Wanted: Blizzard PPC in lowest config: 040LC@25MHz PPC@160MHz no SCSI

[stachu100]

Hi,
I'm looking for card in above mentioned config.
Reason:
I would like to test possibility to change the PPC processor to 300MHz version.
Europe vendors preferred.
Regards, Stan

[stachu100]

there's no need to test,it will work ( both processors ).
 
did you fix the PPC on the Cyberstorm PPC?. if so what was the problem?

I know.
My target was to upgrade PPC procesor (un-solder old one and solder new) just to check my soldering skills
Now it's done:
 
http://picasaweb.google.pl/stacho100/BlizzardPPC_06066MHzPPC300MHz#

http://www.youtube.com/watch?v=ewuwDs6QNgM

Initially this card was 040@25MHz/PPC@160MHz with bus speed 50MHz for PPC processor (lowest config posible).

Now it's the best config possible:
060@66MHz/PPC@300MHz with 60MHz PPC bus speed

060 CPU and 060 oscillator are socketed just in case (some BVision cards refuse to work on 66MHz).

I tried this PPC processor to work with 66MHz bus speed, but cards hung-up after 2 minutes; so I was forced to set 60MHz oscillator for PPC with x5 clock multiplier.

Just missing the heatsink+clips+5Vfan for PPC processor; I bought this card without it.
If you have it as spare I'm ready to buy

[stachu100]

@delshay,

Everything you wrote is true.
What I wanted to do is to make an upgrade/repair for polish Amiga community; that is why I've ordered bigger batch at one shot.
And hopefully for me I wasn't cheated.

With concerns timing and memory:

Well, I've checked this particular card with 60ns standard SIMM.
I have 64MB 50ns SIMM as well (from some old SCSI controller), but to be honest my target was to do this upgrade/repair for people with standard SIMM's available on the market even scarifying performance.

Maybe with better (faster) SIMM's the card might work ok with both 66MHz oscillators, or even 70MHz; who know?

My intention was not to create "super card", but to resurrect broken cards and for PPC@160Mhz users to give them a chance for upgrade at reasonable cost.

by the way:
Your project to install 740 500MHz CPU on the Blizzard PPC is really impressive!
In this case my card is not the best config; your is.

[stachu100]

after buying many faulty Blizzard cards iv found not all faults was on the CPU. in most cases the CPU turn out not to be faulty,i only found one card that had a faulty CPU and in another the CPU had a cracked joint.
 
but i have found faults in places where you think thats impossible.
 
i have rebuilt 2 cards from scratch one i had to add the complete SCSI
 
some companys will not touch old PCB because of the age of the card,and doing BGA re-work if something go`s wrong you must know how to repair it.
 
ACUBE was doing CPU upgrade but if the CPU turns out not to be faulty and the fault was somewhere else on the card how will the card be returned (with new or old CPU) with the card still not working.

 
Thanks a lot for this info!
People from Poland are asking me to repair their broken cards, but my answer was always that repair is not possible; what I can offer is to replace CPU and if this will not help just to look around to find unsure solder joints, broken cooper tracks, defective trough-holes in PCB or even faulty eeprom.

I agree, repair of such cards is like lottery: prize is high, but probability to win is significantly lower than 100%
 
We are not David Copperfield
 
And small update:
I've managed to run PPC at 330MHz (5,5x60MHz):
 
http://www.youtube.com/watch?v=SloiWlPkG18
 
I'm not going to test it at 360MHz, however as card was stable for 1,5h under AOS3.1/4.0, so I can say that it's possible.
And I'm using standard "light" A1200 power supply.

[stachu100]

Well, you are probably right.
But 60MHz is the fastest PPC bus settings with standard RAM SIMM's.
I prefer stable work than additional 10% of power and unsure stability.

[stachu100]

Anybody ever play with this kind of stuff on the CSPPCs? Same idea?

I'm going to try to put 604e 300MHz into Cyberstorm and hope to have 330MHz speed running stable with standard cooling.
Proper processors are one way to me already
I will keep you posted about Cyberstorm upgrade progress.
 
I have no intention to experiment with other processors type as it might be unstable/impossible/kill the processor/kill CSPPC board/kill both.
 
I don't care to kill the processor, but I don't want to sacrifice my CSPPC for sure! :afro:

[stachu100]

Hi All,
I've managed to upgrade another two boards :-)
To eliminate BGA soldering problem to zero I decided to change PPC processor with help of professional service center.
My tools are ok, but why not use better if it's possible?
 
http://picasaweb.google.com/lh/photo/3wtPiX9T0Y6F2oVVhTTnUg?feat=directlink
 
http://picasaweb.google.com/lh/photo/YzF3hOWdUhWk_PIbBItKmQ?feat=directlink
 
http://picasaweb.google.com/lh/photo/derHswCVzK_HGLd5Da5k7w?feat=directlink
 
http://picasaweb.google.com/lh/photo/xKURovEVTWrNhYvGYmfCZg?feat=directlink
 
And another 2 movies:
 
http://www.youtube.com/watch?v=UJw64ZTzGY8
 
http://www.youtube.com/watch?v=BvCUaePyvKo
 
Cards after upgrade:
 
http://picasaweb.google.com/stacho100/BlizzardsPPCAfterUpgrade#
 
I used MC68060RC60 with mask 71E41J (rev 6) working natively at 60MHz taken-out from some industrial CPU boards; on the center of CPU you can see mark from heatsink glued onto 060 (I've took the heatsinks out of course).
 
As it was expected; Blizzard PPC with SCSI version cannot work with 060 clock higher than 50MHz :-(
I was trying even to erase Serial Number  with flashing with "Force" option on one board to see if this will disable the onboard SCSI controller, but unfortunately this is not the case...
 
So bad news for Blizzard PPC with SCSI users (603e+): your card cannot work with 060 clock higher than 50MHz...
 
If someone is really not using SCSI in the desktop case maybe the best option for him is to remove onboard SCSI.
By doing so it's possible to increase 060 clock to 66MHz, and having very good RAM even higher frequencies are possible :-).

[stachu100]

Well, the cards which I have just refuse to work even with 60MHz oscillator and different memory config with even no memory at all (checked several times).
Always get SetPatch error...
Your post is very interesting, I will dig more about it; thank you!

[stachu100]

Yes, 66 Mhz shoud work with SCSI also, but results may differ:
http://xoomer.virgilio.it/giorsign/overclock.html

Thank you very much!
Yes, SCSI is working even with 72MHz
http://www.youtube.com/watch?v=wdOESYNMaGE
 
Just quick playing with SCSI settings is enough to have it working.

[stachu100]

With full memory settings (no read/write waitstates and no precharge) I had to remove the 2nd memory SIMM to make my miggy boot to WB, this was with 66Mhz.

Maybe SIMM was just on the edge of spec ant thus refused to work..
 
I did some performance tests of PPC@366MHz.
It wasn't my intention to do so high overclocking, but I've got the card with one oscillator for PPC and 060...
And this card is don't care about soldered separate oscillaotr for PPC; always clock for PPC part is taken from 060 oscillator.
 
@delshay,
Please find the performance results you asked:
 
http://picasaweb.google.com/stacho100/BlizzardPPC06066MHzPPC366MHz#
 
Memory settings: "60ns"
SIMM USED: 64MB 50ns in single module.
 
Card was working 100% stable with these settings with desktop case without top; card was moderate hot with standard cooling during all tests (I could keep figers on both CPU without burning them).

[stachu100]

looking at the FFTdemo test the first two results is very very poor it does not even beat a CyberStorm 150Mhz.
 
 
looking at my test it beats a CyberStorm PPC 150Mhz ( even with 256Mb ram) but no one with a CyberStorm 233Mhz wiil do the test (FFTdemo) perhaps their think CyberStorm 233Mhz will be outclassed.
 
the result for top end CPU speed is impressive but not the performance.
 
it looks like i have prepare my other Blizzard and upload another test result. it's going to be the Blizzard with a BGA socket with it's Patent Pending cooler,and im sure it's going to take the results even lower.

 
Well, first of all I'm not software geek.
 
Second when I read test guidelines about this test that "lower value are better."
See the statement:
http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=29964&forum=25&start=0&viewmode=flat&order=0
 
And what we have:
 
-------------------------

PPC 604 150 MHZ OS4

Speed test for FFT + iFFT: (float)
time needed 19449ms for 413696 samples, => .241165652871131x speed
@44100Hz/stereo
------------ (5840ms at 500 MHZ)
Speed test for FFT + iFFT: (integer)
time needed 16485ms for 413696 samples, => .28452718257904x speed
@44100Hz/stereo
------------ (4950ms at 500 MHZ)
Speed test for FFT + iFFT: (integer handopmized 68K ASM)
time needed 7108ms for 413696 samples, => .659880518913269x speed
@44100Hz/stereo
------------ (2134ms at 500 MHZ)

----------------------

Blizzard PPC 321Mhz 80.333Mhz bus. OS4.0

Speed test for FFT + iFFT: (float)
time needed 17928ms for 413696 samples, => .261626005172729x speed @44100Hz/stereo
(11566ms at 500MHZ)
Speed test for FFT + iFFT: (integer)
time needed 15377ms for 413696 samples, => .30502900481224x speed @44100Hz/stereo
(9920 ms at 500MHZ)
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 12774ms for 413696 samples, => .367185741662979x speed @44100Hz/stereo
(8221 ms at 500MHZ)
-------------------------------------------------------------

Now look on mine:
 
http://picasaweb.google.com/lh/photo/Vjm139UZgaXferYr8HgBWg?feat=directlink
 
In my opinion .1XXX is lower than .2XXXXX
But I might be wrong...
 
I also compared performance this setup with Efika with one of Polish Amiga Community and he got on Efika:
 
Integral of df(x)/f(x)

98.67

PI from Taylor Series expansion of atan(1.0)

62.92

Integral of sin(x) from 0.0 to PI/3.0 (Trapazoidal)

156.9

Integral of cos(x) from 0.0 to PI/3.0 (Trapazoidal)

111.75

Integral of tan(x) from 0.0 to PI/3.0 (Trapazoidal)

124.48

Integral of sos(x)*cos(x) from 0.0 to PI/4 (Trap.)

193.93

Int. 0 to sa of 1/(x+1), x/(x*x+1) & x*x/(x*x*x+1)

33.86

Int. sin(x)*cos(x)*cos(x) from 0 to PI/3 (Trap.)

184.97

Iteration

128000000

NullTime (usec)

0.0

MFLOPS(1)

78.17

MFLOPS(2)

69.56

MFLOPS(3)

113.70

MFLOPS(4)

164.1

 
http://www.apc74.ppa.pl/PPA/Efika_vs_reszta_swiata.html
 
(in polish, sorry)
 
 
Now look on mine:
http://picasaweb.google.com/lh/photo/h-oja3ik-r36scwtVGrO1g?feat=directlink
 
Doesn't look bad in my opinion.
 
@Delshay,
I really don't want to compete who is the best; I'm just doing my job as best I can.
 
If you feel that I hurt you by any way; I'm sorry.
 
Regards,
Stan

[stachu100]

I'm using AOS3.1, kickstart 3.0 softkicked to 3.1 A1200 1D4 motherboard.
It's pre-configured AmigaSYS 4 AGA.
Blizzard PPC clocked from one 66Mhz oscillator.
Memory set to "60ns" in Blizzard boot menu.
Single 64MB 50ns RAM module.
PPC@366MHz  with bus clock 66MHz.
To be honest the PPC clock is 66MHzx5.5=363MHz, but these 3MHz are not so important...
 
Just one remark:
which is the benchmark result?
Is it this:
time needed 16023ms for 413696 samples, => .292731136083602x speed @44100Hz/stereo
 
or is it this:
time needed 16023ms for 413696 samples, => .292731136083602x speed @44100Hz/stereo
 
After checking the sentence of test result it seems that first is the test result...
As said I'm not software geek...
 
If so, your card is better as running with much higher bus speed 80MHz comparing to 66MHz in my case.
I know that in these conditions you have to use really good and selected RAM; otherwise 80MHz bus clock is just too much for popular RAM modules...

[stachu100]

here is the latest test result which is already in this thread. Blizzard PPC OS4.0
 
Speed test for FFT + iFFT: (float)
time needed 16023ms for 413696 samples, => .292731136083602x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 13450ms for 413696 samples, => .3487309217453x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 10865ms for 413696 samples, => .431700944900512x speed @44100Hz/stereo
 
YOU MUST STATE WHICH OS YOU ARE USING WHEN DOING THE TEST.

@Delshay,
Could you please present the hardware/software config on which you've got this excellent result?
Pictures (or movies) would be the best.
As you know very well I'm using freely available pictures/movies to show Amiga community what and how I'm doing.
 
I'm very curious about RAM used in this experiment.
And is there any other hardware hacks on the Blizzard PPC board (except replacing PPC of course)?
Did you decided to replace original Blizzard PPC logic chips by better and faster versions?
I know that's possible, but as I said before my intention is not to create "super board" but to repair/upgrade and left as much from orginal phase5 design as possible.
 
Thanks in advance!

[stachu100]

Blizzard PPC with BGA socket & patent cooler now tops the time sheet on it's first outing,and now sets the pace.(OS4).
 
Speed test for FFT + iFFT: (float)
time needed 15625ms for 413696 samples, => .300187587738037x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 13146ms for 413696 samples, => .356795281171798x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 10613ms for 413696 samples, => .44195145368576x speed @44100Hz/stereo

 
@Delshay,
Very nice results indeed!
 
Is that mean no other hardware modification than PPC processor socket and PPC oscillator socket?
 
Can we have pictures of your hardware?
What about RAM used?