Home   Forum

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
VISIT APOLLO IRC CHANNEL



All TopicsNewsPerformanceGamesApolloVampireReleases
Documentation about the Vampire hardware

Vampire Hardware Featurespage  1 2 3 4 5 

Thierry Atheist

Posts 311
20 Apr 2017 19:20


Why would you want them to make a

128 MB accelerator
and then
256 or 512 MB  accelerator
and then
1 GB accelerator
AND THEN a 2 GB accelerator????

When everyone REALLY wants a 2 GB accelerator!!!!

(Assuming 4 can't work with anything, except maybe linux.)


Michael R

Posts 140
20 Apr 2017 20:06


While it's true that we shouldn't try to compare Amiga hardware, software and Amiga RAM requirements to modern computers, there is an "adequate" amount of RAM that will work for most Amiga applications, but not all. The Amiga OS can do many things will very little memory. So 128MB of RAM is very good, 256MB is better, and 512MB is better still. More is better, up to a certain point.

Perhaps, if it doesn't draw too much power, 256MB for the V1200 accelerator would be a nice upgrade over the 128MB we currently have on V600/V500. For the standalone motherboard 512MB might work well since power supply wattage isn't a concern.

How fast does Vampire accelerator RAM run? 100MHz or faster? The memory speed is important as well.


Michael R

Posts 140
20 Apr 2017 20:12


Samuel Crow wrote:

 
Peter Heginbotham wrote:
2GB and why not if there is no sodimm connector
    Full PCI Bus with the following using well documented items
      |-Ethernet controller
      |-USB 2.0
      |-SATA or UDMA IDE Controller (Bootable)
      |- Audio chip + deicated outputs
    Expansion connector to allow for PCI slots

 
  First off 2GB is more than the maximum for the Amiga OS.  I/O registers and Kickstart flash memory take up some addressing as well.  The maximum is more in the range of 1GB of RAM or so.
 
  Secondly, why waste board space with PCI when Ethernet, some USB, IDE, and Audio are already planned to be on the motherboard without using PCI for any of it?
 

 
  I believe he said "expansion connector" not actual PCI slots. The connector for future expansion for the standalone board is important. It could take the form of an edge connector or a pin header for a ribbon cable but it might allow the future addition of bus-boards for PCI/PCIe that we don't have at the moment. The concept is similar to that of the Mediator board.
 
  The bus-board expansion connector approach wouldn't take up much space on the mainboard, and the board is so small that if it were installed in a large enough case there would plenty of room for "extras".

But as Gunnar said previously, maybe these things will be considered for mainboard Gen2, because the features for Gen1 board are already set. Gen1 board can be simplistic yet functional. Gen2 can add some components and improve upon the first as far as functionality. One step at a time!


Daniel Sevo

Posts 144
20 Apr 2017 23:23


Thierry Atheist wrote:

Why would you want them to make a
 
  128 MB accelerator
  and then
  256 or 512 MB  accelerator
  and then
  1 GB accelerator
  AND THEN a 2 GB accelerator????
 
  When everyone REALLY wants a 2 GB accelerator!!!!
 
  (Assuming 4 can't work with anything, except maybe linux.)

If there is no significant cost premium, or increased power usage or unnecessary layout/design issues, then logically, yes, you could max out at 2GB and be done with it..

But, good hardware design can hit a "sweet spot" that brings you a good combo of features and performance versus cost.
I'd say that 512MB RAM is probably the sweet spot in this case. And the arguments for going beyond really need to make sure they don't bring *any* other compromises because it is simply so few that would benefit from it that it's not worth it.

The amount of RAM should be in some sort of proportion to the available CPU power. Ask yourself, what would you do on, say, a Pentium 166 with 8GB RAM? Yes, you could dream up some weird scenario, but realistically, its not a useful combo.
As I mentioned before, I work with 3d graphics and *every time* we get more RAM in our machines we increase scene complexity, only to notice we now want a faster CPU. ;-) And so it goes on forever...




Gregthe Canuck

Posts 79
20 Apr 2017 23:51


Michael R wrote:

But as Gunnar said previously, maybe these things will be considered for mainboard Gen2, because the features for Gen1 board are already set. Gen1 board can be simplistic yet functional. Gen2 can add some components and improve upon the first as far as functionality. One step at a time!

Exactly - one step at a time. I think many are expecting too much in one release. There is a lot of work involved just to get *any* sort of standalone board working. Whatever form Gen1 comes in will be amazing.


Xan X-vision

Posts 8
21 Apr 2017 00:31


Daniel Sevo wrote:

 
Thierry Atheist wrote:

  Why would you want them to make a
   
    128 MB accelerator
    and then
    256 or 512 MB  accelerator
    and then
    1 GB accelerator
    AND THEN a 2 GB accelerator????
   
    When everyone REALLY wants a 2 GB accelerator!!!!
   
    (Assuming 4 can't work with anything, except maybe linux.)
 

 
  If there is no significant cost premium, or increased power usage or unnecessary layout/design issues, then logically, yes, you could max out at 2GB and be done with it..
 
  But, good hardware design can hit a "sweet spot" that brings you a good combo of features and performance versus cost.
  I'd say that 512MB RAM is probably the sweet spot in this case. And the arguments for going beyond really need to make sure they don't bring *any* other compromises because it is simply so few that would benefit from it that it's not worth it.
 
  The amount of RAM should be in some sort of proportion to the available CPU power. Ask yourself, what would you do on, say, a Pentium 166 with 8GB RAM? Yes, you could dream up some weird scenario, but realistically, its not a useful combo.
  As I mentioned before, I work with 3d graphics and *every time* we get more RAM in our machines we increase scene complexity, only to notice we now want a faster CPU. ;-) And so it goes on forever...
 
 
 

 
  You answered to yourself without noticing: Memory is CRUCIAL for 3D. If you have a slower processor, then it will take more time to render your scene, but you will render your scene anyway  (like in the old times: go to bed leaving your Amiga rendering). But if you run out of memory, there is no scene.
 
We ask for more memory because it is a realistic demand. Nobody here asked for a faster processor because we respect Gunnar's work and assume it is not feasible at this point, but memory? it is. And there are many situations in which we can use it: 3D, Video, Music, internet,...
 


Michael R

Posts 140
21 Apr 2017 03:15


Michael R wrote:

    While it's true that we shouldn't try to compare Amiga hardware, software and Amiga RAM requirements to modern computers, there is an "adequate" amount of RAM that will work for most Amiga applications, but not all. The Amiga OS can do many things will very little memory. So 128MB of RAM is very good, 256MB is better, and 512MB is better still. More is better, up to a certain point.
     
      Perhaps, if it doesn't draw too much power, 256MB for the V1200 accelerator would be a nice upgrade over the 128MB we currently have on V600/V500. For the standalone motherboard 512MB might work well since power supply wattage isn't a concern.
     
      How fast does Vampire accelerator RAM run? 100MHz or faster? The memory speed is important as well.
   

   
   
    Since no technical people familiar with Vampire 600/500 design answered my question about frequency of the RAM used on these cards, I have searched and found somewhat of an answer on Majsta's site.
   
    EXTERNAL LINK     
    It is in the long paragraph where he talks about LitePlacer PnP machine. He allotted one month for that. He was involved in many things while designing v500 and v1200, such as looking up specs for DDR3 memory. I'm not sure if it was for v500 or v1200. I suppose that answers my question about memory speed.
   
    If these Vampire accelerator cards are using DDR3 that's phenomenal speed and bandwidth for an FPGA!
   
    From Wikipedia:
   
    "The primary benefit of DDR3 SDRAM over its immediate predecessor, DDR2 SDRAM, is its ability to transfer data at twice the rate (eight times the speed of its internal memory arrays), enabling higher bandwidth or peak data rates. With two transfers per cycle of a quadrupled clock signal, a 64-bit wide DDR3 module may achieve a transfer rate (in megabytes per second, MB/s) of up to 64 times the memory clock speed (in MHz). With data being transferred 64 bits at a time per memory module, DDR3 SDRAM gives a transfer rate of (memory clock rate) × 4 (for bus clock multiplier) × 2 (for data rate) × 64 (number of bits transferred) / 8 (number of bits/byte). Thus with a memory clock frequency of 100 MHz, DDR3 SDRAM gives a maximum transfer rate of 6400 MB/s."
   
    DDR3 is Super Fast memory for an Amiga! And now we are talking about increasing the amount to 512MB or 1GB. That's crazy!
   


Thierry Atheist

Posts 311
21 Apr 2017 04:40


I hope a verdict is handed down soon, or at least a loose outline.

2 gigabytes is a HUGE issue here.


Thierry Atheist

Posts 311
21 Apr 2017 04:49


Daniel Sevo wrote:
The amount of RAM should be in some sort of proportion to the available CPU power. Ask yourself, what would you do on, say, a Pentium 166 with 8GB RAM? Yes, you could dream up some weird scenario, but realistically, its not a useful combo.

Hi Daniel,

I dispute that profusely!

RAM "makes it possible", BUT, windowz is an INCOMPREHENSIBLY TERRIBLE JOKE parading around pretending that it's an "operating system"... It is a COMPLETE AND UTTER FAILURE!

RAM is the DEFINING FACTOR in SHOWING OFF just how GREAT AOS IS!!!!

Remember, you could have 100 TERABYTES OF RAM, and YET windows says "I need a swap file". It's GARBAGE.
Daniel Sevo wrote:
As I mentioned before, I work with 3d graphics and *every time* we get more RAM in our machines we increase scene complexity, only to notice we now want a faster CPU. ;-) And so it goes on forever...

But that's,.... okay. ;-) :-D


Mr Niding

Posts 19
21 Apr 2017 05:14


Gunnar von Boehn wrote:

A1200 coder wrote:

  Now, if 68080 has a memory access speed of around 300 MB/sec,
 

 
  68080 can do memory access of 64bit per cycle.
  This means with optimal memory an 68080@100 MHz will do 800 MB/sec.
 
  In theory with prefect prefetching this can even be more.
  Up to 1600 MB/sec ideally.

Incase people has forgotten the memoryspeed quoted by Gunnar.

And I did seem to remember him making a point about BOTH cost and power requirements when you get up into the 2 gigabyte range.
No actual details tho, unless it was laid out in details after I logged off IRC.


Thierry Atheist

Posts 311
21 Apr 2017 05:51


Mr Niding wrote:
Gunnar von Boehn wrote:
A1200 coder wrote:
Now, if 68080 has a memory access speed of around 300 MB/sec,
68080 can do memory access of 64bit per cycle.
This means with optimal memory an 68080@100 MHz will do 800 MB/sec.

In theory with prefect prefetching this can even be more.
Up to 1600 MB/sec ideally.


Incase people has forgotten the memoryspeed quoted by Gunnar.

And I did seem to remember him making a point about BOTH cost and power requirements when you get up into the 2 gigabyte range.
No actual details tho, unless it was laid out in details after I logged off IRC.


Hello Mr Niding,

Errr, yeah.

What in relation to what in relation to what, are we talking about?

"And I did seem to remember him making a point about BOTH cost and power requirements"

Well, IF we're talking about an accelerator board for the A1200, that is pulling electricity from a ~24 year old computer....

If we're talking about the standalone with 100% brand new components on the motherboard, I don't see a problem. We're talking about an additional 512 megabytes of RAM, how could that possibly increase the price drastically?

Then there's the "with optimal memory an 68080@100 MHz will do 800 MB/sec." HOWEVER, that is with the CURRENT FPGA, not the ARRIA 10! That one will be able to do about 1.9 times, to as much as 2.8 times as much, this is from Gunnar, himself.

So, for those concerned about a "power of CPU to RAM capacity" ratio, you don't get it, do you?

The standalone is going to be a MONSTER!


Gregthe Canuck

Posts 79
21 Apr 2017 05:58


Thierry - man you need to lay off the caffeine!! ;)

Actual memory performance of current Vampires is somewhere between 300 to 400MB/s. It varies based on clock speed, measurement type and if built-in display logic is completely disabled or not.

The Arria 10 isn't coming any time soon. That is one very expensive chip ($400 +). My understanding is that the V1200 and standalone boards will be using a Cyclone V which will have some speed benefits over the III. How much remains to be seen.




Thierry Atheist

Posts 311
21 Apr 2017 08:58


gregthe canuck wrote:
Thierry - man you need to lay off the caffeine!! ;)

Hi gregthe canuck,

He he he.

Fair enough.

But, these are things I've read, right here, on this website.


Andrew Copland

Posts 31
21 Apr 2017 10:09


No 2GB vs 128MB isn't a huge issue, nor is switching an Arria 10 and getting 1.9 to 2.8 times the current performance.
 
At that point you'll be approaching the power of a PS2, a machine with 32MB + 4MB of ram but without any of the 3D acceleration hardware or other custom processors.
 
As a developer who released games on that machine I am telling you that for the *given processing power available* you do not need more ram, not even for games.
At best it's a case of "it might be nice not to have load between levels" but you mostly cannot make use it during a frame (i.e. when rendering your game).
 
It might be *nice* for a 3D modelling artist, or other artists doing high res 2D, audio/video but those people are a problem for everyone which is why our artists use PCs/Macs with 32GB to 64GB of ram in them... not 2GB. Because there is NEVER enough ram for them, ever.
 
  If it was quick, cheap & easy, then sure, satisfy the craving of the raving lunatics who will never use all 2GB, or even 1GB. However I'd rather they launch the damned V1200, 128MB ram and HDMI out and that's it. Later I'd like the pipelined FPU to be added as an update, but that's literally everything that I want from it.
 
That would be mind blowing.
 
All this other shit is just distraction and loud yelling about something that does not matter at all, even to those who think it does, because you could launch it with 16GB and artists would STILL never be happy with the amount because they're artists.
 
Also Thierry please layoff the Windows rants it makes you seem crazy and you have no idea how half those things you're slating work under the hood, nor that they're done the same way on all modern operating systems because of what they allow in terms of software and security. Also you can disable the Windows page file  EXTERNAL LINK 2911-4.html) but generally it is a bad idea, it's actually a really useful feature 99.99% of the time.
 
Andy-(sorry I'm in a BAD mood)


Uros Vidovic

Posts 10
21 Apr 2017 14:06


As an Atari user and owner of two Atari clones ( Milan 060 with 128 MB and FireBee with 512 Mb of RAM) I can asure you that 128 MB of RAM is more than enough for most of the tasks you will need your computer! Even for the artists. if you have that amount of data you also need to process that data one way or another. And 060, 080 and Coldfire are VERY slow CPU. It is not fun to process that much data on that slow CPU.

Most usefull use of that much of RAM is to use it as RAM disc. But you still need to fill than much of data and this still needs quite some time on so slow machine.

Looking forward to get any useful computer based on 080 as Atari clone ...


Wawa T

Posts 233
21 Apr 2017 14:19


this thread becomes the usual wall of text kind of keyborad warriors favourite chit-chat fest of cheap demands. just as a regular natami forum thread. if this is gonna progress this way, we are heading exactly where it went there. i hope gunnar and the team will simply impose their sensible decisions upon all of you..


Gunnar von Boehn
(Apollo Team Member)
Posts 2448
21 Apr 2017 16:33


Gentlemen,

please help to focus on this forum.

Please no more posts like "my windows does.."




Olaf Schoenweiss
(Apollo Team Member)
Posts 304
21 Apr 2017 17:14


+1
 
  512 MB RAM would be good. Some time ago I talked to Deadwood who did the Aros port of OWB. He wrote me that 512 MB RAM are needed because OWB needs at least 300 MB. Also Wawa already mentioned MESA/Gallium in Aros that needs 256 MB RAM to run. So if you want something that is enough equipped for even more demanding needs 512 MB are minimum. Of course 1 GB would be ok too propably if not making it too expensive. More is not needed for next future and nobody will try to do high-end graphics or editing of HD pictures on amiga because of not enough processing power so we should not ask for unrealistic things. BTW Gunnar and the other team members will decide anyway what they think is reasonable whatever we talk and dream here. But perhaps they will take one or two things from it :).


Wawa T

Posts 233
21 Apr 2017 18:43


thx, gunnar, much appreciated. ;)

Olaf Schoenweiss wrote:
512 MB RAM would be good. Some time ago I talked to Deadwood who did the Aros port of OWB. He wrote me that 512 MB RAM are needed because OWB needs at least 300 MB. Also Wawa already mentioned MESA/Gallium in Aros that needs 256 MB RAM to run.

yet, as i mentioned as well, both are aros use cases. and it remains to be seen if the usage of any of them makes any sense on (68k) ~100mhz speed range cpu. aros mesa softpipe can certainly serve just as a proof of concept. it doesnt make sense without proper hardware acceleration. and odyssey hasnt even been compiled to 68k yet. in its current state it needs a compiler of in the range of gcc5, so its good that aros provides gcc6, also native on 68k and that there is amiga-m68k gcc6 cross-toolchain being worked on with all bells and whistles.

considering this i see rather a bright future for further 68k development/ports within reach.

;)



Kolbjørn Barmen

Posts 93
21 Apr 2017 18:44


My A1200 currently has 192MB and I manage to run empty every now and then just by using it for "regular" tasks, without even using a web browser. My A3000 only has 128MB and quickly runs out of RAM to the point where it is annoying. One thing people should remember is that when using RTG/P96 sceeen modes, that is also RAM taken from the 128MB that the Vampire has, and the more screens you have open, the more RAM is used. A silly situation may even occure where P96 swaps out screens to fast ram to save "gfx card RAM", or keep copy of screens in Fast RAM - silly on Vampire since there is no "graphics board", and the RAM is the same.

posts 95page  1 2 3 4 5