|
Content Copyright 2003, Jim Shorney, all rights reserved. Edited by Major Tom.
Comparative Benchmarks between the Type 3 486DX-50 and Type 4 P60 Complexes
This page is my attempt to answer the question of which is faster, a "stock"
T4 Pentium 60/66 or a T3 50 MHz complex fitted with an AMD 5x86 clocked at 150
MHz (x3 multiplier).
5x86 - AMD 5x86-133 @ 150 MHz (IBM PS/2 8595, Type 3 complex)
P60 - Pentium 60 (IBM PS/2 8595, Type 4 complex)
Sysbench 0.9.1d Benchmark Results
Operating system: IBM OS/2 Warp 4.0
Test Conditions: test platform was a PS/2 Model 95xp 486 8595-0xF,
48 MB parity RAM, OS/2 timeslice-32,32.
Benchmarks were run on the known stable Type-3 processor complex that
had been in service for several months in this box. The T3 was then removed
and replaced with a Type-4 Pentium 60 complex, and benchmarks re-run.
No changes were made to the operating system or other system hardware.
CPU Integer tests
| CPU |
Cache |
Dhrystone |
Hanoi |
Heapsort |
Sieve |
CpuI-marks |
| 5x86 |
256 K |
29.861 |
8.721 |
23.694 |
24.445 |
23.820 |
| P60 |
256K |
49.604 |
39.171 |
27.145 |
48.708 |
42.973 |
CPU Floating Point tests
| CPU |
Cache |
LinPack |
Flops |
FFT |
CpuF-marks |
| 5x86 |
256 K |
4.512 |
7.641 |
5.516 |
6.348 |
| P60 |
256K |
4.210 |
9.308 |
6.074 |
7.255 |
DIVE tests
XGA-2, 1 MB VRAM
| CPU |
Video Bus Bandwidth |
DIVE fun |
M->S DD 1.00:1 |
Dive-marks |
| 5x86 |
7.430 |
26.562 |
25.323 |
9.554 |
| P60 |
6.997 |
24.514 |
23.849 |
8.963 |
Video tests
1024x768x256
XGA-2, 1 MB VRAM
| CPU |
BitBlt S->S Copy |
BitBlt M->S Copy |
Filled Rect. |
Pattern Fill |
Vert. Lines |
Horiz. Lines |
Diag. Lines |
Text Render |
PM-marks |
| 5x86 |
13.815 |
4.246 |
13.041 |
36.546 |
2.734 |
4.810 |
3.476 |
6.947 |
10.133 |
| P60 |
13.855 |
3.592 |
13.059 |
35.946 |
2.778 |
4.955 |
3.534 |
6.725 |
9.982 |
Disk tests
5x86 (T3), IBM SCSI/Cache w/2 MB
| Drive |
Avg. Data Access time |
Cache/Bus xfer |
Avg. Transfer |
Disk use CPU load |
DiskIO-marks |
IBM 0661
379 MB |
35.100 |
2.950 |
1.417 |
11.000 |
12.164 |
IBM 0661
305 MB |
36.400 |
2.955 |
1.127 |
9.000 |
10.611 |
IBM 62RW100
817 MB |
32.100 |
2.951 |
1.299 |
12.000 |
11.653 |
P60 (T4), IBM SCSI/Cache w/2 MB
| Drive |
Avg. Data Access time |
Cache/Bus xfer |
Avg. Transfer |
Disk use CPU load |
DiskIO-marks |
IBM 0661
379 MB |
34.900 |
2.826 |
1.503 |
15.000 |
12.439 |
IBM 0661
305MB |
35.900 |
2.830 |
1.183 |
5.000 |
10.983 |
IBM 62RW100
817 MB |
32.200 |
2.825 |
1.325 |
6.000 |
11.845 |
Memory
|
5x86 (T3) |
P60 (T4) |
| 5 kB copy |
20.839 MB/s |
201.613 MB/s |
| 10 kB copy |
20.915 MB/s |
102.513 MB/s |
| 20 kB copy |
20.905 MB/s |
57.969 MB/s |
| 40 kB copy |
20.889 MB/s |
54.342 MB/s |
| 80 kB copy |
20.887 MB/s |
53.934 MB/s |
| 160 kB copy |
20.480 MB/s |
42.553 MB/s |
| 320 kB copy |
13.698 MB/s |
16.247 MB/s |
| 640 kB copy |
10.131 MB/s |
10.891 MB/s |
| 1280 kB copy |
10.101 MB/s |
10.878 MB/s |
| 5 kB read |
135.141 MB/s |
73.366 MB/s |
| 10 kB read |
127.885 MB/s |
63.951 MB/s |
| 20 kB read |
69.618 MB/s |
59.476 MB/s |
| 40 kB read |
69.623 MB/s |
59.217 MB/s |
| 80 kB read |
69.254 MB/s |
59.260 MB/s |
| 160 kB read |
53.862 MB/s |
58.331 MB/s |
| 320 kB read |
33.396 MB/s |
33.221 MB/s |
| 640 kB read |
24.125 MB/s |
26.372 MB/s |
| 1280 kB read |
24.127 MB/s |
26.388 MB/s |
| 5 kB write |
21.003 MB/s |
75.102 MB/s |
| 10 kB write |
20.974 MB/s |
75.020 MB/s |
| 20 kB write |
20.973 MB/s |
75.188 MB/s |
| 40 kB write |
20.993 MB/s |
75.038 MB/s |
| 80 kB write |
20.998 MB/s |
74.847 MB/s |
| 160 kB write |
20.956 MB/s |
59.260 MB/s |
| 320 kB write |
20.998 MB/s |
23.049 MB/s |
| 640 kB write |
20.998 MB/s |
16.903 MB/s |
| 1280 kB write |
20.992 MB/s |
16.875 MB/s |
| Total |
28.180 Mem-marks |
45.619 Mem-marks |
Landmark "Speed" (DOS) Benchmark Results
Disclaimer: In today's world of blinding
fast multi-cached CPU's and multi-tasking operating systems, Landmark Speed
numbers are virtually useless for quantifying a system's real-world performance.
They are presented here only as a relative indicator of CPU core performance
when executing multiple timed loops in a single-tasking environment.
The operating system used was IBM PC-DOS 2000, loaded in a "pure" configuration
with no memory managers or device drivers loaded.
5x86 (T3)
| |
Speed v2.0 |
Speed v6.0
|
| CPU |
225.49 MHz |
495.96 MHz |
| FPU |
807.69 MHz |
580.20 MHz |
| Video |
3963 chr/ms |
3947.95 chr/ms |
P60 (T4)
| |
Speed v2.0 |
Speed v6.0
|
| CPU |
346.5 MHz |
625.9 MHz |
| FPU |
1177.37 MHz |
1002.70 MHz |
| Video |
3963 chr/ms |
3963 chr/ms |
Micro2000 "Microscope" v6.5 Benchmark Results
Micro2000 uses their own custom kernel to run the Microscope diagnostic
program. I don't know a whole lot about it, but I present it here
as yet another set of numbers to confuse the issue.
5x86 (T3)
| |
Instructions/Second |
MHz |
| CPU |
66225 K |
134 |
| NPU |
27405 K |
148 |
| System Memory |
5771.9 KBPS |
| Text Mode |
145.0 KBPS |
| CGA |
106.1 KBPS |
| EGA |
220.0 KBPS |
P60 (T4)
| |
Instructions/Second |
MHz |
| CPU |
68015 K |
76 |
| NPU |
30959 K |
60 |
| System Memory |
6765 KBPS |
| Text Mode |
189 KBPS |
| CGA |
85 KBPS |
| EGA |
142 KBPS |
|