Quote:

[b] So lets break down whether Natami has the fastest/most powerful Amiga graphics hardware around: 3. What is the maximum theoretical (device isn't finished) screen resolution/refresh rate on the Natami?

Amiga Mode (PAL2-Super-Hires) in 60Hz (LCD) (1220x1024)
and
This Screen has 1.2 million pixel.
In HiColor you need a bandwith of 120 MB/sec
In Truecoloe you need a bandwidth of 240 MB/sec
A 32 Bit Amiga Chipset using 100Mhz bus has a bandwidth of 400 MB/sec

Amiga Mode (PAL2-Super-Hires + Overscan) in 60Hz (LCD) (1440x1024)
This Screen has 1.4 million pixel.
In HiColor you need a bandwith of 170 MB/sec
In Truecolor you need a bandwidth of 340 MB/sec
A 32 Bit Amiga Chipset using 100Mhz bus has a bandwidth of 400 MB/sec

Quote:

4. What is the most graphically intensive app that classic Amigas can run (preferrably one that requires a GFX card)?

I can think of two demanding examples:
a) A Vector Game using Mip-mapped/Antialiased texture mapped pixels with ALPHA and antialising with calculated lighting. Such a game needs to color combine four read pixels into one written pixel.
All current AMIGA GFX card fail here as the reads are from indepenpendent source addresses.
This means wide buses or prefetching does not help and the speed is fully latency bound.

b) A application making use of AMIGA soft scrolling features, multiple playfields and copperlists.
No GFX besides original AMIGA can do this.
Emulating all the combination of these features with Sprites and collision and Software is very expensive. Stuff that you got for free on AMIGA HW needs huge amounts of resources on emulation.

Quote:

5. How likely is it that Natami could run an app of similar or better graphical quality?

The strength of the NATAMI is that its supports the original AMIGA features
- As super hires scrolling, copperlist , playfields, Sprites
But has a AMIGA chipset that is much faster than the original one (100 times in theory)
You can do the AMIGA typical GFX effects as smooth scrolling, screen down pulling etc, for free.
[b]NO PC in the world is fast enough to emulate these effects that the NATAMI can do in real time.[b/]
The main focus of the NATAMI is supporting the original AMIGA features.

In addition to this NATAMI has the needed low latency and bandwidth to texture map 40 Million hicolor pixel per second. This by using four pixel mixed color correction.
This is worst case and not best case texture map performance!
PC cards do always quote the bets case performance.
This is of course less than PS3 or PC cards can do.
But it means that you can run 3D games in higher quality then you ever saw on AMIGA before.

Another advantage of the NATAMI chipset is that the BUS between CPU and GFX memory to load addition texture or others things is much faster than the bus in old and new Amigas AONE) was and is.

The strength of the NATAMI is the perfect integration of old and new features.
You can drag down your demo running on the 32bit texture mapped screen and see behind the HAM screen using copperbars.

Last edited by BigGun on 25-Jan-2008 at 09:05 AM.

_________________
APOLLO the new 68K : www.apollo-core.com

Quote:

It's an awful example, because it runs a very serious risk of running in cache. But OK, if your heart is set on it... My machine does that at just over 60 million reads per second. Happy now?

I see your Gigaherz Computer runs on local memory and gets 60 M.
You don't even run on GFX memory - where you said yourself you will get MUCH, MUCH less.
The NATAMI chipset gets 100 million.
So a Natami using even a cheap $20 Coldfire can beat your number in this test.


I fear its pointless to answer your other posts.
With halve of your post you verify what I say.

When I stated that 68K emu is expensive you corrected my post
and say that I posted by accident the interpretive numbers.

You verified that:
- interpretive 68k emu needs 70 clock per 68k instruction
The 68060 CPU could execute up to THREE 68k instructions per clock.
So you verify that x86 can emulate on 68060 between 1/70 to 1/200 clockspeed.
Using your numbers a 3 Gigaherz x86 interpretes an 68060 CPU with between 42 - 15 Mhz.

- You state that JIT has on overhead of 4000 clock per 68K instruction.
Which means if a 68k program code is called more than 60 times JIT will be faster then interpretive.
As more often the same code is executed as faster JIT will be - at the end it will run a lot faster then existing 68060 CPUs.
But if the code is called only once a 3 Gigaherz x86 CPU, the x86 CPU will be 60-200 times slower than a native 68060 with 50MHZ.


Its sad, that you are "very selective quoting"
When you compare GFX card resolutions you do not compare 32bit with 32bit bit screen
But you compare a 24bit CV screen with 32bit Natami screen.
Do you do this, as you know the CV can not do any better than NATAMI in 32bit ?

When you quote the CPU card you purpose misquote me.
- Was I said was that CPU cards will have own FAST Mem.
- I said that a CELL CPU card will obviously need expensive XDR MEM.
- I said that a 68060 CPU card can use fast SRAM for removing latency.
- I said a PowerPC CPU could burst to local DDR fastmem.
You mix this points and claim a PPC card can not burst in local SRAM.
I think you try very hard to twist and mix what I say to prove your point.


I would like to ask to stop trying to misinterpreting my points.

The point that I made were:
- The NATAMI is backwards compatible.
It support even reading writing of old AMIGA HD disks
- Its support AGA in improved performance
- I can do AGA like effects in such an increased speed that no PC is able to emulate them.
- It adds new features to AGA: Hicolor, Truecolor, Texture mapping
- It adds these features in a fully harmonised way.
Screen dragging of mixed old/new screen modes.
AGA Spites are on new modes.
Copper work normally on new modes.
- The AGA++ has much lower latency then any used AMIGA GFX extension card
- Ported SDL Games like Robin Hood will run on it faster then on AmigaONE/Pegasos Cards.
- A high clocked 68060 using faster newer memory runs AMIGA OS faster then many will expect.
- The effect that are possible by combination of high clocked 68060 or Coldfire with using of AMIGA AGA features are extremily demanding to emulate and challenging even for Gigaherz x86.
If you take in account the increased speed of the NATAMI AGA++
then these effects are totally impossible to archive in any emulation!
- The NATAMI has a CPU slot.
By using "future" CPU cards you can upgrade the CPU performance (using ie PowerPC)
- The NATAMI AGA chipset is in my believe the fastest AGA implementation
that is possible without having the faps of NVIDIA or ATI.
- Todays PS3 or PC Gfxcards are much faster in max performance.
You can NOT expect that a AGA remake having not their fabrication resources can beat them.
- While the NATAMI chip can NOT be as high clocked as PC GFX chips it performs still very good.
The NATAMI uses very very low latency memory.
- This low latency guarentees that the minimum performance is high compared to other systems.
- There are situations where the NATAMI can even outperform 10 times higher clocked systems.
- For workbench usage the NATAMI is very powerfull.

If you think about these points then I'm sure you agree with them.
For someone wanting to run AMIGA OS the NATAMI is certainly an excellent solution.


Umisef,
I want to ask you one question:
You probably know that there are projects like the DRAGON to bring an Coldfire CPU to the AMIGA.
The COLDFIRE many bot not all AMIGA instructions.
With coldfire lib its possible to execute 68k programs normally in Coldfire.
But using something like 68K JIT compiler would 100% guarantee that any 68k program run perfect and in maximum speed on a Coldfire cards.

You know the UAE very good.
Did you ever think about supporting these projects by helping them to 68 JIT complier running?
Or by helping making the Kickstart Coldfire clean?

_________________
APOLLO the new 68K : www.apollo-core.com

@BigGun

Quote:

BigGun wrote: @Hammer Quote: Hammer wrote: The computing power of the AMD processor stack** makes CELL looks really poor. **AMD RV670(or RV680) + AMD64. RV670 supports IEEE754 and double precision floating point. RV680 has ~1TFLOPs (theoretical for stream/shader processors) on a card i.e. basically two RV670s with an unknown CrossFire management chip. WHAT DOES YOUR POST TO DO WITH AMIGA ? CAN YOU RUN AMIGA OS ON YOUR x86 CPU? We have asked you the question many times now. There are many fast computers out there. Ranking from a simple HP Superdome to IBM Bluegene or to the IBM Roadrunner project. But can you run AMIGA OS on them ? What is the purpose of your constant off-topic posts?

Refer to your own post
Quote:

And if you compare the computing power of the CELL than the [b]AMD looks really poor

Claims like yours are open for counter statement.

Last edited by Hammer on 25-Jan-2008 at 10:58 AM.
Last edited by Hammer on 25-Jan-2008 at 10:56 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@BigGun

This is definitively interesting.
Can you summarize the goal of the project, giving estimated time and price, please?

_________________
retired

@Hammer

Hi, I'm new to Amigaworld, but have been following its forums for several years. Being a former Amiga user, that had to sell it's A2000 ~6 years ago, I was particularly interested in this thread. In spite of having 2 PCs, one running Slackware Linux (P3@900MHz) and another with XP (Athlon 64 X2 4000+), I have a great interest in using an Amiga again.

Personally I find the Natami a great opportunity for someone that needs to replace an aging classic Amiga. The machine might not compete with an Intel/AMD rig, but that's not it's purpose, in the first place.

When BigGun said Quote:

And if you compare the computing power of the CELL than the AMD looks really poor

I think he was comparing general purpose CPUs and not gpGPUs.

There are particular cases where a Cell might be faster than a gpGPU and to see this, you might check this link: http://gametomorrow.com/blog/index.php/2007/09/05/cell-vs-g80/

What one must consider, is the level of complexity/power requirements that a part must have, to reach a certain performance level. In this regard the Cell appears to be a well balanced solution. It might not be faster in all situations, but pure performance is not the only factor to consider.

Regards

_________________
Only one man in a thousand is a leader of men, the other 999 follow women. - Groucho Marx.

@BigGun

I have not read every page on this thread so forgive me if this question has been asked. I think I already know the answer but, are there any plans to have compatabilty with the Video Toaster? Just figured I would throw that out there.

_________________
Questions are a burden to others.
Answers are a prison for oneself.

Two A2000's with no working harddrives.
One Efika with MorphOS 2.2
One HP Netbook running Amikit.

@BigGun

Quote:

The 68060 CPU could execute up to THREE 68k instructions per clock.

The 68060 CPU can issue two (CISC) instructions from it's front end.
A Core 2 Duo (aka 80557) CPU can issue 4 (CISC) instructions from its front end.

As you already know, 68060 doesn't have any support for 128bit wide SIMD i.e. this is important for media processing where you have repetitive data sets.

Quote:

But if the code is called only once a 3 Gigaherz x86 CPU, the x86 CPU will be 60-200 times slower than a native 68060 with 50MHZ.

One shouldn't group X86s in the same bundle i.e. which 3Ghz x86 CPU (e.g. Netburst, Core 2, K8) ?

Quote:

The NATAMI uses very very low latency memory. - This low latency guarentees that the minimum performance is high compared to other systems. - There are situations where the NATAMI can even outperform 10 times higher clocked systems. - For workbench usage the NATAMI is very powerfull.

Refer to Nintendo's Wii with an embedded T1-SRAM. ATI and NEC is aware of this kind of techniques and implemented it accordingly.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Zealot

Quote:

There are particular cases where a Cell might be faster than a gpGPU and to see this, you might check this link: http://gametomorrow.com/blog/index.php/2007/09/05/cell-vs-g80/

A counter example is Fold@Home, where the old Radeon X1900(RV570/R580) beats CELL twice over. The current Radeon HD 3850/3870 (RV670) supports IEEE 754 and DP FP maths.

Note that I specifically noted ATI's Radeon HD GPU in this case since the G80 has some issues i.e. still missing Fold@Home client.

For CELL's deferred shading refer to http://research.scea.com/ps3_deferred_shading.pdf

In practical deferred shading (pixel shading) terms (using Sony's own white papers), 5 SPUs @3.2Ghz is roughly equivalent to 20 pixel shader G7x (e.g. Geforce 7800).

G7x ‘s stream units can't process shader and texture operation at same time.

Example of deferred shader powered title is Unreal Tournament 3 (for PS3 and PC) .
In http://www.anandtech.com/video/showdoc.aspx?i=3128&p=3
You'll notice the G7x takes a major hit in deferred shader enabled Unreal Tournament 3.

Last edited by Hammer on 25-Jan-2008 at 12:16 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Stop trying to hijack the thread gain and again!

This thread is about the NATAMI, as new AMIGA system running AMIGA-OS .
[

_________________
APOLLO the new 68K : www.apollo-core.com

@Zealot

Quote:

Zealot wrote: @Hammer Hi, I'm new to Amigaworld, but have been following its forums for several years. Being a former Amiga user, that had to sell it's A2000 ~6 years ago, I was particularly interested in this thread. In spite of having 2 PCs, one running Slackware Linux (P3@900MHz) and another with XP (Athlon 64 X2 4000+), I have a great interest in using an Amiga again. Personally I find the Natami a great opportunity for someone that needs to replace an aging classic Amiga. The machine might not compete with an Intel/AMD rig, but that's not it's purpose, in the first place.

It's an interesting "what if" project.

Quote:

When BigGun said Quote: And if you compare the computing power of the CELL than the AMD looks really poor I think he was comparing general purpose CPUs and not gpGPUs.

Well, Toshiba’s CELL based SpurEngine doesn’t have a PPE core i.e. just 4 SPUs @1.5Ghz, some hardware decoders/encoders, works like a PC accelerator and 10 to 20 watts TDP.

Anyway, having just a CPU doesn’t make a PC. In HPC math market, AMD targets its RV670 GPUs in the same market as STI’s CELL.

Last edited by Hammer on 25-Jan-2008 at 12:40 PM.
Last edited by Hammer on 25-Jan-2008 at 12:37 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@BigGun

Quote:

BigGun wrote: @Hammer Stop trying to hijack the thread gain and again! This thread is about the NATAMI, as new AMIGA system running AMIGA-OS . [

What a hypocrite. You are the one who introduced OT by making claims in reference to mainstream PC products.

If you bait in a topic, expect a bite.

Last edited by Hammer on 25-Jan-2008 at 12:42 PM.
Last edited by Hammer on 25-Jan-2008 at 12:34 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@umisef

Quote:

Yeah, *reading* from gfx memory is slow, but you have yet to provide a single example where one would have reason to read gfx memory with the CPU.

Reading a (screen or offscreen) bitmap into a pixel buffer, modifying it (add some effect like transparency, alpha blending, smoothing, whatever) and then writing it back is pretty common for Amiga apps and libs. To take advantage of gfx card blitters bitmaps may be in VRAM but then it is still likely that some gfx functions needs to read pixels from VRAM with CPU as doing it with gfx card blitter is not possible and/or the gfx card driver does not support it and/or the transfering data from VRAM to RAM with DMA is no supported and/or even possible (is it possible with PCI?).

@Hammer, BigGun and umisef and few others

Although I agree with Hammer the off topic was not caused by Hammer (or umisef) it seems the technical side is to hot for some to handle.

Feel free to make a new thread about it but I suggest to let it be in this one.

Last edited by Seer on 25-Jan-2008 at 01:09 PM.

_________________
~
Everything you say will be misquoted and used against you..
~

Moved to here. Please continue this discussion there.

Last edited by Zardoz on 25-Jan-2008 at 01:42 PM.
Last edited by Zardoz on 25-Jan-2008 at 01:41 PM.
Last edited by Hammer on 25-Jan-2008 at 01:29 PM.
Last edited by Hammer on 25-Jan-2008 at 01:26 PM.
Last edited by Hammer on 25-Jan-2008 at 01:15 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@wawa

Hi, has anyone else told you anything about this A1000 improvement? I also don't understand German, unfortunately.

_________________
Only one man in a thousand is a leader of men, the other 999 follow women. - Groucho Marx.

@Zealot

The thing on A1k is an A1000 replacement board with the following features:

- 030'er and FPU
- 8MB 32Bit-Fastram without Waitstates
- FliFi (Through Amber Chip)
- HD-Controller
- 2MB Chip
- Zorro-II-Slot
- Frontslot

_________________
<- Amiga 1260 / CD ->
Looking for:
A1200/CF CFV4/@200,256MB,eAGA,SATA,120GB,AROS

@Donar

Thanks!!

_________________
Only one man in a thousand is a leader of men, the other 999 follow women. - Groucho Marx.

@Zealot

i posted it for a reason. u do not need to speak german to interprete screenshot of the intuition benchmark of this replacement board. the memory throuput seems to be well over original comodore machines, be it a3k or a4k. so this is probably a proof that ist is possible to improve upon the original architecture, even if this is just the first and most obvious step.

@wawa

Yes, the screenshots are indeed revealing but I wanted to know if he only planned to improve is own A1K, or if he plans something else, in the future. I should have been more explicit, sorry.

_________________
Only one man in a thousand is a leader of men, the other 999 follow women. - Groucho Marx.

@Zealot

i dont know the guy personally, as far he states on the thread its just his hobby project. that means hes not gonna to produce and distribute the thing in any way but will provide the complete online documentation online charge free. because of current legal situation it seems now impossible in germany to produce short series of hardware without a lot of beaurocracy and additional costs so do not expect miracles here. and anyway the replacement board is only interesting for the a1k-users. but where is the will maybe one will find a way..