The 28MHz clock is in time with the motherboard (i.e. it's a synchronous clock). The 50MHz clock is asynchronous, i.e. the clock signal isn't aligned with the timing of the motherboard. I think that's why it's working out somewhat slower when using the accelerator clock? You're basically missing the odd cycle due to that non-alignment. If you draw the clocks on some graph paper I think it should become apparent what's happening :)
Reminded me of when I had my A1200 with an 030 processer and FPU accelerator around 93. I left it on to render the Enterprise with a 3D package (can't remember which one, but I think it was an Amiga Format cover disk.), I think it had been rendering for around 2 days and was around not far from being competed when I decided to go to the local pub, I came back after a heavy night out and my Mum told me she had switched my Amiga off to save electricity!
Great video =D The reason its not running as fast as you thought with the clock coming from 030 - Don't forget the 030 doesnt run at its full clock speed all the time! When you connect to mobo 28Mhz - that's a constant 28Mhz! When TF1230 slows down to access chip or other things, that means FPU is slowed down too! I suspect the crashing is when it peaks at 60Mhz, that's a bit too much for it over time.
Glad you enjoyed it and thanks for the help on discord. Never occurred to me that the 030 slowing might be the cause of the slower clock but I do think the issue here might be as Stephen suggested and the clock is too fast for the chip. Running AIBB now it thinks the fpu is running at only 22.4 MHz or just a little under half the CPU clock. It's a pitty my scope can't handle frequencies that high or we could take a look at it.
@@CRG The FPU chip select signal come from the AA Gayle chip and will be limited by its maximum decoding speed and it maybe synced to the 14Mhz bus/clock? That may explain why it ran slower on the faster 030 clock. May also explain the .6 reading at 28Mhz as it would get 2x28Mhz cycles and have to wait for 1x14Mhz cycle whereas at 14Mhz it would be 1x1 giving .4. You could also try and cut the GND you added with the twisted pair at the FPU end (just leave it floating) as you may get ground induced noise if current is flowing (ground loop). or add a 10 to 33 Ohm resistor in series with the GND, or you could go to extremes and use a micro coax (WiFI antenna coax from an old notebook PC) Also: I found the best way to test the FPU is AIBB Beach Ball test, it will exit early and not draw the entire ball. (gives impossible results)
28 MHz are generated on the mainboard of the A1200. But because the graphics and the CPU have to access the chip RAM, it is 14 MHz for the CPU (& FPU) and 14 MHz for the graphics (AGA chip set). An external CPU board can access the external Fast RAM synchronously with 50 MHz, but asynchronously access the Chip RAM only with a 14 MHz bus. If you direct 28 MHz into the FPU then these are in sync with the 14 MHz. The restriction that the 50 MHz CPU can only access it asynchronously remains. If you direct 50 MHz to the FPU, the FPU is faster, but no longer synchronous to the 14 MHz bus on which it is attached. The CPU looks even worse. The CPU with 50 MHz can only access the 14 MHz bus asynchronously and can then only reach the 50 Mhz FPU asynchronously from the 14 MHz bus.
Thanks, I have to admit I missed your video and didn't see it until after getting most of this one recorded. It seems to be a popular topic at the minute as a few people have done videos on it. Ultimately I think the separate crystal is the best way to do it and I'll revisit this to undo the clock from the 1230 and use a 40mhz crystal instead.
@@CRG 10Marc has just done a video on the Tsunami 1230 card (New version of the old Phase 5 Blizzard). The Tsunami is like the TF1230 but has space for a onboard FPU socket to be run at the same speed or above the 030 @ 42mhz 👍🏻
Very intetresting results although I would also be very nervous doing that on my Amiga 1200. Fortunately, I have a Memory Expansion card with a 68881 FPU so I'm even more happy with that now, lol. -Mark.
Here is some info to help you: In theory a (e.g. 2.5 watt) laser etcher/engraver (as used for PCB etching) might accurately be able to get that black coating off the 68030 to reveal metal for a heatsink. As an aside that long wire is pulling a relatively heavy current fast and will probably want a (decoupling) capacitor for a current delayed by stray inductance in (e.g. lines for) power, lest you spike/drop the voltage. For the decoupling capacitor, film capacitors better than ceramic, and then electrolytic without additionals will be too slow before they are full with a voltage spike/drop. Maybe shield that long run wire too. At a push you could use a logic-analyzer(or MacGyver one out of an fpga) to see how each end of the wire behaves _(or an oscope kinda like checking an eye pattern)._ What decoupling capacitor do you think you need though? Say 100nF-400nF? Aside from Amiga, it is commonplace on 80s and 90s devices to see decoupling caps soldered atop the VCC and GND pins of IC's. As a thought beyond this Amiga as such, protection diodes (for ECD protection) indicating a maximum current (whatever that is, be it 1mA or whatever) means protection from voltage _(+12V and -5V on inputs for boards on machines but not that amiga, per se)_ as for leakage current, so 10k would protect against all shorts but might render the whole thing unusuable (until taken off again). That is a tangent though. ;D In general, another youtube video fitting an FPU to the Mobo while using a tf1230 is _"Amiga 1200 internal FPU anyone can do this"_ by Chris Edwards Restoration (video 3 March 2022). Just FYI, when not using that tf1230, the 020 on the mobo can do 28MHz stable, and even 33MHz. An MC68882FN40A might take some such jiggery-pokery again though. Upload a Jpeg/Mpeg4 of your FPU close-up to see the writing on it? My comment has no hate in it and I do no harm. I am not appalled or afraid, boasting or envying or complaining... Just saying. Psalms23: Giving thanks and praise to the Lord and peace and love. Also, I'd say Matthew6.
My blizzard a1220 8mb have one on it, so I changed it from the stock 33mhz to a 40mhz and changed the clock cristal to 50mhz. It would only over clock to 44mhz but still interesting to play around with it.
Check the two clock signals with an oscilloscope. You might have ringing or dampening on the longer wire, rise times might be degraded preventing the fpu from deriving a good clock signal. I would personally use a 40MHz separate TTL oscillator just mounted close to the fpu. P. S. : don't connect the ground wire on both sides.
Unfortunately my scope won't go anywhere near those frequencies. A new scope is in my list of things to get but at present I don't have the funds for it.
That is almost certainly a remarked 16MHz part. I bought a few a while back - only some would run at 40MHz reliably. Even though there is only one mask the 16MHz ones may have been marked as such because they failed at higher frequencies. I would see if you can get a genuine freescale marked one or a known genuine one out of an old accelerator board.
@@georgehunter2124 You use it to melt the solder, when its molten you place the Socket on top and then move away. I mean those Sockets got to be installed in a Factory somehow. ANd that might be an Oven or so.
Solder paste may have been an option but having never used it before and seeing other people struggle with it I didn't want to take any chances and risk damaging the 1200.
@@CRG since you're in UK, you could ask "TheCod3r" for example. He seems to do some more extreme Soldering Jobs. Like reballing your BGA thing. Without a Stencil, just Balls from a can.... I think he did some Solder Paste Jobs as well, so he might know how to use it.
I honestly don't like using an Amiga 1200 without an FPU. I used an Apollo 68040 for the last 20 years, and when I upgraded to my TerribleFire 1260 board and used a 68LC060 chip with no FPU, half my software whined and complained (and stopped working) since I had the FPU versions installed! I know I am the exception, not the rule, but I need an FPU for the stuff I do. I will get a full 68RC060 later this month.
For my A1200, my TF1260 was shipped with a low-cost 68LC060 Rev 4 and I later purchased the full 68060 Rev 1 (overclocked to 62.5Mhz). I use to own an A3000 with 25 Mhz 68030/68882.
It's likey impedance matching issues, or the split needs a buffer. I'm not sure how the fpu works but I assume it uses an interrupt to signal calculation is done. If the cpu is running async it takes a few more clock cycles to sync up. Maybe there are bus errors and the data needs to be resent. I think the _BEER signal is bus error, if you put an oscilloscope on that, see if it triggers.
have you by any chance checked what freq you are getting from the Accel board? I have seen chips/cpu's that take a system clock and multiply the signal to run faster internally. This may not be that, but why trust it... check it with an Oscope or freq counter.... that may explain why the main board is running faster... it may only be 14 Mhz.
Watched your video-very interested-but don't those demo software runs include software that utilizes disk transferring not direct memory transfers, and is the fpu saving 3 out of 4 clock cycles doing math arithmetic and is there conflict with 2 types of processing one type being disk to memory and the other processing floating point number arithmetic.Greater speed obtained by using dma transfers!
I wonder what was Commodore's thinking in provisioning for one and how much it would have cost Commodore to factory fit one. Perhaps they were going to but decided against it at some point.
Just a guess but ... Engineers: Of course it should at least have the option for an FPU and we have the board space so why not? Bean counters: If we don't fit that footprint with an empty FPU socket we can shave some tens of cents from the BoM.
I asked this question over on EAB. The thoughts over there are that engineers designed the board and give the sales guys options of different configurations. They could have ordered the machines to be made with an FPU but never did.
Anything spare on the edge connector ? You could use a spare to connect to then pick up for your FPU from the Amiga side. Then you could remove when needed a bit more easily. - Just a thought.
To be honest I'm not happy with how this was left. I've now got a 40mhz crystal and I'm going to wire that up beside the fpu. Ultimately I think that is the better option to run the chip at full speed. I might do a quick follow up video to show the install of the crystal.
i would remove the pin on the socket prior soldering it on the board....and solder the wire without bending the pin :). Isnt there any pin on the board connector for the clock signal? Just to have a shorter wire
I suppose in hindsight I could have removed the pin from the socket after initially trying it with the 14mhz clock, but its done now. I don't believe the accelerators CPU clock is present on the edge connector so the wire had to run to the card itself. I've since further modified this and placed a 40mhz crystal beside the FPU solely for it. It's running happy at that speed.
Just a warning, don’t mess this up like I did. I managed to damage some traces on my Re-Amiga board and had to send it off to a mate if mine for repair 🙄 Also this won’t work with the TF1260 LC or EC 060 processors.
68060 design assumes an integrated FPU, hence off-chip FPU wouldn't work. Intel 80487 "FPU" upgrade is just full 80486DX with a slightly different pin layout.
I know this is not the 'proper' method of doing such a job, but I would have been tempted to just clamp the socket in place on the motherboard with the legs on the solder pads followed by using hot air to melt the solder on the pads.
I paid a mortgage and got an old school 1230 with a slot for 68882 - got that separate and put it on. like you , why , because lol. 030 runs at 55.80 mhz 9.01 mips 1.35 mflops and 8630 dhrystones.
The fpu chip is a 40 mhz part so running the clock at 28mhz gives it plenty of time to do its work. Yet trying to run at 54 or 60 mhz the chip does not have time to do its stuff
While I appreciate your thinking, that's not how it works. The chip is rated at 40mhz by the manufacturer as a speed at which it will run but it has no internal clock generator, it relies on an external clock signal to drive it. Consider a vehicle with a cruising speed of say 40MPH. The vehicle can move slower than that or faster but go to fast and it may become unstable. The same is true (mostly) for a CPU or FPU. It has its designed clock speed, it will run slower than that and might run faster. Go to fast and it will become unstable and usually crash.
Would this work for an ACA1233n? I know indivision say they don't support putting an FPU on the accelerator itself, even though there is a section for it. Does anyone know if an FPU on the motherboard will still keep working just like it does with the TF card? Thanks.
I don't think it will to be honest. I expect the aca will disable the onboard FPU like it does the onboard CPU. I no longer have the ACA card or I would test it for you.
I've seen situations where chips go too far past their rated speed only see every 2nd or 3rd edge.. Eg. some of my 1260s clocked at 125mhz look like they are running at about half that.
@@TerribleFire couldn't he just lift the clock signal pin on the FPU and not connect it to a clock signal at all? I think I've heard somewhere that if you don't feed the FPU a clock signal, it will run at it's own rated clock speed.
@@MallaganVloggs Thats a myth... A simple read of the MC68882 user manual will tell you this is not true. Especially when you consider these chips are tested and rated after manufacture and get sorted by how they perform (binned). How could they know beforehand what speed they would be rated to? It always amazes me how these rumors get started.
Been there, done that. You cannot o.c an FPU, you will have crashes and graphics corruption. Sysinfo is the wrong tool. You need AIBB65 and run all the benchmarks that are typed with yellow fonts. Also you must enable 68020+FPU next to your machine and others you compare on the list. You should do that on every step you do to check the results without the fpu and between 28 and whatever from the accelerator. Jesus is your demo for that things.
@@CRG in AIBB there are cyclebuttons next to your machine and others which selects cpu and/or fpu to test. What it does is, if you don't select fpu then AIBB is clever enough to perform the test with forcing the floating points onto the CPU, enable the FPU and you will see a great speed-up. Also, you could always try the Real3d 1.3 demo picture..... Try Beachball with and without the fpu selected.....
@@brynjarborgersen8131 got it working now and yes there is some difference when the fpu is enabled or disabled. The results are coming in similar to a 25mhz A3000 which makes sense as AIBB sees the fpu clock at 22.4mhz. I think a quick follow up video with a separate 40mhz crystal is the way to go!
Just as - Vincent GR - correctly say , You can not overclock a FPU , they will run slower . Those chips simply doesnt like the higher speed and wont see all the clock cycles and they are also much more likely to get dammaged too , even with just slight overclocking . I have also being there and done that too , including dammaging a FPU so it was really unstable . But what you can do other than reconnect it to the 28 mhz clock is to make a seperate little board with a 40 mhz clock and feed the FPU from there . This will work and also give you both faster and stable FPU .
As I said at the end of the video fitting a separate 40mhz crystal is the next thing I want to try. Ultimately it's probably the best option just right now I don't have a 40mhz crystal in stock. I'll order one this evening and maybe do a quick follow up video.
The 28MHz clock is in time with the motherboard (i.e. it's a synchronous clock). The 50MHz clock is asynchronous, i.e. the clock signal isn't aligned with the timing of the motherboard. I think that's why it's working out somewhat slower when using the accelerator clock? You're basically missing the odd cycle due to that non-alignment. If you draw the clocks on some graph paper I think it should become apparent what's happening :)
Reminded me of when I had my A1200 with an 030 processer and FPU accelerator around 93. I left it on to render the Enterprise with a 3D package (can't remember which one, but I think it was an Amiga Format cover disk.), I think it had been rendering for around 2 days and was around not far from being competed when I decided to go to the local pub, I came back after a heavy night out and my Mum told me she had switched my Amiga off to save electricity!
That's even worse than a games console you left on getting turned off at a point before you can save... ouch
Great video =D The reason its not running as fast as you thought with the clock coming from 030 - Don't forget the 030 doesnt run at its full clock speed all the time! When you connect to mobo 28Mhz - that's a constant 28Mhz! When TF1230 slows down to access chip or other things, that means FPU is slowed down too! I suspect the crashing is when it peaks at 60Mhz, that's a bit too much for it over time.
Glad you enjoyed it and thanks for the help on discord. Never occurred to me that the 030 slowing might be the cause of the slower clock but I do think the issue here might be as Stephen suggested and the clock is too fast for the chip. Running AIBB now it thinks the fpu is running at only 22.4 MHz or just a little under half the CPU clock. It's a pitty my scope can't handle frequencies that high or we could take a look at it.
@@CRG Its definitely due to cpu clock speed varying! That 28mhz is always 28Mhz
@@GadgetUK164 I'm going to fit a separate 40mhz crystal. That should sort it out and allow me to crank the CPU back up to 60.
@@CRG also do the acetone test to check you don't have a remarked 16MHz part, which I'm guessing you do :)
@@CRG The FPU chip select signal come from the AA Gayle chip and will be limited by its maximum decoding speed and it maybe synced to the 14Mhz bus/clock? That may explain why it ran slower on the faster 030 clock. May also explain the .6 reading at 28Mhz as it would get 2x28Mhz cycles and have to wait for 1x14Mhz cycle whereas at 14Mhz it would be 1x1 giving .4.
You could also try and cut the GND you added with the twisted pair at the FPU end (just leave it floating) as you may get ground induced noise if current is flowing (ground loop). or add a 10 to 33 Ohm resistor in series with the GND, or you could go to extremes and use a micro coax (WiFI antenna coax from an old notebook PC)
Also: I found the best way to test the FPU is AIBB Beach Ball test, it will exit early and not draw the entire ball. (gives impossible results)
FPU - Fantastic Parallel Universe
28 MHz are generated on the mainboard of the A1200. But because the graphics and the CPU have to access the chip RAM, it is 14 MHz for the CPU (& FPU) and 14 MHz for the graphics (AGA chip set).
An external CPU board can access the external Fast RAM synchronously with 50 MHz, but asynchronously access the Chip RAM only with a 14 MHz bus.
If you direct 28 MHz into the FPU then these are in sync with the 14 MHz. The restriction that the 50 MHz CPU can only access it asynchronously remains.
If you direct 50 MHz to the FPU, the FPU is faster, but no longer synchronous to the 14 MHz bus on which it is attached. The CPU looks even worse. The CPU with 50 MHz can only access the 14 MHz bus asynchronously and can then only reach the 50 Mhz FPU asynchronously from the 14 MHz bus.
I did the initial video on this. But nice job tapping into the 28. I just use the crystal
Thanks, I have to admit I missed your video and didn't see it until after getting most of this one recorded. It seems to be a popular topic at the minute as a few people have done videos on it.
Ultimately I think the separate crystal is the best way to do it and I'll revisit this to undo the clock from the 1230 and use a 40mhz crystal instead.
@@CRG 10Marc has just done a video on the Tsunami 1230 card (New version of the old Phase 5 Blizzard). The Tsunami is like the TF1230 but has space for a onboard FPU socket to be run at the same speed or above the 030 @ 42mhz 👍🏻
Interesting! I thought about adding FPU, running at 14Mhz, never thought of bending up clock pin. Thanks for another great video :)
The faster results may be due to less wait states needed to deal with the asynchronous clocks
Very intetresting results although I would also be very nervous doing that on my Amiga 1200. Fortunately, I have a Memory Expansion card with a 68881 FPU so I'm even more happy with that now, lol. -Mark.
Great video man! Very interesting to watch! Keep on! 👍💪
Here is some info to help you: In theory a (e.g. 2.5 watt) laser etcher/engraver (as used for PCB etching) might accurately be able to get that black coating off the 68030 to reveal metal for a heatsink. As an aside that long wire is pulling a relatively heavy current fast and will probably want a (decoupling) capacitor for a current delayed by stray inductance in (e.g. lines for) power, lest you spike/drop the voltage. For the decoupling capacitor, film capacitors better than ceramic, and then electrolytic without additionals will be too slow before they are full with a voltage spike/drop. Maybe shield that long run wire too. At a push you could use a logic-analyzer(or MacGyver one out of an fpga) to see how each end of the wire behaves _(or an oscope kinda like checking an eye pattern)._ What decoupling capacitor do you think you need though? Say 100nF-400nF? Aside from Amiga, it is commonplace on 80s and 90s devices to see decoupling caps soldered atop the VCC and GND pins of IC's.
As a thought beyond this Amiga as such, protection diodes (for ECD protection) indicating a maximum current (whatever that is, be it 1mA or whatever) means protection from voltage _(+12V and -5V on inputs for boards on machines but not that amiga, per se)_ as for leakage current, so 10k would protect against all shorts but might render the whole thing unusuable (until taken off again). That is a tangent though.
;D
In general, another youtube video fitting an FPU to the Mobo while using a tf1230 is _"Amiga 1200 internal FPU anyone can do this"_ by Chris Edwards Restoration (video 3 March 2022).
Just FYI, when not using that tf1230, the 020 on the mobo can do 28MHz stable, and even 33MHz. An MC68882FN40A might take some such jiggery-pokery again though. Upload a Jpeg/Mpeg4 of your FPU close-up to see the writing on it?
My comment has no hate in it and I do no harm. I am not appalled or afraid, boasting or envying or complaining... Just saying. Psalms23: Giving thanks and praise to the Lord and peace and love. Also, I'd say Matthew6.
you could have left the pstic part in soldered the pads on the motherboard then applied the socket using hot air
Possibly but I didn't want to risk any mistakes so I just put it on as I know how.
@@CRG Check out some of the Console Repair channels. Things just float into place due to surface tension. Amazing to watch.
Some impressive soldering! Interesting results changing the clock
Excellent Glenn. Sounded like a Colin Furze video that demo!!
Thanks, is that the JCM demo? It is a fun demo, looks great and the sound track is banging.
Fascinating stuff!
Happy hello everyone
My blizzard a1220 8mb have one on it, so I changed it from the stock 33mhz to a 40mhz and changed the clock cristal to 50mhz. It would only over clock to 44mhz but still interesting to play around with it.
Check the two clock signals with an oscilloscope. You might have ringing or dampening on the longer wire, rise times might be degraded preventing the fpu from deriving a good clock signal. I would personally use a 40MHz separate TTL oscillator just mounted close to the fpu.
P. S. : don't connect the ground wire on both sides.
Unfortunately my scope won't go anywhere near those frequencies. A new scope is in my list of things to get but at present I don't have the funds for it.
Great video. I wonder... is it possible to pipe the 28mhz clock straight in to the CPU, as well?
Might be worth getting some good quality Kapton tape instead of relying on the very easily melted (especially under sockets) Blu Tack.
Use a Wifi cable, not a twisted cable for a low freq. hogwash sensor.
Shielded wifi wire that it
That is almost certainly a remarked 16MHz part. I bought a few a while back - only some would run at 40MHz reliably. Even though there is only one mask the 16MHz ones may have been marked as such because they failed at higher frequencies. I would see if you can get a genuine freescale marked one or a known genuine one out of an old accelerator board.
What about Solder Paste or pretinning it, adding flux and hot air from the back to get the Socket on?
I dont think they sockets like a lot of heat. I thought along the same lines
@@georgehunter2124 You use it to melt the solder, when its molten you place the Socket on top and then move away.
I mean those Sockets got to be installed in a Factory somehow. ANd that might be an Oven or so.
Solder paste may have been an option but having never used it before and seeing other people struggle with it I didn't want to take any chances and risk damaging the 1200.
@@CRG since you're in UK, you could ask "TheCod3r" for example. He seems to do some more extreme Soldering Jobs. Like reballing your BGA thing. Without a Stencil, just Balls from a can....
I think he did some Solder Paste Jobs as well, so he might know how to use it.
I honestly don't like using an Amiga 1200 without an FPU. I used an Apollo 68040 for the last 20 years, and when I upgraded to my TerribleFire 1260 board and used a 68LC060 chip with no FPU, half my software whined and complained (and stopped working) since I had the FPU versions installed! I know I am the exception, not the rule, but I need an FPU for the stuff I do.
I will get a full 68RC060 later this month.
For my A1200, my TF1260 was shipped with a low-cost 68LC060 Rev 4 and I later purchased the full 68060 Rev 1 (overclocked to 62.5Mhz). I use to own an A3000 with 25 Mhz 68030/68882.
It's likey impedance matching issues, or the split needs a buffer. I'm not sure how the fpu works but I assume it uses an interrupt to signal calculation is done. If the cpu is running async it takes a few more clock cycles to sync up. Maybe there are bus errors and the data needs to be resent. I think the _BEER signal is bus error, if you put an oscilloscope on that, see if it triggers.
OMFG DROOOOOOL!!!! love it
have you by any chance checked what freq you are getting from the Accel board? I have seen chips/cpu's that take a system clock and multiply the signal to run faster internally. This may not be that, but why trust it... check it with an Oscope or freq counter.... that may explain why the main board is running faster... it may only be 14 Mhz.
I know this video is 7 months old buuut, iirc plcc sockets are to be soldered to the mobo with a heatgun
Watched your video-very interested-but don't those demo software runs include software that utilizes disk transferring not direct memory transfers, and is the fpu saving 3 out of 4 clock cycles doing math arithmetic and is there conflict with 2 types of processing one type being disk to memory and the other processing floating point number arithmetic.Greater speed obtained by using dma transfers!
I wonder what was Commodore's thinking in provisioning for one and how much it would have cost Commodore to factory fit one. Perhaps they were going to but decided against it at some point.
Just a guess but ...
Engineers: Of course it should at least have the option for an FPU and we have the board space so why not?
Bean counters: If we don't fit that footprint with an empty FPU socket we can shave some tens of cents from the BoM.
I asked this question over on EAB. The thoughts over there are that engineers designed the board and give the sales guys options of different configurations. They could have ordered the machines to be made with an FPU but never did.
Worst still, at one time we were to get DSP included in all machines, but so that they could save a few bucks they didn't add it
Anything spare on the edge connector ? You could use a spare to connect to then pick up for your FPU from the Amiga side. Then you could remove when needed a bit more easily. - Just a thought.
To be honest I'm not happy with how this was left. I've now got a 40mhz crystal and I'm going to wire that up beside the fpu. Ultimately I think that is the better option to run the chip at full speed. I might do a quick follow up video to show the install of the crystal.
i would remove the pin on the socket prior soldering it on the board....and solder the wire without bending the pin :). Isnt there any pin on the board connector for the clock signal? Just to have a shorter wire
I suppose in hindsight I could have removed the pin from the socket after initially trying it with the 14mhz clock, but its done now. I don't believe the accelerators CPU clock is present on the edge connector so the wire had to run to the card itself. I've since further modified this and placed a 40mhz crystal beside the FPU solely for it. It's running happy at that speed.
@@CRG mhh so how does it look like now..?
@@CRG ok found the second part :)
@@klofisch honestly its a bit of a mess, it was a rushed 2nd part to be honest and I need to revisit it to tidy the soldering. But at least it works 😆
Just a warning, don’t mess this up like I did. I managed to damage some traces on my Re-Amiga board and had to send it off to a mate if mine for repair 🙄 Also this won’t work with the TF1260 LC or EC 060 processors.
68060 design assumes an integrated FPU, hence off-chip FPU wouldn't work. Intel 80487 "FPU" upgrade is just full 80486DX with a slightly different pin layout.
I know this is not the 'proper' method of doing such a job, but I would have been tempted to just clamp the socket in place on the motherboard with the legs on the solder pads followed by using hot air to melt the solder on the pads.
I paid a mortgage and got an old school 1230 with a slot for 68882 - got that separate and put it on. like you , why , because lol. 030 runs at 55.80 mhz 9.01 mips 1.35 mflops and 8630 dhrystones.
The fpu chip is a 40 mhz part so running the clock at 28mhz gives it plenty of time to do its work. Yet trying to run at 54 or 60 mhz the chip does not have time to do its stuff
While I appreciate your thinking, that's not how it works. The chip is rated at 40mhz by the manufacturer as a speed at which it will run but it has no internal clock generator, it relies on an external clock signal to drive it.
Consider a vehicle with a cruising speed of say 40MPH. The vehicle can move slower than that or faster but go to fast and it may become unstable.
The same is true (mostly) for a CPU or FPU. It has its designed clock speed, it will run slower than that and might run faster. Go to fast and it will become unstable and usually crash.
Hi !!
Hi freq signals must be connected with shielded cables !! The same as RF cables !!
Would this work for an ACA1233n? I know indivision say they don't support putting an FPU on the accelerator itself, even though there is a section for it. Does anyone know if an FPU on the motherboard will still keep working just like it does with the TF card? Thanks.
I don't think it will to be honest. I expect the aca will disable the onboard FPU like it does the onboard CPU.
I no longer have the ACA card or I would test it for you.
I've seen situations where chips go too far past their rated speed only see every 2nd or 3rd edge.. Eg. some of my 1260s clocked at 125mhz look like they are running at about half that.
I think ultimately that's the problem. I've ordered a 40mhz crystal and I'll install that beside the fpu.
@@CRG make sure there is a decoupling cap near the xtal
@@TerribleFire couldn't he just lift the clock signal pin on the FPU and not connect it to a clock signal at all? I think I've heard somewhere that if you don't feed the FPU a clock signal, it will run at it's own rated clock speed.
@@MallaganVloggs Thats a myth... A simple read of the MC68882 user manual will tell you this is not true. Especially when you consider these chips are tested and rated after manufacture and get sorted by how they perform (binned). How could they know beforehand what speed they would be rated to? It always amazes me how these rumors get started.
@@TerribleFire alright, thanks for clearing that up
pffft... nothing missing :)
Nothing useful anyway 😆
i tink its the tuppa shaft its always the tuppa shaft.
You could just get the desired clock frequency using a counter chip can't you?
Ultimately I think what I'll do it place a separate 40mhz crystal beside the fpu and wire that in directly.
Odd… my 50mhz 1230 has an fpu onboard…
What would you do with that FPU.
Math lol
Been there, done that.
You cannot o.c an FPU, you will have crashes and graphics corruption.
Sysinfo is the wrong tool.
You need AIBB65 and run all the benchmarks that are typed with yellow fonts. Also you must enable 68020+FPU next to your machine and others you compare on the list.
You should do that on every step you do to check the results without the fpu and between 28 and whatever from the accelerator.
Jesus is your demo for that things.
What do you mean by enable 68020+fpu?
I'll try AIBB see how the results pan out. Sysinfo isn't the best tool but I just had it to hand.
@@CRG in AIBB there are cyclebuttons next to your machine and others which selects cpu and/or fpu to test. What it does is, if you don't select fpu then AIBB is clever enough to perform the test with forcing the floating points onto the CPU, enable the FPU and you will see a great speed-up.
Also, you could always try the Real3d 1.3 demo picture.....
Try Beachball with and without the fpu selected.....
@@brynjarborgersen8131 got it working now and yes there is some difference when the fpu is enabled or disabled. The results are coming in similar to a 25mhz A3000 which makes sense as AIBB sees the fpu clock at 22.4mhz.
I think a quick follow up video with a separate 40mhz crystal is the way to go!
@@CRG Sorry for the late reply.
What Brynjar said.
Great that you made it working.
Just as - Vincent GR - correctly say , You can not overclock a FPU , they will run slower .
Those chips simply doesnt like the higher speed and wont see all the clock cycles
and they are also much more likely to get dammaged too , even with just slight overclocking .
I have also being there and done that too , including dammaging a FPU so it was really unstable .
But what you can do other than reconnect it to the 28 mhz clock is to make a seperate
little board with a 40 mhz clock and feed the FPU from there .
This will work and also give you both faster and stable FPU .
As I said at the end of the video fitting a separate 40mhz crystal is the next thing I want to try. Ultimately it's probably the best option just right now I don't have a 40mhz crystal in stock. I'll order one this evening and maybe do a quick follow up video.
Edit: if the fpu gets damaged it's not the end of the world, they were only cheap. I think the 2 cost maybe £5