LC520 Overclock

[JDW]

The logical guess is that it's related to the clock control ...

Oh yes, I know.
But we don't know this:

1. Are all 3 resistors (R67/75/79) truly used by the clock circuit on the LC520 board?
2. To what exact chips and pins do each of the 3 resistors lead?

So we are all just "assuming" the 3 need to be removed at this point. Same for folks in Japan back in 1997 when Doping Mac was released. They simply read the book saying that 4 resistors (R80 included) needed to be removed, and they didn't investigate further and simply did that. Seems to work, so no one looked deeper at it.

For me personally, I am filming an LC520 video right now and need to figure out what I will say on the subject. I will point people back to this thread, and mention what Doping Mac says, while also saying that removing the 3 resistors allow SHOULD WORK FINE, but I cannot offer hard evidence to prove it only because we don't have a schematic and don't know where R67/75/79 lead.

 

[dougg3]

Based on looking at the 520/550 board, it looks like an equivalent to R75 doesn't exist. But thank you for confirming the general suspicion!

Happy to help, it's a fun little puzzle. Here's a little mini deep dive further into it. I thought about it some more, and I think I have a reasonable guess about why they got rid of it on the 520/550 board. Looking at the CC schematic, populating only R75 would keep VPP at 12V at all times, just like they said all those other machines did.

But...if you look at the signal that's going to the /WE pin of the ROM SIMM socket on the CC, it's IOW* which is also the write enable for the 85C80 Serial/SCSI combo chip. I bet the 520/550 is the same way.

This means that any time you are writing to the Serial/SCSI chip, you are also sending a write cycle to the ROM SIMM socket at the same time. Keeping VPP at 5V during normal operation forces the flash chips on the ROM SIMM to ignore those write cycles, thus preventing serial/SCSI traffic from unintentionally reprogramming the ROM.

I think that's the reasoning behind this whole "controllable VPP" circuit. Because the write enable line is shared with SCSI/serial, you only put 12V onto VPP when you want to reprogram the ROM SIMM. Then you make sure you don't do any SCSI/serial stuff until you're done writing the SIMM and VPP is back to 5V.

I bet with all that in mind, they realized it would be error-prone to populate R75 rather than the more complicated circuit, and stripped it out going forward.

 

[kefkafloyd]

Oh yes, I know.
But we don't know this:

1. Are all 3 resistors (R67/75/79) truly used by the clock circuit on the LC520 board?
2. To what exact chips and pins do each of the 3 resistors lead?

So we are all just "assuming" the 3 need to be removed at this point. Same for folks in Japan back in 1997 when Doping Mac was released. They simply read the book saying that 4 resistors (R80 included) needed to be removed, and they didn't investigate further and simply did that. Seems to work, so no one looked deeper at it.

For me personally, I am filming an LC520 video right now and need to figure out what I will say on the subject. I will point people back to this thread, and mention what Doping Mac says, while also saying that removing the 3 resistors allow SHOULD WORK FINE, but I cannot offer hard evidence to prove it only because we don't have a schematic and don't know where R67/75/79 lead.

Right, what I meant was that the LC III's behavior could be used as a basis to investigate further on the LC 520/550 (like the Color Classic was for this circuit). There are at least third party schematics for the LC III; as well as the LC III reloaded project. They could be used as a starting point to start toning things out.

Happy to help, it's a fun little puzzle. Here's a little mini deep dive further into it. I thought about it some more, and I think I have a reasonable guess about why they got rid of it on the 520/550 board. Looking at the CC schematic, populating only R75 would keep VPP at 12V at all times, just like they said all those other machines did.

But...if you look at the signal that's going to the /WE pin of the ROM SIMM socket on the CC, it's IOW* which is also the write enable for the 85C80 Serial/SCSI combo chip. I bet the 520/550 is the same way.

This means that any time you are writing to the Serial/SCSI chip, you are also sending a write cycle to the ROM SIMM socket at the same time. Keeping VPP at 5V during normal operation forces the flash chips on the ROM SIMM to ignore those write cycles, thus preventing serial/SCSI traffic from unintentionally reprogramming the ROM.

I think that's the reasoning behind this whole "controllable VPP" circuit. Because the write enable line is shared with SCSI/serial, you only put 12V onto VPP when you want to reprogram the ROM SIMM. Then you make sure you don't do any SCSI/serial stuff until you're done writing the SIMM and VPP is back to 5V.

I bet with all that in mind, they realized it would be error-prone to populate R75 rather than the more complicated circuit, and stripped it out going forward.

Excellent theory. We also know this ROM programming feature was refined further in later 68040 machines.

 

[JDW]

@dougg3 , @Fizzbinn , and you other LC550 owners...

Sorry for the trouble, but could you please post your LC550 board benchmarks in the following apps?

• Norton System Info 3.5.3
• Norton System Info 3.2.1
• Speedometer 4.0
• Speedometer 3.0.6
• MacBench 3.0 (all tests)

It could be that no single one of you have run benchmarks in all the apps above, but perhaps a combination of you might have them all.

My LC520 Recap and Overclock is done, and I've run my own benchmarks in the above apps. I did that before the Overclock too. I simply don't have any LC550 benchmark results with which to compare. My board's results should be exactly the same as the 550 board results, but I'd like to compare.

Interestingly, my FPU is only rated for 16MHz, but it overclocks to 33MHz just fine. But I've heard that the 68882 overclocks nicely like that. Very glad I didn't have to buy a new one. That FPU was actually in the stock Color Classic board, and I extracted it for the LC520, since an FPU makes more sense in the faster board.

Prior to the Overclock, the About this Macintosh in System 7.1 mentioned it was an LC520, but after my Overclock, it just says Macintosh. Do you 550 board owners find that to be true? (My board is inside a Color Classic.) Even so, TattleTech is calling it an LC550, so it's clearly smarter than System 7.1.

I have filmed everything, and tomorrow night after work will be my filming of the ending segment. I still need to construct some screencasts and voiceovers, but that will need to wait for the weekend. Hopefully, I get can the video released next week, then send my GH5 camera off for maintenance. Bought it back in 2017, and now the rubber parts are starting to come off.

Thanks!

 

[kefkafloyd]

Prior to the Overclock, the About this Macintosh in System 7.1 mentioned it was an LC520, but after my Overclock, it just says Macintosh. Do you 550 board owners find that to be true? (My board is inside a Color Classic.) Even so, TattleTech is calling it an LC550, so it's clearly smarter than System 7.1.

That's expected for About This Macintosh. This is because in System Enabler 403 the model names for the gestalt values for the 550 (and CC II) are blank, and in that case About This Mac will simply say "Macintosh" (even though it will display the proper icon for the chassis the board is installed in). It was likely never updated for their model numbers (see earlier post). If you want the proper name to show up, edit the resource values in the Enabler file.

My 550 board says "MacintosH" in my Hook chassis (with a Hook icon) and in my Color Classic (with a CC icon) with the stock enabler.

I can try running some benchmarks on my LC 550 board in my CC.

 

[Fizzbinn]

That's expected for About This Macintosh. This is because in System Enabler 403 the model names for the gestalt values for the 550 (and CC II) are blank, and in that case About This Mac will simply say "Macintosh" (even though it will display the proper icon for the chassis the board is installed in). It was likely never updated for their model numbers (see earlier post). If you want the proper name to show up, edit the resource values in the Enabler file.

My 550 board says "MacintosH" in my Hook chassis (with a Hook icon) and in my Color Classic (with a CC icon) with the stock enabler.

I can try running some benchmarks on my LC 550 board in my CC.

Is it possible your not using the last version of System Enabler 403?

System Enabler 403

------------------

1.0 First release for Macintosh LC 520.

1.0.1 Manufacturing release only.

1.0.2 Added support for Macintosh LC 550.

Source:

System Enablers: Current Versions & Change History (12/1994)

Apple Computer, Inc. -- Technical Information Retrieval System Document Retrieved: /wais/TIL/Macintosh Software/System 7 Information/System 7.1/Sys Enablers-Versions - History (Note: This article was extracted from the Apple Tech Info Library, Apple's official technical support database. Look...

wiki.preterhuman.net

Looks like you can get version 1.0.2 here if not:

Macintosh System Enablers for OS 7.1 to 7.5.5: Current Versions and Change History – Steve's Blog

 

[kefkafloyd]

Is it possible your not using the last version of System Enabler 403?

System Enabler 403

------------------

1.0 First release for Macintosh LC 520.

1.0.1 Manufacturing release only.

1.0.2 Added support for Macintosh LC 550.

Source:

System Enablers: Current Versions & Change History (12/1994)

Apple Computer, Inc. -- Technical Information Retrieval System Document Retrieved: /wais/TIL/Macintosh Software/System 7 Information/System 7.1/Sys Enablers-Versions - History (Note: This article was extracted from the Apple Tech Info Library, Apple's official technical support database. Look...

wiki.preterhuman.net

Looks like you can get version 1.0.2 here if not:

Macintosh System Enablers for OS 7.1 to 7.5.5: Current Versions and Change History – Steve's Blog

My enabler's version string says 1.0.2, it was installed from the LC 550 image from the Apple Legacy Recovery CD-ROM. Every enabler I've downloaded from multiple sources labeled 1.0.2 (the garden, MacGUI, old FTPs, etc) has blank strings.

While I can't 100% definitively say there's no 1.0.2 enabler shipped from Apple that had the strings populated, I haven't found one after extensive searching across multiple sources and system install sets.

There are other enablers which added support for later systems and didn't include a string for the gestalt value, so it's not unheard of to me.

 

[kefkafloyd]

@JDW Here's a stuff-it file of test result files from benchmarks I ran on my CC with my LC 550/CC II board. I pulled them off my disk image via Infinite Mac. I skipped the disk benchmarks on MacBench because they took forever and aren't relevant to CPU performance. Plus, MacBench crashed after the first complete run and I was not about to do it again.

 

[JDW]

@kefkafloyd
Thank you. I downloaded your file and opened the Speedometer folder and I see Records "3 & 4" but do you which specific version of each?

As you can see below, I have my own 3.06 and 4.01 folders (top left). I believe the creator of Speedometer changed this testing methodology somewhere during version 3. I can't remember if 3.06 results differed from 3.23 or if it was another version, so anyway, that's why I ask.

I went ahead and merged your "Mac Color Classic II" (LC550 board) record with my main Machine Records file for v3.06. Here's the comparison, with my overclocked and recapped LC520 board shown at left...

You're getting more "Graf" speed than my board, enough to where it doesn't seem to be regular testing error. Disk is faster too, but that is conceivable because I used a spinner drive and you might have used something else. Math results differ by what I would consider to be "normal testing error". Here are the exact percentage differences between our results:

• Graf: 9.37 & 9.78 → 4.33% difference (testing error difference?)
• Disk: 5.11 & 5.81 → 13.69% difference (are you using BlueSCSI?)
• Math: 31.29 & 31.05 → 0.77% difference (definitely a testing error difference)

My LC520 board has the 256K VRAM SIMM installed, with 32MB RAM SIMM (for 36MB total). The combined 768K of VRAM (Soldered + SIMM) yields 16-bit color capability, but I had the monitor set to 256 colors at the time of testing. Disk Cache was set to 128K, because that is what Norton System Info prefers.

The reason I was seeking benchmark results from a tru LC550 board (CCII) is because some people told me the LC520 overclock boosts BUS SPEED too, but others said it does not. Benchmarks don't lie, so I am looking at differences to see what the truth is.

 

[JDW]

Hmmm...

I just did the side-by-side with your Speedometer 4 benchmark results (shown at right below — I renamed your record), and my overclocked LC520 board at left. All numbers (except Disk) are clearly within "testing error" difference. So I guess this gives strong evidence that the BUS SPEED is very likely overclocked along with the CPU no my LC520 board...

And here is the Norton System Info 3.2.1 comparisons (showing only CPU/Video/FPU because your Disk scores were missing)...

 

[JDW]

And here are the MacBench 3.0 results, showing that CPU/FPU are essentially the same for your board and mine, but your board still shows a bit higher on the Graphics...

Even so, if you consider how much lower the Graphics scores are with the 25MHz LC520 board, I don't think it can be said that the BUS SPEED is not being accelerated with the LC520 overclock to 33MHz. Your graphics scores are a bit higher in all of the benchmark apps except Speedometer 4, where they are identical. Hmmm...

Anyway, thank you very much for helping me do this speed comparison!

 

[kefkafloyd]

Sorry for not clarifying on Speedometer, the versions I used were Speedometer 3.2.3 and 4.0.2 (and I labeled them 3 and 4 as you surmised).

I also have a 33MHz FPU, your overclock on your 16MHz one matches perfectly. And yes, I was using an external BlueSCSI v2 as my host drive on the latest firmware.

Consider that my graphics scores were probably higher because I was running at the CC's native 512x384 60Hz resolution, and you're running at 640x480 at a higher refresh rate. I have the VRAM SIMM installed but 768K VRAM probably doesn't help on a 512x384 screen. I figure both of our machines have maximum standard RAM (36 MB).

 

[JDW]

I did not use Speedometer 3.23, unfortunately. When checking the Speedometer 3 page on Macintosh Garden, it says this:

Note: Algorithm changes occurred at v3.01 and v3.1; so absolute benchmark results across those version boundaries are not strictly comparable.

That implies that the 3.06 version I used should differ from the 3.32 version you used, and yet, the results are quite "comparable," so maybe that statement on the Garden is in error. I do know versions 3 and 4 differ in their methodology. I just can't remember which 3.x versions mark the change.

Thanks also for clarification on screen resolution differences! That explains the numbers perfectly!

 

[kefkafloyd]

No worries, I'll rerun it with 3.06.

 

[JDW]

Thank you for your kindness, but you may not need to re-run it in v3.06.

I just downloaded v3.23 and checked its version history, which indicates the only change impacting 040 processors (not the 030 like you and I are testing)...

This also means none of us should use Speedometer versions lower than 3.1 when testing any 040 machine.

 

[kefkafloyd]

Here's a record for 3.0.6.

Here's a screengrab of the results. The 3.2.3 run is on the left, 3.0.6 on the right. Clearly some difference between the two, so use the 3.0.6 one for your comparisons (or do a run in 3.2.3 to see if you can reproduce it).

 

[JDW]

I just released the full LC520 recap and overclock walkthrough video...