huh that's a cool case!

Was a Cyrix 6x86 166 MHz originally when I got it. 😁

More pictures.

Mainboard:

Dallas RTC socket:

That looks like a very clean and stock system. I dig it, especially all those period correct components. The case looks fresh too, was it new old stock or just kept in good condition over the years?

Nothing NOS, I just gave all parts a good clean. The case was a bit dirty when I got it but was in great condition. 😀

Nice build, and nice case.

derSammler wrote:

The turbo switch overclocks the CPU to 90 MHz with the display set up accordingly.

How did you achieve that ? is that a feature of the Plato mainboard ?

There's only a single jumper to change the FSB (ignoring 100 MHz, which this revision does not support). That's where I connected the turbo switch. Works even while the system is running.

Real turbo mode exists as well, but that is triggered by a hot-key.

What's performance like when you de-turbo (using the hotkey)?

Very nice. You have any more interior pics?

gdjacobs wrote:

What's performance like when you de-turbo (using the hotkey)?

Did not use that feature yet, also no benchmark tools installed on it so far. Will test that later.

wirerogue wrote:

You have any more interior pics?

I have to make some more. I somehow missed making pictures when the hardware was done and the case still open. 🤣

But I have to take out the CF card for backing it up anyway, so pictures of the interior will be made.

gdjacobs wrote:

What's performance like when you de-turbo (using the hotkey)?

On a P5-90 the deturbo mode scores around 68 on Topbench, which is in the neighborhood of a higher-end 386. In practice this is good for older 386-era games but still too fast for anything that really wants that XT/AT timing.

derSammler wrote:

There's only a single jumper to change the FSB (ignoring 100 MHz, which this revision does not support). That's where I connected the turbo switch. Works even while the system is running.

That's clever. I would not have thought that FSB could be changed on the fly (at least w/o fine tuning), but thinking twice the turbo on my 386 does simply do FSB/2 and can be changed dynamically.

Finished software installation today and pulled the CF card for a backup. And made pictures during that, of course. 😉

I approve of this build, and extra points for going with a gold cap Pentium. I think not too many people build with a P75. The only thing I would change to make it more period correct is going with a Mach64 GX or a S3 964. The Mach64 GT is more of a 1996 thing.
I'm trying to put together something similar, only I want a Neptune board with dual P100 gold caps. I have everything except the board.

It's a Mach64 CT actually, did not notice that typo. Corrected. 😀

JudgeMonroe wrote:

gdjacobs wrote:

What's performance like when you de-turbo (using the hotkey)?

On a P5-90 the deturbo mode scores around 68 on Topbench, which is in the neighborhood of a higher-end 386. In practice this is good for older 386-era games but still too fast for anything that really wants that XT/AT timing.

Hmm... I wonder how that stacks with cache tricks.

gdjacobs wrote:

JudgeMonroe wrote:

gdjacobs wrote:

What's performance like when you de-turbo (using the hotkey)?

On a P5-90 the deturbo mode scores around 68 on Topbench, which is in the neighborhood of a higher-end 386. In practice this is good for older 386-era games but still too fast for anything that really wants that XT/AT timing.

Hmm... I wonder how that stacks with cache tricks.

It doesn't. The deturbo mode on this board *is* (L2) cache tricks.

Ah, damn. Disappointing.

gdjacobs wrote:

Ah, damn. Disappointing.

The BIOS has separate settings for "Boot Speed" (Turbo or De-Turbo) and "System Cache" (Enabled or Disabled) but in my testing, either and both of these settings result in identical benchmarks, as does the deturbo hotkey. The documentation describes Deturbo specifically in terms of disabling the cache and "adding increased refresh cycles" but I didn't see any evidence of those extra cycles in the benchmark.

Sounds like they added some memory latency.