small performance comparison IIe vs 2x II

[Riccardo Mottola <riccardo.mottola%libero.it@localhost>]

Hi!

just to share a comparison, now that I upgraded my Ultra2 to 10.1 I can 
compare "apples with apples".

How does a Netra T1 compare to Ultra2?
General setup is similar. I "lost" some RAM modules for both, so RAM 
situation is a little anemic.

Netra T1:
[     1.000000] total memory = 1536 MB
[     1.000000] avail memory = 1486 MB
[     1.000000] mainbus0 (root): SUNW,UltraAX-i2 (Netra T1 200): hostid 
8305d938
[     1.000000] cpu0 at mainbus0: SUNW,UltraSPARC-IIe @ 500 MHz, CPU id 0
[     1.000000] cpu0: manuf 17, impl 13, mask 14
[     1.000000] cpu0: system tick frequency 11.111 MHz
[     1.000000] cpu0: 16K instruction (32 b/l), 16K data (32 b/l), 256K 
external (64 b/l)

[     3.490015] sd0 at scsibus0 target 0 lun 0: <QUANTUM, 
ATLAS10K2-TY367J, DA40> disk fixed
[     3.585683] sd0: 35020 MB, 17338 cyl, 10 head, 413 sec, 512 
bytes/sect x 71721820 sectors
[     3.690023] sd0: sync (50.00ns offset 31), 16-bit (40.000MB/s) 
transfers, tagged queueing



Ultra2:
[     1.000000] total memory = 1024 MB
[     1.000000] avail memory = 984 MB
[     1.000000] mainbus0 (root): SUNW,Ultra-2 (Sun Ultra 2 UPA/SBus): 
hostid 80c04df4
[     1.000000] cpu0 at mainbus0: SUNW,UltraSPARC-II @ 295.992 MHz, CPU id 0
[     1.000000] cpu0: manuf 17, impl 11, mask 20
[     1.000000] cpu0: 16K instruction (32 b/l), 16K data (32 b/l), 2048K 
external (64 b/l)
[     1.000000] cpu1 at mainbus0: SUNW,UltraSPARC-II @ 295.992 MHz, CPU id 1
[     1.000000] cpu1: manuf 17, impl 11, mask 20
[     1.000000] cpu1: 16K instruction (32 b/l), 16K data (32 b/l), 2048K 
external (64 b/l)


[     3.230023] sd0 at scsibus0 target 0 lun 0: <FUJITSU, MAJ3182M 
SUN18G, 0804> disk fixed
[     3.324193] sd0: 17274 MB, 7508 cyl, 19 head, 248 sec, 512 
bytes/sect x 35378533 sectors
[     3.430026] sd0: sync (100.00ns offset 15), 16-bit (20.000MB/s) 
transfers, tagged queueing

never got if the IIe is somehow a "cheap" version and how it does 
compare to the original II processor. It has only 256K cache compared to 
2MBytes.

300MHz x 2 = 600Mhz ? 20% faster?

The Ultra2 sure has a slower disk and less RAM.



Let's try a couple of things.
1. Compilation of GNUstep base library - should fit in RAM decently. I 
exclude the configure part since that is not paralle and slow anyway.
2. Running tests of GNUstep base.


NetraT1. (make -j1)
gnustep-base compile:
2605.78 real      2452.17 user       142.33 sys

gnustep-base tests:
5599.20 real      4640.70 user       854.28 sys


Ultra2: (make -j2)
gnustep-base compile:
1786.32 real      3201.10 user       178.55 sys

gnustep-base tests:
4973.47 real      6726.23 user      1158.18 sys


It shows that in raw compilation real time the IIi is ca. 1.4x slower 
than the 2x II!
Interesting is the full user time, might suppose that a single CPU is 
1.28 slower... but the clock ismuch less, so maybe the 2MB cache do 
wonders in compilation?
For running tests instead the difference is smaller. Still the dual CPU 
wins.

The T1 consumed for sure less electricity, just or the amount of 
whirring fans. Should meter it too once!


Riccardo