First the good news: the user will ge hardware aiming not only
at the current 68060 + SuperAGA environment, but designed and
tuned for the future with SoftCore N070. The 68060 will now
reside on a CPU card, like the CPU does for the Natami-030
prototype.

The reason is easy to explain: Having the 68060 onboard, even the
socket alone, adds a lot cost and complexity. Making another
design directly  after the 1st one would have meant paying a
huge sum the 2nd time for the process from the netlist to
the gerber-data, not to talk about maintaining firmware upgrades
for 2 designs with different chips and RAM-technologies.

We also felt, the price for the dev-series was going too high
with the originally selected components. Now we could lower it
thanks to these changes and the move from SRAM to DDR2.

The reason we could not announce this more early is, that we
needed the new specifications to be settled. Thomas had a lot of
work making the complete change on many parts with now changed
voltage-potentials.

The board specifications, completed as design, are now:
NEW:
-Big Performance Plus: 2 independent 64bit busses with symmetrical
  256MB CHIP and 256MB FAST DDR2 memory, both clocked with FPGA speed
-68060 on CPU-card via PCI (optionally local SRAM)
-Altera Cyclone 3 FPGA with interal SRAM for 3d texturemapper speed-up

-DVI-output for flatscreens
-usb1.1 for mouse/keyboard

MAINTAINED:
-2x dma-capable IDE
-VGA output (with integrated scandoubler)
-15kHz Video Out
-floppy connector, supporting Amiga-diskettes
-parallel and serial
-ps2-connector for AMIGA-keyboard or PC ps2-keyboard (autodetect)

Outlook on our timeline:
Basically we are back to the point of August, when Thomas
had completed the original 060-onboard design. This delays
the new expected date for the 1st board to November. Sorry to tell
you this, but in the end the change is very important to us, to
avoid Natami being a quick-shot without a future.

Have you thought about adding some of the last couple of months' news to the news section? Not the news section of the forum, but this one: CLICK HERE - For people just dropping by the site it could be misunderstood as if the project was frozen... just my 2 cents.

Oh, and are there any guesstimates about the price-range of the developer version?

Thanks

I agree with you, Laust, that maintaining the news section you've mentioned is important. I also join Przemek wondering how switching to DDR2 is going to work in terms of latency and overall performance?

Take care.

However, losing the SRAM is horrible! It would have made a HUGE difference to 1920*1080 resolution graphics screens!!

Now, are we at ONE spare PCI slot? I think we need 3 or 4.

It's an add in card, no chance of leaving an empty socket for SRAM for people that are adventurous??

I spent $800 buying 2 * 1 Gigbyte PC133 MHz registered sdram for my AmigaOne XE.

The new speccs are totally cool although... Now that you decided to move to a new design, CAN'T we have a little sneakpeak on the old board as a picture for the fun of it? I think all people here that monitor the progress of Natami are waiting SO much for a couple more pr0n pictures - at least of the work in progress or the old board!
Don't call me hungry! We all are!!! (lol)

Cheers

"However, losing the SRAM is horrible! It would have made a HUGE difference to 1920*1080 resolution graphics screens"

Sorry, this is much more complicated: SRAM makes a difference
on random access - for the bandwith on streamed data we
will get more performance with DDR2, otherwise we
would not have accepted this as a team-decision.

Gunnar will write the attainable transfer rates in another
technical thread. Of course there will be cases, where SRAM
wins because of access-latencies: but for the 3d performance the design showed,  that the algorithm will win with the chosen bigger
Altera, that keeps texture data inside his in-chip-SRAM.

This Amiga will so totally rule!!!!!

All those things I knew that Amiga could be, could do, if only this, that and the other thing!

The programs are SO small compared to today's GARGANTUAN INSANE overhead systems!!

I can still use "more" and read off of a FLOPPY DISK a file faster than can be read by me, screen after screen displyed! While, on my 2.26 GHz P4, it can take 10 seconds and MORE to save an internet page IN RAM (not ram: drive) to the hard drive in striped RAID! And the computer FREEZES UP until it's FINISHED doing this!!!!

AWeb will START faster than it takes to save a 50 K file!!!!!!

Sorry for the ranting. :-D

Some questions though:

1) The obvious, as SRAM was mentioned so many times, what impact will the DDR2 have on the speed?

2) USB is a cool adition, but no Ethernet yet! I believe that an onboard ethernet adapter is essential, and since you are in the "redesing" process it would be wise to add it now.

3) Since many user groups are in the discussion of making "group pre - orders" we need to know the final size of Natami to get the cases. Will it be MicroATX, MiniATX, ATX.. ?

And some comments:

DVI integration was a very smart idea. Most 32-42" TFT TVs come with HDMI/DVI input and DVI->HDMI adapters are cheap enough! Great idea indeed!
Having 256 Chip ram will help programms like ClariSSA PRO to load a HUGE animation into chip ram and play it back smooth enough to compete with any professiona HW based player. It will also help programms like SCALA and Dpaint which suffer from chip ram limitations. It was the right thing to do :-)

Keep up the good work guys :)

The throughput of SRAM and DDR memory is quite similar.
For example: A 64bit 200Mhz DDR2 interface (also known as called DDR400) has a maximum throughput of 3.2 GB/sec.

SRAM has of course a much lower latency.
DRAM in general has a higher latency.

So how important is throughput and latency for the system performance?

1) Screen resolution is depending on the throughput only.
2) Blitting screen or moving windows is depending on throughput only.
3) Doing 3D operations depends on both latency and throughput.

For the new design we go for an FPGA with more internal SRAM.
We have two different types of SRAM that we mention here:
- SRAM that is inside the FPGA
- and SRAM that is external

Internal SRAM has the lowest latency of all memory.
Internal SRAM has also the very highest throughput.
External SRAM has higher latency than internal SRAM but lower than external DRAM.

We have changed our 3D core design to rely on internal SRAM much more. This is qwuite a drastic change but it will improve performance a lot.
The FPGA that we are going for has more internal SRAM now.
By changing the design by exchanging the external SRAM with internal SRAM and more external DRAM we are doing a trade off but a good one.

The amount internal SRAM is lower but it's a lot faster.
And the total amount total memory is now a lot more.

The chip memory will now be 256 MB.
This should satisfy all wishes for more chip memory.

The internal SRAM is more than before and the 3D core relies more on it. The 3D core will work in chunks. It will burst from DDR2 memory chip memory to the internal SRAM do the processing and then burst the results out again.

Cheers
Gunnar