Fastest running CyberStorm Mk III?

[delshay]

So you mean you want to overclock it to the extreme

Suggestions for stretching your overclocking chances:
1-Make a volt mod, the 68060 can tolerate an absolute maximum 4.5 volts as supply
2-Put heatsinks and coolers working at high speed, not only on the processor, but also to the accompanying glue chipset of your CyberStorm. The more extra chips you cool the more likely the CyberStorm will tolerate overclocking.
3-Get a stable PSU, a PC one with at least 2 times the watts than the power consumed by your entire system, to ensure voltage stability at all times.
4-Overclock bit by bit. This means, dont go for 100mhz at one shot. Try 80, 85, 90, 95 and then 100mhz
5-When overclocking a 68060 the first component that will fail will be the built in FPU. So get a proper fpu test/benchmark and use it while you clock up your CyberStorm to ensure you are getting a stable system.
6-Dont forget to have your setup in a proper ventilated case, otherwise cooling will be useless.
7-If you got the money and the will, you can also try some water cooling setup
8-If you have plenty of money, use phase change cooling for the best results. If not you may also hack a beer mini-fridge for your CyberStorm. When using this type of cooling, remember to protect the cpu/chips by putting a heater near the cpu/chip to avoid condensation.

Bottomline: Overclocking to the extremes is half science and half art. Your mileage will allways vary.

Good luck!

you forgat a couple of things

it's not all about PSU watts,it's also the PSU connections. as you lose voltage if all of the GND and 5v wires are not connected and should be kept short as possible to the PCB.

anyway here with EXTREME speeds the logic chips need voltage near their operating voltage ie 5v if it jrops to say 4.92 4.93 it starts to lock-up to be safe it must read 4.98 and above. the voltage reading must be take from the PCB ie the GND probe must not be connected to the frame but on the PCB itself along with positive probe to get a TRUE reading of the voltage on the PCB.

SIMMS

this also plays a important factor and most simms will fail even if it reads 50ns.

someone here on amiga.org said whats marked on the outside does not mean what on the inside,well he was right (he knows who he is).

i have here SIMM which is marked 60ns,but it is *not true*,it's infact 43ns which allows me to overclock pass 80MHz.

this may or may not effect all PHASE5 cards but i can say for the first time Blizzard PPC cards can operate at 80Mhz+.

68060 on **my** Blizzard PPC card has reached 80Mhz and passed 24Hr test and i don't thing this is the limit and im almost sure it will go faster as i have the correct simm for this kind of speed.

there are other things thats also stopping cards operating at EXTREME speeds which im not going into.

[delshay]

that link shows how to build a socket,but i dont like that idea.

the problem is when the CPU is going to be removed it can put a lot of stress on the outer pins and can or may removed the pads from the PCB.

a proper socket will add strength to all the pins and work together but if you are going to fit the latest mask why bother with a socket.

from what iv seen and my point of veiw 060 sockets is a bad idea and cause more problems,so any of my future projects will not include 060 sockets.

[delshay]

Unfortunately, there's no way to solder an m68060 directly since the circuit board isn't through-pin - the socket is surface mounted. If you read the repair link, you'll see that they recommend building up the socket because there's no way to get underneath it.

If you think soldering the chip directly is a better idea, though, why would removal of the m68060 from the socket be a reason to support that? If the chip is soldered, then you can't swap it for any reason. If you have a socket, you just have to be very careful if you do swap it... but I don't have any reason whatsoever to ever change it since it's the latest / last mask.

this is why the CPU should be solder in direct by a expert,to stop users changing the CPU via a socket as it's surface mounted. if you keep changing CPU sooner or later you will damage the pads on the PCB.

if a socket must be mounted removing the bottom part of the pin (thin part) and solder a proper complete socket just like how the old socket was removed. but i don't recommend a socket for this PCB as to many changes of CPU may damage the pads.

that type of PCB is better off with the latest mask soldered in direct.

[delshay]

when the pads go though a heat cycle (ie solder/resolder) it's not the same strength/bonding to the PCB.

when a new CPU is inserted force will be put on all the pins left,right,up,down and can damage the pads as there may be movement not on the pin(s) but the suface mounted pads itself in-effect tearing it from the PCB,this is why the *plastic housing* is inportant as all pins will work together.

so it's not just when you removing a CPU it's also when you insert it and it applies to all the pins not just the outer.

[delshay]

Unfortunately, there's no way to solder an m68060 directly since the circuit board isn't through-pin - the socket is surface mounted. If you read the repair link, you'll see that they recommend building up the socket because there's no way to get underneath it.

If you think soldering the chip directly is a better idea, though, why would removal of the m68060 from the socket be a reason to support that? If the chip is soldered, then you can't swap it for any reason. If you have a socket, you just have to be very careful if you do swap it... but I don't have any reason whatsoever to ever change it since it's the latest / last mask.

there are other reasons why a socket should not be used which im not going into.but why do need a socket if you have the latest mask,what would be the reason to swap a 68060. how offten do their go faulty.

from my point of veiw 060 socket should be left for TESTING purpose only.

[delshay]

i do have a improvment for the socket but i don't recommend a socket as said before.

first check that the plastic housing is not too wide,try this on a non working 68060/68040 add strip of pins socket on all of the 68060/040 pins,if all ok then try below.  

when each row of pins (with plastic housing)as you are soldering along add a small amount of glue in between each row of  the plastic casing,this will add strength to the hole socket.

the glue should be added first before soldering in the pins.

care must be taken that glue must not come into contact with the pins above and below. if glue does show it can be cleaned before it drys,but here again care must be taken as glue can go down the pin hole which can block the path for the CPU. very strong glue must be used to hold the socket together.

if glue does show but is NOT IN CONTACT with the pins it can be left as it is to dry,as it can be clean after or just leave it as it is. (sometimes it's best to leave it alone if glue shows above the socket).

very little glue must also be used in between the row of plastic housing,if too much glue is used when it comes to doing the last few rows of pins you may find it not in alignment with the pads.

SO KEEP A EYE ALSO ON ALIGNMENT WITH THE PADS AS YOU GO ALONG.
 
alignment can be done by adding two two pins at each end (in other words four pin socket inserted at each end) this will hold the single row of pin in place. after it has dryed do the same to the next row and so on until socket is complete. the glue must be completely set in-order to start the next row of pins and don't forget to solder the pins in as you go along.

NOTE: to complete the socket it may take one to two days (depending on glue used).

all above has never being tested on surface mounted PCB

with all the advised iv given in this thread im going to say it for the last time. socket is not recommened for that PCB.

***ALL MODIFICATION AT YOUR OWN RISK***

[delshay]

im not going to give any advice on overclocking,but here is a workbench screenshot of my overclocked Blizzard PPC and i don't have any problems whatsoever (this is the second fastest Blizzard card in the world)

http://www.amiga.org/gallery/index.php?n=3039

i also own the fastest

[delshay]

The right way to do the overclocking with the CSMK3 is to change the jumper and add a 2nd oscillator. This overclocks the 68060 only. This should be safe enough as this should be the way Phase5 would have supported faster 68060's. 75MHz is very fast considering the memory speeds up too and it was already faster than most other 68060 accelerators. Plus we have ultra wide SCSI. I wouldn't trade for an Apollo 68060@90MHz with half speed memory and broken SCSI. If I want a faster 68060, I'll wait for the Natami with 68060.

CYBERSTORM VS BLIZZARD


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)
-------------------------------------------------------------


***  any photo/benchmark of your apollo @90Mhz? ***

[delshay]

Blizzard PPC card but may apply to Cyberstorm PPC/Mk III card

**depending on settings** anything over 64Mhz can damage the card. i discovered this many years ago.

the problem has been fixed on my cards which also has a hack which allows me to have a FPU on the A1200 motherboard,without me having to remove it. so when the Blizzard PPC card is disable by holding down the 2 key the 020/FPU on the A1200 motherboard comes active

anyway cards overclocked over 64Mhz you will know if you have a problem,as you will get many lock-ups,BUT IT CAN GO WITHOUT WARNING.

one of the fault is that memory and PCI will stop working,but the 68k processor may continue to work (depending on type of card). SCSI may also stop working (i did not check this part)

so care must be taken when overclocking over 64Mhz bus.

[delshay]

so it`s back to benchmark.

[delshay]

I got 20fps most of the time. It flashed a higher or lower fps from time to time but it wasn't displayed long enough to read it. I had AWeb & Genesis going in the background and my memory was probably pretty well fragmented. I decompressed to ram and did a lot of compiles and disassembles first. The speed seamed fast and responsive but the gfx were messed up. The foreground objects were multi-colored. The background was fine.

err what is the speed of your processor(s) and what card do you have this would help if we are doing benchmark.

do you own a apollo @90Mhz?

[delshay]

i did the test with sound 8 FPS most of the time,but when everthing went mutunt it jumped to around 20 - 24 FPS..no problems with GFX.

[delshay]

Nope. Just a CSMK3 with 68060@75MHz, 50ns SIMMs running with the fastest memory settings of the CS, Mediator 3000T/4000T with Voodoo4 and AmigaOS 3.9 with updates in a 3000T.

WOW WOW WOW thats fast.

lets see if i can get anywhere near it.

the last test was a trick (8 FPS) it's real but it was done on my Blizzard PPC TEST CARD which has a 200Mhz PPC processor overclocked to 240Mhz with 060 clocked @50Mhz with the slowest memory setting.

will upload test to this thread soon from one of my faster cards.

[delshay]

i will upload test on monday. note my Bvision is overclocked from the standard 83Mhz to 100Mhz i don't know how much of a effect (if any) it will have on the test result.

[delshay]

i wiil try on my card when i have time,as at the moment i have my test card plug-in testing out new modifcations,which may make my card faster.

i can also say the latest test shows the Blizzard PPC card will not go 100Mhz (to be comfirmed)