CPU speed discrepancies oddity

Troubles with your machine? Just want to speak about the latest improvements? This is the place!

Moderators: Mug UK, Zorro 2, Greenious, spiny, Moderator Team

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

CPU speed discrepancies oddity

Postby exxos » Fri Apr 29, 2016 9:42 pm

While building up a couple of V2 boosters and testing out GB6, I found something odd.

For starters, I've been having issues where in GB6 I could benchmark a machine and get 100% scores. Then next benchmark it would start throwing up 99% scores. Next lot of tests more scores would show up as 99%. Its been a long outstanding issue where it seemed tasks didn't always take the same amount of time.

Yesterday, I built a V2 and re-tested it and got 100% scores. 20+ benchmarks later, still all 100%, great I thought! Then I tested a second V2 and it was chucking out 95-97% scores. A big WTF there. The results varied about 1% which was the original issue I was having with GB6.

So, I put the CPU out of the 100% reliable V2, into the not so reliable 95% scored V2, and then I got 99% scores on all tests which seemed stable. I ran out of time to do more benchmarks but it proves a odd point, that the CPU is somehow not as fast as others. But that does not make any sense to me.

How can one CPU be faster than another ?! They are both 68HC000FN16 CPU's. I've even seen these odd results on every V2 I've built so far , where the benchmark results are never constant from one V2 to the next. Though with current tests swapping the CPU for a difference one, gave different results.

So thought I would post my findings here, as It makes no sense to me....
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

Zarchos
Captain Atari
Captain Atari
Posts: 263
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: CPU speed discrepancies oddity

Postby Zarchos » Sat Apr 30, 2016 7:13 am

1st idea is that they don't use the same microcodes ... or do your tests use non official mnemonics in which case from a revision to another one these haven't been implemented the same way (thus, again : different microcodes are used).

Additonally when I read on the ST with the GLUE etc ... you've got some waitstates etc ... and their 'state' is more or less unknown couldn't this add, even if the tests routines are very well written, some extra cycles anyway ?

Sorry I'm not a real Atarian so all these ideas can be BS but from I've been reading here since I joined that's what I think.
Falcon, Atari 1040 STE, 520ST, 800xl, xegs, Amigas, Archimedes, RISC PCs, Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Russian clones, Sam Coupe, V6Z80P, QLs inc. Q68, and more !
2200 m2 museum on its way https://youtu.be/xjB6_Ez-3BA
Shorter video here : https://youtu.be/UEZisfkcN1Y

Currently porting SOTB to the Archie : https://www.youtube.com/user/Archimedes75009/featured

User avatar
troed
Atari God
Atari God
Posts: 1438
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: CPU speed discrepancies oddity

Postby troed » Sat Apr 30, 2016 7:26 am

Same machine = the CPUs are equally fast. Else a lot of demos would fail instantly.

Different machines might have different CPU clocks. 8.007, 8.01, 8.02 and 8.05

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sat Apr 30, 2016 7:41 am

Yes, this is for sure a result of GEMBench CPU benchmark routines. I have never tried any version of GEMbench that gives a 100% result on any of my machines. It's always 99% or 101%.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Sat Apr 30, 2016 8:23 am

joska wrote:Yes, this is for sure a result of GEMBench CPU benchmark routines. I have never tried any version of GEMbench that gives a 100% result on any of my machines. It's always 99% or 101%.


That is true, BUT, one CPU gave 100% results for over 20 benchmarks. This suggests gembench code is fine. I even rebooted several times during that and always solid 100% figures. When I changed to a different V2, I got 95% results which were more or less solid. So this isn't a software issue. GB6 has more advanced timer code aswell.

Problem is, for some reason tests can take a few timer ticks more or less, this seems random. GB6 can only report figures based on the timer, which never has returned the same amount of ticks per test before. Again considering it was rock solid with one V2 suggests there is something going on in hardware not software.

Zarchos wrote:1st idea is that they don't use the same microcodes ... or do your tests use non official mnemonics in which case from a revision to another one these haven't been implemented the same way (thus, again : different microcodes are used).

Additonally when I read on the ST with the GLUE etc ... you've got some waitstates etc ... and their 'state' is more or less unknown couldn't this add, even if the tests routines are very well written, some extra cycles anyway ?




That is basically what I am wondering. Something seems to "stall" the CPU, but this doesn't explain why 1 CPU worked fine.

troed wrote:Same machine = the CPUs are equally fast. Else a lot of demos would fail instantly.

Different machines might have different CPU clocks. 8.007, 8.01, 8.02 and 8.05



This is all done on the same machine, so clock speed is constant. I'm talking about a couple of timer ticks over a 20 second test. I doubt any software is going to notice that aside from gembench which is timing how long tests take.

I'll say it again, changing the CPU alone gave either 95% results or 99% results, thats with a DIFFERENT CPU on the SAME machine. No matter about rebooting or retesting, always 95% on one CPU, 99% on another, 100% on the "faster" CPU and the first V2 booster I tried. This ISN'T the same issue as GB showing 101% or 99% scores which can vary at random each time the test is done. If that was true, why would one booster always show a rock solid 100% in over 20 full benchmark tests, and yet, simple change the CPU and speed drops 5%, put the other CPU back, and back to 100% scores again. Clearly the CPU is doing that, but WHY?!
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sat Apr 30, 2016 8:40 am

If it's a hardware issue it would be related to the booster.

Here's a simple test that will confirm that the CPU speed does not vary:

1. Install CPU 1, the one that gave 100% consistently.
2. Run a really timing-sensitive demo, like SYNC's {closure}. Confirm that it works.
3. Install CPU 2.
4. Repeat (2). The demo still works.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Sat Apr 30, 2016 9:23 am

Its a good thought, though the hardware combo which gives 100% is on a V2 booster which has sold. The CPU alone didn't take it to 100% on another V2 booster only 99%. So the GAL logic must also contribute to 1% of the speed loss. I've gotta ship them out as they been on back order for some weeks. Though I am building up a batch of about 15 V2's currently. So I can do more tests when I have "stock" to play with.

There is also a issue that this is running in 16mhz mode. So if the demo is graphics related then it won't work anyway ?. Though I haven't done much testing in 8mhz modes with GB6. I have a feeling the results will still drift. But will confirm next week. Overall, its like 1 timer tick every 10 seconds the "drift" is. Though its not constant.

If you have a link for the demo can you post it. I'm about to leave for the weekend, so will do some more tests on Monday in 8mhz mode.. Actually I have about 20 CPU's on my test V2 board, so I will try every CPU in my test board and see what the speed variations are between them all.

Not sure if it helps or not, but the results are always rock solid in steem SSE :)
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sat Apr 30, 2016 10:08 am

You can download {closure} at demozoo. It should work on your ST/STE in 8MHz mode.

Results on STEEM is only relevant on STEEM :) You'll see that no emulator emulates the hardware closely enough to run this demo. It's a really good compatibility-test for your booster.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
shoggoth
Nature
Nature
Posts: 970
Joined: Tue Aug 01, 2006 9:21 am
Location: Halmstad, Sweden
Contact:

Re: CPU speed discrepancies oddity

Postby shoggoth » Sun May 01, 2016 1:28 pm

Benchmarks are the devil. Everyone loves them, not everyone can interpret them properly. Usually the results can't be compared to any real life scenario.

To get a snappy user experience in GEM, all you need to do is to accelerate filled rectangles and blits. Benchmarks will however tell you that your machine is still painfully slow, because pie charts (!) and justified text aren't accelerated.
Ain't no space like PeP-space.

User avatar
Eero Tamminen
Atari God
Atari God
Posts: 1949
Joined: Sun Jul 31, 2011 1:11 pm

Re: CPU speed discrepancies oddity

Postby Eero Tamminen » Sun May 01, 2016 6:20 pm

One possibility is that interrupts are interfering with the test in specific way. Either due to timing (which can be different from boot onwards), or machine configuration being different e.g. 50Hz vs 60Hz.

If performance benchmark is affected by timing of background things like interrupt handling, it's badly done. Either it should disable interrupts (preferable) or run long enough to average the interrupt handler load out.

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sun May 01, 2016 7:26 pm

One way to measure things like memspeed is to set interrupt level to MFP interrupts only, and then disable all MFP interrupts except for a timer. Then start the timer, and start the test. When the interrupt happens after X amount of time you can see how much work the CPU did or how much memory was shuffled around. Restore interrupts to normal. This way the test itself runs completely uninterrupted.

The CPU/mem-tests in GEMBench 4 runs with interrupts enabled, but I don't know about GEMBench 6. So these comments are about GEMBench 4.

GEMBench's CPU/mem-tests are not really usable. E.g. the ROM-speed results depends heavily on whether GEMBench runs from ST-RAM or alt-RAM, which suggests that this test doesn't measure ROM access at all. Also, CPU results are not correct, and the faster the CPU the more incorrect the results are. It's probably due to interrupts - as the CPU gets faster fewer interrupts occurs during the tests, and the interrupts that do occur runs faster.

I find GEMBench a usable tool to run AES/VDI-tests, but not really usable to benchmark pure CPU/FPU/memspeed.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Sun May 01, 2016 8:39 pm

Interrupts are disabled on GB6. d.m.l did some very cool timing code to make it ultra accurate. Also 16mhz speeds gave 100% results, If it was interrupts causing it, it would be random results all the time.

170% is as fast as ROM speed can be at 16mhz, as the test routines run in ST RAM at stock speed. Thats all explained in the docs. That has nothing to do with the issues I am seeing anyway.

I think people are missing my point a lot here. 16mhz speeds constant at 100% UNTIL the CPU is changed. I have ran the tests for almost 2 full days. I can get 100% scores running for endless hours. This rules out software issues as things only change when the CPU does. So will everyone please stop blaming the software! If people don't like benchmarking apps then dont use them!

I will explain one more time, with one CPU I can get 100% scores for hours on one hardware setup running 16mhz, I have rebooted many times to rule out any possible odd start up situations. Then if everything is kept the same other than the CPU, the results slow to 95%. If I put the other CPU back in, it goes to 100% and I can run for hours again at 100%. Put the other CPU in and back to 95% scores again.

The only possible thing could be the CPU misses DTACK and something is borderline timing somewhere. I've seen a lot of timings on the ST where even 2ns can cause issues. So tolerances on the CPU can possibly cause this issue. As the the exact mechanism to why this happens is yet unknown, hence this thread.
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sun May 01, 2016 8:44 pm

exxos wrote:I think people are missing my point a lot here. 16mhz speeds constant at 100% UNTIL the CPU is changed. I have ran the tests for almost 2 full days. I can get 100% scores running for endless hours. This rules out software issues as things only change when the CPU does. So will everyone please stop blaming the software! If people don't like benchmarking apps then dont use them!


If fullscreen-demos works with both CPU's then it is a software problem. If it doesn't work then it's a hardware issue, almost certainly related to the booster or perhaps a hardware problem with your particular test-machine.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Sun May 01, 2016 8:47 pm

joska wrote:If fullscreen-demos works with both CPU's then it is a software problem. If it doesn't work then it's a hardware issue, almost certainly related to the booster or perhaps a hardware problem with your particular test-machine.


Would a fullscreen demo show up really show up a couple of timer ticks drift over 10-30 seconds though ?
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Sun May 01, 2016 8:59 pm

"Timer tick drifts" has nothing to do with the CPU. The timers in the ST runs off the MFP. I find it unlikely that changing the CPU affects the timing of the MFP. I also find it unlikely that two different CPU's (of the same type) will run at different speeds in an unmodified ST. Running this demo will confirm that they behave 100% identically.

If there's a hardware problem then start looking at your 16MHz booster. Depending on the design it might even not be able to run this demo in 8MHz mode since some control signals goes through the extra logic.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

Zarchos
Captain Atari
Captain Atari
Posts: 263
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: CPU speed discrepancies oddity

Postby Zarchos » Sun May 01, 2016 9:00 pm

What's in the benchmark routines anyway ?
You've got some details of what and how 'this or that' is tested ?

Could you have strange things like testing arithmeticql copro ?
And then again from one series of 68000 at 16 Mhz to another the 'abort' because none is present doesn't take the same time ?
And in real use, with either 680000 #1 or 68000 #2 you won't see any difference because no games or demos used arithmetical copros on the ST.

Sorry the ST isn't my universe, just thinking aloud in case this could help.
Falcon, Atari 1040 STE, 520ST, 800xl, xegs, Amigas, Archimedes, RISC PCs, Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Russian clones, Sam Coupe, V6Z80P, QLs inc. Q68, and more !
2200 m2 museum on its way https://youtu.be/xjB6_Ez-3BA
Shorter video here : https://youtu.be/UEZisfkcN1Y

Currently porting SOTB to the Archie : https://www.youtube.com/user/Archimedes75009/featured

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Sun May 01, 2016 9:30 pm

joska wrote:"Timer tick drifts" has nothing to do with the CPU. The timers in the ST runs off the MFP. I find it unlikely that changing the CPU affects the timing of the MFP. I also find it unlikely that two different CPU's (of the same type) will run at different speeds in an unmodified ST. Running this demo will confirm that they behave 100% identically.

If there's a hardware problem then start looking at your 16MHz booster. Depending on the design it might even not be able to run this demo in 8MHz mode since some control signals goes through the extra logic.


I don't mean the CPU effects the timing on the MFP, I mean the CPU is taking longer to run the tests where the timer ticks increase.

I tried the closure demo, I just get 11 bombs when the first line of text comes up as it loads.
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

ijor
Hardware Guru
Hardware Guru
Posts: 3790
Joined: Sat May 29, 2004 7:52 pm
Contact:

Re: CPU speed discrepancies oddity

Postby ijor » Mon May 02, 2016 1:40 am

exxos wrote:Would a fullscreen demo show up really show up a couple of timer ticks drift over 10-30 seconds though ?


Yes. Some demos use the so called "kernel" mode. They are constantly tied in sync with the video electronic beam. A single video cycle drift and they would fail.

Now, video cycle is one thing, and timer tick is another. Since the MFP timers run off a separate crystal, there might be variations. That could explain a difference across multiple computers, or even across multiple runs in the same system. But not across two different CPUs in the same system, and certainly not by such a big (5%) margin.

The situation you describe is pretty weird. And if it wasn't because we are talking about an accelerated setup, it would be hard to imagine a reason ...

The only possible thing could be the CPU misses DTACK and something is borderline timing somewhere. I've seen a lot of timings on the ST where even 2ns can cause issues. So tolerances on the CPU can possibly cause this issue. As the the exact mechanism to why this happens is yet unknown, hence this thread.


That sounds as a reasonable explanation. If your accelerator is violating the setup and hold timing specifications of DTACK (or any other signal), then yes, it is expected than one CPU would show more tolerance than the other. It is not much different than trying to overclock a CPU beyond its nominal maximum frequency. Some would take it more than the other.

It doesn't matter that you aren't overclocking the actual clock frequency. Reducing setup margins have very similar requirements.

Of course, I have no idea if this is the actual reason of the behavior you are describing. But so far, it is the only reasonable one.

joska
Hardware Guru
Hardware Guru
Posts: 4376
Joined: Tue Oct 30, 2007 2:55 pm
Location: Florø, Norway
Contact:

Re: CPU speed discrepancies oddity

Postby joska » Mon May 02, 2016 7:30 am

exxos wrote:I tried the closure demo, I just get 11 bombs when the first line of text comes up as it loads.


So with the booster installed the ST is no longer 100% ST-compatible, even in 8MHz. Not the first accelerator to suffer from this, AFAIK this demo does not work correctly on the MSTE either.

It would be interesting if you could test another timing-sensitive demo as well, like We were @.
Jo Even

VanillaMiNT - Firebee - Falcon060 - Milan060 - Falcon040 - MIST - Mega ST - STM - STE - Amiga 600 - Sharp MZ700 - MSX - Amstrad CPC - C64

User avatar
troed
Atari God
Atari God
Posts: 1438
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: CPU speed discrepancies oddity

Postby troed » Mon May 02, 2016 7:56 am

exxos wrote:I tried the closure demo, I just get 11 bombs when the first line of text comes up as it loads.


There are some reports of such issues but none that I've been able to reproduce. Closure is extremely timing dependent (as an example, it finds out a jitter sequence dynamically at boot at then stays in sync every VBL throughout the whole demo without missing a beat. If a beat is missed the display will go haywire .. ) but I cannot guarantee that there's no single hw revision where even without modifications it couldn't fail ...

/Troed

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Mon May 02, 2016 8:28 am

joska wrote:
exxos wrote:I tried the closure demo, I just get 11 bombs when the first line of text comes up as it loads.


So with the booster installed the ST is no longer 100% ST-compatible, even in 8MHz. Not the first accelerator to suffer from this, AFAIK this demo does not work correctly on the MSTE either.

It would be interesting if you could test another timing-sensitive demo as well, like We were @.


The booster wasn't fitted, I took it out to try it on a stock machine first to see what it did before trying the booster.. I will try the other demo later..
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

User avatar
troed
Atari God
Atari God
Posts: 1438
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: CPU speed discrepancies oddity

Postby troed » Mon May 02, 2016 9:57 am

I wish someone would give _me_ a machine where Closure doesn't boot ;) A Mega STE is on its way, should already be at the post office ... ;)

I "only" have 2 STFMs, 1 Mega ST (8.01) and 2 STEs to test with myself :P

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Mon May 02, 2016 10:07 am

troed wrote:I wish someone would give _me_ a machine where Closure doesn't boot ;) A Mega STE is on its way, should already be at the post office ... ;)

I "only" have 2 STFMs, 1 Mega ST (8.01) and 2 STEs to test with myself :P


When I get more time I will test some more STF/M boards I have like 40 of them ;) My test machine where it doesn't boot, I will try swapping some chips around see if anything changes.

I have had issues with the GLUE where only 1 GLUE worked with 32mhz, I never found a second GLUE to work. So there are differences in the chips specs. Similar with the CPU, only 1 CPU gave 100% results, others seem more random/slower. So its working out what causes all this which is the problem..

I have had to change the 10K pullups to 2.2K because not all hardware combos work otherwise. So likely there could be some other signal which needs a better pullup for reliable operation.
4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

User avatar
exxos
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4933
Joined: Fri Mar 28, 2003 8:36 pm
Location: England
Contact:

Re: CPU speed discrepancies oddity

Postby exxos » Mon May 02, 2016 6:43 pm

For the disbelievers... Using the Closure demo :)

4MB STFM 1.44 FD- VELOCE+ 020 STE - Falcon 030 CT60 - Atari 2600 - Atari 7800 - Gigafile - SD Floppy Emulator - PeST - various clutter

http://www.exxoshost.co.uk/atari/ All my hardware guides - mods - games - STOS
http://www.exxoshost.co.uk/atari/last/storenew/ - All my hardware mods for sale - Please help support by making a purchase.
http://ataristeven.exxoshost.co.uk/Steem.htm Latest Steem Emulator

User avatar
troed
Atari God
Atari God
Posts: 1438
Joined: Mon Apr 30, 2012 6:20 pm
Location: Sweden

Re: CPU speed discrepancies oddity

Postby troed » Mon May 02, 2016 7:04 pm

First boot (first CPU) was in wakestate 2 - where I cannot detect in software which Shifter substate the machine is in so one code path will have "Every other 16 pixels black". Next boot (second CPU) was in wakestate 1, 3 or 4.

At least that's my guess. You can use my wakestate detection program to see this, as long as you only reset the machine it stays in the same wakestate.

If Closure lost video cycles the demo would look very unstable - not just black bands :)

/Troed


Social Media

     

Return to “Hardware”

Who is online

Users browsing this forum: No registered users and 4 guests