Overview Features Instructions Performance Forum Downloads Products Reseller Contact

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 TopicsNewsPerformanceGamesDemosApolloVampireReleases
Performance and Benchmark Results!

Vampire 4, All the Same Speed?page  1 2 3 

Thierry Atheist

Posts 612
28 Dec 2017 09:57


Are there going to be different speed FPGAs available to get on the standalone Vampire 4 motherboard?


Mallagan Bellator

Posts 238
28 Dec 2017 10:16


Thatís a good question actually.
Since it IS the standalone, maybe they could do that. Maybe even on the A500/A600A1200 versions, I guess, but not sure if they would need some kind of redesign to make that happen


Michal Warzecha

Posts 203
28 Dec 2017 11:23


I'm affraid changing speed it's not only faster chip. Probably You need change something with timmings of some core parts. And You need to get almost same chip with higher clock speed. Even small change in FPGA construction (pinout or something) will push them to redesing board or core.


Vojin Vidanovic

Posts 742
28 Dec 2017 16:56


Thierry Atheist wrote:

  Are there going to be different speed FPGAs available to get on the standalone Vampire 4 motherboard?
 

All v4 are same Cyclone 5 FPGA. Extended space to 3 (v2 Vamps) exists, but is more used to new features then to speeding up the core.
 
  Maybe at levels of "black editions" (x11,x13 ... multipliers) but not much. Benefits should be larger cache, faster RAM and full hard FPU that should give performance difference (to current v2, even with GOLD3).


Thierry Atheist

Posts 612
28 Dec 2017 18:09


Michal Warzecha wrote:
I'm affraid changing speed it's not only faster chip. Probably You need change something with timmings of some core parts. And You need to get almost same chip with higher clock speed.

Hi Michal,

Yes, that's what I mean. Is there a pin compatible faster FPGA available? And yes, they would need to recompile the core for the faster timings of the higher frequency FPGA....

Vojin Vidanovic wrote:
All v4 are same Cyclone 5 FPGA. Extended space to 3 (v2 Vamps) exists, but is more used to new features then to speeding up the core.

Maybe at levels of "black editions" (x11,x13 ... multipliers) but not much.


Hi Vojin,

Adding to what I wrote above, they did do that for the "black editions". So I'm basically asking, are they going to make a few different speed Vampire 4's?

I'm willing to pay up to $200US more for a faster FPGA, if there is some reasonable speed increase over the one they are making available early in 2018. Say, 30% or more speed increase for US$200.

I'm aware that the increased speed FPGAs steeply increase in price.


Roy Gillotti

Posts 99
28 Dec 2017 19:15


Well I have to wonder with the A1200 version of the V4, the base clock is faster, so would a x11 multiplier with a 14Mhz Clock give you approximately a 154 Mhz Vampire? or will the multiplier have to be lowered to make things stable?


Andy Hearn

Posts 36
28 Dec 2017 19:23


I must admit, that's something I wondered too.

I kinda figured the V4 is going to be a standard set of features across the board. no need to re-flash to be a500/a1200 compatible as it's all completely supported in the 68080 core anyway.

and I did wonder about the 14Mhz base clock of the 1200 vs the 7Mhz 5/600. just I never put the two together.

I guess we'll know as soon as the Apollo genius's have it sorted :)



Vojin Vidanovic

Posts 742
28 Dec 2017 19:23


Roy Gillotti wrote:

  x11 multiplier with a 14Mhz Clock give you approximately a 154 Mhz Vampire? or will the multiplier have to be lowered to make things stable?
 

 
  Speed was relative to A500/A600 >:)
 
 
Thierry Atheist wrote:

  I'm willing to pay up to $200US more for a faster FPGA, if there is some reasonable speed increase over the one they are making available early in 2018. Say, 30% or more speed increase for US$200.
 

   
  I suppose such experiments need tweaked core and (even) more work, not that it is impossible. Lets see what team says. Cyclone 5 rates at 77K LE, I wonder what could fit there. But I do understand 2.7 and 3 releases (and USB and LAN drivers and full SAGA ones supporting AGA modes and Pamela) are more important. And overall v4 release. I suppose such "black editions" could come later!


Thierry Atheist

Posts 612
28 Dec 2017 20:41


Vojin Vidanovic wrote:
I suppose such experiments need tweaked core and (even) more work, not that it is impossible. Lets see what team says.

Yes, it's absolutely up to them. It's just that, since Arria 10 didn't work out, and that's now something in the distant future, it would be good in my opinion to still have a flagship standout better system on show.

Vojin Vidanovic wrote:
  Cyclone 5 rates at 77K LE, I wonder what could fit there. But I do understand 2.7 and 3 releases (and USB and LAN drivers and full SAGA ones supporting AGA modes and Pamela) are more important. And overall v4 release. I suppose such "black editions" could come later!

Need every extra % possible to decompress video!!!!


Vojin Vidanovic

Posts 742
28 Dec 2017 20:59


Thierry Atheist wrote:

  Need every extra % possible to decompress video!!!!

Currently that would have to be done elsewhere.

First, we ll see what can be done with "ordinary" v4.

Maybe in the future, some after-Cyclone-FPGA design for speed (and TherrY), like that mention Arria 10 to bang the FPGA limit.
But both core (with chipset) and drivers have to be mature for that game "of nearly-ASIC pushing the FPGA limit" game.



Michal Warzecha

Posts 203
29 Dec 2017 07:48


In my opinion We should stop asking for more. Original V4 is not produced yet and ppl still want more. It's problem to split model to different revisions, one faster, one more advanced, and another more XYZ. For now there is no any existing software to fully use V4. We don't need more for now.


Gunnar von Boehn
(Apollo Team Member)
Posts 3088
29 Dec 2017 13:35


Let my answer with some information.

The VAMPIRE V2 and V4 use CYCLONE FPGAs.
The VAMPIRE are shipped with x11 clockrate ~(78 Mhz).
The 68080 can reach in the CYCLONE a clockrate between ~ 80-110 MHz.

CPU clockrate is one factor in the speed/performance of the card.
Another factor is memory speed and also CPU efficiency.

68080 is of all 68K family CPUs the most efficient.
This means at the same clockrate 68080 is faster than 68060 and much faster than 68040, and very much faster than 68030 and lower.

The V4 will come out with the same clockrate at which the V2 are shipped. The V4 main advantages are bigger caches and a lot more memory.

I think for all Vampire users there is huge gain in software tuning.

If you saw the TEXTURE DEMO Videos lately posted here, then you might have seen that the texture mapping had a quality level - never seen on AMIGA before. Such high quality texture mapping needs a lot CPU power - to create the same texture quality with an 68060 CPU - you would have needed around 1 GIGAHERZ clockrate of the 68060.

This clearly shows the huge performance advantage that AMMX offers.



Roy Gillotti

Posts 99
29 Dec 2017 15:30


@Gunnar, So likely an A1200 Vampire with 14Mhz clock will run a multiplier that will put it somewhere in the 80-110Mhz range?


Stefano Briccolani

Posts 82
29 Dec 2017 16:42


the clock of your Amiga have nothing to do with the x11 multiplier. This multiplier is related to the vampire fpga. Going over x11 is like overclocking the fpga. Gunnar said that even with same clock multiplier a v4 will be faster than v2 thanks to memory bandwith of v4 boards..


Vojin Vidanovic

Posts 742
29 Dec 2017 16:48


Roy Gillotti wrote:

  @Gunnar, So likely an A1200 Vampire with 14Mhz clock will run a multiplier that will put it somewhere in the 80-110Mhz range?
 

 
  x11 clockrate ~(78 Mhz).
 
  Again, Vamp has its own "clockrate" relative to A500/A600 speed. If Vamp goes to A3000 it will not alter the clockrate to 030 16Mhz CPU.
 
  78Mhz is its current "x11 standard rate" with several special editions that do x13-x15. This had special cores and were NEVER widely released, so far reaching 107,4Mhz. (thus as Gunnar mentions, current 080 is up to 110Mhz).
 
  However, its not all in CPU speed, as mentioned. Larger cache, faster DDR5 RAM, AMMX/2 instructions, fast IDE and expected full hardware FPU and improved SAGA drivers all do their share. Overall, they should combined bring +30% speed "feel" to v2 GOLD2 Vamps that users are looking too, but not instant CPU operations speed increase.


Roy Gillotti

Posts 99
29 Dec 2017 17:31


So it ignores the Amiga clock signal completely? I figured it was just multiplying the rate set by the Amiga's internal clock as I figured keeping the CPU in sync with the other devices needing a clock signal would be important.



Gunnar von Boehn
(Apollo Team Member)
Posts 3088
29 Dec 2017 17:35


All users will benefit from improved software.
 
SAGA FASTIDE can provide HD speed of 26 MB/sec.
This speedf is not reached on AMIGA OS on IDE yet.
But the hardware is NOT the limit but the AMIGA OS device driver.
The old AMIGA OS 3.x IDE driver reaches ~ 12 MB/sec.
 
Some software tuning can easily get 100% speedup for all users.

This is just one example, in many ways the performance of the VAMP is not fully utilized by the software.


Mallagan Bellator

Posts 238
29 Dec 2017 18:25


Michal Warzecha wrote:

  For now there is no any existing software to fully use V4. We don't need more for now.

Doom? Quake? Quake II?


Gunnar von Boehn
(Apollo Team Member)
Posts 3088
29 Dec 2017 18:38


Mallagan Bellator wrote:

Michal Warzecha wrote:

  For now there is no any existing software to fully use V4. We don't need more for now.
 

  Doom? Quake? Quake II?

Correct all this software only scratches the tip of the iceberg of possible performance.
Tuned this could run many times better.


Mallagan Bellator

Posts 238
29 Dec 2017 18:53


Gunnar von Boehn wrote:

Mallagan Bellator wrote:

 
Michal Warzecha wrote:

    For now there is no any existing software to fully use V4. We don't need more for now.
 

  Doom? Quake? Quake II?
 

  Correct all this software only scratches the tip of the iceberg of possible performance.
  Tuned this could run many times better.

Iím not sure many know that you can actually increase the render resolution of Doom, but you can. I think I read that vampire reaches the limit of 35 fps that doom normally has (?), but if you wanna run Doom in 1280x720 in a good framerate... yaaaa, youíre gonna need horse power. Or AMMX/AMMX2 optimization, with said horse power :) quake is another fine example, so there is definitely software that requires powaaahhh!

posts 48page  1 2 3