A1200 FASTRAM Vs CHIPRAM

[NovaCoder]

Why do people say that adding FASTRAM to a 1200 makes it faster?

If you've got an accelerator then it makes sense for my game to run in fastram and then only use chipram when I have no choice (eg updating the screen).

But what about the standard 020 users with some fastram, in this case wouldn't the game actually be slower if it resided in FASTRAM? I guess it depends on the bus speed vs the chipram speed.

Anyone know the answer?

[Zetr0]

@NovaCoder

I figured you would of know this,

with Fast Ram the CPU doesn't have to fight its way through the custom chips to get at ITS data.

As you know the CHIP data-path is shared, as such the CPU must wait inline like a circular dinner cue.

Once you have FAST you stop all that and the CPU has VIP access to RAM, nothing else can touch it... from here you can build up screens / rasters in fast ram and then dump (copy) them back to chip a lot faster than actually building them in chip!

The CPU is free of waiting in line for memory access and hence Opperates a lot more efficiently, in effect FASTER, twice as fast on the 020EC

True, there are some *hardcoded* games that force to run from chip, mainly old games, and most other games have timers based on the chipset which in effect controlls the speed the game runs (for the most part)

But calculations can be done instantly, the CPU doesnt have to wait to fetch a value from chip, and then wait to get another value from chip, and then to process that value only to have to wait to put it back again...

Fast Ram baby! its what ALL CPU's need =D

[gulliver]

Adding FASTRAM to an A1200 makes it faster because that memory does not need to work at the same speed as the Amiga bus, it can work faster and chips are faster from 60 to 80ns access time. The only bottleneck, as always, is when it needs to communicate with the system bus, that communication will always be at the Amiga bus speed. A DMA controller could, if present on the FASTRAM, speed things even further.

Chipram is slower, the 2MB chip access time is bout 250ns, it is always under siege of the system bus, so it cannot go faster than it, and also remember that on Chipram, Amigas native graphics and sounds are carried out, meaning the most intensive computing aspects of the A1200 depend on this slow chips. If you have FASTRAM, you can easily redirect most of these "system critical" structures to faster FASTRAM.

I hope it helps

[NovaCoder]

Yep, I guess my point is that what you need to compare is the time it takes for a command to get executed via the Zorro2 vs chipram speed. I know the chipram is very slow, but I what kind of hit takes place across the Zorro2?

[coze]

erm sorry but you're not making much sense here.

Amiga 1200 doesn't have a Zorro2 bus.
Zorro2 memory is fast ram (A500, 2000, 3000 etc) since chipset doesn't have access to it.
So there's no 'hit' across Zorro2 if you"re talking about somekind of penalty ?

[gulliver]

@NovaCoder

Well let me see if i can understand you. You mention comparing speeds between Chipmem and zorro2. An A1200 does not come with zorro2 slots, it has a slightly different system bus than say an A2000 with zorro2. Of course you can adapt signals con the A1200 bus to make a zorro2 converter. The point is, despite A1200 bus or even zorro2 having many performance bottlenecks, it is always faster having FASTRAM over there, than having to cope with dead slow Chipram.

[NovaCoder]

Woops, my bad. What's the 1200's 150 pin slot called then if it's not a type of Zorro? Maybe it's just called a 'slot'

Anway, thanks for the feedback...it looks like FASTRAM gets the thumbs up even SANS accelerator

[BrooksterMax]

Just to second this in practice - I had a stock a1200 and got a Blizzard 1200/4 expansion card, it has 4mb ram onboard and an empty FPU socket.

So its just 4mb ram, sysinfo reports a massive speed increase and the machine zips along nicely with whdload.

So why bottleneck the potential of the A1200 at launch? I guess cost was the factor for Commodore....

[StingRay]

Originally Posted by NovaCoder

But what about the standard 020 users with some fastram, in this case wouldn't the game actually be slower if it resided in FASTRAM?

Why should it be slower in Fastram? Zetr0 basically said all. Good example is Wing Commander, try it on a ChipRam only a1200 and then try it on a machine with Fastram, doesn't run slower, does it? =P

[meega]

Originally Posted by BrooksterMax

So why bottleneck the potential of the A1200 at launch? I guess cost was the factor for Commodore....

If you are old enough to have been buying RAM in the early/middle 90s then you know why the base machine was "unexpanded" (a 4MB card with FPU was about half the price of the original machine) - and besides, if they'd put "something" in there then you'd only have to replace it...

[BrooksterMax]

I was old enough and got a 1mb A500+ expansion in the 90s, but I was thinking of why not have 1mb fastram on board integrated into the design of the machine.

Not 4mb because that was some £200 without FPU - agreed very expensive at the time.

[meega]

I paid £180 for a 1MB trapdoor chip RAM expansion for my A500Plus... when they were still a fairly new release.

[BrooksterMax]

ouch, I got my A500+ 1mb chip for £30 from an Amiga Format offer.

[Toffee]

If the A1200 had 1 meg of chip and 1 meg of fast ram with an updated 32 blitter with a chunky screen we would have been much better off!

[pmc]

Originally Posted by Zetr0

Once you have FAST you stop all that and the CPU has VIP access to RAM, nothing else can touch it... from here you can build up screens / rasters in fast ram and then dump (copy) them back to chip a lot faster than actually building them in chip!

Hmmm, interesting. So, taking this as a starter, if I was going for a nice quick C2P routine it would be better to manipulate the chunky graphics in fast RAM and then convert / copy the data into planar format in chip RAM ready for display?

[Leffmann]

PMC, yes it will be noticeably faster. That's why you should always put code and data in external memory when you can, set aside the things that of course must be in chip memory.

All internal devices share the bus to the chip memory and only 1 can access it at any given time. For the Amiga to f.ex display the bitplanes and sprites and to play audio it needs to read this data continously, every scanline of every frame. It's not enough that it just sits in the chip mem. These three types of DMA have the highest priority to guarantee that they can read the data in time so there's no interruption in the video or audio signal.

In this scheme the CPU has the lowest priority of all, and so with many large bitplanes, sprites, audio and perhaps the disk drive and blitter running there will be so much contention that the CPU have to wait more and more often to access the memory. This is less noticeable on AGA because both the width of the data bus and its clock rate have been doubled. Bitplanes in particular benefit a lot from this, but contention can still be significant.