Slowdown in higher vid. modes

Hardware, coding, music, graphic and various applications

Moderators: Mug UK, moondog/.tSCc., [ProToS], lp, Moderator Team

ppera

Slowdown in higher vid. modes

Postby ppera » Sun Feb 03, 2008 3:51 pm

As I see, Falcon works slower in video modes with more res. , colors. At least when accessing slow RAM.
Can someone give exact slowdown depending on video mode(s)?

User avatar
Nyh
Atari God
Atari God
Posts: 1496
Joined: Tue Oct 12, 2004 2:25 pm
Location: Netherlands

Re: Slowdown in higher vid. modes

Postby Nyh » Sun Feb 03, 2008 4:24 pm

ppera wrote:As I see, Falcon works slower in video modes with more res. , colors. At least when accessing slow RAM.
Can someone give exact slowdown depending on video mode(s)?


No, that is not possible. Bottleneck is the bus to ST RAM. This bus is more occupied by the video chip. The processor sometimes has to wait to get to the data in ST RAM. Slowdown is high when moving big chunks of data in ST RAM (lot of bus time needed). Slowdown is almost zero when heavy computing is done in tight loops that fit in the 68030's processor cache (not much bus access needed).

Hans Wessels

ppera

Re: Slowdown in higher vid. modes

Postby ppera » Mon Feb 04, 2008 3:31 pm

That's all normal.
Asking it little more specific: how many % of CPU access (time) to slow RAM is blocked in higher video modes?

Or: What is slowdown by intensive access to slow RAM - for instance clearing display, or fill it ? (in higher resols. )

User avatar
Cyprian
10 GOTO 10
10 GOTO 10
Posts: 1699
Joined: Fri Oct 04, 2002 11:23 am
Location: Warsaw, Poland

Re: Slowdown in higher vid. modes

Postby Cyprian » Mon Feb 04, 2008 8:55 pm

on Czuba-tech site you can find detailed information about Videl's memory consumption:

F030 @ 16 MHz (fitting WITHOUT solders)
All ST-RAM WRITE accesses are done with 3 cycles instead of 4 !

VGA 16c (Ref : F030 VGA 16c) :
Linear 32bit read (ST-Ram) 5.893 MB/s 110 %
Linear 32bit write (ST-Ram) 8.171 MB/s 126 %
Linear 32bit copy (ST-Ram) 3.452 MB/s 106 %
Linear burst copy (ST-Ram) 3.385 MB/s 104 %

VGA 256c (Ref : F030 VGA 16c) :
Linear 32bit read (ST-Ram) 4.488 MB/s 84 %
Linear 32bit write (ST-Ram) 6.188 MB/s 95 %
Linear 32bit copy (ST-Ram) 2.627 MB/s 81 %
Linear burst copy (ST-Ram) 2.570 MB/s 79 %

440x480 TC (with CENTscreen 3) (Ref : F030 VGA 16c) :
Linear 32bit read (ST-Ram) 3.463 MB/s 65 %
Linear 32bit write (ST-Ram) 4.787 MB/s 74 %
Linear 32bit copy (ST-Ram) 2.046 MB/s 63 %
Linear burst copy (ST-Ram) 2.002 MB/s 62 %
Lynx II / Jaugar / TT030 / Mega STe / 800 XL / 1040 STe / Falcon030 / 65 XE / 520 STm / SM124 / SC1435
SDrive / PAK68/3 / Lynx Multi Card / LDW Super 2000 / XCA12 / SkunkBoard / CosmosEx / SatanDisk / UltraSatan / USB Floppy Drive Emulator / Eiffel / SIO2PC / Crazy Dots / PAM Net
Hatari / Steem SSE / Aranym / Saint
http://260ste.appspot.com/

mikro
Hardware Guru
Hardware Guru
Posts: 2013
Joined: Sat Sep 10, 2005 11:11 am
Location: Kosice, Slovakia
Contact:

Re: Slowdown in higher vid. modes

Postby mikro » Wed Mar 05, 2008 10:29 pm

I've written an article years ago based on Czuba's numbers: http://ihrisko.cyberio.org/~mikro/docs/ ... stram.html

Maybe it will be useful for you.

jd
Captain Atari
Captain Atari
Posts: 355
Joined: Thu Nov 09, 2006 12:38 pm
Location: Ruislip, UK

Re: Slowdown in higher vid. modes

Postby jd » Thu Mar 06, 2008 9:51 am

Does increasing the clock to the Videl help with the 'slowdown' in higher resolutions/colours or you just need a bus speeder?

Or even a seperate graphics card. Are these still available for the falcon?

ppera

Re: Slowdown in higher vid. modes

Postby ppera » Thu Mar 06, 2008 12:08 pm

Very useful, Mikro.

I don't want to make any mods to Falcon. Just wanted to clarify some slowdowns of SW what I observed in higher video modes.

User avatar
Cyprian
10 GOTO 10
10 GOTO 10
Posts: 1699
Joined: Fri Oct 04, 2002 11:23 am
Location: Warsaw, Poland

Re: Slowdown in higher vid. modes

Postby Cyprian » Thu Mar 06, 2008 12:52 pm

jd wrote:Does increasing the clock to the Videl help with the 'slowdown' in higher resolutions/colours or you just need a bus speeder?


definitely bus speeder is needed in this case
Lynx II / Jaugar / TT030 / Mega STe / 800 XL / 1040 STe / Falcon030 / 65 XE / 520 STm / SM124 / SC1435
SDrive / PAK68/3 / Lynx Multi Card / LDW Super 2000 / XCA12 / SkunkBoard / CosmosEx / SatanDisk / UltraSatan / USB Floppy Drive Emulator / Eiffel / SIO2PC / Crazy Dots / PAM Net
Hatari / Steem SSE / Aranym / Saint
http://260ste.appspot.com/

mikro
Hardware Guru
Hardware Guru
Posts: 2013
Joined: Sat Sep 10, 2005 11:11 am
Location: Kosice, Slovakia
Contact:

Re: Slowdown in higher vid. modes

Postby mikro » Thu Mar 06, 2008 8:36 pm

jd wrote:Does increasing the clock to the Videl help with the 'slowdown' in higher resolutions/colours or you just need a bus speeder?

Or even a seperate graphics card. Are these still available for the falcon?


On the contrary, it could have even opposite effect -- with increased videl clock you will probably want better resolution or higher framerate => bus will be even more used. Bus speeder is way to go, even 16 -> 25 MHz will give you fantastic speedup.

Graphics card is another story but unfortunately, until the arrive ctpci and/or supervidel there's nothing widely available.

simbo

Re: Slowdown in higher vid. modes

Postby simbo » Fri Mar 07, 2008 12:40 am

U43 U42 74LS244 buffers
if you use fast page ram {14mb ram board}
replace these two chips with 74HCT244
only later revision used HC or HCT{high speed cmos ttl} chips
older falcons rev1,2 {a,b even some c} use LS
and fast page wont like this
this is also true of a few other areas where 74 logic is employed in its series

always when avalible use HCT as new chips with any ST TT series

three types of falcon pcb ive seen are

pale green REV A light green REV -B dark green REVC-D
uk pcb is
rev-A uk
CA401332-000
rev-B
CA401332-001
rev-C
CA401332-002
rev-D
CA401332-003

simbo

Re: Slowdown in higher vid. modes

Postby simbo » Fri Mar 07, 2008 1:02 am

over the last while

ive wanted to find a project to do

in st land it proves to be hard with experienced minds in the software side
reflecting problems i see as hardware
and dont grasp fully as software

at last i see one
many members will love

i designed recently a single micro controlled 1hz resolution 50mhz ddc / a direct digital synth used as a clock
its a single ic
program set by i2c from any simple micro pic and an lcd
the lcd can display the setting and allow for control boundrys and display
the chips auto recal from power on and only change when programmed by the lcd or data to it
as fly control over clock {upset a few people why not}

i can expand this to five or more outputs to yield
independant centraly settable clock to 1hz for each chip

is this usefull ???

or what

any takers for a joint project ????

mikro
Hardware Guru
Hardware Guru
Posts: 2013
Joined: Sat Sep 10, 2005 11:11 am
Location: Kosice, Slovakia
Contact:

Re: Slowdown in higher vid. modes

Postby mikro » Fri Mar 07, 2008 9:02 am

Sorry for my inexperience but what is "50mhz ddc / a direct digital synth used as a clock" ? I use / code on falcon a lot but I really have no idea what to imagine under this label ;)

simbo

Re: Slowdown in higher vid. modes

Postby simbo » Fri Mar 07, 2008 1:23 pm

each ic has clock inputs
a ddc on each pin will allow any clock speeds to be supplied to each chip
like 32mhz for the cpu ... 48mhz for the dma etc etc
its simply a single chip with a filter on a board
above each clock input
all the boards connect thru two wires to a little micro and lcd
the micro has preset functionality
1} to stop you doing stupid things and set bourndrys
2} to allow for preset setups of the clocks
nemesis and phantom etc are ok at this but only allow two or three acceleration choice
doing this would allow fast processing of dma or fpu while steady system processing would occur at a normal or boosted main bus speed

ive used 32mhz 40mhz and 55mhz as the clocks
the only chip that complained was the fpu -33n type and i simply changed it for another and 55mhz is stable
{added heatsinks to the chips}

but there are a few other speeds of chipset that would be handy

over clocking the combel and videl on there own would be handy
the unit would probably be maybe 5ddc chips and one controller

User avatar
Nyh
Atari God
Atari God
Posts: 1496
Joined: Tue Oct 12, 2004 2:25 pm
Location: Netherlands

Re: Slowdown in higher vid. modes

Postby Nyh » Fri Mar 07, 2008 2:13 pm

simbo wrote:each ic has clock inputs
a ddc on each pin will allow any clock speeds to be supplied to each chip
like 32mhz for the cpu ... 48mhz for the dma etc etc

I think 32 mhz is not a useful frequency for a processor. Who would want a processor doing a single step every 31.25 seconds. Most people would appreciate a 32 MHz processor much more, doing 32 million steps every second. :wink:

But still you didn't explain what ddc is supposed to mean.
DDC = Display Data Channel (in video connectors)
DDC = Digital Down Conversion (in measurement)

But DDC in clocking processors? I really don't know.

Hans Wessels

User avatar
stimpy
Captain Atari
Captain Atari
Posts: 305
Joined: Mon Aug 22, 2005 2:47 pm
Location: Somerset, UK
Contact:

Re: Slowdown in higher vid. modes

Postby stimpy » Fri Mar 07, 2008 2:41 pm

simbo wrote:U43 U42 74LS244 buffers
if you use fast page ram {14mb ram board}
replace these two chips with 74HCT244
only later revision used HC or HCT{high speed cmos ttl} chips
older falcons rev1,2 {a,b even some c} use LS
and fast page wont like this


I'm sure all Falcons used FPM memory didn't they? Also, how does changing those chips improve anything, as I think they are only accessed once by TOS to check configuration?
Netus-Bee,Repairs,Upgrades,EtherNEC,Eiffel

simbo

Re: Slowdown in higher vid. modes

Postby simbo » Sat Mar 08, 2008 12:18 am

stimpy wrote:
simbo wrote:U43 U42 74LS244 buffers
if you use fast page ram {14mb ram board}
replace these two chips with 74HCT244
only later revision used HC or HCT{high speed cmos ttl} chips
older falcons rev1,2 {a,b even some c} use LS
and fast page wont like this


stimpy wrote:I'm sure all Falcons used FPM memory didn't they? Also, how does changing those chips improve anything, as I think they are only accessed once by TOS to check configuration?
??? oh ye??? you sure its not every time it want to read a bank of ram??

ok well dont change them then and have very slow bank access u42 is more important than u43 to change

however forinstance if you change monitor type a slow 74LS will not be able to keep up with the change and the ram can be selected at the wrong time
causing video frame loss { packets go missing} scsi 2 dma transfers are effectored so is dma transfers to sound

so does it each time the monitor type is checked by the system
u42 each time a floppy is inserted and removed and also the sound dma settings etc

replaceing both chips cures the ram problem i got anyway
it was my solution do it or not

later memory has a faster access time and with acceloration 74ls looses its edge
most people use some form of it
and 74LS has a 9ns propigation delay in any one gate
74HCT is much faster and provides cleaner fast page dram bank switching


ok its simple read this carefully

when revision A and B were released they used 74LS chips
and both these builds of machine are unstable for scsi transfers
however most rev c and D machines are VERY stable for everything
and most people used to them just cant see why this is

except atari themselfs used in rev C and D 74HCT chips for this sensitive area
I WONDER WHY
or maybe you will always wonder why ????
its good at this point to add 74LS chips are far cheeper in these years of making falcons
so why use 74hct other than for the plain fact it was needed

becouse ram had become an issue and 74LS isnt as good at performing the needed tasks
Last edited by simbo on Sat Mar 08, 2008 12:57 am, edited 3 times in total.

simbo

Re: Slowdown in higher vid. modes

Postby simbo » Sat Mar 08, 2008 12:33 am

Nyh wrote:
simbo wrote:each ic has clock inputs
a ddc on each pin will allow any clock speeds to be supplied to each chip
like 32mhz for the cpu ... 48mhz for the dma etc etc

I think 32 mhz is not a useful frequency for a processor. Who would want a processor doing a single step every 31.25 seconds. Most people would appreciate a 32 MHz processor much more, doing 32 million steps every second. :wink:

But still you didn't explain what ddc is supposed to mean.
DDC = Display Data Channel (in video connectors)
DDC = Digital Down Conversion (in measurement)

But DDC in clocking processors? I really don't know.

Hans Wessels


hans stay out the cafe on thursdays/fridays
:megaphone: dont be a small m yourself and grow up ive got a spare cb radio i can send you for amusement :P
you know fine well what a DDC is
for your benfit its simply Direct Digital Synthesis in the form of a clock output

ie a chip with a range of 0-50mhz in 1hz steps that can be i2c programmed
a ddc chip is a directed digital clock look for the chip AD9851 http://www.analog.com/UploadedFiles/Dat ... AD9851.pdf

hehe 32mips from a 680030 lol
680030 processors provide's a throughput ranging from 10 to 25 Mips,
or are we talking about steps ??? what steps ??? it depends what mode its in

mikro
Hardware Guru
Hardware Guru
Posts: 2013
Joined: Sat Sep 10, 2005 11:11 am
Location: Kosice, Slovakia
Contact:

Re: Slowdown in higher vid. modes

Postby mikro » Sat Mar 08, 2008 10:40 am

I second Hans, still no idea what is it :)

And about your idea -- hm, it looks nice but practically there's no use for it. Falcon chipset needs to preserve some ratios, i.e. when you clock DSP at 50 MHz and CPU at 20, 90% of falcon demos will stop to work. Phantom & co have the reason why they use frequency they use -- it's the highest stable frequency and still maintain 1:2 ratio (16:32 -> 25:50 MHz)

EDIT: ups, posted in the same time as simbo, ok, now I do have idea, what is that DDC :)


Social Media

     

Return to “Professionals”

Who is online

Users browsing this forum: No registered users and 0 guests