Atari Disk Speed vs C64 (stock and modded)

[Mazzspeed]

It won't handshake any higher than 19200 B/sec no matter what divisor I use. I have seen it handshake higher, but it quickly drops back down if you try to transfer data.

 

There's the cable I'm using:

 

[Faicuai]

12 minutes ago, Mazzspeed said:

It won't handshake any higher than 19200 B/sec

Ok. that's the door-stop right there, and we will need to trace down why the handshake refusal at any other divisor (!)

 

I really doubt the problem is on the SIO-host (600XL), if you ask me.

 

(UPDATE: the reason I doubt the 600XL is failing is because on XLs, divisors 5/4 are virtually guaranteed, even on a "naked", no-hops SIO chain, assuming Pokey is healthy).

 

 

Edited April 19, 2021 by Faicuai

[Mazzspeed]

I briefly got it to work @ 110598.

 

[Faicuai]

That's VERY good (divisor 1) !

 

 

[Mazzspeed]

7 minutes ago, Faicuai said:

That's VERY good (divisor 1) !

 

 

Except it won't sustain it...

 

As soon as you disconnect and reconnect you're beck down to 19200B/sec

[Faicuai]

Just now, Mazzspeed said:

As soon as you disconnect and reconnect

Logically at RespeQt, or do you mean traffic-stopping while RespeQt still showing connection?

 

 

[Mazzspeed]

14 minutes ago, Faicuai said:

Logically at RespeQt, or do you mean traffic-stopping while RespeQt still showing connection?

 

 

All I have to do is mount another disk and change to that disk via SDX and RespeQt drops sync sync speed. Happy compatibility seems to increase the speed I can sync at, but then I get errors regarding device not responding.

[Mazzspeed]

Yep, just as I thought. Tried on a Windows laptop as well as my RPi400 and the highest stable speed I can achieve is divisor 5. Any higher that that and sync speed drops to 19200 B/sec. Tried with a USB lead roughly 30cm long.

 

 

[Mazzspeed]

OK, just as I thought.

 

I remove caps C89 and C90 on the SIO port and like that divisor 1 is stable, divisor 0 still isn't stable. So you may get maximum speeds via SIO, but then again you may not? I don't know what else to check.

 

 

 

[Faicuai]

Divisor 1, directly on SDrive NUXX (NO RespeQt) via SIO chain described above, STOCK q983 GTIA 800, no mods or surgery of any kind on host SIO interface:

 

 

 

 

+7100 Bytes/sec writing, +8,000 Bytes/sec reading.

 

Again, no mods or surgery neeeed (at all) for attaining such speeds. 

Edited April 19, 2021 by Faicuai

[Mazzspeed]

28 minutes ago, Faicuai said:

Divisor 1, directly on SDrive NUXX (NO RespeQt) via SIO chain described above, STOCK q983 GTIA 800, no mods or surgery of any kind on host SIO interface:

 

 

 

 

+7100 Bytes/sec writing, +8,000 Bytes/sec reading.

 

Again, no mods or surgery neeeed (at all) for attaining such speeds. 

*However such speeds aren't guaranteed..

 

Looking into this issue, I found many people that couldn't achieve such speeds without modification. Even FJC himself had issues on this 600XL at around 36.19:

 

 

Edited April 19, 2021 by Mazzspeed

[Faicuai]

20 minutes ago, Mazzspeed said:

*However such speeds aren't guaranteed..

 

Looking into this issue, I found many people that couldn't achieve such speeds without modification.

Yes they are, sir. Your 600XL can and will get there, no HW mods., and reliably.

 

Trust me on this one (I would not be stating it here, in vane). Get your SIO-hub from Lot harem and you will never look back.

 

Check this out:

 

 

 

 

Edited April 19, 2021 by Faicuai

[Mazzspeed]

4 minutes ago, Faicuai said:

Yes they are, sir. Your 600XL can and will get there, no HW mods., and reliably.

 

Trust me on this one (I would not be stating it here, in vane). Get your SIO-hub from Lot harem and you will never look back.

 

Check this out:

 

 

 

 

See my edited post above.

 

I'm happy with divisor 1. If I'm only running one SIO device off the SIO port, how will Lotharek's device help me? I'm genuinely curious.

 

EDIT: Is it buffered?

Edited April 19, 2021 by Mazzspeed

[Mazzspeed]

Even FujiNet runs divisor 1 no worries now!

[flashjazzcat]

11 hours ago, Mazzspeed said:

Does the 600XL have the caps that need to be removed?

Yes, generally, but whether they need to be removed or not is another matter. My own daily driver 600XL still has them and they cause no problems at all with divisor 0. A recent customer machine was the same until I replaced POKEY with PokeyMAX, at which point SIO at higher than stock speeds became barely usable until the caps on the signal lines were removed. Divisor 0 then worked perfectly (I see you already found a video where this is documented).

11 hours ago, Mazzspeed said:

Valid point. I doubt the difference would be in any way noticeable in real world usage, I just found it interesting. In fact, I assume that due to it's DMA implementation, in normal real world usage the SIDE3 should have a slight edge shouldn't it?

Correct: variations of a few KB/s are not noticeable in daily usage, and this is true of benchmarks in general on any platform. The differences in latency between one CF card or SD card and another are definitely interesting, and - as mentioned - the only possible opportunity for variance is the time taken for the controller to respond to commands (since transferring the data buffer is accomplished in a tight and/or unrolled loop of code which runs at a completely determinate speed). In both cases, we are essentially polling for the controller to become ready; IDE is ready for data as soon as the busy bit clears, but SD cards (controlled via SPI on the SIDE3) first provide an 'R' response, and then - assuming we're reading a sector - a data token which prefixes the actual data. The code required to run SD cards in SPI mode is - despite months of optimisation - quite a bit more verbose than IDE driver code (which is extremely simple), and I initially found performance disappointing using the Sony 8GB SD card I had been using for testing. Only later when I began testing a much wider selection of cards did it become apparent that other media would match or even best the performance seen with CF cards using the older drivers. I have seen swings in read speeds between different SD cards of almost 10KB/s.

Edited April 19, 2021 by flashjazzcat

[flashjazzcat]

10 hours ago, Faicuai said:

But the funny thing is that there is not (any) removal of (anything) happening, here.

 

Aiming at stable, Divisor 1/0, all-day-long, 7x24, on a COMPLETELY STOCK machine is the name of the game!

 

All #s posted above come from stock (SIO host interface) 1980 CTIA and 1983 GTIA 800's. Surgery-free babies... 8-)

But this may not be possible if the machine does not want to cooperate without removal of the SIO signal caps. As I understand it, the caps are there because of the cumbersome FCC RF regulations, and don't really serve any useful purpose. Moreover, the caps were not always fitted at the factory, and on some machines where they are present, they are bodged onto the back of the SIO connector legs. So they may perhaps be considered a manufacturing error, and thus their removal a 'repair'.

[Mazzspeed]

29 minutes ago, flashjazzcat said:

But this may not be possible if the machine does not want to cooperate without removal of the SIO signal caps. As I understand it, the caps are there because of the cumbersome FCC RF regulations, and don't really serve any useful purpose. Moreover, the caps were not always fitted at the factory, and on some machines where they are present, they are bodged onto the back of the SIO connector legs. So they may perhaps be considered a manufacturing error, and thus their removal a 'repair'.

After pushing this little 600XL as hard as I can at various divisors most of the day, testing what divisors were stable and what divisors weren't. I can state without a doubt that divisor 5 was the limit with the caps in place irregardless of whether I used a long USB lead, a short USB lead, a Windows based RespeQt server or a Linux based RespeQt server. As soon as I removed those caps, divisor 1 was rock stable under both RespeQt and Fujinet-PC.

 

But, no matter what I do, divisor 0 is just not stable. I'm not stressed, in fact I'm quite pleased and found the whole testing process quite interesting (as always, I'm learning plenty as I go).

 

I'm interested in Lotharek's hub, I really want to know just how it's capable of cleaning up data signals to the point that divisor 0 is stable on old Pokey variants?

[R.Cade]

1 minute ago, Mazzspeed said:

After pushing this little 600XL as hard as I can at various divisors most of the day, testing what divisors were stable and what divisors weren't. I can state without a doubt that divisor 5 was the limit with the caps in place irregardless of whether I used a long USB lead, a short USB lead, a Windows based RespeQt server or a Linux based RespeQt server. As soon as I removed those caps, divisor 1 was rock stable under both RespeQt and Fujinet-PC.

 

But, no matter what I do, divisor 0 is just not stable. I'm not stressed, in fact I'm quite pleased and found the whole testing process quite interesting (as always, I'm learning plenty as I go).

 

I'm interested in Lotharek's hub, I really want to know just how it's capable of cleaning up data signals to the point that divisor 0 is stable on old Pokey variants?

I had a similar experience. My 800XL will do it, but my 600XL won't.

 

[Mazzspeed]

1 minute ago, R.Cade said:

I had a similar experience. My 800XL will do it, but my 600XL won't.

 

You should try removing the caps in my image and see what happens? I think you'll be surprised. I was.

[R.Cade]

1 minute ago, Mazzspeed said:

You should try removing the caps in my image and see what happens? I think you'll be surprised. I was.

I did- it didn't help for me. I mean, it was better, but no divisor 0.

 

Edited April 19, 2021 by R.Cade

[Mazzspeed]

Just now, R.Cade said:

I did- it didn't help for me. I mean, it was better, but no divisor 0.

 

Ah, I understand. So your findings were identical to mine?

 

This hub has me intrigued.

[flashjazzcat]

1 hour ago, Mazzspeed said:

After pushing this little 600XL as hard as I can at various divisors most of the day, testing what divisors were stable and what divisors weren't. I can state without a doubt that divisor 5 was the limit with the caps in place irregardless of whether I used a long USB lead, a short USB lead, a Windows based RespeQt server or a Linux based RespeQt server. As soon as I removed those caps, divisor 1 was rock stable under both RespeQt and Fujinet-PC.

So - given the fact my 600XL will do divisor 0 with the caps still in place - there are obviously other variables. What kind of SIO2PC adapter are you using, out of interest? Apologies if you have already provided that info, but the thread is becoming quite busy.

 

EDIT: I see now that you posted a photo of it above.

 

It would be really interesting to nail exactly what the limiting factors are, since it's not only my own machines which work well at divisor 0, but every single one which comes through my hands for an U1MB install (including the couple of 600XLs which needed SIO caps removing first). I update the original firmware at 3x SIO (which works with the standard drivers), then enable HSIO in the new firmware, and apply further updates as needed. If the machine already has the HSIO enabled firmware installed and simply needs an upgrade (which many have recently, since I have been flashing pre-release updates to customer devices), I just whack RespeQt up to 127Kb/s and it 'just works'. The USB cable is only 1M long, but I used to have good results with the built-in FTDI SIO2PC adapter on the 1088XEL as well using a 2M cable.

 

BTW: I use Windows 10. I have not had such a flawless experience with RespeQt at divisor 0 on macOS; haven't tried Linux recently. If we're looking for answers, we might have to look no further than the host machine and FTDI drivers.

Edited April 19, 2021 by flashjazzcat

[R.Cade]

5 hours ago, Mazzspeed said:

Ah, I understand. So your findings were identical to mine?

 

This hub has me intrigued.

I've had about 4 x 600XLs. I still own two of them. I replace the ROM with HiaS-patched HSIO version. I also have had a couple of 800XLs (and still own one). All have had the caps removed.  

 

The 800XL is 100% no problem at divisor 0. No stutters ever, that I can recall (you can easily hear it).

 

The 600XLs (all of them) are not reliable at less than divisor 3. They will work to some extent, but they stutter and sometimes fail to load.

 

It would be nice to know what other difference there could be, since I love the smaller footprint of the 600XL.

 

Also, it's been a couple of years since I played with them (now I use Incognito 800) so maybe I need to go back and test again to be sure. I seem to also remember that the 800XL worked fine at divisor 0 with my SD2SIO variant, but the 600XL didn't.

 

Edited April 19, 2021 by R.Cade

[MrMartian]

17 hours ago, Mazzspeed said:

I'm just benchmarking an SIO attached disk at divisor 0 using high speed SIO BIOS using an ATR at 512 bytes/sec and I get nowhere near the speeds Facuai is getting via RespeQt.

 

 

I don't know what I'm doing wrong here:

 

 

 

 

 

I see in this, you have handshake method as 'NONE'. You will need to have the command signal hooked up to one of the serial signals, and select that in the settings to have proper full speed communicatons.

 

[Mazzspeed]

8 hours ago, flashjazzcat said:

BTW: I use Windows 10. I have not had such a flawless experience with RespeQt at divisor 0 on macOS; haven't tried Linux recently. If we're looking for answers, we might have to look no further than the host machine and FTDI drivers.

I did try Windows 10 thinking that there may have been an issue with my Linux RespeQt server, the problems were identical under both Windows 10 and Linux. It definitely wasn't an OS issue, and as far as I can tell my cable as pictured in one of my earlier posts is perfect. It's even using the shield as GND.

 

2 hours ago, MrMartian said:

I see in this, you have handshake method as 'NONE'. You will need to have the command signal hooked up to one of the serial signals, and select that in the settings to have proper full speed communicatons.

 

I eventually did enable handshaking via CTS, I still have it enabled. I tested with handshaking on and off, it didn't make a difference. Even enabling DTR doesn't make a difference.

 

As far as I can tell, certain 600XL's simply struggle at divisors below 5. Although, as mentioned, since removing the caps I have got it rock stable at divisor 1.

Edited April 19, 2021 by Mazzspeed