Apollo 1240/40 Accelerator Card Review Part 2

===========================================================================
              Review: Apollo 1240/40 Accelerator Card Part 2
                            By: @{" Jason Compton " link JASON
===========================================================================

Recently, we took a look at the Apollo 1240/40 card.  Now it's time to take
a look at the numbers.

We ran three benchmark routines as well as the AIBB comparison suite.  The
benchmarks were run with FastExec installed, which moves exec.library to
RAM and gives a significant performance increase to the Apollo card.

The first test was a raytrace render of the Staircase example from
Cinema4D.  It was rendered in 320x400 resolution on the following machines:

An Amiga Tech A4000T with 14 megs of memory and a standard 3640 040/25 card
A Macrosystem Draco with 32 megs of memory and the Draco 060/50 card
An A1200 with 10 megs of memory and the Falcon 040/25 card
An A1200 with 10 megs of memory and the CyberStorm 060/50 card
An A1200 with 10 megs of memory and the Apollo 040/40 card

The time breakdown is as follows (in minutes:seconds)

Cyberstorm-- 1:26
Apollo--     2:31
Draco--      5:05
Falcon--     5:05
4000T--      6:01

The Cyberstorm wins the test, as well it should as it is an 060 card.
Furthermore, the Cyberstorm was running CyberPatcher, which enhances the
FPU performance of a select number of programs (of which Cinema4D is one).
The Apollo clocks in with performance far better than that of an 040/25
machine (either the Falcon or the 4000T).  It is worth noticing that the
Draco under Cinema4D is exactly as fast as the Falcon.  We ran this test a
number of times with the same result.  The answer seems to lie with
CyberPatcher saving the day for the Cyberstorm.  Draco users do not have a
similar recourse.

The Apollo, in this application, offers a great deal more performance than
anything else in its price class.

Moving along to the ImageFX tests, we ran two effects on the included
"guy.jpg" picture.  The first was the Sobel Edge Detect algorithm, run on
the 4000T, the Apollo, and the CyberStorm.  The time breakdown here, in
seconds.hundredths of seconds:

Cyberstorm-- 2.87
Apollo--     6.63
A4000T--     10.68

Here, the Apollo is working a great deal faster than the stock 040/25
A4000T but is not on the same pace as the 060/50 Cyberstorm.

The final test was run using the Crystallize effect with default settings.
Time breakdowns:

Cyberstorm-- 3.36
Apollo--     6.46
4000T--      10.73

As we saw in the first test, under this particular situation the Apollo
performs at better than 50% of the 060 speed.  (ImageFX, incidentally, is
not currently supported by the version of Cyberpatcher we used)  It is also
nearly twice as fast as the 4000T.

By these standards, then, the Apollo is a significantly speedy card.  The
AIBB benchmarks offer additional insight as well.

By and large, of course, in the comparison between the Cyberstorm 1260, the
Apollo, the Falcon, and a 4000T, the Cyberstorm rules the day.  But in most
tests, the Apollo outperforms its 040 cousins (the Falcon and 4000T) by at
least 20% and sometimes over 100%.  The MemTest, in particular, is won by
the Apollo, with the 4000T's poor memory system rating only 22% of the
Apollo's score.  Some graphically-oriented tests are fastest on the Apollo
as well, notably the WritePixel and EllipseTest.  But overall, the
CyberStorm 060 outscores the Apollo on most intense-mathematics tests,
usually at a level of 50-125%, with the 040/25 machines far behind.

There are a couple of concerns with the Apollo.  One is the lack of disable
option for pesky software that is not 040-friendly.  The other is the fact
that towards the end of our tests here at AR, the Apollo fan failed, making
it difficult to complete our tests.  This may of course just be an isolated
case.

The Apollo is retailing for well under US$500 these days, which in
comparison to other 040 options (the overpriced Falcon), slower 030 options
(which can reach over $300), and the high-end 060 choices (at least
US$800), is a good performance/price choice.  If you're looking for a
serious (but not end-all be-all) speed increase and are opposed to the
concept of spending more than you bought your 1200 for just to speed it up,
the Apollo is worth a serious look.

Provided for review by Datakompaniet

Datakompaniet AS
Trondheim Innovation Centre
N-7030 Trondheim
Norway
+47 7354 0375 voice
+47 7354 3861 fax
datakompaniet@interlink.no e-mail
http://www.interlink.no/datakompaniet (accepting online orders)

Amiga Report Main Page Amiga Web Directory