I get from the power meter, and the other one is reduced by 50W, to account for the Quadro and the bunch of SSDs I have on my G5.

The G5 was eating 223W out of the wall with two cores out on reduced performance mode during the benchmarks.
The G4 was using 132W.

Of course, I know many other things make difference. My G5 has 8 RAM sticks, the G4 has 4. The expansion cards are different, so are the power supplies. But it gives an overall picture. I'm attaching here the benchmark results.

I'd love to hear your thoughts.

Edit: typo

results.zip
(3.3 KiB) Downloaded 7 times


Code: Select all

Geekbench overall - G4

0.11W watt per Geekbench

CPU only
Integer single threaded 985 -> 0.13
Integer multi threaded 1968 -> 0.07
Single threaded FP 846 -> 0.15
Multi threaded FP 1688 -> 0.07

XBench overall

1.73 watt per XBench

GCD loop -> 20W per Mops per second
Floating point basic -> 126W per Gflop per second
AltiVec -> 13W per Gflop per second
vecLib -> 53W per Gflop per second
FP library -> 18W per Mops per second


Code: Select all

Geekbench overall - G5

0.18 Geekbench Watt per Geekbench
0.14 Geekbench Watt per Geekbench

CPU only
Integer single threaded 843 -> 0.26 / 0.20
Integer multi threaded 1686 -> 0.13 / 0.10
Single threaded FP -> 918 -> 0.24 / 0.18
Multi threaded FP -> 1776 -> 0.12 / 0.09

XBench overall

2.63 watt per XBench
2.00 watt per XBench

GCD loop -> 74W/56W per Mops per second
Floating point basic -> 113W/86W per Gflop per second
AltiVec -> 34W/26W per Gflop per second
vecLib FFT -> 103W/78W per Gflop per second
FP library -> 21W/16W per Mops per second


Code: Select all

G5 vs. G4 delta of power consumption
(considering a G5 without a Quadro and without extra storage)

127% x G4 per Geekbench point

CPU only (Geekbench)
Integer single threaded -> 153%
Integer multi threaded -> 142%
Single threaded FP -> 120%
Multi threaded FP -> 128%

XBench overall

115% x G4 per XBench point

GCD loop -> 280%
Floating point basic -> 68% (G5 wins!)
AltiVec -> 200%
vecLib FFT -> 147%
FP library -> 89% (G5 wins!)

Average: 142%
G5 wins 2 out of 11 tests.
ImageImage

User avatar
ClassicHasClass
Donor
Donor
Posts: 2221
Joined: Wed Jul 25, 2012 7:12 pm
Location: Sunny So Cal
Contact:

Re: Efficiency - PowerPC G4 vs. G5

Unread postby ClassicHasClass » Sat Mar 24, 2018 4:17 pm

I don't find this in any way surprising. The 970 is just big and power-hungry.

But it's the fastest Power Mac, so.
smit happens.

:Fuel: bigred, 900MHz R16K, 4GB RAM, V12 DCD, 6.5.30
:Indy: indy, 150MHz R4400SC, 256MB RAM, XL24, 6.5.10
:Indigo2IMP: purplehaze, 175MHz R10000, Solid IMPACT
probably posted from Image bruce, Quad 2.5GHz PowerPC 970MP, 16GB RAM, Mac OS X 10.4.11
plus IBM POWER6 p520 * Apple Network Server 500 * RDI PrecisionBook * BeBox * Solbourne S3000 * Commodore 128 * many more...

Shiunbird
Donor
Donor
Posts: 497
Joined: Fri May 06, 2016 1:43 pm
Location: Czech Republic

Re: Efficiency - PowerPC G4 vs. G5

Unread postby Shiunbird » Sun Mar 25, 2018 6:34 am

Yes, definitely.

I guess it's the same case with the Intels of the time, NetBurst and all.
I have a dual 1.4 GHz Tualatin back across the ocean and I will give it a shot next time I fly there. I bet it's going to be the same case.
ImageImage

User avatar
guardian452
Donor
Donor
Posts: 3567
Joined: Tue Aug 21, 2007 10:12 pm
Location: United States

Re: Efficiency - PowerPC G4 vs. G5

Unread postby guardian452 » Mon Mar 26, 2018 4:42 am

Shiunbird wrote:The G5 was eating 223W out of the wall with two cores out on reduced performance mode during the benchmarks.
The G4 was using 132W...

I'd love to hear your thoughts.

Power consumption when idle or even standby/off is going to be more important. I'm guessing you're not using these old ladies for 24x7 number crunching? What's the kwh/month of each machine as it is used?

P3 was much more efficient than P4, that's why Intel scrapped Netburst and went back to the old designs for Pentium M and Core, especially with the popularity of portables at the time. Who knows about the rest of the machine though.

User avatar
shutitalldown
Posts: 177
Joined: Sat Feb 10, 2018 3:28 am

Re: Efficiency - PowerPC G4 vs. G5

Unread postby shutitalldown » Mon Mar 26, 2018 5:38 am

The first gen of PowerMacG4 (aka "graphite") was the best in terms of efficiency.
They were silent and passive cooled, and they didn't eat a lot of electricity.
Of course, they were PPC7410 @ 400/500Mhz.
Good for servers. Not good for dense computing.

Removing the videoCard (it requires two hacks on MDD) has saved 40Watt of electricity.
I am tattooed in reverse, cause if you break rules you don't know where it leads.
But, say, all the venus in furs are so manically depressed and manically dressed.

Shiunbird
Donor
Donor
Posts: 497
Joined: Fri May 06, 2016 1:43 pm
Location: Czech Republic

Re: Efficiency - PowerPC G4 vs. G5

Unread postby Shiunbird » Tue Mar 27, 2018 2:20 am

shutitalldown wrote:Removing the videoCard (it requires two hacks on MDD) has saved 40Watt of electricity.


So you managed headless booting in the end?
What was the trick?
(sorry if I missed it in some other thread)
ImageImage

User avatar
shutitalldown
Posts: 177
Joined: Sat Feb 10, 2018 3:28 am

Re: Efficiency - PowerPC G4 vs. G5

Unread postby shutitalldown » Tue Mar 27, 2018 4:23 am

Shiunbird wrote:So you managed headless booting in the end?


yes. The AGP card consumes 50Watt and it's not requested in a server configuration

Shiunbird wrote:What was the trick?


OForth patch, console redirection, plus Yaboot hack.

Yaboot is mysteriously broken at the moment, due to a strange bug that involves the ecosystem and it's not clear which part is causing the failure. Recompiling it fails with a common recipe.
I am tattooed in reverse, cause if you break rules you don't know where it leads.
But, say, all the venus in furs are so manically depressed and manically dressed.

User avatar
ClassicHasClass
Donor
Donor
Posts: 2221
Joined: Wed Jul 25, 2012 7:12 pm
Location: Sunny So Cal
Contact:

Re: Efficiency - PowerPC G4 vs. G5

Unread postby ClassicHasClass » Wed Mar 28, 2018 3:48 am

I'm guessing you're not using these old ladies for 24x7 number crunching?


My quad G5 actually *is* on 24/7, largely because there are files on it I need to sometimes remotely access, and I can do some work on the browser remotely over coffee breaks.
smit happens.

:Fuel: bigred, 900MHz R16K, 4GB RAM, V12 DCD, 6.5.30
:Indy: indy, 150MHz R4400SC, 256MB RAM, XL24, 6.5.10
:Indigo2IMP: purplehaze, 175MHz R10000, Solid IMPACT
probably posted from Image bruce, Quad 2.5GHz PowerPC 970MP, 16GB RAM, Mac OS X 10.4.11
plus IBM POWER6 p520 * Apple Network Server 500 * RDI PrecisionBook * BeBox * Solbourne S3000 * Commodore 128 * many more...

User avatar
guardian452
Donor
Donor
Posts: 3567
Joined: Tue Aug 21, 2007 10:12 pm
Location: United States

Re: Efficiency - PowerPC G4 vs. G5

Unread postby guardian452 » Wed Mar 28, 2018 4:14 am

ClassicHasClass wrote:
I'm guessing you're not using these old ladies for 24x7 number crunching?


My quad G5 actually *is* on 24/7, largely because there are files on it I need to sometimes remotely access, and I can do some work on the browser remotely over coffee breaks.


Exactly, you’re not using 100% CPU, so idle and/or standby consumption is much more important.

User avatar
ClassicHasClass
Donor
Donor
Posts: 2221
Joined: Wed Jul 25, 2012 7:12 pm
Location: Sunny So Cal
Contact:

Re: Efficiency - PowerPC G4 vs. G5

Unread postby ClassicHasClass » Wed Mar 28, 2018 2:45 pm

Well, either way, under "sitting around" load it pulls around 220W. That's with 16GB of RAM, an SSD and a spinning disk, CPUs at Reduced performance and a Quadro FX4500 video card.
smit happens.

:Fuel: bigred, 900MHz R16K, 4GB RAM, V12 DCD, 6.5.30
:Indy: indy, 150MHz R4400SC, 256MB RAM, XL24, 6.5.10
:Indigo2IMP: purplehaze, 175MHz R10000, Solid IMPACT
probably posted from Image bruce, Quad 2.5GHz PowerPC 970MP, 16GB RAM, Mac OS X 10.4.11
plus IBM POWER6 p520 * Apple Network Server 500 * RDI PrecisionBook * BeBox * Solbourne S3000 * Commodore 128 * many more...

Shiunbird
Donor
Donor
Posts: 497
Joined: Fri May 06, 2016 1:43 pm
Location: Czech Republic

Re: Efficiency - PowerPC G4 vs. G5

Unread postby Shiunbird » Thu Mar 29, 2018 6:13 am

Power consumption when off doesn't give any readings on my meter: 0W. The G4 uses 8W connected to mains.
I never leave it on standby, so I don't know.

I use mine 3-4 hours per day.
ImageImage

User avatar
johnnym
Donor
Donor
Posts: 229
Joined: Sun Sep 04, 2016 9:53 pm

Re: Efficiency - PowerPC G4 vs. G5

Unread postby johnnym » Thu Mar 29, 2018 1:03 pm

Shiunbird wrote:Yesterday eve I was super bored, and decided to force my G5 Quad on two cores only (both cores on the same CPU) and reduced performance mode and benchmark it against the G4 dual 1.25. Even though we all know the G4s are more efficient, some numbers still surprised me.

Interesting, I didn't anticipate this considerable efficiency difference but looking into a few of my older benchmark runs on related machines - running Debian GNU/Linux Sid though - these indeed show really astonishing differences.

I have results for OpenSSL and 7z for an Xserve G4 (Slot Load), an Xserve G5 and a Power Macintosh G4 (Quicksilver 2002):

Xserve G4 w/2 x 1.33 GHz PowerPC 7455

Code: Select all

root@xserve-g4:~# uname -a
Linux xserve-g4 4.12.0-2-powerpc-smp #1 SMP Debian 4.12.12-2 (2017-09-11) ppc GNU/Linux

root@xserve-g4:~# lscpu
Architecture:          ppc
CPU op-mode(s):        32-bit
Byte Order:            Big Endian
CPU(s):                2
On-line CPU(s) list:   0,1
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             2
Model:                 3.3 (pvr 8001 0303)
Model name:            7455, altivec supported
BogoMIPS:              83.07
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              2048K

root@xserve-g4:~# cat /proc/cpuinfo
processor   : 0
cpu      : 7455, altivec supported
clock      : 1333.333328MHz
revision   : 3.3 (pvr 8001 0303)
bogomips   : 83.07

processor   : 1
cpu      : 7455, altivec supported
clock      : 1333.333328MHz
revision   : 3.3 (pvr 8001 0303)
bogomips   : 83.07

total bogomips   : 166.14
timebase   : 41537183
platform   : PowerMac
model      : RackMac1,2
machine      : RackMac1,2
motherboard   : RackMac1,2 MacRISC3 Power Macintosh
detected as   : 128 (XServe rev. 2)
pmac flags   : 00000000
L2 cache   : 256K unified
pmac-generation   : NewWorld
Memory      : 2048 MB

root@xserve-g4:~# time openssl speed -elapsed
[...]
OpenSSL 1.1.0f  25 May 2017
built on: reproducible build, date unspecified
options:bn(64,32) rc4(char) des(long) aes(partial) blowfish(ptr)
compiler: gcc -DDSO_DLFCN -DHAVE_DLFCN_H -DNDEBUG -DOPENSSL_THREADS -DOPENSSL_NO_STATIC_ENGINE -DOPENSSL_PIC -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DVPAES_ASM -DPOLY1305_ASM -DOPENSSLDIR="\"/usr/lib/ssl\"" -DENGINESDIR="\"/usr/lib/powerpc-linux-gnu/engines-1.1\""
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
md2                  0.00         0.00         0.00         0.00         0.00         0.00
mdc2                 0.00         0.00         0.00         0.00         0.00         0.00
md4               6977.38k    26864.09k    91383.30k   228075.86k   405020.67k   428305.07k
md5              22918.05k    68216.38k   152853.08k   221644.80k   255112.53k   257906.01k
hmac(md5)         7870.65k    27949.72k    84613.29k   171508.74k   244785.15k   252455.59k
sha1             21759.58k    63419.20k   138473.64k   196239.36k   223272.96k   225476.61k
rmd160            5585.13k    18550.06k    48064.34k    79762.77k    98790.06k   100444.84k
rc4              79883.41k    93961.79k    97705.47k    99198.98k    99732.14k    99729.41k
des cbc          22149.96k    23927.21k    24395.43k    24507.39k    24515.93k    24477.70k
des ede3          7938.62k     8149.85k     8205.82k     8217.94k     8224.77k     8224.77k
idea cbc             0.00         0.00         0.00         0.00         0.00         0.00
seed cbc         31295.77k    35536.38k    36626.43k    36872.53k    36975.96k    36973.23k
rc2 cbc          19214.17k    20566.51k    20919.89k    21010.09k    21042.52k    21037.06k
rc5-32/12 cbc        0.00         0.00         0.00         0.00         0.00         0.00
blowfish cbc     44495.85k    51624.19k    53868.12k    54445.40k    54629.72k    54580.57k
cast cbc         36695.17k    41403.97k    42360.58k    42733.57k    43444.91k    43423.06k
aes-128 cbc      31792.54k    35433.60k    36516.69k    36766.72k    36891.31k    36896.77k
aes-192 cbc      27845.01k    30586.39k    31432.02k    31609.51k    31703.04k    31703.04k
aes-256 cbc      24781.93k    26932.33k    27583.57k    27723.43k    27792.73k    27705.83k
camellia-128 cbc    36850.73k    41959.27k    43476.22k    43856.21k    43988.31k    43963.73k
camellia-192 cbc    30753.64k    34229.27k    35236.86k    35483.99k    35569.66k    35553.28k
camellia-256 cbc    30755.09k    34233.81k    35279.96k    35530.41k    35594.24k    35547.82k
sha256           13110.17k    32676.37k    62889.98k    82106.37k    90122.92k    90761.90k
sha512            4056.44k    16155.73k    23693.82k    32712.36k    36839.42k    37164.37k
whirlpool         2772.11k     5860.76k     9657.09k    11557.55k    12255.23k    12298.92k
aes-128 ige      30968.89k    34507.22k    35547.31k    35769.00k    35883.69k    35831.81k
aes-192 ige      27226.16k    29928.13k    30708.48k    30871.89k    30954.84k    30916.61k
aes-256 ige      24289.46k    26420.50k    27025.66k    27153.75k    27219.29k    27186.52k
ghash            45286.49k    53686.02k    56329.13k    57032.02k    57232.04k    57245.70k
[...]
real   19m10.473s
user   19m8.748s
sys   0m0.250s

root@xserve-g4:~# time 7z b -mmt1

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,32 bits,2 CPUs BE)

BE
CPU Freq:  1319  1326  1317  1326  1325  1324  1322  1324  1324

RAM size:    2015 MB,  # CPU hardware threads:   2
RAM usage:    435 MB,  # Benchmark threads:      1

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

22:        801   100    782    779  |      20012   100   1713   1709
23:        754   100    771    769  |      19408   100   1685   1680
24:        721   100    778    775  |      18731   100   1648   1644
25:        699   100    800    799  |      17543   100   1562   1561
----------------------------------  | ------------------------------
Avr:             100    783    781  |              100   1652   1649
Tot:             100   1217   1215

real   1m40.891s
user   1m36.412s
sys   0m4.125s

root@xserve-g4:~# time 7z b

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,32 bits,2 CPUs BE)

BE
CPU Freq:  1320  1325  1325  1326  1325  1324  1324  1323  1324

RAM size:    2015 MB,  # CPU hardware threads:   2
RAM usage:    441 MB,  # Benchmark threads:      2

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

22:       1578   189    813   1536  |      39723   200   1699   3392
23:       1401   181    791   1428  |      38747   200   1681   3354
24:       1358   183    797   1460  |      37237   199   1643   3269
25:       1382   192    821   1579  |      34807   199   1558   3098
----------------------------------  | ------------------------------
Avr:             186    805   1501  |              199   1645   3278
Tot:             193   1225   2389

real   0m59.317s
user   1m43.935s
sys   0m4.821s


Xserve G5 w/2 x 2.0 GHz PowerPC 970FX

Code: Select all

root@xserve-g5:~# uname -a
Linux xserve-g5 4.12.0-2-powerpc64 #1 SMP Debian 4.12.12-2 (2017-09-11) ppc64 GNU/Linux

root@xserve-g5:~# lscpu
Architecture:          ppc64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Big Endian
CPU(s):                2
On-line CPU(s) list:   0,1
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             2
NUMA node(s):          1
Model:                 3.0 (pvr 003c 0300)
Model name:            PPC970FX, altivec supported
CPU max MHz:           2000.0000
CPU min MHz:           1800.0000
L1d cache:             32K
L1i cache:             64K
L2 cache:              512K
NUMA node0 CPU(s):     0,1

root@xserve-g5:~# cat /proc/cpuinfo
processor       : 0
cpu             : PPC970FX, altivec supported
clock           : 2000.000000MHz
revision        : 3.0 (pvr 003c 0300)

processor       : 1
cpu             : PPC970FX, altivec supported
clock           : 2000.000000MHz
revision        : 3.0 (pvr 003c 0300)

timebase        : 33333333
platform        : PowerMac
model           : RackMac3,1
machine         : RackMac3,1
motherboard     : RackMac3,1 MacRISC4 Power Macintosh
detected as     : 339 (XServe G5)
pmac flags      : 00000000
L2 cache        : 512K unified
pmac-generation : NewWorld

root@xserve-g5:~# time openssl speed -elapsed
[...]
OpenSSL 1.1.0f  25 May 2017
built on: reproducible build, date unspecified
options:bn(64,64) rc4(char) des(int) aes(partial) blowfish(ptr)
compiler: gcc -DDSO_DLFCN -DHAVE_DLFCN_H -DNDEBUG -DOPENSSL_THREADS -DOPENSSL_NO_STATIC_ENGINE -DOPENSSL_PIC -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM
 -DAES_ASM -DVPAES_ASM -DPOLY1305_ASM -DOPENSSLDIR="\"/usr/lib/ssl\"" -DENGINESDIR="\"/usr/lib/powerpc64-linux-gnu/engines-1.1\""
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
md2                  0.00         0.00         0.00         0.00         0.00         0.00
mdc2                 0.00         0.00         0.00         0.00         0.00         0.00
md4               8723.67k    31527.13k    92332.89k   178112.51k   244271.79k   250839.04k
md5              23901.43k    60959.06k   114535.85k   146770.94k   159864.15k   160262.83k
hmac(md5)         8866.84k    28802.26k    75131.82k   125627.73k   156235.09k   159017.64k
sha1             28473.09k    76915.18k   152388.18k   203088.21k   225312.77k   227076.78k
rmd160            5889.41k    17200.68k    37206.87k    52586.15k    59785.22k    60375.04k
rc4             188974.87k   216784.09k   226998.44k   230111.91k   230787.75k   230883.33k
des cbc          32278.11k    33505.32k    33832.36k    33915.90k    33917.61k    33909.42k
des ede3         12275.23k    12427.48k    12499.54k    12517.72k    12517.38k    12517.38k
idea cbc             0.00         0.00         0.00         0.00         0.00         0.00
seed cbc         36486.93k    37780.16k    38347.09k    38497.28k    38565.21k    38567.94k
rc2 cbc          15808.10k    16173.48k    16264.19k    16291.84k    16299.35k    16237.03k
rc5-32/12 cbc        0.00         0.00         0.00         0.00         0.00         0.00
blowfish cbc     54222.31k    57660.37k    58779.39k    59031.89k    59135.32k    59140.78k
cast cbc         50577.90k    53187.80k    54059.35k    54277.46k    54359.38k    54351.19k
aes-128 cbc      49054.66k    51025.66k    52328.87k    52755.46k    52759.21k    52822.02k
aes-192 cbc      39605.24k    42533.93k    43823.62k    44264.45k    44414.29k    44346.03k
aes-256 cbc      36460.65k    37242.88k    37943.98k    38157.99k    38259.37k    38245.72k
camellia-128 cbc    50629.19k    56996.03k    59131.56k    59627.18k    59801.60k    59823.45k
camellia-192 cbc    41773.81k    44992.85k    46348.80k    46685.87k    46828.20k    46814.55k
camellia-256 cbc    41766.90k    44977.79k    46364.93k    46736.04k    46836.39k    46852.78k
sha256           17182.59k    41151.53k    74566.23k    95128.92k   103721.64k   104278.70k
sha512           14564.63k    58182.68k    93964.20k   136922.45k   158272.17k   159836.84k
whirlpool         6707.26k    14405.76k    24225.54k    28861.44k    30436.01k    30605.31k
aes-128 ige      48831.15k    50712.36k    51492.01k    51437.91k    51623.25k    51904.51k
aes-192 ige      41554.01k    42800.75k    43431.51k    43402.58k    43551.40k    43843.58k
aes-256 ige      36180.04k    37126.42k    37575.94k    37536.77k    37724.16k    37907.11k
ghash            88201.57k    95960.19k    99611.99k   100863.66k   101291.35k   101122.05k
[...]
real    18m56.975s
user    18m53.693s
sys     0m1.904s

root@xserve-g5:~# time 7z b -mmt1

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs BE)

BE
CPU Freq:   989   991   990   991   991   990   990   990

RAM size:    2958 MB,  # CPU hardware threads:   2
RAM usage:    435 MB,  # Benchmark threads:      1

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

22:        908   100    884    884  |      18356   100   1568   1567
23:        855   100    873    872  |      17911   100   1551   1550
24:        825   100    889    888  |      17383   100   1527   1526
25:        806   100    922    921  |      16649   100   1483   1482
----------------------------------  | ------------------------------
Avr:             100    892    891  |              100   1532   1531
Tot:             100   1212   1211

real    1m26.931s
user    1m25.105s
sys     0m1.705s 

root@xserve-g5:~# time 7z b

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs BE)

BE
CPU Freq:   989   991   979   986   991   991   990   989

RAM size:    2958 MB,  # CPU hardware threads:   2
RAM usage:    441 MB,  # Benchmark threads:      2

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

22:       1912   198    939   1861  |      36544   200   1562   3120
23:       1727   198    889   1760  |      35630   200   1545   3084
24:       1630   198    887   1753  |      34548   200   1518   3033
25:       1588   198    917   1814  |      33065   200   1475   2943
----------------------------------  | ------------------------------
Avr:             198    908   1797  |              200   1525   3045
Tot:             199   1216   2421

real    0m49.824s
user    1m33.205s
sys     0m2.123s


I currently don't have results for the STREAM benchmark on these machines - which would surely favour the G5 because of its much faster memory interface - but OpenSSL and 7z already show some interesting facts. E.g. the Xserve G4 is much faster for older hash algorithms than the Xserve G5. It's slower in e.g. AES encryption but in 7z decompression on the Xserve G4 is faster or on par with decompression on the Xserve G5 despite the lower clock rate. Compression speeds are close, but the PPC7455 also only runs at 1.33 GHz instead of 2.0 GHz for the PPC970FX.

Unfortunately 7z seems to not correctly detect the clock speed of the PPC970FX. Or do these CPUs clock down when idle?

The QS 2002 with upgraded CPU - it's the one I described in viewtopic.php?f=9&t=16730294&start=75#p7398989, a dual PowerPC 7447A running at 1800 MHz currently - also outperforms the Xserve G5 in OpenSSL for older hashing algorithms and decompression speeds but is slower in e.g. AES encryption and 7z compression (don't mind the prompt, I used the disk from the Xserve G4 in the QS 2002):

Code: Select all

root@xserve-g4:~# uname -a
Linux xserve-g4 4.12.0-2-powerpc-smp #1 SMP Debian 4.12.12-2 (2017-09-11) ppc GNU/Linux

root@xserve-g4:~# lscpu
Architecture:          ppc
CPU op-mode(s):        32-bit
Byte Order:            Big Endian
CPU(s):                2
On-line CPU(s) list:   0,1
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             2
Model:                 1.1 (pvr 8003 0101)
Model name:            7447A, altivec supported
BogoMIPS:              66.43
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K

root@xserve-g4:~# cat /proc/cpuinfo
processor   : 0
cpu      : 7447A, altivec supported
clock      : 466.666665MHz
revision   : 1.1 (pvr 8003 0101)
bogomips   : 66.43

processor   : 1
cpu      : 7447A, altivec supported
clock      : 466.666665MHz
revision   : 1.1 (pvr 8003 0101)
bogomips   : 66.43

total bogomips   : 132.86
timebase   : 33217116
platform   : PowerMac
model      : PowerMac3,5
machine      : PowerMac3,5
motherboard   : PowerMac3,5 MacRISC2 MacRISC Power Macintosh
detected as   : 69 (PowerMac G4 Silver)
pmac flags   : 00000010
L2 cache   : 256K unified
pmac-generation   : NewWorld
Memory      : 1536 MB
root@xserve-g4:~# time openssl speed -elapsed
[...]
OpenSSL 1.1.0f  25 May 2017
built on: reproducible build, date unspecified
options:bn(64,32) rc4(char) des(long) aes(partial) blowfish(ptr)
compiler: gcc -DDSO_DLFCN -DHAVE_DLFCN_H -DNDEBUG -DOPENSSL_THREADS -DOPENSSL_NO_STATIC_ENGINE -DOPENSSL_PIC -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DVPAES_ASM -DPOLY1305_ASM -DOPENSSLDIR="\"/usr/lib/ssl\"" -DENGINESDIR="\"/usr/lib/powerpc-linux-gnu/engines-1.1\""
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
md2                  0.00         0.00         0.00         0.00         0.00         0.00
mdc2                 0.00         0.00         0.00         0.00         0.00         0.00
md4               9171.49k    34947.75k   119111.25k   301748.91k   544765.27k   578185.90k
md5              31132.03k    92601.81k   206951.51k   299971.24k   344978.77k   348706.13k
hmac(md5)        10695.56k    37957.59k   114751.15k   232317.95k   331096.06k   341442.56k
sha1             29528.46k    86126.12k   187430.49k   265537.19k   301929.81k   304911.70k
rmd160            7372.60k    24312.02k    63802.97k   107080.70k   133406.72k   135828.82k
rc4             107984.95k   127050.03k   132083.97k   134124.89k   134834.86k   134856.70k
des cbc          29966.86k    32341.31k    32969.39k    33128.45k    33174.87k    33166.68k
des ede3         10722.86k    11010.73k    11085.82k    11105.62k    11111.08k    11108.35k
idea cbc             0.00         0.00         0.00         0.00         0.00         0.00
seed cbc         42385.25k    48061.29k    49521.66k    49903.96k    49998.51k    49998.51k
rc2 cbc          25966.19k    27807.10k    28285.35k    28407.81k    28445.35k    28448.09k
rc5-32/12 cbc        0.00         0.00         0.00         0.00         0.00         0.00
blowfish cbc     60134.91k    69762.67k    72801.96k    73598.63k    73839.96k    73804.46k
cast cbc         49612.39k    56157.27k    58093.48k    58602.15k    58758.49k    58731.18k
aes-128 cbc      42920.04k    47844.76k    49373.10k    49774.59k    49872.90k    49883.82k
aes-192 cbc      37628.11k    41353.92k    42491.22k    42786.82k    42860.54k    42866.01k
aes-256 cbc      33490.61k    36409.58k    37291.18k    37519.02k    37576.70k    37454.59k
camellia-128 cbc    49819.20k    56728.62k    58778.45k    59314.86k    59468.46k    59452.07k
camellia-192 cbc    41577.30k    46279.96k    47636.82k    47989.42k    48087.04k    48076.12k
camellia-256 cbc    41578.31k    46280.75k    47635.88k    47980.20k    48084.31k    48076.12k
sha256           17752.78k    44251.03k    85044.99k   110900.22k   121856.00k   122721.62k
sha512            5501.31k    21861.40k    32041.90k    44251.14k    49821.01k    50266.11k
whirlpool         3752.06k     7898.37k    13068.20k    15647.40k    16575.15k    16635.22k
aes-128 ige      41876.18k    46656.45k    48049.07k    48429.74k    46729.90k    47546.37k
aes-192 ige      36809.90k    40460.89k    41517.65k    41791.83k    40523.09k    41123.84k
aes-256 ige      32851.75k    35719.06k    36538.28k    36750.68k    35763.54k    36235.95k
ghash            62367.78k    72927.49k    76248.23k    77130.07k    77359.79k    77398.02k
[...]
real   19m10.918s
user   19m10.615s
sys   0m0.073s

root@xserve-g4:~# time 7z b -mmt1

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,32 bits,2 CPUs BE)

BE
CPU Freq:  1776  1783  1790  1789  1789  1789  1790  1783  1788

RAM size:    1507 MB,  # CPU hardware threads:   2
RAM usage:    435 MB,  # Benchmark threads:      1

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

22:        758   100    739    738  |      24411   100   2085   2084
23:        717   100    732    731  |      23517   100   2036   2036
24:        690   100    743    742  |      22174   100   1947   1947
25:        671   100    767    766  |      20829   100   1855   1854
----------------------------------  | ------------------------------
Avr:             100    745    744  |              100   1981   1980
Tot:             100   1363   1362

real   1m46.457s
user   1m42.096s
sys   0m4.309s

root@xserve-g4:~# time 7z b

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,32 bits,2 CPUs BE)

BE
CPU Freq:  1786  1790  1790  1789  1788  1790  1790  1790  1790

RAM size:    1507 MB,  # CPU hardware threads:   2
RAM usage:    441 MB,  # Benchmark threads:      2

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

22:       1402   195    700   1364  |      47159   198   2030   4026
23:       1319   194    692   1344  |      44962   199   1951   3892
24:       1252   193    698   1347  |      42791   199   1890   3757
25:       1197   193    709   1368  |      39618   199   1776   3526
----------------------------------  | ------------------------------
Avr:             194    700   1356  |              199   1912   3800
Tot:             196   1306   2578

real   1m7.031s
user   2m0.805s
sys   0m6.010s


@Shiunbird:
Would be interesting to compare these to results for your machines. 7z and OpenSSL 1.1.x would be available from Homebrew.
:Indigo: :Indy: :O2: :Octane: :Octane2: :O200: = :O200: - :O200: = :O200: (O200 cluster w/2 GIGAchannel cabinets)
[ ( hp ) ] 712/80 c3000 (dead) :hpserv: (J5600) c3700 c3750 c8000 rp2470 rp3440 :rx2600: (rx2620) rx2660 rx4640
| d | i | g | i | t | a | l | AXPpci33 AlphaStation 200 AlphaStation 255 PWS 500au AlphaServer DS20E AlphaServer DS25
C O B A L T Qube 2 Qube 3 RaQ RaQ 2 RaQ 4r RaQ XTR

Shiunbird
Donor
Donor
Posts: 497
Joined: Fri May 06, 2016 1:43 pm
Location: Czech Republic

Re: Efficiency - PowerPC G4 vs. G5

Unread postby Shiunbird » Fri Mar 30, 2018 1:27 am

johnnym wrote:@Shiunbird:
Would be interesting to compare these to results for your machines. 7z and OpenSSL 1.1.x would be available from Homebrew.


Challenge accepted.
I wonder how much of that is software optimization vs. architecture differences.
ImageImage

User avatar
shutitalldown
Posts: 177
Joined: Sat Feb 10, 2018 3:28 am

Re: Efficiency - PowerPC G4 vs. G5

Unread postby shutitalldown » Fri Mar 30, 2018 2:48 am

The real bottleneck of the MDD seems to be on the PCI-X arbitration.
I am tattooed in reverse, cause if you break rules you don't know where it leads.
But, say, all the venus in furs are so manically depressed and manically dressed.

User avatar
johnnym
Donor
Donor
Posts: 229
Joined: Sun Sep 04, 2016 9:53 pm

Re: Efficiency - PowerPC G4 vs. G5

Unread postby johnnym » Sat Mar 31, 2018 1:00 pm

Shiunbird wrote:Challenge accepted.

Added results for a Power Mac G5 type 7,3 with two slightly faster PPC970FX running at 2.3 GHz and newer Linux kernel and OS parts. Now also includes results for STREAM using this OpenMP version of STREAM:

Code: Select all

root@powermac-g5-2:~# uname -a
Linux powermac-g5-2 4.15.0-2-powerpc64 #1 SMP Debian 4.15.11-1 (2018-03-20) ppc64 GNU/Linux

root@powermac-g5-2:~# lscpu
Architecture:        ppc64
CPU op-mode(s):      32-bit, 64-bit
Byte Order:          Big Endian
CPU(s):              2
On-line CPU(s) list: 0,1
Thread(s) per core:  1
Core(s) per socket:  1
Socket(s):           2
NUMA node(s):        1
Model:               3.0 (pvr 003c 0300)
Model name:          PPC970FX, altivec supported
CPU max MHz:         2300.0000
CPU min MHz:         1800.0000
L1d cache:           32K
L1i cache:           64K
L2 cache:            512K
NUMA node0 CPU(s):   0,1

root@powermac-g5-2:~# cat /proc/cpuinfo
processor   : 0
cpu      : PPC970FX, altivec supported
clock      : 2300.000000MHz
revision   : 3.0 (pvr 003c 0300)

processor   : 1
cpu      : PPC970FX, altivec supported
clock      : 2300.000000MHz
revision   : 3.0 (pvr 003c 0300)

timebase   : 33333333
platform   : PowerMac
model      : PowerMac7,3
machine      : PowerMac7,3
motherboard   : PowerMac7,3 MacRISC4 Power Macintosh
detected as   : 336 (PowerMac G5)
pmac flags   : 00000000
L2 cache   : 512K unified
pmac-generation   : NewWorld

root@powermac-g5-2:~# time openssl speed -elapsed
[...]
OpenSSL 1.1.0h  27 Mar 2018
built on: reproducible build, date unspecified
options:bn(64,64) rc4(char) des(int) aes(partial) blowfish(ptr)
compiler: gcc -DDSO_DLFCN -DHAVE_DLFCN_H -DNDEBUG -DOPENSSL_THREADS -DOPENSSL_NO_STATIC_ENGINE -DOPENSSL_PIC -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DVPAES_ASM -DPOLY1305_ASM -DOPENSSLDIR="\"/usr/lib/ssl\"" -DENGINESDIR="\"/usr/lib/powerpc64-linux-gnu/engines-1.1\""
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
md2                  0.00         0.00         0.00         0.00         0.00         0.00
mdc2                 0.00         0.00         0.00         0.00         0.00         0.00
md4              14360.83k    49128.00k   131499.95k   225878.02k   285343.74k   290914.30k
md5              27650.25k    70777.56k   132338.26k   169063.42k   183956.82k   185090.05k
hmac(md5)         9381.37k    31095.66k    82897.41k   141931.18k   179271.00k   182692.52k
sha1             32880.21k    90747.41k   177921.96k   236055.89k   261275.65k   262821.21k
rmd160            8783.47k    23635.84k    47015.85k    62471.51k    69077.67k    69610.15k
rc4             217511.43k   248822.53k   260875.78k   263839.40k   265030.31k   264885.59k
des cbc          36054.70k    38025.41k    38783.57k    38976.51k    39029.42k    39010.30k
des ede3         14170.11k    14317.21k    14342.31k    14417.24k    14439.77k    14434.30k
idea cbc             0.00         0.00         0.00         0.00         0.00         0.00
seed cbc         41119.69k    43211.43k    44093.53k    44319.40k    44389.72k    44378.79k
rc2 cbc          18177.78k    18605.29k    18712.83k    18736.81k    18746.03k    18748.76k
rc5-32/12 cbc        0.00         0.00         0.00         0.00         0.00         0.00
blowfish cbc     62402.83k    66424.75k    67758.08k    68086.10k    68182.02k    68179.29k
cast cbc         58139.14k    61140.89k    62172.93k    62434.99k    62409.39k    62455.81k
aes-128 cbc      53884.93k    57891.75k    59896.83k    60465.15k    60678.14k    60620.80k
aes-192 cbc      46468.83k    48766.06k    50294.19k    50879.15k    51049.81k    50970.62k
aes-256 cbc      40459.98k    42598.72k    43591.08k    43865.09k    44127.57k    44001.96k
camellia-128 cbc    61844.65k    65810.79k    68454.14k    69008.73k    69323.43k    69200.55k
camellia-192 cbc    49216.42k    52042.60k    53448.87k    53800.62k    53862.40k    53837.82k
camellia-256 cbc    48861.71k    52077.18k    53491.71k    53819.73k    53911.55k    53908.82k
sha256           19273.62k    47301.59k    85548.29k   109242.71k   118631.08k   119313.75k
sha512           16485.27k    67408.79k   107878.31k   157378.90k   181687.64k   183670.10k
whirlpool         7565.41k    16429.80k    27602.01k    33305.60k    35378.52k    35826.35k
aes-128 ige      55015.02k    58055.55k    59648.60k    59923.80k    59891.71k    59665.07k
aes-192 ige      46973.67k    49086.40k    50274.13k    50508.12k    50479.10k    50315.26k
aes-256 ige      41033.22k    42663.25k    43357.70k    43444.22k    43412.14k    43319.30k
ghash           102211.16k   110282.65k   114720.00k   115703.81k   115979.61k   115867.65k
[...]
real   18m54.469s
user   18m52.221s
sys   0m1.536s

root@powermac-g5-2:~# time 7z b -mmt1

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs BE)

BE
CPU Freq:  1137  1139  1139  1139  1138  1136  1139  1138  1138

RAM size:    3470 MB,  # CPU hardware threads:   2
RAM usage:    435 MB,  # Benchmark threads:      1

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

22:       1027   100   1000   1000  |      21085   100   1801   1800
23:        980   100    999    999  |      20598   100   1783   1783
24:        946   100   1018   1018  |      19957   100   1752   1752
25:        925   100   1057   1056  |      19080   100   1698   1698
----------------------------------  | ------------------------------
Avr:             100   1018   1018  |              100   1759   1758
Tot:             100   1389   1388

real   1m19.044s
user   1m17.632s
sys   0m1.365s

root@powermac-g5-2:~# time 7z b

7-Zip 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs BE)

BE
CPU Freq:  1137  1139  1139  1139  1139  1139  1135  1137  1138

RAM size:    3470 MB,  # CPU hardware threads:   2
RAM usage:    441 MB,  # Benchmark threads:      2

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

22:       2071   197   1025   2015  |      42007   200   1796   3587
23:       1920   198    990   1957  |      40922   200   1774   3542
24:       1811   197    987   1948  |      39651   200   1742   3481
25:       1775   198   1023   2027  |      37870   200   1688   3371
----------------------------------  | ------------------------------
Avr:             197   1006   1987  |              200   1750   3495
Tot:             199   1378   2741

real   0m47.646s
user   1m27.985s
sys   0m1.714s

root@powermac-g5-2:~/git-projects/STREAM# time ./stream_c.exe
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
 The *best* time for each kernel (excluding the first iteration)
 will be used to compute the reported bandwidth.
-------------------------------------------------------------
Number of Threads requested = 2
Number of Threads counted = 2
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 53181 microseconds.
   (= 53181 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function    Best Rate MB/s  Avg time     Min time     Max time
Copy:            2048.5     0.078803     0.078106     0.079628
Scale:           2025.8     0.079501     0.078980     0.081083
Add:             2156.0     0.111508     0.111317     0.111805
Triad:           2162.5     0.111251     0.110985     0.111733
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------

real   0m4.113s
user   0m7.824s
sys   0m0.256s

root@powermac-g5-2:~/git-projects/STREAM# export OMP_NUM_THREADS=1
root@powermac-g5-2:~/git-projects/STREAM# time ./stream_c.exe
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
 The *best* time for each kernel (excluding the first iteration)
 will be used to compute the reported bandwidth.
-------------------------------------------------------------
Number of Threads requested = 1
Number of Threads counted = 1
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 50636 microseconds.
   (= 50636 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function    Best Rate MB/s  Avg time     Min time     Max time
Copy:            2050.8     0.078542     0.078018     0.080427
Scale:           2036.0     0.078683     0.078586     0.078777
Add:             2105.1     0.114454     0.114010     0.116834
Triad:           2097.6     0.114602     0.114416     0.115324
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------

real   0m4.234s
user   0m4.058s
sys   0m0.171s

I should add that STREAM has a high variance in results on this specific machine, e.g. the following box and whisker plot was created from the results of 500 consecutive runs on both CPUs:
plot.jpg


Will add additional STREAM results for the other machines when time allows.
:Indigo: :Indy: :O2: :Octane: :Octane2: :O200: = :O200: - :O200: = :O200: (O200 cluster w/2 GIGAchannel cabinets)
[ ( hp ) ] 712/80 c3000 (dead) :hpserv: (J5600) c3700 c3750 c8000 rp2470 rp3440 :rx2600: (rx2620) rx2660 rx4640
| d | i | g | i | t | a | l | AXPpci33 AlphaStation 200 AlphaStation 255 PWS 500au AlphaServer DS20E AlphaServer DS25
C O B A L T Qube 2 Qube 3 RaQ RaQ 2 RaQ 4r RaQ XTR


Return to “Apple”

Who is online

Users browsing this forum: No registered users and 1 guest