WinUAE and Atom

[DyLucke]

Hi there folks, i've recently acquired an Acer AspireOne, and i'm trying to set an emulated Amiga system on it using WinUAE.

However i've noticed WinUAE has some kind of problem on Atom, despite i can emulate easilly a relativelly high spec system, some audio sync problems appear on games and demos, specially using JIT. So i have to emulate with no JIT and using a 030 CPU, to supress the most of these problems.

Is this normal? Is anybody having the same issue? How could i fix that?

I'm using the last WinUAE release 1.6.1

Is Atom 1,6ghz powerful enough? ot maybe it's just a problem within WinUAE code? I'm emulating other systems and i don't have this issue but just on WinUAE.

Thanks!

[lilalurl]

http://eab.abime.net/showthread.php?t=43882

Turning off vsync seemed to have helped gaula88 who had similar issues.

[mds]

Hi, i have WiNUAE running on my Aspire ONE and it's running like a charme.
Much faster than on my older Celeron 2.8 GHZ Notebook.

[Photon]

Atom has plenty of power to emulate Amigas. What makes you suspect it? Do you have some info you're not sharing? The standard thing to do is attach the config that is failing and a bootlog.

Other netbook users having "the same problem as you" won't help you. Giving some #¤%&%&/( info will.

[DyLucke]

Originally Posted by lilalurl

http://eab.abime.net/showthread.php?t=43882

Turning off vsync seemed to have helped gaula88 who had similar issues.

Yep i've noticed that too, but it's not enough. Even turning off JIT helps a bit, but not enough either.

In fact the problem here comes with audio sync not the cpu nor video emulation itself. I can emulate without problems when no audio is involved in the programs used, but if you start a game or a demo, you will see the audio sync starts to get out of range, the more specs needed to run the program itself by the emulated Amiga the most out of sync tones you get.

Maybe it's just something related with the Acer drivers?
Audio drivers could be?

About the question about "if someone is having the same problem using atom", is as simple as this, i don't think more info is needed. I've read in another thread about the "power-hungry" features of certain releases of WinUAE, i don't know what's wrong here, drivers, or maybe emulator release.

So my question is directed to "someone" who could have the same problem and got it fixed.

Simple. XD

[Photon]

Well, if you had attached a config we could have seen your settings for the audio, like buffer size and if you use '100% accurate'.

Vsync is disabled on the gfx "card" drivers I hope? (Or set to application-controlled)

[Toni Wilen]

Delete/rename default.uae, run winuae, do not load any config, select a1200 quickstart, middle compatibility (which should enable jit), start some game (name MUST be included)

QS config is always good starting point without any extra settings. If it works, start modifying the config slowly until the problem happens again.

[DyLucke]

I have the same problem using Quickstart configs... Windowed is impossible to run, (but that happens with other emus also), however when i turn it to fullscreen i get the same issue.

And i've installed newer drivers already. I'm starting to think it's a configuration problem.
I do use WinUAE to run a virtual image of my real 1200, running the same hardware config. But it's possible settings here are not suitable for the Atom.

Here are my WinUAE display settings.

SCREEN:
--------
Fullscreen: mobile Intel (R) 945 express Chipset family (1024x600)
1024x600 32bit default refresh rate

Windowed: 680x580


SETTINGS:
-----------

Native mode: Fullscreen
RTG mode: Fullscreen

Black than black: unchecked
Filtered low res: unchecked
Remove interlaced artifacts: unchecked

Resolution: Hires (normal)

Refres: every frame

FPS adjust: 50


CENTERING:
-----------

Horizontal: checked
Vertical: checked

LINE MODE:
-----------
Double


These are my current settings, and they do work on my desktop computer, plus on another laptop i have (more powerful than the Atom). Actually no JIT, no GFX card emulation, just like an ordinary 030 1200... Any idea?

[TCD]

Have you deleted 'default.uae'? Any filter settings set?

[DyLucke]

No filters, and i've deleted the default.uae config.
I'm trying to set a proper config from zero and starting to figure out something... Let's see

If i don't use filters, i can achieve to get a nicelly displayed Hi-res Laced PAL desktop. Means an emulated 650x512 Amiga screen shown perfectly with borders, but i will have some out of sync audio values if i play something that features sound. Games, mods, or demos.

Out of sync values shown by "on-screen leds" are numbers that can vary from 01 to 10, no negative values here, so it means there's a delay between video and audio that can reach to the 10% using this WinUAE render mode (i assume GDI).


On the other hand if i use some filters, that means having D3D aid on rendering, i managed to suppress these sound problems but the problem is, all the stuff related with the screen scaling that uses 3D acceleration will distort any hi-res screen displayed by the emulator... Just lo-res are working properly. PLUS if you just turn off the scaling here, you're turning off 3D accel, so having the same problems than before...



It seems AspireOne has a really inneffective GDI mode, and D3D mode under WinUAE can't achieve a proper hi-res display on the 1024x600 AspireOne screen because unavoidably it will need to do scaling in order to use 3D accel. That's what i'm starting to think.

[TCD]

Originally Posted by Toni Wilen

Delete/rename default.uae, run winuae, do not load any config, select a1200 quickstart, middle compatibility (which should enable jit), start some game (name MUST be included)

Guess you should name one game (and which version) and let Toni have a look.

[Gaula92]

Please, DyLucke, try ASIO audio, too. With the smallest possible buffer (1) and 22050Hz sound, you shouldn't have any sound distortion. DirectSound doesn't work right for me in any realtek-equiped pc, be it an Atom or Core2DUO machine, causing distorted sound with very strange sound pitch changes. There's still another little sound problem if you activate vsync and use ASIO sound output, but that's another story and maybe you won't notice it.

[Toni Wilen]

Nothing uses GDI. GDI would be extremely slow. You can try changing display buffer setting in misc-panel (remember to restart emulation after changing it). Some display drivers do weird things..

D3D filter does not scale anything if you don't use any scaling (1x, sliders at middle etc..) All other filters use CPU for everything = much slower than D3D (as long as display card is not ancient)

"distort", "properly" -> always include screenshot.

-10% to 10% is perfectly fine SND% value. What exactly you mean by "sound problems"?

[DyLucke]

Then if D3D is working for SURE when 1x, and believe me, ive tested all the variables there, then the problem is more than probably caused by the audio driver used by winUAE.

Because Toni, believe there's no visible performance difference between using D3D pixelx1 or pixelx2 and not using D3D after several tests. That leads to the conclussion, Atom is powerful enough to render the screen without 3D accel aid, so the problem is not within the rendering itself but as Gaula pointed up here, some unefficient audio driver.

-10% ti 10% is not a perfectly fine sound value, allow me to disagree... It's more than enough to find strange reverberations that will annoy you a lot. Something above 5% of difference is clearly noticeable.

About what i said about distort and properly, just think about a TFT monitor trying to show something on lower resolution than the TFT native one. You know that, scattered doubled pixels around, just the same that happens when you try to scale a font to a size that will not fit perfectly.

[Photon]

I envy Toni's patience.

[john4p]

@DyLucke: I wouldn't be too surprised if everything runs ~perfect with WinUAE 1.4.6. Please try that. After 1.4.6 there were lots of code-changes. I'd like to know how well 1.4.6 works for you:

http://www.winuae.net/files/InstallWinUAE1460.exe

[Toni Wilen]

Originally Posted by DyLucke

Because Toni, believe there's no visible performance difference between using D3D pixelx1 or pixelx2 and not using D3D after several tests. That leads to the conclussion, Atom is powerful enough to render the screen without 3D accel aid, so the problem is not within the rendering itself but as Gaula pointed up here, some unefficient audio driver.

Another useless test. Of course you won't see any difference as long as there is enough CPU power left. (50% vs 95% = you still have full frame rate but nearly 2x CPU power usage difference..)

I can guarantee you see at least 10% difference between null filter and d3d.

-10% ti 10% is not a perfectly fine sound value, allow me to disagree... It's more than enough to find strange reverberations that will annoy you a lot. Something above 5% of difference is clearly noticeable.

It is perfectly fine if sound driver/card does not suck

About what i said about distort and properly, just think about a TFT monitor trying to show something on lower resolution than the TFT native one. You know that, scattered doubled pixels around, just the same that happens when you try to scale a font to a size that will not fit perfectly.

Reason why I always ask for windowed mode results. There is always reason for default testing configuration.

I envy Toni's patience.

"Buy a new PC" is most probable answer when I stop caring.

I'd like to know how well 1.4.6 works for you

I don't. Only exception if answer includes version that was last working with EXACT SAME configuration including both CPU% and SND% difference if any.

[john4p]

Originally Posted by Toni Wilen

I don't.

Of course not. But it might be interesting for DyLucke - if it works better on his netbook.

I'm sure WinUAE 1.6.1 is currently THE most accurate Amiga emulator available. Still 1.4.6 maybe shows better performance on weak PCs...

[Toni Wilen]

Originally Posted by john4p

Still 1.4.6 maybe shows better performance on weak PCs...

So far this has been a myth. "Everyone" says older versions are supposedly faster but nobody has proven it. Ever.

This is what annoys me. Either shut up or include hard facts (for example: version x uses X% of CPU in task manager but version y uses Y% of CPU in multiple tests with identical configuration without using newer versions features.)

[john4p]

Well, I'm not sure if that older version is faster. But couldn't the different source-code have an effect on DyLucke's sound-synch-problems? Or...umm...okay, I'll shut up.