Click Here
home features news forums classifieds faqs links search
5788 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
Home
Features
News
Forums
Classifieds
Links
Downloads
Extras
OS4 Zone
IRC Network
AmigaWorld Radio
Newsfeed
Top Members
Amiga Dealers
Information
About Us
FAQs
Advertise
Polls
Terms of Service
Search

IRC Channel
Server: irc.amigaworld.net
Channel: #Amigaworld
Channel Policy and Guidelines

(Uses JAVA Applet and Port 1024)
Visit the Chatroom Website

Who's Online
 66 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 outlawal2:  7 mins ago
 newlight:  15 mins ago
 robbhar:  18 mins ago
 remotenemesis:  37 mins ago
 Jasper:  38 mins ago
 cip060:  49 mins ago
 Robert:  1 hr 9 mins ago
 Mason:  1 hr 54 mins ago
 g0blin:  1 hr 54 mins ago
 Gothmog:  1 hr 55 mins ago

/  Forum Index
   /  Classic Amiga Hardware
      /  News about Vampire and Apollo
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 Next Page )
PosterThread
megol 
Re: News about Vampire and Apollo
Posted on 3-Jan-2017 20:59:41
#961 ]
Regular Member
Joined: 17-Mar-2008
Posts: 256
From: Unknown

@Samurai_Crow

Yes but the thing with Altivec was that it was designed to be generally useful (within limitations) and it took Intel ages to make something comparable (IIRC until SSE3). More registers are useful but it doesn't help when the instruction set chosen is obviously inferior. With that said for basic SIMD support a 3 operand MMX version isn't too bad, I just hope there are spaces in the encoding for improving the SIMD support in later versions. Judging by a post by Gunnar (that was quickly removed) describing one (potential?) encoding there should be...

 Status: Offline
Profile     Report this post  
Overflow 
Re: News about Vampire and Apollo
Posted on 4-Jan-2017 2:41:55
#962 ]
Super Member
Joined: 12-Jun-2012
Posts: 1536
From: Norway

A more general comment/copy regarding the improvements in the Vampire core below:
Another comment was the IDE speed for the upcoming V1200 IDE being quoted to be 12mb/sec, which is about 4 times the original speed.

The comment about ADoom not being optimized for the Vampire, yet, is a bit meh. If any ID games should be optimized it should be Quake 1 or Quake 3 Arena since they are actually competative games that are still played today. It would be the perfect games to port for a lively gaming scene on the Vampire.
I realise Quake 3 Arena would be a massive undertaking tho. Quake 1 is already running, altho slow (until FPU is incorporated).

Quote:
Gunnar von Boehn
(Apollo Team Member)
Posts 2288
03 Jan 2017 19:52

Quote:
Jan Vonka wrote:

ADoomWOS "timedemo demo1":

CyberstormPPC 604e/333 + 060/50
320x200: 77,03 fps
800x600: 16,79 fps

Sonnet Crescendo 7200 G3/400 (512k L2)
320x200: 110,63 fps
800x600: 25,35 fps


Can anybody benchmark Vampires 600 and 500+ right now?
Latest Vampire2 Core reaches at DOOM in 800x600 = 19.8 FPS.
So its somewhat faster than CyberstormPPC with PPC 604e@333MHz.
This is the normal 68k DOOM executable, its not tuned for AMMX or for APOLLO (yet).

With video playback the Vampire is even better.
RIVA video player which has some AMMX tunings, the VAMPIRE2 is more than twice as fast in video playback than the CyberstormPPC 604e@333MHz.

Both the Sonnet and Vampire play our Testvideo equally well.
And the Sonnet PCI with PPC-G3@400Mhz scores today a little bit faster than the Vampire in Doom.

Of course the tuning of the APOLLO core has just started.
And after GOLD1 comes GOLD2 and the there will come a GOLD3 release.

And is likely that the Vampire will beat the Sonnet PCI also in DOOM in the future.

Last edited by Overflow on 04-Jan-2017 at 02:42 AM.

 Status: Offline
Profile     Report this post  
OneTimer1 
Re: News about Vampire and Apollo
Posted on 4-Jan-2017 17:28:24
#963 ]
Regular Member
Joined: 3-Aug-2015
Posts: 380
From: Unknown

@Thread

When the Intel came up with MMX it was recommended using some kind of code asking first if this extension is available or not, so you could use the same code on CPUs with or without MMX.

Maybe Intel wasn't clear if they would keep it, the original MMX features became obsolete fast, because they where limited to integer processing. Today most GPUs could do more complex operations faster than MMX could, I don't even know if todays 64Bit CPUs support this poorly used command set.

https://en.wikipedia.org/wiki/MMX_(instruction_set)

Last edited by OneTimer1 on 04-Jan-2017 at 05:30 PM.

 Status: Offline
Profile     Report this post  
pavlor 
Re: News about Vampire and Apollo
Posted on 4-Jan-2017 18:55:54
#964 ]
Elite Member
Joined: 10-Jul-2005
Posts: 8815
From: Unknown

@Overflow

Doom is good application benchmark for Vampire: shows integer and 2D performance. Vampire can outshine even 603e/604e + PCI GFX there (thanks to really fast memory for both CPU and GFX). Quake 3 is out of question - requires 3D acceleration.

 Status: Offline
Profile     Report this post  
Seiya 
Re: News about Vampire and Apollo
Posted on 4-Jan-2017 21:58:11
#965 ]
Super Member
Joined: 19-Aug-2006
Posts: 1412
From: Italia

@pavlor

so, at this point it needs some doom benchmarks with standard processor (not overclocked)

Last edited by Seiya on 04-Jan-2017 at 09:58 PM.

_________________

 Status: Offline
Profile     Report this post  
pavlor 
Re: News about Vampire and Apollo
Posted on 5-Jan-2017 16:35:17
#966 ]
Elite Member
Joined: 10-Jul-2005
Posts: 8815
From: Unknown

@Seiya

I have ADoom WOS results for 320x200 (classic demo3 and shareware WAD):

Amiga 3000, Cyberstorm PPC604 (68060 60 MHz) 604e 233 MHz, VooDoo3/2000, 144? MB RAM Fast: 56.5 FPS
Amiga 1200T, BlizzardPPC (68060 50 MHz) 603e 150 MHz, Voodoo 3, 256 MB RAM, AmigaOS 3.9 BB2: 29.6

Now ADoom (1.3 I think; 68k):
Amiga 3000, Cyberstorm PPC604 (604e 233 MHz) 68060 60 MHz, VooDoo3/2000, 144? MB RAM Fast: 36.6 FPS
Amiga 4000D, Cyberstorm MKIII 060 @ 50MHZ, 128 MB Fast RAM, Cybervision 64/3D: 24.4 FPS

As you see 68k ADoom seems to be more optimized than WOS version (or there are other bottlenecks).

 Status: Offline
Profile     Report this post  
Seiya 
Re: News about Vampire and Apollo
Posted on 6-Jan-2017 0:27:50
#967 ]
Super Member
Joined: 19-Aug-2006
Posts: 1412
From: Italia

Quote:

pavlor wrote:


I have ADoom WOS results for 320x200 (classic demo3 and shareware WAD):

Amiga 3000, Cyberstorm PPC604 (68060 60 MHz) 604e 233 MHz, VooDoo3/2000, 144? MB RAM Fast: 56.5 FPS

Now ADoom (1.3 I think; 68k):
Amiga 3000, Cyberstorm PPC604 (604e 233 MHz) 68060 60 MHz, VooDoo3/2000, 144? MB RAM Fast: 36.6 FPS


604e:
56 FPS WOS
36 FPS 68k

wos version is much faster than 68k version

and the ohter case is maybe 603 very worst for slow memory access.

Last edited by Seiya on 06-Jan-2017 at 12:31 AM.
Last edited by Seiya on 06-Jan-2017 at 12:29 AM.

_________________

 Status: Offline
Profile     Report this post  
pavlor 
Re: News about Vampire and Apollo
Posted on 6-Jan-2017 15:31:28
#968 ]
Elite Member
Joined: 10-Jul-2005
Posts: 8815
From: Unknown

@Seiya

Quote:
wos version is much faster than 68k version


FPS/MHz:

604e/ADoomWOS - 0.24
68060/ADoom 68k - 0.61

 Status: Offline
Profile     Report this post  
gregthecanuck 
Re: News about Vampire and Apollo
Posted on 6-Jan-2017 17:55:51
#969 ]
Cult Member
Joined: 30-Dec-2003
Posts: 825
From: Vancouver, Canada

@pavlor

That FPS/MHz ratio is highlighting the difference in performance/clock of 68K vs. PPC. PPC has simpler instructions (hence RISC) but can be clocked higher.

With the "68080" Apollo core this is becoming even more apparent. Clock for clock the 68K architecture is far more performant than PPC.

Last edited by gregthecanuck on 06-Jan-2017 at 05:56 PM.

 Status: Offline
Profile     Report this post  
WolfpackN64 
Re: News about Vampire and Apollo
Posted on 6-Jan-2017 18:06:21
#970 ]
Regular Member
Joined: 8-Oct-2016
Posts: 299
From: Unknown

@gregthecanuck

Depends, the core the Apollo core is compared against is quite old.
CISC chips are for complex instructions indeed faster then RISC chips, but RISC chips will be faster when performing simple, shorter instructions that the architecture was build for.

The 68080 is one hell of a chipdesign for such a small team, but I wouldn't say the 68k architecture is surperior to PPC, especially modern POWER chips would hammer the 68080 on a perf/clock basis.

I think in many cases, it's not really a good comparison to make. How much software is there on OS3 that even uses the PPC accelerators?

 Status: Offline
Profile     Report this post  
pavlor 
Re: News about Vampire and Apollo
Posted on 6-Jan-2017 18:15:02
#971 ]
Elite Member
Joined: 10-Jul-2005
Posts: 8815
From: Unknown

@gregthecanuck

Quote:
That FPS/MHz ratio is highlighting the difference in performance/clock of 68K vs. PPC. PPC has simpler instructions (hence RISC) but can be clocked higher.


In CPU intensive tasks 604e shows better performance/MHz than 68060. Difference in ADoom performance/MHz means simply more optimized 68k port, I think.

 Status: Offline
Profile     Report this post  
gregthecanuck 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 9:26:55
#972 ]
Cult Member
Joined: 30-Dec-2003
Posts: 825
From: Vancouver, Canada

@WolfpackN64

Sortbench seems to be used as a benchmark recently.

Here is a result from a more modern PPC chip from http://www.apollo-core.com/sortbench/results/result_ppc_7447.txt

CPU PPC 7447
Clock 1.42 GHz
L2 512 KB
System ibook

SORTBENCH 1.1 (Gunnar von Boehn)
-------------------------------------------------------------
1 K Element : 1750.24 MB/sec
2 K Element : 1784.18 MB/sec
4 K Element : 1784.84 MB/sec
8 K Element : 1779.10 MB/sec
16 K Element : 1782.16 MB/sec
32 K Element : 1765.12 MB/sec

That works out to just over 1.2 MB/sec / MHz on 32K element test.

Here are some recent POWER (Power 7) results from http://www.apollo-core.com/sortbench/results/result_ppc_power7.txt:

./sort.aix
-------------------------------------------------------------
SORTBENCH 1.1 (Gunnar von Boehn)
Its a CPU benchmark that stresses CPU, DCache and branch prediction.
-------------------------------------------------------------
1 K Element : 4418.00 MB/sec
2 K Element : 4353.62 MB/sec
4 K Element : 4429.73 MB/sec
8 K Element : 4464.52 MB/sec
16 K Element : 4278.30 MB/sec
32 K Element : 4223.28 MB/sec

Processor Implementation Mode: POWER 7
Processor Version: PV_7_Compat
Number Of Processors: 9
Processor Clock Speed: 3864 MHz

That works out to 1.09 MB/sec / MHz. Slightly less performant than 7447 on this benchmark. But made up for by many cores, higher clock and threading.

Recent benchmark on Sonnet PPC G3/500 from http://www.apollo-core.com/knowledge.php?b=2e=3369&x=0&z=EHTMx6 ...

The WarpOS optimized SortBench gives the following corrected numbers.

1k = 634.82 MB/sec
2k = 635.04 MB/sec
4k = 635.16 MB/sec
8k = 635.18 MB/sec
16k = 538.31 MB/sec
32k = 515.31 MB/sec

That works out to 1.03MB/sec / MHz on 32K test.

Recent results from November against GOLD2 Apollo core (currently WIP but due "soon") from http://www.apollo-core.com/knowledge.php?b=2e=2923&x=2 ...

1k = 341.79 MB/sec
2k = 344.56 MB/sec
4k = 345.69 MB/sec
8k = 346.17 MB/sec
16k = 313.34 MB/sec
32k = 292.73 MB/sec

The clock speed on this core is somewhere around 100MHz. Getting roughly 3MB/sec / MHz. *Significantly* faster than any PPC/Power core shown above.

For grins here are results from an Intel Xeon E5 class http://www.apollo-core.com/sortbench/results/result_x86_xeon.txt:

Intel(R) Xeon(R) CPU E5-2680 0
Clockrate 2.70GHz
cache size : 20480 KB
bogomips : 5399.21

SORTBENCH 1.0 (Gunnar von Boehn)
-------------------------------------------------------------
1 K Element : 7659.10 MB/sec
2 K Element : 7750.36 MB/sec
4 K Element : 7756.81 MB/sec
8 K Element : 7813.89 MB/sec
16 K Element : 7612.14 MB/sec
32 K Element : 7568.92 MB/sec

Roughly 2.8 MB/sec / MHz on 32K test.

In summary, what this shows is that the enhancements being integrated into the Apollo core are resulting in a very efficient 680x0 design. Also keep in mind the core has no L2 cache and is running on a low-end FPGA. The V1200 is rumoured to be designed for a newer FPGA (Cyclone V vs. III) with more/faster memory.


 Status: Offline
Profile     Report this post  
WolfpackN64 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 10:42:18
#973 ]
Regular Member
Joined: 8-Oct-2016
Posts: 299
From: Unknown

@gregthecanuck

The benchmark looks to be pretty accurate for Amiga use. But I have my doubts when a Core Duo outperforms a POWER7 chip.

In threaded workloads, the POWER chip should outperform the Core by a significant margin (but on the other hand, not a single Amiga platform to date really makes use of multiple cores and threads).

 Status: Offline
Profile     Report this post  
gregthecanuck 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 10:46:57
#974 ]
Cult Member
Joined: 30-Dec-2003
Posts: 825
From: Vancouver, Canada

@WolfpackN64

The benchmarks are testing single-core/thread performance from what I understand. There is no internal multi-threading.

Last edited by gregthecanuck on 07-Jan-2017 at 10:48 AM.

 Status: Offline
Profile     Report this post  
WolfpackN64 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 11:00:37
#975 ]
Regular Member
Joined: 8-Oct-2016
Posts: 299
From: Unknown

@gregthecanuck

Then the tests seem to be quite reasonable. The Apollo's perm/watt is quite an achievement for such a small team. They certainly put Intel's chip design team to shame.

 Status: Offline
Profile     Report this post  
gregthecanuck 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 11:14:40
#976 ]
Cult Member
Joined: 30-Dec-2003
Posts: 825
From: Vancouver, Canada

@WolfpackN64

Yes this is one smart team. They have been working on this for many years.

Here is a nice wiki page showing the features in the core: http://apollo-accelerators.com/wiki/doku.php?id=apollo_core#features

 Status: Offline
Profile     Report this post  
WolfpackN64 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 11:21:34
#977 ]
Regular Member
Joined: 8-Oct-2016
Posts: 299
From: Unknown

@gregthecanuck

I hope they get to an ASIC design one day. I can see a lot of embedded applications for something like this (next to the classic systems of course).
If only something similar existed for embedded PPC :)

 Status: Offline
Profile     Report this post  
Overflow 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 13:52:28
#978 ]
Super Member
Joined: 12-Jun-2012
Posts: 1536
From: Norway

Gunnar von Boehn
(Apollo Team Member)
Posts 2298
07 Jan 2017 10:48

Quote:
Ian Parsons wrote:

Is that confirmation of a USB controller on the Vampire 1200? If so it sounds like the board will make (almost) everybody very happy.
The V1200 has 2 expansion ports.
These ports are much faster than old Clockport.


There will be USB, Network, and Wifi expansion for them.
We are right now testing / developing drivers for those expansions.

 Status: Offline
Profile     Report this post  
pavlor 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 14:46:53
#979 ]
Elite Member
Joined: 10-Jul-2005
Posts: 8815
From: Unknown

@gregthecanuck

Sorry, but this sort of benchmarks is meaningless for us. Real world application benchmarks would be far more interesting.

Quote:
The V1200 is rumoured to be designed for a newer FPGA (Cyclone V vs. III) with more/faster memory.


I hope it will not rise price of the board.

 Status: Offline
Profile     Report this post  
gregthecanuck 
Re: News about Vampire and Apollo
Posted on 7-Jan-2017 21:03:15
#980 ]
Cult Member
Joined: 30-Dec-2003
Posts: 825
From: Vancouver, Canada

@WolfpackN64

Quote:
I hope they get to an ASIC design one day. I can see a lot of embedded applications for something like this (next to the classic systems of course).


ASIC comes up all the time. Everyone wants more speed. :)

However the core still has a long way to go before ASIC is considered. In the meantime even on a low-end FPGA this is a very cool project.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright 2000 - 2017 Amigaworld.net.

Amigaworld.net was originally founded by David Doyle