@Seiya

Quote:

with Dosbox your X1000 emulate a 386/25 or 386/33

Well, it is faster than my 486SX 25 MHz notebook.

Are you sure you didn´t use 486? (Am386DX 40 MHz tested by author who compiled that benchmark scores 13.7 DMIPS)

@pavlor

This 386 use 128k L2 cache and for some benchmarks, L2 cache are very important.

howver, now i try with dosbox with 386 setting to see how works this drystone.

Last edited by Seiya on 14-Jul-2012 at 07:56 PM.

_________________

with 10000 cycles i have the same result of 386..

now we can confirm that 10000 is = 386@25 mhz

with a x86 cpu at 3,2 ghz dual core.

with 20000 cycles should be about 486 66 mhz:

19.06
52459
29,86

_________________

@Seiya

Quote:

now we can confirm that 10000 is = 386@25 mhz

I don´t know what magic your 386 use, but Am386DX 40 MHz scores 13.7 DMIPS with the same benchmark. Your 386 would be faster even than my own 486!

Quote:

with 20000 cycles should be about 486 66 mhz:

Speed increase in DosBox is linear with CPU cycles. According to your own logic, 10000 CPU cycles is equal to 486DX 33 MHz.

I highly suspect your 386 is 486.

@pavlor
there is a more simle test: Quake.

Quake in dosbox with 10000 cycles and 20000 cyles.
i rember my old 486dx2@66 mhz, with Tseng LAB ET4000/W32 512k and quake the timedemo take about 6 fps without sound.

_________________

@pavlor

Edit: there was a mistake in results. Frameskip = 1 was used to make benchmark. Here are results with fs=0:

http://sintonen.fi/pics/dhry2od_powerbook167.png

PowerBook G4 1.67 GHz, MorphOS 3, DosBox 0.73

Following values with frameskip=0
Dhrystone 2.1:
Microseconds 1 loop: 31.94
Dhrystones / second: 31311
VAX MIPS rating: 17.82

(Was following values with frameskip=1
Dhrystone 2.1:
Microseconds 1 loop: 29.38
Dhrystones / second: 34043
VAX MIPS rating: 19.38)

No idea if Piru is going to do Doom benchmark.

Last edited by itix on 15-Jul-2012 at 08:20 PM.
Last edited by itix on 15-Jul-2012 at 03:38 PM.

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@itix

Thanks!

That confirms my expectations: DosBox with similar optimalisations would have nearly the same speed on PA6T 1.8 GHz and G4 1.5 GHz. As we see here, CPU emulation speed difference between this (0.73) and previous (0.72) MorphOS version is clearly visible (60% !).

Could Piru try also DoomBenchmark?

Edit: I see Piru used frameskip 1, X1000 result was with FS 0 (however, I don´t think there would be big difference - around 1-2 DMIPS).

Last edited by pavlor on 15-Jul-2012 at 10:50 AM.

Quote:

PowerBook G4 1.67 GHz, MorphOS 3, DosBox 0.73

@Seiya

Quote:

there is a more simle test: Quake.

DosBox 40039 CPU cycles:
15.5 FPS (no sound)

DosBox 20000 CPU cycles:
7.8 FPS (no sound)

DosBox 10039 CPU cycles:
3.9 FPS (no sound)

@pavlor

Slow enough to try to play "Prisoner of Ice" on a real PC (after a failed attempt to share the amiga volume) !

Last edited by Cass on 15-Jul-2012 at 11:37 AM.

_________________
Ordell Robbie: Is she dead, yes or no?
Louis: Pretty much.

i confirm that dosbox emulation is the same both we use PPC or X86 cpu
Maybe is different how a cpu can handle cycles, however here my test on
intel core2 Duo E8500 (3.2 ghz)

with 20000 cycles, quake timedemo demo2 = 7,9 fps (with sound)
with 40000 cycles, quake timedemo demo2 = 15,9 (with sound)

_________________

@Seiya

Of course it should be the same with fixed CPU settings (only faster 2D GFX speed could reach a little bit better result).

@pavlor

you should try with max setting now :)

_________________

@Seiya

Quote:

you should try with max setting now :)

Core i5 of my brother will be faster than your Core 2.


Judging by Dhrystone benchmark results posted by Epsilon, I think Quake benchmark result of X1000 in DosBox with CPU cycles = max could be close to our 10000 CPU cycles settings.

I think now is safe to state that PA6T 1.8 GHz (and G4 1.5 GHz) are as fast in most tasks. I estimate SPEC numbers of both CPUs around 65 SpecInt95 (or 585 SpecInt2000 or 4 SpecInt2006).

@pavlor

with max setting = 114 fps

_________________

@pavlor

"PA6T 1.8 GHz (and G4 1.5 GHz) are as fast in most tasks"

Yes, especially in single core + CPU bound tasks.
Longer pipeline of PA6T shows.

(wider memory bandwidth etc. show up perhaps mainly after second core is taken in use, like the good performance scaling in blender test)

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

PowerMac G5, 970 1.8 GHz, MorphOS 3.2 (thanx venox!)

DOSBOX 0.74.1

VAX MIPS = 17.48
MICROSECOND FOR ONE RUN TROUGHT DHRYSTONE = 32.56
DHRYSTONES PER SECOND = 30710

POWER PC G5 DUAL 1.8 768 RAM DDR


Interesting. PowerBook G4 1.666 GHz seems to be a little bit faster.
If my estimates are right, G5 2.7 GHz could reach 26.22 DMIPS (486DX 50 MHz performance).

Simple graph with few posted results (DHRY2OD benchmark, see post 57 for more info):

@pavlor

howerver these test could be interesting to see how could be run the games.
will be interesting to see now how run with at high resolution not only like games, but also video and audio

_________________

@Seiya

Quote:

will be interesting to see now how run with at high resolution not only like games, but also video and audio

I don´t think lower 486DX performance is enough for high resolution games.