From: Johan Klockars <>
Date: Fri, 26 Sep 1997 16:11:07 +0200 (MET DST)

> > > He are two Gembench results running the A/B at 32Mhz.
> >
> > It can be interesting to compare these results to an AB040 machine
> > using standard monochrome mode.
> Yes it would in some extent, but comparing a 16-bit mode in 1024x768
> with a 2-bit mode in 640x480 isn't realy fair. :-)

That was actually a 1-bit mode. ;-)

Anyway, it's the only mode it's reasonable to compare with, IMO, since the
computer has about the same amount of work to do.
If you like the Falcon speed in monochrome, the numbers show that it'll for
the most part be just as fast in 8 or 16 bit mode with a NOVA...

Since GEMBench compares with a Falcon without NVDI, which noone in their
right mind would use, the usual numbers are quite a bit inflated. I don't
think very many people use their Falcon's in 8 bit mode anyway, but if it
should be timed, it should definitely be with NVDI installed.

> I guess the screen resolution in 16-bit mode doesn't realy matter?
> Drawing a line in 1024x768 would be as fast as drawing it in 640x480.

If GEMBench doesn't draw longer lines or open larger windows.
I never checked that.

> > > VDI Text Effects: 0.510 2872%
> > > VDI Text Effects: 0.530 2764%
> > 0.370
> >
> > Rather similar results, for the same reason as above.
> > I don't quite see why the NOVA would lose here, though.
> Bad drivers, or what? If NVDI builds the text as a mono bitmap it
> should realy be the same result as for VDI text, or? The nova card

I would have thought so, but something a little different must be going on.

Hmm, perhaps NVDI does the effects while blitting (it's been a while since
I examined that code), but one would have hoped that the NOVA could have
still have done it faster.

> > > VDI Graphics: 0.920 1815%
> > > VDI Graphics: 0.950 1757%
> > 0.840
> >
> > I would have expected the NOVA to be much faster here, since most of the
> > work should be things that the hardware would be good at.
> > The drivers might not be as good as they should be, or perhaps there's
> > simply too much overhead sending commands to the board.
> So guess we need some new drivers here! Or might it be that the Nova card
> needs more calculations to draw the vdi graphics as it's using a much higher
> colour depth? On the other hand, there is realy not a big difference between
> 8-bit and 16-bit mode.

Those VDI graphics are some lines, ellipses and boxes. Nothing that would
require any extra calculations due to colours.

Perhaps the NOVA needs blitter setup for every line of the ellipse or box
that is drawn. It wouldn't be too surprising if that took longer than
simply throwing a few words onto the mono screen.

> It might be interresting to compare the result from the nova using 2-bit
> graphics and a falcon without a nova in mono to see what is realy going on.

Yes, if it's possible to run the NOVA in monochrome.

> > > GEM Window: 0.360 1041%
> > > GEM Window: 0.400 937%
> > 0.440
> >
> > Most of the work here should be window clearing and background redraw, IMO,
> > which should be fast blitter operations on the NOVA.
> > The small speed difference between 8 and 16 bit mode suggests that something
> > else is happening, though. Strange.
> If you use a background picture, this must slow down it alot as you have to
> blit the new background from ST-RAM/Fast-RAM to the Nova card. It can't blit it
> from one area on the screen to another which should be faster (if you are not
> clever and do just that but it can't be done in all cases), right?

I think a background image would have had far more impact on the numbers.
Perhaps there is some mono->8/16 bit expansion going on here for whatever
reason. That certainly would explain the numbers.

> > > VDI Scroll: 0.375 4000%
> > > VDI Scroll: 0.645 2325%
> > 1.160
> >
> > Ah, finally the NOVA hardware shows that it can do fast blits!
> Yeah, this is nice.

Looks like an internal bandwidth of something around 200 Mbyte/s or so
to use for the blitting.
(Eight times more memory to move than my mono test and done in a quarter
 of the time. The AB040 has 5-6 Mbyte/s in mono mode, IIRC.)

(The STB Velocity 128 (nVIDIA RIVA 128 chip) I'm thinking about bying for
 my PC has 1600 Mbyte/s total, which should leave far more than that for
 blitting... I'll test it when/if I buy one.)

> So what could differ from a Nova with DRAM and VRAM?

The things that seem to be limited by the bus bandwidth of the Falcon, like
the text tests, should not show much difference at all. The scrolling tests
might, but I wouldn't be surprised if 640x480x8bit is too little to make
a difference, unless extremely high refresh frequencies are used.
In 1024x768x16bit it's more likely that the output to the screen is going
to take it's toll on the DRAM version.

> You will be able to get higher refresh rates at higher resolutions, but

Uhm, you usually get the highest refresh rates at low resolutions.
You _need_ higher frequencies at higher resolutions, but it will be more
difficult for the card to give it to you.

What's the maximum refresh rates at different resolutions with the NOVA?
How about maximum dot clock frequency?
What chip is it using, by the way?

> are they much faster to actually display the screen? I don't realy care
> if I get 75Hz insteed of 110Hz at the same colour depth and resolution.
> Or will the actaull drawing to the screen be that much faster as well?

After a certain point, drawing to the screen will slow down when you increase
the refresh rate, since you're stealing bus bandwidth from the blitter etc.

> I guess a GEM-Bench test will tell us some of the truth.

Don't bet on it. ;-)

  Chalmers University   | Why are these |  e-mail:
     of Technology      |  .signatures  |  
                        | so hard to do |  WWW/ftp:
   Gothenburg, Sweden   |     well?     |            (MGIFv5, QLem, BAD MOOD)
Received on fr. sep. 26 1997 - 16:31:00 CEST

This archive was generated by hypermail 2.3.0 : ti. nov. 03 2015 - 20:07:53 CET