| Happy Birthday Arne! | page 1 2 3 4 5
|
|---|
|
|---|
Mr. Derp
USA
| |
Posts 41
08 Jun 2011 09:53
| I agree. People have been willing to spend a lot of money for 060 accelerator boards and judging by these benchmarks the Natami blows them away.
Great job!
| |
Lord Aga
| |
Posts 129
08 Jun 2011 09:54
| Gunnar von Boehn wrote:
|
I think this is a not bad.
|
This is far from bad actually. It's fantastic :)
| |
Tito Amigang
France
| | (Natami Team)
Posts 29
08 Jun 2011 11:05
| yes very good and as said gunhar : "the CPU is not the strongpoint of the classic AMIGA/NATAMI" but the goal is not to challenge with PC , these cpu perfs are very good for this project.
make always better is never the better choice :)
the more important now is to finalize super AGA , for sell us the first stable Natami model asap , because we are waitiiiiiiiiing on it ! :)
if i'm right, you'll can enhance and fix bugs of super aga and 68050/70 after this step , and we'll download your "service pack" :) for update our FPGA ?
| |
Angel of Paradise
Germany
| |
Posts 61
08 Jun 2011 11:18
| Gunnar von Boehn wrote:
|
BTW the current memcopy SysSpeed results of the NATAMI are already a little better than a Pegasos 1 could do!
|
Is this true?
How sure are you that the 68060 can beat a PowerPC in memory performance?
| |
Marcel Verdaasdonk
Netherlands
| |
Posts 3976
08 Jun 2011 11:26
| Angel of paradise it's real simple by using better Memory
| |
André Jernung
Sweden
| | (MX-Board Owner)
Posts 988
08 Jun 2011 11:33
| Tito Amigang wrote:
|
if i'm right, you'll can enhance and fix bugs of super aga and 68050/70 after this step , and we'll download your "service pack" :) for update our FPGA ?
|
Yes, this is correct, and one of the benefits of using a FPGA-based system. We can fine tune and enhance the system after release.
But before selling to the public, we must finish support for all parts of the system in both hardware and software. There are still quite some things to do there.
We cannot possibly source components for a big production run anytime soon anyway. The FPGA model we use is hard to get hold of currently. But such quantities of components will most likely be available during the latter half of 2011. Then it would be possible to make a big production run. Stay tuned.
In the meantime, we have enough components for making some more boards for team developers to accelerate the development process. We are working on production specifics for this small batch of boards now.
| |
Christian Kummerow
Germany
| |
Posts 314
08 Jun 2011 12:04
| Angel of Paradise wrote:
|
Is this true?
How sure are you that the 68060 can beat a PowerPC in memory performance?
|
Later maybe, i dont know what Pegasus types exists but long time ago i got a Screenshot of one of my Programms it showed more than 148MB/s.
I dont have the Screenshoot anymore, so i googled it and found screenshoots of it showing 155MB/s. That combined read/write.
The CSPPC@60MHZ BUS/604e@233MHZ already reach 116MB/s with Sysspeed.
| |
Tito Amigang
France
| | (Natami Team)
Posts 29
08 Jun 2011 12:20
| @André Jernung : i'm developer and i'm working on a little program connected to a web site (on Aros ) . Do you expect to launch "beta test program", like on X1000, for developers which would develop program on natami ?
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 12:48
| Angel of Paradise wrote:
|
Gunnar von Boehn wrote:
|
BTW the current memcopy SysSpeed results of the NATAMI are already a little better than a Pegasos 1 could do!
|
Is this true?
How sure are you that the 68060 can beat a PowerPC in memory performance?
|
200% sure.
Some examples:
Some time ago, people did extensive benchmarks of PPC systems.
Out of my head, the glibc memcopy results for typical AMIGA PPC systems where:
- Pegasos 600 MHz G3 reaches 120 MB/sec
- AMIGA ONE with 800 MHz G3 reaches 150 MB/sec
- AMIGA ONE with 933 MHz G4 reaches 178 MB/sec
- I think EFIKA scores around 100 .
- and SAM score around 180 MB/sec
It might sound unbelievable but the 68060 is able to beat all of those.
It might not sound breathtaking but a 68060 is also able to beat many old PC systems e.g the 68060 can beat a typical 600 MHz AMD K6-2 system.
The NATAMI 68060 CPU card can beat many of the above PPC systems.
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 12:49
| Christian Kummerow wrote:
|
The CSPPC@60MHZ BUS/604e@233MHZ already reach 116MB/s with Sysspeed.
|
In which test exactly?
| |
Christian Kummerow
Germany
| |
Posts 314
08 Jun 2011 13:11
| Gunnar von Boehn wrote:
|
Christian Kummerow wrote:
|
The CSPPC@60MHZ BUS/604e@233MHZ already reach 116MB/s with Sysspeed.
|
In which test exactly?
|
Readb 95,99
Readh 110,12
Readw 116,36
Writeb 96,81
Writeh 85,31
Writew 75,56 Gunnar von Boehn wrote:
|
>Pegasos 600 MHz G3 reaches 120 MB/sec
|
And if there are no other Pegs with higher MHZ this must be wrong.
If you get 155MB/s as output from my Programm that means Writingtime+Readingtime(including compare and outputtext)/2.
Thats standard C memset and then Read&Compare bytes.
Exept the MOS WOS Emulation dont work right. I may verify it later with MOS on CSPPC, my Amiga is busy..
| |
Thomas Clarke
United Kingdom
| |
Posts 286
08 Jun 2011 13:11
| Gunnar von Boehn wrote:
|
Out of my head, the glibc memcopy results for typical AMIGA PPC systems where:
- Pegasos 600 MHz G3 reaches 120 MB/sec
- AMIGA ONE with 800 MHz G3 reaches 150 MB/sec
- AMIGA ONE with 933 MHz G4 reaches 178 MB/sec
- I think EFIKA scores around 100 .
- and SAM score around 180 MB/sec
It might sound unbelievable but the 68060 is able to beat all of those. |
If that's true, what are the design mistakes that PPC designers made that the 68060 designers did not, with regards to memory access?
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 13:14
| Christian Kummerow wrote:
|
Gunnar von Boehn wrote:
|
Christian Kummerow wrote:
|
The CSPPC@60MHZ BUS/604e@233MHZ already reach 116MB/s with Sysspeed.
|
In which test exactly?
|
Readb 95,99
Readh 110,12
Readw 116,36
Writeb 96,81
Writeh 85,31
Writew 75,56
|
Actually we were talking about Fast2Fast memcopy.
What score do you get for Fast2Fast copy? Christian Kummerow wrote:
|
Gunnar von Boehn wrote:
|
>Pegasos 600 MHz G3 reaches 120 MB/sec
|
And if there are no other Pegs with higher MHZ this must be wrong.
If you get 155MB/s as output from my Programm that means Writingtime+Readingtime(including compare and outputtext)/2.
Thats standard C memset and then Read&Compare bytes.
Exept the MOS WOS Emulation dont work right. I may verify it later with MOS on CSPPC, my Amiga is busy..
|
But this is not what we talk about here.
What do you get for MEM-COPY ?
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 13:20
| Thomas Clarke wrote:
|
Gunnar von Boehn wrote:
|
Out of my head, the glibc memcopy results for typical AMIGA PPC systems where:
- Pegasos 600 MHz G3 reaches 120 MB/sec
- AMIGA ONE with 800 MHz G3 reaches 150 MB/sec
- AMIGA ONE with 933 MHz G4 reaches 178 MB/sec
- I think EFIKA scores around 100 .
- and SAM score around 180 MB/sec
It might sound unbelievable but the 68060 is able to beat all of those. |
If that's true, what are the design mistakes that PPC designers made that the 68060 designers did not, with regards to memory access?
|
I would phrase it the other way around. The 68060 is a good CPU.
And if you really push it to its maximum - as the NATAMI CPU card tries to do - then it can still impress a little bit. To prevent any misunderstandings its the combination of NATAMI CPU card with the 68060 which makes it so fast. But a 68060 on a normal AMIGA turbo card has no chance against the PPC systems.
| |
Ayodele Stephenson
USA
| |
Posts 83
08 Jun 2011 13:55
| Gunnar von Boehn wrote:
|
Angel of Paradise wrote:
|
Gunnar von Boehn wrote:
|
BTW the current memcopy SysSpeed results of the NATAMI are already a little better than a Pegasos 1 could do!
|
Is this true?
How sure are you that the 68060 can beat a PowerPC in memory performance?
|
200% sure.
Some examples:
Some time ago, people did extensive benchmarks of PPC systems.
Out of my head, the glibc memcopy results for typical AMIGA PPC systems where:
- Pegasos 600 MHz G3 reaches 120 MB/sec
- AMIGA ONE with 800 MHz G3 reaches 150 MB/sec
- AMIGA ONE with 933 MHz G4 reaches 178 MB/sec
- I think EFIKA scores around 100 .
- and SAM score around 180 MB/sec
It might sound unbelievable but the 68060 is able to beat all of those.
It might not sound breathtaking but a 68060 is also able to beat many old PC systems e.g the 68060 can beat a typical 600 MHz AMD K6-2 system.
The NATAMI 68060 CPU card can beat many of the above PPC systems.
|
Yes, I remember all the Red/Blue arguments about memory performance... when ultimately it was all limited by the Artica(s) Northbridge. So the Peg I and Amiga One's were never able to get close to the optimal memory performance they should've. Plus I think people forget how good the 68k series was. It's a technological shame that it wasn't promoted and continued during the early 90's. But... back in 91 it was all about the new "AIM" alliance... and now that's history.
| |
Christian Kummerow
Germany
| |
Posts 314
08 Jun 2011 14:06
| Gunnar von Boehn wrote:
|
Actually we were talking about Fast2Fast memcopy.
What score do you get for Fast2Fast copy?
|
copyb 24,97
copyh 39,14
copyw 46,85 i dont know what h means, but if w means WORD thats still faster
than Fast2Fastw 44,65. And thats only a 604e@233 MHZ with PS/2 Ram. Gunnar von Boehn wrote:
|
But this is not what we talk about here.
What do you get for MEM-COPY ?
|
There must be a reason that Sysspeed dont have only MEM-Copy.
For MOS i have no MEM-Copy results cause "PPC-Memtest" does what it does. Writing/Reading/Compare/Output. If i would have a PEG i would write a quick programm but now i can only compare Screenshoots.
And the WOS Emulation of MOS seems working right i get 74MB/s on MOS Classic and 75MB/s on OS3.9 WOS Classic.
It may possible that memcopy is faster on Natami than on MOS but
Read/Write is not.
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 14:20
| Christian Kummerow wrote:
|
Gunnar von Boehn wrote:
|
Actually we were talking about Fast2Fast memcopy.
What score do you get for Fast2Fast copy?
|
copyb 24,97
copyh 39,14
copyw 46,85
i dont know what h means, but if w means WORD thats still faster
than Fast2Fastw 44,65. And thats only a 604e@233 MHZ with PS/2 Ram.
|
h is the name for 16bit on PPC.
w is the name for 32bit on PPC. This means: You look at the wrong row. :-D
46 for the PPC is actually lower than 56 or 63 for the 68K.
The 32bit Copy on 68k is :
Fast2Fastl which score 56.06 Mb/sec on NATAMI.
This means your 233MHz 604 scores 2x46,85 = 93.7 MB/sec
And the NATAMI 68060 at 120 MHz scores 2x56.06 = 112.1 MB/sec
As you see the NATAMI is 10% faster than the 233 MHz PPC.
And we are not even finished tuning the CPU card. :-D
All clear now?
| |
Christian Kummerow
Germany
| |
Posts 314
08 Jun 2011 14:51
| Gunnar von Boehn wrote:
|
h is the name for 16bit on PPC.
w is the name for 32bit on PPC.
This means your 233MHz 604 scores 2x46,85 = 93.7 MB/sec
And the NATAMI 68060 at 120 MHz scores 2x56.06 = 112.1 MB/sec
As you see the NATAMI is 10% faster than the 233 MHz PPC.
And we are not even finished tuning the CPU card. :-D
All clear now?
|
Ok, but thats only memcopy. And a 604e@233MHZ is not the same like the G3-PPC with 600MHZ and newer Memory.
| |
Gunnar von Boehn
Germany
| | (Moderator)
Posts 5775
08 Jun 2011 15:01
| Christian Kummerow wrote:
|
Gunnar von Boehn wrote:
|
h is the name for 16bit on PPC.
w is the name for 32bit on PPC.
This means your 233MHz 604 scores 2x46,85 = 93.7 MB/sec
And the NATAMI 68060 at 120 MHz scores 2x56.06 = 112.1 MB/sec
As you see the NATAMI is 10% faster than the 233 MHz PPC.
And we are not even finished tuning the CPU card. :-D
All clear now?
|
Ok, but thats only memcopy. |
I don't know what you mean with only memcopy. Add your own PPC scores for Read and write up and compare it with 68k then you see that even not fully tuned the 68060 can beat the 233 MHZ 604 already: Your PPC
Readw 116,36
Writew 75,56 68060:
ReadFastl 113.88
WriteFastl 100.0
Christian Kummerow wrote:
|
And a 604e@233MHZ is not the same like the G3-PPC with 600MHZ and newer Memory.
|
If your above numbers for the 604 are correct than obviously it can very well match the G3.
The "Pegasos I G3@600" does reaches 120 MB/sec memcopy - which is only insignificantly more than your 604.
| |
Angel of Paradise
Germany
| |
Posts 61
08 Jun 2011 15:12
| Marcel Verdaasdonk wrote:
|
Angel of paradise it's real simple by using better Memory
|
Ok thanks!
What is the absolute maximum a 68060 can reach with a normal memcopy?
| |
|
The team will post updates and news here