SGI Performance Comparisons

Image Processing Tests With Neko GIMP V2.2.8

Last Change: 04/Aug/2010

NOTE: I have not yet run these tests on older R4K systems. If you wish to compare GIMP performance on R4K systems, or between an R4K and a non-R4K system, then consult the pages for the GIMP V1.0.4 tests.
GIMP is a good program to use for a more complex benchmark task because it's easy to install, eays to use, and I can construct simple, repeatable tests. I use the Nekoware release of GIMP V2.2.8.

The tests are of two types:

Note that these tests are not described as specifically integer or floating-point; rather, they are referred to as image processing tests. This is because it's hard to know exactly what each test is doing in terms of int vs. fp calculations - this is especially true of the scripts which make use of multiple processing modules.

Together, this combination of tests offers a broader picture of how a system will perform for different imaging tasks.

For multi-CPU systems, only one CPU is used for these tests. Since some disk I/O is involved, the model of disk used for each system is shown after the tables below, though eventually I intend to redo the TYPE A tests to use the same model of disk in every system, namely a Seagate Cheetah ST336704LCV U160 36GB 1" 10000rpm SCA (I've done this recently for the Lynx compilation test). All O2, Octane and Fuel systems tested with 6.5.26m, while all Indy and Indigo2 systems tested with 6.5.22m. TYPE A tests were done with Tile Cache and New Image Size settings both set to 128MB. Note that GIMP does not support hardware accelerated processing, so in reality some tasks would be better done with a different application such as Pegamento or DJV, but such tasks still serve ok as CPU benchmarks.

All TYPE A tests are done using the 48MB image file that comes with the O2 'roam' demonstration program, called 'sanFran4k.tif' (a large aerial photograph of San Fransisco, USA), but for the purposes of the test I convert the image to RGB format using imgworks. The test file is thus a 4096 by 4096 RGB image of size 50,332,160 bytes (ignoring the image file header, exactly 48MB). You can download the file from the Depot Resources page as a gzip archive, but note that the file size is 39MB.

Here are the results. All times shown in the tables are in minutes and seconds, timed by hand with a stopwatch, ending when the image is fully redrawn.

TYPE B: Script-Fu -> Render -> Lava

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:23
Fuel      R16000   900     8MB     2      00:27
Fuel      R16000   800     4MB     2      00:30
Fuel      R16000   700     4MB     7      00:35
Fuel      R14000   600     4MB     2      00:40
Octane2   R14000   600     2MB     5      00:40
Octane2   R12000   400     2MB     1      01:01
O2        R7000    600  256K/1MB   2      01:08
O2        R12000   400     2MB     3      01:14
Octane    R12000   300     2MB     2      01:16
O2        R12000   380     2MB     2      01:18
Octane    R10000   250     1MB     1      01:31
O2        R7000    350  256K/1MB   6      01:42
Octane    R10000   195     1MB     1      02:08
Indigo2   R10000   195     1MB     4      02:14
O2        R5200    300     1MB     2      02:21
O2        R10000   195     1MB     2      02:22


TYPE B: Script-Fu -> Render -> Circuit

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:26
Fuel      R16000   900     8MB     2      00:29
Fuel      R16000   800     4MB     2      00:33
Fuel      R16000   700     4MB     7      00:38
Fuel      R14000   600     4MB     2      00:44
Octane2   R14000   600     2MB     5      00:44
O2        R7000    600  256K/1MB   2      01:05
Octane2   R12000   400     2MB     1      01:08
O2        R12000   400     2MB     3      01:12
O2        R12000   380     2MB     2      01:17
Octane    R12000   300     2MB     2      01:27
Octane    R10000   250     1MB     1      01:41
O2        R7000    350  256K/1MB   6      01:43
Octane    R10000   195     1MB     1      02:13
Indigo2   R10000   195     1MB     4      02:13
O2        R10000   195     1MB     2      02:15
O2        R5200    300     1MB     2      02:26


TYPE A: Filters -> Artistic -> Cubism

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      03:16
Fuel      R16000   900     8MB     2      03:54
Fuel      R16000   800     4MB     2      04:19
Fuel      R16000   700     4MB     7      05:09
Octane2   R14000   600     2MB     5      05:41
Fuel      R14000   600     4MB     2      05:47
Octane2   R12000   400     2MB     1      09:00
O2        R7000    600  256K/1MB   2      09:34
O2        R12000   400     2MB     3      11:00
O2        R12000   380     2MB     2      11:21
Octane    R10000   250     1MB     1      12:58
O2        R7000    350  256K/1MB   6      14:08
Octane    R10000   195     1MB     1      18:58
O2        R5200    300     1MB     2      19:21
Indigo2   R10000   195     1MB     4      19:38
O2        R10000   195     1MB     2      21:18


TYPE A: Filters -> Distorts -> Ripple

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:26
Fuel      R16000   900     8MB     2      00:31
Fuel      R16000   800     4MB     2      00:35
Fuel      R16000   700     4MB     7      00:39
Octane2   R14000   600     2MB     5      00:45
Fuel      R14000   600     4MB     2      00:46
Octane2   R12000   400     2MB     1      01:10
O2        R12000   400     2MB     3      01:32
O2        R7000    600  256K/1MB   2      01:35
O2        R12000   380     2MB     2      01:36
Octane    R10000   250     1MB     1      01:49
O2        R7000    350  256K/1MB   6      02:17
Octane    R10000   195     1MB     1      02:32
Indigo2   R10000   195     1MB     4      02:41
O2        R10000   195     1MB     2      02:46
O2        R5200    300     1MB     2      03:07


TYPE A: Filters -> Enhance -> Sharpen

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:07
Fuel      R16000   900     8MB     2      00:07
Fuel      R16000   800     4MB     2      00:07
Fuel      R16000   700     4MB     7      00:07
Octane2   R14000   600     2MB     5      00:09
Fuel      R14000   600     4MB     2      00:12
Octane2   R12000   400     2MB     1      00:16
Octane    R10000   250     1MB     1      00:24
O2        R12000   400     2MB     3      00:30
O2        R7000    600  256K/1MB   2      00:32
Octane    R10000   195     1MB     1      00:34
O2        R12000   380     2MB     2      00:35
O2        R7000    350  256K/1MB   6      00:35
Indigo2   R10000   195     1MB     4      00:41
O2        R10000   195     1MB     2      00:42
O2        R5200    300     1MB     2      00:59


TYPE A: Filters -> Colours -> Map -> AlienMap2

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:11
Fuel      R16000   900     8MB     2      00:13
Fuel      R16000   800     4MB     2      00:13
Fuel      R16000   700     4MB     7      00:14
Fuel      R14000   600     4MB     2      00:16
Octane2   R14000   600     2MB     5      00:17
Octane2   R12000   400     2MB     1      00:25
Octane    R10000   250     1MB     1      00:37
O2        R12000   400     2MB     3      00:39
O2        R12000   380     2MB     2      00:44
O2        R7000    600  256K/1MB   2      00:45
Octane    R10000   195     1MB     1      00:52
O2        R7000    350  256K/1MB   6      00:55
Indigo2   R10000   195     1MB     4      00:57
O2        R10000   195     1MB     2      00:57
O2        R5200    300     1MB     2      01:29


TYPE A: Filters -> Noise -> ScatterHSV

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:36
Fuel      R16000   900     8MB     2      00:40
Fuel      R16000   800     4MB     2      00:44
Fuel      R16000   700     4MB     7      00:52
Fuel      R14000   600     4MB     2      00:59
Octane2   R14000   600     2MB     5      01:01
Octane2   R12000   400     2MB     1      01:33
O2        R7000    600  256K/1MB   2      01:36
O2        R12000   400     2MB     3      01:46
O2        R12000   380     2MB     2      01:52
Octane    R10000   250     1MB     1      02:14
O2        R7000    350  256K/1MB   6      02:19
O2        R10000   195     1MB     2      03:10
Octane    R10000   195     1MB     1      03:11
Indigo2   R10000   195     1MB     4      03:11
O2        R5200    300     1MB     2      03:37


TYPE A: Filters -> Edge Detect -> Edge

System    CPU              L2/    Disk    Time
Name      Model    MHz     L3     Ref     mm:ss

Tezro     R16000  1000    16MB     2      00:25
Fuel      R16000   900     8MB     2      00:29
Fuel      R16000   700     4MB     7      00:30
Fuel      R16000   800     4MB     2      00:32
Octane2   R14000   600     2MB     5      00:33
Fuel      R14000   600     4MB     2      00:44
Octane2   R12000   400     2MB     1      00:54
Octane    R10000   250     1MB     1      01:18
O2        R12000   400     2MB     3      01:18
O2        R7000    600  256K/1MB   2      01:31
O2        R12000   380     2MB     2      01:40
O2        R7000    350  256K/1MB   6      01:42
Octane    R10000   195     1MB     1      01:55
Indigo2   R10000   195     1MB     4      02:01
O2        R10000   195     1MB     2      02:14
O2        R5200    300     1MB     2      02:33


Disk References:

 1: IBM DXHS36D 36GB 1.6" 7200rpm SCA
 2: Seagate Cheetah ST336704LCV U160 36GB 1" 10000rpm SCA
 3: IBM DDYS-T36950M 36GB 10000rpm SCA
 4: SGI SEAGATE ST3182032710
 5: SGI SEAGATE ST318406LC
 6: SGI IBM DDYS-T18350M 18GB 10000rpm SCA
 7: HITACHI HUS151414VL3600 147GB 15000rpm SCA
 8: IBM-ESXSBBD073C3ESTT0ZFNJP86 73GB 15000rpm SCA (Maxtor Atlas 15K II)
 9: SGI 9GB HVD 7200rpm 68pin


Feel free to send me results for other systems! I'll add more when I can.