NX benchmark results wanted

[hotpaw4]

Hi,

I'm trying to update my benchmark results. Could someone with an NX try the latest release of yCPUBench and post the results (both armlet and emulated 68k)? The yCPUBench benchmark utility is on this web site:

http://www.hotpaw.com/rhn/palm

There's also a new beta release of the cbasPad5 armlet accelerated Basic interpreter on the above web site. Could someone try this benchmark on a NX?

10 t1 = fn timer(100)
20 x = 100000
30 y = 0
40 while (x > 0): x -= 1: y += 2: wend
70 t2 = fn timer(100)
80 print x, y, (t2 - t1)/100, "seconds"
90 end
run

The cbasPad beta on an NX might be over 50x faster than HotPaw Basic on a T615C.

Thanks!

[cbulock]

Heres the results from yCPU benchmark:

sieve8192 (68K)

Time: 43.80 mS/loop
Loops: 50 times
48 bogo_68k_MHz
result=1899

sieve8192 (armlet)

Time: 2.215 mS/loop
Loops: 2000 times
951 bogo_68k_MHz
result=1899

cplxifft256 (68k)

Time: 66.40 mS/loop
Loops: 25 times
60 bogo_68k_MHz
result=0

cplxifft256 (armlet)

Time: .945 mS/loop
Loops: 2000 times
4237 bogo_68k_MHz
result=0

scrn_read (68k)

Time: 49.90 mS/loop
Loops: 100 times
2052 kB/sec
102400 bytes/loop

scrn_read (armlet)

Time: 2.70 mS/loop
Loops: 1000 times
37925 kB/sec
102400 bytes/loop

scrn_write (68k)

Time: 40.30 mS/loop
Loops: 100 times
2540 kB/sec
102400 bytes/loop

scrn_write (armlet)

Time: 1.110 mS/loop
Loops: 1000 times
92252 kB/sec
102400 bytes/loop

mem_read (68k)

Time: 100.10 mS/loop
Loops: 100 times
2045 kB/sec
204800 bytes/loop

mem_read (armlet)

Time: 5.720 mS/loop
Loops: 1000 times
35804 kB/sec
204800 bytes/loop

mem_write (68k)

Time: 80.70 mS/loop
Loops: 100 times
2537 kB/sec
204800 bytes/loop

mem_write (armlet)

Time: 2.460 mS/loop
Loops: 1000 times
83252 kB/sec
204800 bytes/loop

ManBrot32 (68k)

Time 890.0 mS/loop
Loops: 1 time
170 kFlop/s
result=135

ManBrot32 (armlet)

Time: 637.0 mS/loop
Loops: 10 times
238 kFlop/s
result=135

ManBrot32v9 (68k)

Time 890.0 mS/loop
Loops: 1 time
170 kFlop/s
result=135

ManBrot32v9 (armlet)

Time 108.40 mS/loop
Loops: 100 times
1402 kFlop/s
result=135

ifft256v9 (68k)

Time: 66.40 mS/loop
Loops: 25 times
60 bogo_68K_MHz
result=0

ifft256v9 (armlet)

Time: 1.275 mS/loop
Loops: 2000 times
3140 bogo_68k_MHz
result=0

[cbulock]

The output from the cbasPad 5 app was:

0 200000 1.7 seconds

[hotpaw4]

Interesting. According to these results, the NX seems around 28% faster than a TT(m550) on the Codewarrior v9 compiled floating point Mandlebrot armlet test, but the TT was around 34% faster on the Codewarrior v9 compiled integer fft armlet test. The NX was faster on all the 68k tests (emulated by PACE).

On the cbasPad5 test, the NX seems to be around 15% faster, but I can't seem to get very reliable measurements on the TT. The reported times vary from 1.8 to 2.3 seconds from run to run (exiting cbasPad5 to some other app, and then returning and rerunning the benchmark again, imported from the MemoPad). Do the NX timings show any variation on successive runs? Or are is this benchmark more stable on the NX?

Thanks!

[cbulock]

I have tried running different programs and going back in, and the only variation I have noticed is from 1.69 to 1.74 seconds. I will try it some more later on and see if I get any different results.