Sonnet 7200 PCI programming

Originally Posted by nexus

hope everyone has money left over to buy the final version which doesnt need carriers just FYI this project wasnt cheap as in time and money spent researching deving programming etc

I started the project like yours. Abbaddoned it when i counted how much cost parts, multilayer PCB of PCI size routing and the market size... Stil got two MPC7410 and TSI107 for use but no PCB yet... Even got Sodimm sdram slot cause it is quite small.

But good luck! When you draw PCB I am first in line to assemble one with 7410 cpu. I can even order few pcb.

[nexus]

Originally Posted by zomos

I started the project like yours. Abbaddoned it when i counted how much cost parts, multilayer PCB of PCI size routing and the market size... Stil got two MPC7410 and TSI107 for use but no PCB yet... Even got Sodimm sdram slot cause it is quite small.

But good luck! When you draw PCB I am first in line to assemble one with 7410 cpu. I can even order few pcb.

we will be using 7447a and 745X why i said this card is the slow slow ver
PS i have a 7457 ready to test just things just got hard for me as my father just died

Originally Posted by nexus

we will be using 7447a and 745X why i said this card is the slow slow ver
PS i have a 7457 ready to test just things just got hard for me as my father just died

Sorry to hear that... Still first in line to assemble.

[Magic]

Originally Posted by nexus

PS i have a 7457 ready to test just things just got hard for me as my father just died

I am sorry for your loss

[nexus]

Originally Posted by Magic

I am sorry for your loss

yeah gonna slow things down a tad as i have a bunch of crap to do with that atm but this isnt stopping so dont worry

[grelbfarlk]

I don't suppose someone with an A4000d Mediator, the "black version" with 3.3V support can look on the voltage regulator and tell me which one is on it? LMxxx? Or take a picture?

I think this is it right here:

[amigoun]

Htc
lm1085
3.3 311

[grelbfarlk]

Originally Posted by amigoun

Htc
lm1085
3.3 311

Well that's going to be a problem. So basically on the Sonnet you have 5V powering the RAM and 3.3V runs the CPU and Northbridge (not really but the voltages from this are derived from the 3.3V), best guess was that the Sonnet pulled something like 10W total maybe a little more. That LM1085 is rated for 3A, that's not going to cut it for the Rapture G4. An LT1084 would probably be fine or an LT1083 would be more than enough. Sucky. I guess Elbox only intended for the 3.3V to power the Radeon.

*Sigh* now I have to save up for a 3.3V Mediator to do some testing on.

*EDIT* Maybe it's not as bad as all of that. 7410@450MHz 7.43W at full mode, plus a little more drivers. I guess it depends on how much the RAM and other bits and bobs use.

[Hedeon]

Getting the card to work on my A4000T also resulted in a failure. Where on the 3000 at least the card was recognized, but stopped working shortly after starting the PPC, here I get a ramlib error as soon as I peek into the config space of the card. Or when pci.library is started.

More cards come in today and a new carrier in 2 weeks or so to rule out hardware issues. Also hopefully getting someone to mod an A3000 PSU for more power.

[nexus]

well the PSU is here modding has begun just remember that the Rapture is using all 3.3v for cpu ram etc so 3amps is not enough what you can try is lowering the cpu speed see if that can save you some amps and see if you can get something to run in low power mode

[grelbfarlk]

Q3 with Lightmap, Geometric detail on high and texture quality on max, only scores 24 fps.

[Cowcat]

For comparison:

My system with same maxed out settings & bilinerar with mac-mini G4 1.4:

timedemo four.dm_68 : 50. something fps. Not a real difference with trilinear texture filter.

I know, not the fastest Q3 version ever, but is what I got now

Edit: 640x480 display.

[grelbfarlk]

Originally Posted by Cowcat

For comparison:

My system with same maxed out settings & bilinerar with mac-mini G4 1.4:

timedemo four.dm_68 : 50. something fps. Not a real difference with trilinear texture filter.

I know, not the fastest Q3 version ever, but is what I got now

Well not so bad then, your CPU is 3-4x faster.

[Cowcat]

7447A / 1417 Mhz against 450 Mhz.

[grelbfarlk]

Originally Posted by Cowcat

7447A / 1417 Mhz against 450 Mhz.

And the Radeon 9200 is probably faster too, maybe.

[Cowcat]

Better no to be so critical: Those results can be daunting but compared with a "modernized" system are pretty reasonable. And a sucky Q3 version

[Hedeon]

Got success with running the card in my A3000D. :-) It needs to stabilize a few minutes but then it runs without crashing. The crashing on cold boot is actually the 68K in combination with PPC memory (doing code or data). The PPC itself does not crash during this period.

[jack-3d]

Originally Posted by grelbfarlk

And the Radeon 9200 is probably faster too, maybe.

BTW do you see any change how to port Radeon Warp3D drivers to OS3.X?

[grelbfarlk]

Originally Posted by jack-3d

BTW do you see any change how to port Radeon Warp3D drivers to OS3.X?

Ah well that's complicated. I'm not the best person to ask about this, someone was working on it but I think they needed some Elbox help.

Also we have the same problem as we do with Voodoo 4 that they don't support Big Endian modes.

[grelbfarlk]

Hedeon compiled an Altivec version of Dnetc. Results are a lot higher, like a LOT LOT higher.

Updating the results:
RC5-72
4,785,523 keys/sec ---> WinUAE(AOS4.x Intel i5 3570k)
4,438,031 keys/sec ---> NON AltiVec ---> Pegasos2(AOS4.1FEu1 G4@1131/Radeon9000pro)
4,112,400 keys/sec ---> Altivec ---> (Rapture G4@450MHz)
2,193,253 keys/sec ---> WinUAE(AOS4.1FEu1 Corei3-5005U@2.0GHz/IntelHD5500)

OGR-NG
11,616,256 nodes/sec ---> NON AltiVec ---> Pegasos2(AOS4.1FEu1 G4@1131/Radeon9000pro)
8,892,416 nodes/sec ---> Altivec ---> (Rapture G4@450MHz)
8,647,111 nodes/sec ---> Amithlon(AOS3.9 Intel Core2Duo@4.0 Ghz)
5,218,781 nodes/sec ---> Sam440ep(AOS4.x AMCC440@667MHz)
5,051,323 nodes/sec ---> NON AltiVec ---> (Rapture G4@450MHz)

8.Mark:dnetc/Dnetc> dnetc_wos -bench

distributed.net client for AmigaOS Copyright 1997-2016, distributed.net
RC5-72 PowerPC assembly by Malcolm Howell and Didier Levet
Enhancements for 604e CPUs by Roberto Ragusa
RC5-72 Altivec and OGR assembly by Didier Levet
Please visit http://www.distributed.net/ for up-to-date contest information.
Start the client with '-help' for a list of valid command line options.


dnetc v2.9112-521-CTR-16021317 for AmigaOS (OS 3.9, WarpOS 17.8).
Please provide the *entire* version descriptor when submitting bug reports.
The distributed.net bug report pages are at http://bugs.distributed.net/

[Mar 31 00:01:02 UTC] Automatic processor type detection found
                      a PowerPC 7400 (G4) processor.
[Mar 31 00:01:02 UTC] OGR-NG: using core #0 (KOGE 3.1 Scalar).
[Mar 31 00:01:21 UTC] OGR-NG: Benchmark for core #0 (KOGE 3.1 Scalar)
                      0.00:00:16.80 [5,054,464 nodes/sec]
[Mar 31 00:01:21 UTC] OGR-NG: using core #1 (KOGE 3.1 Hybrid).
[Mar 31 00:01:41 UTC] OGR-NG: Benchmark for core #1 (KOGE 3.1 Hybrid)
                      0.00:00:17.04 [8,892,416 nodes/sec]
[Mar 31 00:01:41 UTC] OGR-NG benchmark summary :
                      Default core : #1 (KOGE 3.1 Hybrid) 8,892,416 nodes/sec
                      Fastest core : #1 (KOGE 3.1 Hybrid) 8,892,416 nodes/sec
[Mar 31 00:01:41 UTC] RC5-72: using core #0 (MH 2-pipe).
[Mar 31 00:02:00 UTC] RC5-72: Benchmark for core #0 (MH 2-pipe)
                      0.00:00:16.20 [1,471,616 keys/sec]
[Mar 31 00:02:00 UTC] RC5-72: using core #1 (KKS 2-pipe).
[Mar 31 00:02:18 UTC] RC5-72: Benchmark for core #1 (KKS 2-pipe)
                      0.00:00:16.17 [1,490,976 keys/sec]
[Mar 31 00:02:18 UTC] RC5-72: using core #2 (KKS 604e).
[Mar 31 00:02:36 UTC] RC5-72: Benchmark for core #2 (KKS 604e)
                      0.00:00:16.19 [1,342,177 keys/sec]
[Mar 31 00:02:36 UTC] RC5-72: using core #3 (KKS 7400).
[Mar 31 00:02:54 UTC] RC5-72: Benchmark for core #3 (KKS 7400)
                      0.00:00:16.17 [4,112,400 keys/sec]
[Mar 31 00:02:54 UTC] RC5-72: using core #4 (KKS 7450).
[Mar 31 00:03:14 UTC] RC5-72: Benchmark for core #4 (KKS 7450)
                      0.00:00:16.19 [2,970,725 keys/sec]
[Mar 31 00:03:14 UTC] RC5-72: using core #5 (MH 1-pipe).
[Mar 31 00:03:32 UTC] RC5-72: Benchmark for core #5 (MH 1-pipe)
                      0.00:00:16.21 [1,524,029 keys/sec]
[Mar 31 00:03:32 UTC] RC5-72: using core #6 (MH 1-pipe 604e).
[Mar 31 00:03:51 UTC] RC5-72: Benchmark for core #6 (MH 1-pipe 604e)
                      0.00:00:17.17 [1,478,872 keys/sec]
[Mar 31 00:03:51 UTC] RC5-72 benchmark summary :
                      Default core : #3 (KKS 7400) 4,112,400 keys/sec
                      Fastest core : #3 (KKS 7400) 4,112,400 keys/sec
[Mar 31 00:03:51 UTC] Compare and share your rates in the speeds database at
                      http://www.distributed.net/speed/
                      (benchmark rates are for a single processor core)