Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Performance and Benchmark Results!

160 MIpspage  1 2 3 

Roger Shimada

Posts 30
30 Aug 2017 00:50


Gunnar von Boehn wrote:
Sysinfo claims to measure Dhrystone results but does not implement the dhrystone test.

That's too bad.

This is the K&R UNIX source that I used EXTERNAL LINK 
Sorry, I'm not (yet?) an Amiga programmer.




Thierry Atheist

Posts 644
30 Aug 2017 21:37


Gunnar von Boehn wrote:
SYSINFO has some selfmade instruction test -

it measures how long it takes and prints out the Score as MIPS and dhrystone - but this test is not Dhrystone at all and has nothing to do with it.

The Sysinfo numbers are fake - and can not be compared to any real MIPS or DHRYSTONE numbers


Hi Gunnar,

"Fake"? Well, who cares?

We are comparing "apples to apples" when we run (the same) Sysinfo program on all of the systems, and the Vampire scores quite higher and now MUCH higher than the highest result of any ordinary unoverclocked 680x0 CPUs!!!!!

The highest (Vampire 2) benchmark I saw was 104.14 MIPS, and NOW it's hit 160.97!!! That's 54.6% higher than the previous best!!!!!

And now, even the Raspberry Pi 3's number of 136.38 has been bested by 18%!!!!!!

And that's ONLY a 1.2 GHz CPU. I suppose, though, only one of the 4 cores is being used. To match the new result it would have to run at 1.42 GHz... but even then, no one is probably EVER going to make an UAE that also does S-AGA, extra registers, 64 bit assembler instructions, AMMX, and the new advanced FPU !*1,000

Funny, hilarious?, thing is, if one DID make that version of UAE, it would run CONSIDERABLY SLOWER than the current Amiga emulations that are on any other computer platforms! HA HA HA!!!!


David Wright

Posts 373
31 Aug 2017 00:21


Yeah, because I just wasn't enjoying my Amiga at just 104 mips.
At some point speed is irrelevant or at least should take a back seat to other needs.

The Rgb through Hdmi and aga compatibility are improvements that will make this new Amiga experience much more enjoyable.

That and new software which I know is not the Apollo teams purpose but for others to do.


Thierry Atheist

Posts 644
31 Aug 2017 07:40


David Wright wrote:
Yeah, because I just wasn't enjoying my Amiga at just 104 mips.

At some point speed is irrelevant or at least should take a back seat to other needs.


Weeeellllllllll, you DID read the interview of Artur Jarosik?
CLICK HERE 
He's saying "Amiga isn't fast enough for web browsing". Now, there are levels of web browsing that are "adequate", but AMIGA can do MORE than what it does now WITH that extra speed!!!!

We would not have 1920x1080 output without the "excessive speed" we are getting from the Vampire Apollo Core made for the FPGAs they are using!!!!!

"Odyssey is the best NG browser. It supports new technologies but for the classic it is too demanding.
Only fast FPGAs could pull it off like the announced Vampire card on Arria 10, which is only planned. As I wrote above, the next Vampire will be based on Cyclone V, which is at least 20% faster than Cyclone 3." -- Artur Jarosik

Artur said that NetSurf works pretty good, but OWB needs more horsepower. When you go to a website, and it has 20 to 30 JPEGs.... YOU NEED ANY EXTRA SPEED YOU CAN GET to decode and display it "quickly".

Now, IF we were able to get an (64 bit) assembler level (using S-AGA and the new FPU and AMMX) coded web browser for the Vampire, well, that's ANOTHER STORY ALTOGETHER! (Ported software is GARGANTUAN, however they have access to 8 gigabytes of RAM and (256 bit GPU) video cards with 4 gigs and 1 terabyte hard drives.)

AMIGA DOES have the power to be "pretty ____ GOOD"!!!!!!!


Eric Gus

Posts 477
31 Aug 2017 07:49


Thierry Atheist wrote:

David Wright wrote:
Yeah, because I just wasn't enjoying my Amiga at just 104 mips.
 
  At some point speed is irrelevant or at least should take a back seat to other needs.

  Weeeellllllllll, you DID read the interview of Artur Jarosik?
  CLICK HERE 
  He's saying "Amiga isn't fast enough for web browsing". Now, there are levels of web browsing that are "adequate", but AMIGA can do MORE than what it does now WITH that extra speed!!!!
 
  We would not have 1920x1080 output without the "excessive speed" we are getting from the Vampire Apollo Core made for the FPGAs they are using!!!!!
 
  "Odyssey is the best NG browser. It supports new technologies but for the classic it is too demanding.
  Only fast FPGAs could pull it off like the announced Vampire card on Arria 10, which is only planned. As I wrote above, the next Vampire will be based on Cyclone V, which is at least 20% faster than Cyclone 3." -- Artur Jarosik
 
  Artur said that NetSurf works pretty good, but OWB needs more horsepower. When you go to a website, and it has 20 to 30 JPEGs.... YOU NEED ANY EXTRA SPEED YOU CAN GET to decode and display it "quickly".
 
  Now, IF we were able to get an (64 bit) assembler level (using S-AGA and the new FPU and AMMX) coded web browser for the Vampire, well, that's ANOTHER STORY ALTOGETHER!
 
  AMIGA DOES have the power to be "pretty ____ GOOD"!!!!!!!
 
 

Gonna have to side with David a bit on this Thierry, speed yes, its important but not everyone wants to browse the web on their Amigas..! there is a point of diminishing returns, once you hit ludicrous speed on the Amiga, I do think other things probably should get more focus for a while, such as the items David mentions or say how about work on wifi networking support with an inexpensive ESP8266 module (instead of SDNET) etc., As an Vampire user those things would be far more beneficial to me NOW than a 10% speed increase, (I know im just itching to get 2mb chip into my 512Kb chip A500, I can't run whdload or anything with that, speed is great but if you are hamstrung its pointless (fast engine you can't actually drive anywhere yet because you got go-cart wheels) .. its not like they can't go back and give us a firmware update 6 months from now that gives us the extra speed.. Personally Id like to see other features get a bit more focus (granted they may doing that now and we just dont know ..) .. just my 2 cents.



Thierry Atheist

Posts 644
31 Aug 2017 08:02


Hi Eric,

(I slightly altered my post in the short time it took you to reply.)

Yes, I agree with you. But (lack of) speed is a very contentious issue with many who denigrate us for wanting to STAY "old school".


Mr Niding

Posts 459
31 Aug 2017 09:38


Well, there is major gains to be had with regards to browser thru AMMX support, to improve the jpeg encoding etc.
And then you have the download speed of the pages itself.
And fast enough memory etc.

V4 will takes care of several aspects, but the browsers got alot of ground to cover actually utilizing the features of the core to maximize performance.


Sebastian Blanco

Posts 148
31 Aug 2017 21:17


The vampire is getting very fast each day, you guys are awesome!.
Wish my vampire arrive so i can try it.



Benutzer A4K

Posts 24
12 Sep 2017 13:14


Gunnar von Boehn wrote:

  MINIBENCH is unfortunately very inaccurate
  on a fast machine when called with small test parameter size (default) please set to biggest to get correct scores.


 
:) I just wanted to beat the highest Minibench CPU Score from that link EXTERNAL LINK 
Amiga 600 Vampire V2-128 Apollo Core 3053 (SILVER2).  106.35MHz (x15) CPU: 9856 MEM: 5146

Olli.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
12 Sep 2017 13:27


Benutzer A4K wrote:

Gunnar von Boehn wrote:

 
    MINIBENCH is unfortunately very inaccurate
    on a fast machine when called with small test parameter size (default) please set to biggest to get correct scores.
 

   
  :) I just wanted to beat the highest Minibench CPU Score from that link EXTERNAL LINK 
  Amiga 600 Vampire V2-128 Apollo Core 3053 (SILVER2).  106.35MHz (x15) CPU: 9856 MEM: 5146
 
  Olli.

Olli the default "testrun-time" is to short for 68080 and 68060
Please run it with STRONG or STRONGEST parameter then the measurment is correct



Mallagan Bellator

Posts 393
12 Sep 2017 22:33


eric gus wrote:

Thierry Atheist wrote:

 
David Wright wrote:
Yeah, because I just wasn't enjoying my Amiga at just 104 mips.
 
  At some point speed is irrelevant or at least should take a back seat to other needs.

  Weeeellllllllll, you DID read the interview of Artur Jarosik?
  CLICK HERE   
  He's saying "Amiga isn't fast enough for web browsing". Now, there are levels of web browsing that are "adequate", but AMIGA can do MORE than what it does now WITH that extra speed!!!!
 
  We would not have 1920x1080 output without the "excessive speed" we are getting from the Vampire Apollo Core made for the FPGAs they are using!!!!!
 
  "Odyssey is the best NG browser. It supports new technologies but for the classic it is too demanding.
  Only fast FPGAs could pull it off like the announced Vampire card on Arria 10, which is only planned. As I wrote above, the next Vampire will be based on Cyclone V, which is at least 20% faster than Cyclone 3." -- Artur Jarosik
 
  Artur said that NetSurf works pretty good, but OWB needs more horsepower. When you go to a website, and it has 20 to 30 JPEGs.... YOU NEED ANY EXTRA SPEED YOU CAN GET to decode and display it "quickly".
 
  Now, IF we were able to get an (64 bit) assembler level (using S-AGA and the new FPU and AMMX) coded web browser for the Vampire, well, that's ANOTHER STORY ALTOGETHER!
 
  AMIGA DOES have the power to be "pretty ____ GOOD"!!!!!!!
 
 
 

 
  Gonna have to side with David a bit on this Thierry, speed yes, its important but not everyone wants to browse the web on their Amigas..! there is a point of diminishing returns, once you hit ludicrous speed on the Amiga, I do think other things probably should get more focus for a while, such as the items David mentions or say how about work on wifi networking support with an inexpensive ESP8266 module (instead of SDNET) etc., As an Vampire user those things would be far more beneficial to me NOW than a 10% speed increase, (I know im just itching to get 2mb chip into my 512Kb chip A500, I can't run whdload or anything with that, speed is great but if you are hamstrung its pointless (fast engine you can't actually drive anywhere yet because you got go-cart wheels) .. its not like they can't go back and give us a firmware update 6 months from now that gives us the extra speed.. Personally Id like to see other features get a bit more focus (granted they may doing that now and we just dont know ..) .. just my 2 cents.
 

When you're working on a programming project of any kind that takes alot of code, you're better of finishing what you're working on, because coming back to it later, you might have a hard time getting your head back in that specific game. When you already have your head in the game, you know what the code looks like and where stuff is. In short, it's better to finish what you started, and then get into another part of the project


Mallagan Bellator

Posts 393
20 Sep 2017 21:38


One thought...

Arn't cpu instructions in different lengths?
Saying "Million Instructions Per Second", isn't that in a way somewhat misleading, in general? Or does one use the most common instructions and try to push as many of them as possible into a cycle over a couple of seconds to determine a MIPS count?

For example, if the shortest instruction would be half the length of the longest, and the longest one was used more times in a cycle, wouldn't that practically lower the MIPS count in that situation?


Nixus Minimax

Posts 416
21 Sep 2017 11:56


Mallagan Bellator wrote:
Arn't cpu instructions in different lengths?
  Saying "Million Instructions Per Second", isn't that in a way somewhat misleading, in general?

Yes, it is totally misleading! A CISC instruction often will do much more work than a RISC instruction. Processors like MIPS or ARM who didn't even have a DIV instruction had to execute dozens of other instructions instead. And yes, for marketing reasons a processor company wouldn't use the slower instructions taking more than one cycle for counting MIPS. So basically a MIPS number is as meaningful as a MHz number, it can't give you more than a rough estimate. This is why this is about "Sysinfo MIPS", a well-known (stupid) standard in the Amiga world that actually disadvantages superscalar CPUs like the 060 and the 080.



Mallagan Bellator

Posts 393
21 Sep 2017 12:08


Still, we need something to go by, I guess. I always thought real performance was a better measure stick. Like how many fps can you get in Doom or Quake. That's real


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
21 Sep 2017 12:21


Nixus Minimax wrote:

And yes, for marketing reasons a processor company wouldn't use the slower instructions taking more than one cycle for counting MIPS.

Lets try to clear this up a little.
a) Yes 68K supports instruction of different length

EXAMPLE


                  LENGTH      CLK on 68000    CLK on 68080
MOVEQ  #0,D0    2 BYTE      4                0.5
MOVE.w #0,D0    4 BYTE      8                0.5
MOVE.l #0,D0    6 BYTE      12              0.5
MOVE.l #0,(A0)  6 BYTE      20              0.5

So yes longer instructions took longer on 68000 -
but the 68080 is generally unaffected by instruction length


Mallagan Bellator

Posts 393
21 Sep 2017 18:17


Ya, but if you compare it to other CISC cpus (to be fair), like intel I5, it would be diffferent instructions, right?
Yes, they are clocked a thousand times faster, but still


Thierry Atheist

Posts 644
21 Sep 2017 18:18


Mallagan Bellator wrote:
Still, we need something to go by, I guess. I always thought real performance was a better measure stick. Like how many fps can you get in Doom or Quake. That's real

I'm going to take a "leap of faith" here, on Amiga you can disable multitasking, on win-dos, who knows WHAT the hell is going on "behind the scenes"?

Advantage: AMIGA!


Mallagan Bellator

Posts 393
21 Sep 2017 18:31


In windows, yes, in MsDos, not much...


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
21 Sep 2017 21:44


Mallagan Bellator wrote:

  Ya, but if you compare it to other CISC cpus (to be fair), like intel I5, it would be diffferent instructions, right?
  Yes, they are clocked a thousand times faster, but still
 

 
Clockrate is only a question of production as ASIC.
FPGA are always slow.
As FPGA can by nature not reach the same clock as on ASIC.
 
But as a matter of fact, IBM PPC or Intel CPU put in an FPGA will run slower than APOLLO.
   
Its common practice for CPU development companies during development
to put their next CPU core in an FPGA.
 
For example go to ARM developer side and look up the ARM test-cores for FPGA - they reach 50 MHz only, even in more expensive FPGA than VAMP. In ASIC the same core reaches 2 GHz.
 
Put 68080 in ASIC and Gigaherz is no problem.


Crow Mohikan

Posts 78
21 Sep 2017 22:56



Fpga to asic conversion:
 
  EXTERNAL LINK

posts 43page  1 2 3