Trying a 68010 in my 14MHz Amiga 500
Вставка
- Опубліковано 2 чер 2024
- Remember the new A500 14MHz accelerator I showed a few weeks back? Well I thought it would be fun to try a simple upgrade by swapping out the 68000 PLCC CPU for a 68010 PLCC CPU. This chip was ordered on eBay from a seller based in China and it looks as if it might be refinished. We have two questions that need answering. Firstly is it a fake chip? If not is upgrading to a 68010 pointless?
- Наука та технологія
A big giveaway that the chip has been remarked is the manufacturing date, Motorola were long out of the CPU making business in 2010 (they spun it off as Freescale), so no way was that chip made in the 25th week of 2010 by Motorola. You can see the genuine chip was made in 1996. But Chinese sellers do that to make it look new or faster than the original chip (which may be the case here, this could be originally an 8MHz chip), even going as far as laser etching as they realize buyers are getting smarter and can check.
I can't believe I didn't pick up on this. To busy looking at the finish of the chip to actually read the text on it. Well spotted though.
Few years ago I got 68882 FPU from china, I think it was 40MHz, it could barely run at 25MHz. It had LASER etching on it. Later on, I got the same chip from AmigaKit, and it runs fine at 50MHz.
Sometimes it's better to spend more money and get it from reliable source than try your luck from china. I still have somewhere in my chip pile DIP version of 68010, and yes, it did not help much in speed.
The mottled finish can be seen on many chips and the slanted text too. But you are correct to be sceptical. Good info video.
I remember swapping out the CPU in my Amiga 1000 with a 68010, the mythical 10% performance boost never materialised, not even in pure CPU based polygon games. The incompatibility problems however were not a myth.
I did this as well back in the day. (exactly the same. 10mhz-rated 68010 into my unaccelerated 7.16Mhz A1000. Didn't see any noticible improvement at all. swapped it back. :}
Great video, love how the Amiga still has such a good community and new kit coming out all the time. I love dusting off the old A1200 and playing those old games. It run a lot better now than it did 30 year ago, with the newer expansion cards and WHDLoad on a CF hard drive.
Yep its great to see all these new expansions coming out for our classic computers.
Not too pointless, it's just good fun tinkering around with Amiga stuff.
Or with computers, or digital electronics, in general? I personally never had an Amiga but tinker with a few dozens other machines using 68K, 8086, Z80, 6809, 6502, and even TTL CPUs. A couple of them designed by myself in the 1980s. Totally "pointless" in a way (just like gaming, reading books, talking with friends, or playing sports or musical instruments).
Just stopped by to say I recently found your channel and I love your videos!
Very kind of you to say, glad to hear you're enjoying the channel.
Speaking of CF cards, I went the ACA500plus route. The difference a mass-storage device makes to an Amiga 500 is incredible! It doesn't feel like a pokey little machine any more :)
It's great to see all these modern upgrades for the old systems. Only thing I don't like about the ACA500 is that it sits external. Wouldn't suit me since I rarely have these machines in a permanent setup. That's not to take away from it though, it's a great card.
@@CRG i printed a case for my ACA500 plus that is in the shape of the GVP HD8+ hdd add on. Looks really good. And hides the card nicely.
Yes it was thing only selling without HD ,and yes it works,with floppy, ,you can boot PC from floppy, imaging PC sales would have went if PC's where sold as only needing a floppy to work?
all amiga upgrade vids are good, even if it doesn't go right at the time. thanks for the video
My 3rd year project at college 30 years ago was to design a scsi ram disk using a 68010 as the main controller. It's a lovely chip to design with. The benefit there was that the async dtack could be routed straight from the ack from SCSI to terminate SCSI transfers as memory cycles.
Another problem with a 68010 is that some games won't run anymore. WHDLoad is fine, but some disk games won't work (SWIV for example).
That's true and I probably should have mentioned it in the video. Thanks for reminding me.
Yeah, some games do stupid tricks with the stack during exceptions, but stack frames on the 68010 are sized differently than the 68000. If a game won't work on an '020, it will likely also have problems on an '010.
The one thing people have to understand, swapping out the chips for a faster one of the same class will help overall. However, unless you use the additional ops codes that are in the 68010 vs. that of original 68000, it's not going to make a hill of beans difference in speed. I know this because I was a programmer back in day that worked on the MC68k and MC88k series CPUs and MMUs.
The most useful aspect of the 68010 is with whdload you can exit back to workbench (normally ESC or F10) without the .slave file having to support quit to wb. This is a 68010 feature is called the VBR, vector base register and it allows a program to have its own isolated environment. The only other feature is a slightly faster loop mode for certain types of machine code loops. The HC' 68000 chip uses CMOS technology and runs much cooler, CMOS only draws power when it is switching and can be over clocked more. The 010 is NMOS? and consumes power all the time thus running hotter and being less overclockable.
Another great thing about the 68010 is that the VBR (vector base register) is movable - in the 68000, it's fixed at address $0; in the 68010 it's set in a register (but defaults to $0). On the Amiga, location $0 is chip RAM so it's contended and not as fast as fast RAM. All interrupts hit the VBR to find the location of the ISR (interrupt service routine), so that includes vertical blank, hardware events, timers, etc. Moving that VBR out of $0 to somewhere in fast RAM was a huge speed boost. Sadly, some games used to write into the vector table at location $0, so if the VBR was moved, or the game didn't respect where the VBR was relocated to (overwrote the table), it would cause the game to fail. So, arguably, yes, you _could_ see a significant speed boost for games with a 68010, if you've relocated the VBR before starting the game. ColdCapture was a system register which could be intercepted so you could execute code on reboot, so a VBR relocator which sat in a ColdCapture hook would have achieved this.
I got a 68882 pga for my A1200 card on ebay. When I got it you can see where the top had been refinished. It went back I did not fit it. I found a OK one later.
The thing is all of the chips are over 30 years old there might be some damage to the top text.
With real fastmem you would want the 010 upgrade to have the VBR in fastmem if you do serial port transfers. It should really help on upping serial speed.
The IPA and nail polish remover tests are red herrings. I believe they sand blast the top of the chips. This also explains the change in texture.
The 68010 was a common "upgrade" in the 80's. Some people claimed it was slightly faster, but realistically, it seemed hard to believe it would/
The main potential performance gain was due to its optimized loop prefetch feature, but I guess that unmodified software written for the 68000 would not be able to take much advantage of it. Well exploited, you could get up to 50% faster for typical memory copy loops. But in the end, too much was already based on the 68000 and probably few (except for some niche applications) were willing to take the risk of writing 68010 software only (not compatible with the 68000), so it just missed its market. The other benefits were of interest more to workstations and server stuff than "consumer" computers.
I can dimly remember that the 68010 made some changes to the instruction set to do with interrupt handling - it's not just a 68000 with a small instruction cache added. So it could be that some programs will run fine and some will choke, depending on how they were compiled.
Yeah, stack frames are sized differently, and poorly programmed games that try to manually manage the stack will blow up. The instruction set isn't the problem. Almost all compatibility problems with the '020 and up are due to variations of the stack frame issue.
When I worked with the 68k family in the late 1980s, I seem to remember the 68010 to be only a variant of the 68k that had some inconsistencies in the instruction set corrected in line with the later 68020 + 68030 chips. So, unless you needed complete code compatibility between computer model variants and deployed the whole family the 68k was the better choice. I ran an Atari ST with a 68020 CPU upgrade at the time and remember that as a noticeable improvement in speed. But I got rid of it almost 30 years ago. Real shame that.
I switched from the MC68000 to the MC68010 many moons ago on my AMIGA 2000HD. I have the 64-pin DIP type and never have had any of the problems that you guys mention. Am I lucky or full of it? Only my computer knows. I wish I could find my old chip and try the same test cause there is a slight difference. I also run the 1.3 to 2.04 keyboard selectable ROM if it makes a difference. Might have to dust off the old beast.
Looking forward to your expansion card for the venerable A500. It'll give me a reason to resurrect my own dead A500.
I don't feel that it's fake or a slower one that they spent a lot of time on trying to sell it as a faster one... not for something like a low-end 010. You hit it pretty well with the realizing that you just can't assume you can overclock, based on luck (I put a 010 in the CMI Processor Accelerator when I was working on that in the 80s, and it was very sensitive to speed input. You know these are binned parts of higher crystal speeds that failed, so it gets marked as the one that it passes their internal fab tests at so there is a likelyhood that 12MHz really was all it could do for extended times. And the amount of time that you got it to run before it hung was millions and millions of instructions per second, so if you get a 16MHz part you'll likely be winning.
Now, for the reason you put an 010 in: not for games. What you very much wanted to use with an 010 was this:
"Loop mode" which accelerates loops consisting of only two instructions, such as a MOVE and a DBRA. The two-instruction mini-loop opcodes are prefetched and held in the 6-byte instruction cache while subsequent memory read/write cycles are only needed for the data operands for the duration of the loop. It provided for performance improvements averaging 50%, as a result of the elimination of instruction opcodes fetching during the loop.
You did see an improvement over the original 68k in sysinfo, but as you pointed out, sysinfo isn't a great way of testing it. I used the 010 loop mode on the CMI Multiport Board SCSI code with a 4-way set associative cache, but had to yank it because I needed to code space in the boot ROM. It did yield at least a 10% improvement... which doesn't sound like much, but it was not expensive to get, and was fairly easy to install [sorry about your PLCC carrier.. I did that far too many times even with the extractor] (at the sake of some timing-critical games or cycle-specific code would break).
Cheers, and great video as always.
Thanks for the detailed reply. I have to admit I often wonder why anyone would bother relabelling chips, especially low value parts like this. It may well be the case that it isn't refinished but I can't help but wonder with the feel of the top of it and the fact that the nail polish remover did take something off.
I may yet get another one and try to look for a 16mhz part so I can do some more thorough testing but for my use, for games, I don't think there will be much if any improvement.
@@CRG Nail polish remover is just rough on certain plastics in general and you were really rubbing hard on that and only got a very small bit of stuff, so it might've just been getting the bits of rough plastic off. When I've seen this being demonstrated, it was just immediately a bunch of stuff coming off, turning the tip completely black. The differences could certainly be chalked up to different production lines even in the past. Just doesn't seem very conclusive.
@@Aeduo that may well be the case but there is still the fact that the labelling does feel etched into the surface. Of course nothing to say it wasn't etched in the first place. It's hard to conclusively say either way on if it it's genuinely a 12mhz chip. All I can say for certain is that it doesn't work at 14mhz. That in itself isn't enough to prove it's a relabelled part but certainly from experience most CPUs will overclock a small percentage.
I was unaware there was ever any 16MHz 68k10 parts made. Could not find any back in the early 1990s, when I was trying to wring every last scrap of speed out of my already aging A500 with AdSpeed accelerator. Ended up buying two DIP package 12MHz '010s (one after the other) and they both crashed immediately when you switched into 14MHz mode, sadly. Not sure if that's because the CPUs wouldn't overclock, or if it was some quirk of the AdSpeed itself; some hardware timing or whatever being different. *shrug*
Pity, because the VBR feature of the '010 was quite nifty when doing serial port I/O as I recall (dial-up modems being an important part of my life back then.)
bit late but,... it looks not much of a deal to overclock a 12MHz chip to 14.6MHz. But this is almost 22% overclock! Imagine to push a 5GHz i9 to 6.1GHz...
Having done a quick Ebay search I have seen 68000 CPU's listed as working at frequencies as high as 20Mhz. One of these would probably give a much more noticeable speed increase than a 68010. Having said that these listings were from China so I have many doubts about the veracity of them. Looking at the cost of other 68k chips that you could use to build an accelerator around the costs seem to shoot up extremely quickly with 68060's listed in the £400+ price range. It looks like the best value CPU upgrade for an Amiga is essentially a PiStorm.
The 68010 is only good for things using tight loops but that comes at the cost of compatibility. Expect to see speed ups in printing, serial transfers, stuff like an sdbox. My guess is you have an older 010 meant to run at a lower frequency like you suspected. Nice vid. People learn more from the journey than the destination.
Iirc, the 68010 had slightly optimized 32-bit instructions compared to the 68000. But what are the compatibility problem with it? I though they were 100% binary compatible. Is it because of timing? Some code than depends on exact 68000 instruction timing?
@@herrbonk3635 the compatibility issues arise from software expecting 68000 behaviour handled differently by the 010. The cache can affect timing messing with some demos and games for example.
@@imalebowski So it was what I guessed, code depending on exact instruction timing.
I recall back in the day games magazines advised against using a 68010 because it is not entirely code compatible and caused games to crash. The point being why use one when some software won't work with it. Been decades since I've delved into 68K processor types but from my limited knowledge, is it possible its a 12.5Mhz part? I don't recognise the part number from the ones I've seen but the 12 on the end usually meant 12.5.
As far as I know, it's 100% compatible in user mode, but in supervisor mode the stack frame format is different, and only the OS needs to care about that. Games tended to manage the stack manually and use supervisor exceptions for things like copy protection, so things would often blow up. The processor is just fine, but game programmers are never known for doing things according to spec, especially when it comes to anti-piracy. 8)
@@Waccoon Yeah, games programmers trying to squeeze as much out of it as they can, what are they like ;) Recall when I upgraded to Amiga A1200. Having to turn cache off on '020 and later '030 accelerators because of self modifying code. Often used in copy protection, it would crash it if the instruction cache was enabled. Or where they use the redundant 8bits on the address registers, something I admit trying myself before realised it wouldn't run on a full 32bit system.
I once saw someone on a channel barely removing the 'paint' (or whatever it is..) from a refinished IC using acetone. IPA did not touch the finish at all, it was just too resistant. (Don't remember exactly, but probably it was Adrian or David Murray).
PS: Just saw you later tried it xD
Yes, the paint cover will go off with acetone. With a bit of luck, you can reveal the real markings hidden underneath.
The solder dipping of pins generally only happens with pins that get soldered to the board. These are almost always in sockets so no need for them to have done that.
Also, the counterfeiters in China have gotten smarter about their toppings. IPA doesn't always do it now. You need to try acetone or similar.
Edit: I see you tried that later.
Just got me a 68882 50mhz from ebay. This was to put in the empty FPU socket of an old DCE typhoon accelerator. The moment I removed it from the jiffy bag. You could see where the fakers had masked the gold top and respray it. The ink came straight of it with a wee drop of IPA. It was a poor attempt at making it look genuine. Saying that it does run OK at 40mhz.
Oh, didn't know that WHDLoad doesn't require an 68010 now anymore. Just bought one some weeks, ago .. xD
That 68010 may not be even a 12MHz part. There is also some incompatibility between the 68000 and the 68010 on how some stack frames are formed, but I doubt that has any impact in the Amiga.
In my A500 i've just replaced the rectangular 68000 with pin compatible 68010.
4€ shipped...works fine.
The 68 and 88 series from Motorola during that era was like this, the text wasn't always straight and some of the chips did have a texture to them. One of the reason is that it holds on to a TIM better, should you use a heatsink with it for really warm devices in a fairly hostile environment, the Commodore Amigas are not such an environment.
I used to work with the 680x0 CPUs in the late 1980s.
The 68k was a nice CPU that had some limitations and the next iterations improved the situation gradually.
To my memory the main advantage of the 68010 was that it had the full instruction set, but only very little I.provment in performance compared to the normal 68k. That situation only improved with the 68020 and 68040 with the FPU and MMU co-processors.
I did not gather any experience with the 68040.
But I think it is fair to say that an upgrade with a 68010 is not worth the cost and effort.
So, yes it is useless.
Reminds me of the NEC V20 chip upgrade for the 8088.
68010 wasnt really much of a bump and with the hassles it brought it wasnt worth doing, I remember those chips being in dip format back in the day on those old ''rock lobster' boards, if I remember correctly, I maybe wrong as it was over 30 years ago, a pin was lifted and shorted to ground via a switch for compatibility reasons, once you flipped the switch you had to reboot or power down
Yes I remember something similar, ' I think' there was a mod to lift the clock pin and use a different oscillator perhaps, so if you had a stock 68k and got a faster 010 + oscillator there was a difference, but I might be mistaken. I think also 7% speed boost whilst it isn't much, if this was the latest laptop, it would be described as a whopping 7% increase in speed :-)
look at the chip numbers, the last N16 and the 10 chip is N12, its probably the mhz rating.
many years ago my friend who works now at ARM put a 68010 in my Amiga 500. I really cant say i noticed much improvement ( maybe 5%) even when i used the old rendering software "imagine" etc i didn't see improvement that was worth it even tho the chip was free.
@ 6.34 "100% genuine"! Well "yes" it is a 68010, but definitely remarked. God knows the original speed grade. IPA won't reveal anything with a part that has been sanded down and laser etched. At 8:49 it probably the fine dust residue from the sanding process.
Hi G. that 68k says 16Mhz on it. Any scope for clocking it higher ?
Not without a redesign of the accelerator. It takes the 7mhz original clock and doubles it but the 68k chip should be stable with the clock tripled of not quadrupled.
@@CRG so it doesn't have a clock crystal or a tuning app for AigaDOS ?
The '010 was also made by a few other companies so I wonder if it's a rebadged non first party chip.
I must confess that my overall knowledge of the '010 is pretty slim, I know how the speedups work, but not enough to actually help with diagnosing the actual real speed or fab of a remarked unit...
-Dx
Why would that be a problem? They all used identical masks from Motorola. And even actual reverse engineered CPUs often works flawlessly. (Such as NEC, Sharp, Toshiba, Hitachi, and Goldstar (LG) clones of the Z80, for instance.)
@@herrbonk3635 Who said it would be a problem? I was speculating if that's why it was remarked.
your Amiga 500 is well kept - or is it brand new ?? thanks.................
Yeah it sounds like an 8 that's pushed too hard.
I have a 20mhz plcc 010 if you're interested, i was going to use it in my wicher 508i but it didn't like it (the firmware on the wicher couldn't hack it) I'd be happy to send it along if you're interested like.
I'd certainly be interested in getting the chip, if anything it would let me do some more thorough testing.
Drop me an email if you like, casualretrogamer@outlook.com
@@CRG sorry it's taken so long to get back, I've been going through my chips and I can't find the 010 I had, i found 2 68000's which i didn't know were there but no sign of the 010 I'll keep you posted.
Even if it was a totally legit 14mhz 68010 it's usually not worth the cost and trouble to replace the 68000.* I'm not aware of a single accelerator that used one. IIRC, the '010 was mainly used in support hardware like external drives, printers and the like. Adrian Black could talk your ear off about how many fake chips he's bought from Chinese sellers over the years...
* Unless you're increasing the actual clockspeed too.
Portadown, my home town! 😁
I'm up in Belfast myself and this machine came from a person in Lisburn :)
Speedup would be negligible, the instruction cache is very small (6 bytes) on '010, only would apply to super tight loops.
Also if it is genuine it should allow you to quit out of games on WHDLoad, whole '000 wont.
Whdload was updated back in July to allow exiting on the 68000. I only learnt that while recording this.
Yes the speed increase is very small, it's a topic I've seen debated several times on forums so thought I'd try it for myself. It's just a shame the chip wasn't stable as I had planned to do more tests.
@@CRG It was marked N12, seems those 2 extra MHz were too much for it 😉
Yep seems I was asking a bit much 😂
Looks like a phone microprocessor. We used in the mount 11 phone chassis.
Can you set the clock lower and see if it even runs?
Unfortunately not, all I can do is disable the accelerators ram but I can't see that making any difference.
I have had lengthy conversations with a seller on Aliexpress about chips being resurfaced and they told me that they do it to maintain asset tracking. They basically copy one to all of them so that they can track what they sell for returns. If they left them all original they would know what was sold by them or not.
Interesting, but could they not just make note of the existing part numbers.
while that may be partially true - i've bought a lot of rebadged chips that actually were what they claimed to be - they almost **always** take it as an opportunity to mark it as higher speed grade. just ask anybody who has attempted to buy an '040 or '060 off a chinese seller..
In a stock A500 a 68010 runs as fast as 68000 with fast ram. With Fast Ram installed there is no difference.
I was one of those that tried a 68010 upgrade, on my Amiga 1000. There was the tiny speed increase, but it introduced so many software issues that I went back to the 010. Some games worked, some did not. Same with productivity software. There were a few differences in the opcodes. I don't remember the details, but it just lead to to many crashes.
I remember there was a program called decegel that patched around some of the differences in the 010. But I thought by WB 1.3 support for 010/020 was baked in?
@@neozeed8139 Oh yeah! I forgot about decigel. Yep, by 1.3 the OS was fine with the other CPUs. But many of the games out at the time still had problems. Eventually things got better.
68010 may not be CMOS version, hence the heat
I never thought of that, although the chip does get hot, not burn you hot but just a bit more than you'd be comfortable leaving.
AFAIK, the 68010 was never popular enough to warrant a CMOS conversion. The 68000 CMOS version was designed by Hitachi.
can it be swaped from a mega drive?
I have read a few things about people using trying the 68010 in a mega drive. I've not seen anything confirmed though and I don't know if it would make much difference.
But a cool collectible :)
The 68010 was a controller version on the 68000 I think, not a true successor.
I would not have the expectation of a speed difference in any meaningful way.
The 68020 is the true successor and the first true 32 bit chip (hardware wise).
I still remember, the swap did no good for my experiance 😭
Aaaaaaah, I tried the same and to my shock, almost nothing ran! :D
cotton buds and acetone, isnt it?
Possibly, IPA maybe isn't strong enough but you'd like to think nail polish remover would shift whatever was put on the chip. It did take something off but there might be more to come.
@@CRG Sometimes the chip tops are over-painted, sometimes they're sanded-back. Reduced residue and a rough texture could well be a sanded top.
Time spent tinkering with any Amiga can hardly be called pointless. Think of it more as a calling!
ok in the world of chips, chips are made to do a certain job. If the chip does the job that qualifies it as a real working chip. So are they even really fake chips?
I used a 68010 and was not really much faster at all.
Meanwhile back in reality, you can run an amiga emulator on a laptop pc faster than 68040 chip.
You must be absurdly boring at parties...
Next time. Soldering the 68010 in a A600
would need to get a 600 first :)
A PiStorm would probobly be a real upgraded.
I have an A500+ with the pistorm, its a great piece of kit although I enjoy tinkering with period correct hardware too.
Don't use IPA. Use acetone instead
test it Apollo 68080 CPU
id look for a 25 mhz of 33 instead
If you had done your research you'd found this wikipedia article about the 68010...
en.wikipedia.org/wiki/Motorola_68010
If I am not mistaken the 68010 was primarily used in the A1000...
No it wasn't. If you'd done your research you'd have seen there was a non-Commodore article about using it as an upgrade, but the A1000 had a 68000 CPU.
And with a 68010 you can actually quit whdload games .. so not totally pointless
It's not pointless. And throwing the chip like that was just childish and wasteful. First, that chip was probably a 12.5MHz or slower version of the 68010. Try running it at that clock speed. Second, did you research beforehand to see if any of the games you were planning to test were 68010 compatible from the start or modded since their release to work with it? The chip has improvements over the 68000 that cause [mainly] games to not always work if they aren't programmed to identify a 68010. That happens whether the platform is an Amiga, an Atari ST, an Apple Mac, or a Sega Genesis/MegaDrive.
As for the performance differences between the 68000 and the 68010, I'd suggest looking at the Atari Games Corp's arcade titles from the mid to early-late 80s era. There's a reason why "Atari System 1" games like Gauntlet used a 68010.
the finish.... man i seen sweet real fake chinese chips in way worse condition... with like 2mhz to spare and all! 😆👍
thanks for the share anyhow mate! amiga 😆💛
Why do people think that the Chinese are no able to laser-engrave a chip with the data on it of another BIGGER and FASTER chip? Everyone is testing a chip with some isopropanol and a cotton tab. In China this can be seen as well. Therefore a counterfeiters have improved their act, no longer white ink is used to print the name of the manufacturer on on it, for a while now they have improved the etching, scraping, filing off the original letters and cyphers to re-do it with the names and serial-numbers of the very very wanted little gems.
China does not give a shit. Train wagon loads of rejected Aspirin is rebranded as Penicillin and sold to Africa. And if you think this is the worst, think again, almost all ingredients of western medicine have nowadays their origin in the "Great Country of the Middle" because it is cheaper. Does anybody know what gutter-oil is and why it is an ingredient of food?
Please don't call chips fake if they are real chips that have been desoldered and recycled.
Fake branding, of real ICs :)
Fake News Chips !!! Gone stale.
The chip has some evidence of remarking, and does not perform as per marked. I'd call that fake.