PDA

View Full Version : Is SysInfo Accurate?



Gouldin
24th March 2012, 15:10
Just been setting up my amigas and using sysinfo to see the specs etc. Thought I'd run a benchmark on my CD32/SX-1. The results suggest that it's twice as fast as a stock A1200. But aren't they pretty much identical in spec?

Hence my question, is sysinfo even accurate? :blink:

D.B
24th March 2012, 15:13
In a word, "no". :thumbsup:

J.T.Kirk
24th March 2012, 15:14
Well, SysInfo is an old program, so you will find some inaccurates when it comes to 060 processors.

But, does the SX-1 comes with extra Fast RAM? A stock A1200 has only 2MB of Chip RAM available.

fitzsteve
24th March 2012, 15:15
Your SX1 has fast Ram? That make all the difference so yea it probably is almost twice as fast at many operations.

As for the accurancy, I'm sure it's not spot on but should give you a good guide :)

Just sticking an 4/8mb Ram expansion in an A1200 gives it a nice little boost :thumbsup:

mfilos
24th March 2012, 15:22
SysInfo is okeish only till 030 CPUs.
For 040/060 it's better to use either SysSpeed or AIBB.

As Steve said, a fast RAM expansion really boosts the performance. SysInfo for your CPU will be accurate enough imho.

Zetr0
24th March 2012, 15:24
Sysinfo is great for an 040 or earlier CPU. its pretty reasonable in terms as long as you know HOW and WHAT its testing.

a Stock 020@14Mhz is (theoretically) capable of 3 MIPS, however without FAST RAM it is crippled and achieves about 1.1MIPS .

This is because the CPU has to wait its turn to read / write data to CHIP RAM, however, an A1200 with FAST RAM will achieve between 2.4 - 2.8 MIPS pending the card / adapter

Your CD32 with SX-1 Has FAST RAM, this is why its processing is nearly twice as fast as a stock machine.

Sysinfo scores differently to Sysspeed as it also uses a different algorythm to test - this does have some limitations for the 060 as the 060 has infact TWO ALU's not one ALU like the rest of the 68000 series CPU's. to ALU's (or Arithmetic Logic Units) - consider this the place where it processes the data - it has two of them... thus is twice as fast as Sysinfo can count as it only expects ONE ALU.

The Score for an 060@50Mhz using Sysinfo is about 35.8 MIPS - when in actual its 65MIPS under SysSpeed - this uses a different algorythm whereby the CPU utilises its full core of two ALU's.

All in all as long as you know about the testing program you can put that into perspective of what the results mean - I find both programs very usable - in fact I use two more testing programs

MIPS (for CPU integer) and RSCP (for SCSI / IDE host speed)

Gouldin
24th March 2012, 15:41
Nice info from everyone there :thumbsup:

I always assumed that the speed remained the same, kinda like an x86 pc. :oops:. Only reason I've ever uprgraded ram on these systems was because whdload wanted it for preload reasons.

I'm still a young padawan when it comes to understanding all things 68k it seems :lol:

BrooksterMax
24th March 2012, 15:50
It is nuts how much of a performance boost you get from FAST RAM. Shame many ordinary users back in the 90s would not have known this fact (me included) and had bottlenecked stock A1200s.

SpeedGeek
24th March 2012, 16:54
I'm getting tired of repeating my answer to this question.

http://www.amibay.com/showpost.php?p=149575&postcount=10

For the 060 the only practical choice is SysSpeed. But for 040 and below AIBB rules!

rkauer
24th March 2012, 17:08
Yes, the expected boost in performance is around 2.3X faster than a stock A1200 with trapdoor memory added.

Don't know the boost on the CD32 as I never had one.

Even on plain 68000 units (A500/2000/600) the boost is 33% (1.33X) with real* FAST RAM.

*You can't count trapdoor-added RAM as "real" because it is in fact loaded to the Ranger memory area and is dependent of the Gary/Agnus clock generation, crippling the speed.

External RAM (side cars, Zorro card RAM, RAM that is plugged directly on the CPU socket) is true 16bit RAM on this case.

PCMCIA RAM on the A600 will also boost the A600 performance, but on the A1200 it will cripple even more the performance, but will work.