Video chip Vicky capability question

Description of your first forum.
bzuidgeest
Posts: 41
Joined: Wed May 01, 2019 6:51 pm

Re: Video chip Vicky capability question

Post by bzuidgeest » Fri May 10, 2019 10:51 am

tomxp411 wrote:
Thu May 09, 2019 11:46 pm
stef wrote:
Tue May 07, 2019 5:55 pm
Now, trying to do anything in Bitmap with the Foenix will be very challenging and if I were you I would try to stay away. The Power of the CPU and the size and the speed of the screen refresh is too disproportionate. You just don't have enough bandwidth to peek and poke in the video memory @ full frame rate. THis is not happening. THis is the reason, there is a gazillions amount of sprites and tiles, this is to give the developer the tools to avoid going and start moving shit around in the bitmap memory. If the Video Resolution would be 320 x 200 then it would be fine, but not @ 640 x 480 @ 60FPS. And yes, you can move the video memory pointer with the Video Memory ($B0:0000 - $EF:FFFF) since you have 4Megs of it and yes you can do double/triple buffering, but the way VICKY's core is implemented to support all these tiles and Sprites makes the access to the video memory during playback very limited.

Stef, It's funny how the same problems keep cropping up on different machines. A few weeks ago, I read one of the updates on the Commander 16... and sure enough, David and Lorin are talking about memory bandwidth and GPU issues - the same kinds of discussions we were having last year.
I think that is not surprising. Everytime I do the math on graphics data I am astounded about how much data has to be moved around. And it grows exponentially when resolution and color depth and frame-rate increase. I once thought sound would be the more complex one to do, but sound is very low bandwidth to video data.
ZeroByte
Posts: 13
Joined: Thu Apr 25, 2019 5:13 pm

Re: Video chip Vicky capability question

Post by ZeroByte » Fri May 10, 2019 8:05 pm

In general, sound tends to be time/latency critical whereas video tends to be bandwidth critical. Video can drop a frame here and there and the impact is not bad except for the framerate purists. If the audio pipeline experiences either underflow or overflow, though, the results can be downright unintelligible. These rules of thumb are from a networking perspective, but still hold true within the system itself. My old 386 couldn't keep up with a couple of 32-track screamtracker modules I had that featured stereo panning - basically it amounts to 64 streams getting mixed. My Windows player would go over 100% cpu load and the audio would break a lot during sections of the music with tons of effects in use simultaneously. My DOS player could keep up though - but that's understandable. ;)
User avatar
stef
Posts: 64
Joined: Thu Jan 01, 1970 12:00 am
Location: Somewhere in the Universe
Contact:

Re: Video chip Vicky capability question

Post by stef » Fri May 10, 2019 10:58 pm

tomxp411 wrote:
Thu May 09, 2019 11:46 pm
stef wrote:
Tue May 07, 2019 5:55 pm
Now, trying to do anything in Bitmap with the Foenix will be very challenging and if I were you I would try to stay away. The Power of the CPU and the size and the speed of the screen refresh is too disproportionate. You just don't have enough bandwidth to peek and poke in the video memory @ full frame rate. THis is not happening. THis is the reason, there is a gazillions amount of sprites and tiles, this is to give the developer the tools to avoid going and start moving shit around in the bitmap memory. If the Video Resolution would be 320 x 200 then it would be fine, but not @ 640 x 480 @ 60FPS. And yes, you can move the video memory pointer with the Video Memory ($B0:0000 - $EF:FFFF) since you have 4Megs of it and yes you can do double/triple buffering, but the way VICKY's core is implemented to support all these tiles and Sprites makes the access to the video memory during playback very limited.

Stef, It's funny how the same problems keep cropping up on different machines. A few weeks ago, I read one of the updates on the Commander 16... and sure enough, David and Lorin are talking about memory bandwidth and GPU issues - the same kinds of discussions we were having last year.
Tom,

It would surprising if they didn't have the same issue. 14Mhz is not very fast and we are kind of force to use the VGA resolution since this is the only one that survived the longest. If we could overclock the 65C816 to 28 or 56Mhz, then it would be way more acceptable. However, I would need to change the static RAM to Dynamic Ram which would bring other issues.

Hence the reason why has so many sprites and tiles...

Cheers

Stefany
Mistress of All Villainy
User avatar
stef
Posts: 64
Joined: Thu Jan 01, 1970 12:00 am
Location: Somewhere in the Universe
Contact:

Re: Video chip Vicky capability question

Post by stef » Fri May 10, 2019 11:02 pm

ZeroByte wrote:
Fri May 10, 2019 8:05 pm
In general, sound tends to be time/latency critical whereas video tends to be bandwidth critical. Video can drop a frame here and there and the impact is not bad except for the framerate purists. If the audio pipeline experiences either underflow or overflow, though, the results can be downright unintelligible. These rules of thumb are from a networking perspective, but still hold true within the system itself. My old 386 couldn't keep up with a couple of 32-track screamtracker modules I had that featured stereo panning - basically it amounts to 64 streams getting mixed. My Windows player would go over 100% cpu load and the audio would break a lot during sections of the music with tons of effects in use simultaneously. My DOS player could keep up though - but that's understandable. ;)
ZeroByte,

yes the sound is very interrupt driven although, I am not concerned with the Foenix when it comes the OPL2s and SID but I begin to be a bit more concerned about the CODEC that could be very limited in its usage since even it is setup @ its slowest capture rate or playback rate like 32Khz @ 16Bits, it is going to take a lot of data and to fill the buffer memory for the DAC and ADC which in terms will need a lot of transfer doing the vertical blanking.

Anyway will see...

Stefany
Mistress of All Villainy
Post Reply