VOGONS


First post, by pentiumspeed

User metadata
Rank Oldbie
Rank
Oldbie

Subject says all?

Just wondering.

Cheers,

Great Northern aka Canada.

Reply 2 of 12, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
pentiumspeed wrote on 2020-04-13, 20:17:

Subject says all?

Just wondering.

Cheers,

Yeah, I was testing this several years ago. I managed to get a POD200MMX running at 1× multiplier and 16 MHz FSB on a Socket 5 board with working Turbo switch. Some discussion in the Setmul thread:

Re: SetMul - Multiplier control for VIA C3 / AMD K6 Mobile / Cyrix 5x86

Essentially the Pentium was running at 16 MHz and could use de-turbo to slow it down to very slow 286 speeds. Multi 1x kicks in when the on-chip fan is disabled that comes with the POD200MMX. You could rig a switch to toggle this behaviour.

However, this setup didn't go as slow as an IBM XT 4.77 MHz 8088, or slower IBM PCjr, achieved on my Socket 7 Turbo- switched boards, so I didn't pursue this endeavor any further...

Reply 3 of 12, by dionb

User metadata
Rank l33t
Rank
l33t

That's common to all the POD CPUs, I pulled the same trick on my PODP83 to get it running at 1x whatever I set the board at (16MHz was lowest iirc, but that was with 32b 486 bus)

Reply 4 of 12, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
dionb wrote on 2020-04-14, 21:36:

That's common to all the POD CPUs, I pulled the same trick on my PODP83 to get it running at 1x whatever I set the board at (16MHz was lowest iirc, but that was with 32b 486 bus)

Nice thing about this Socket 5 setup and POD200MMX was that it also ran stable at 3x multiplier and 80 MHz FSB, meaning that it had a clock range from 16- 240 MHz, and still had de-turbo, which made it very flexible with a smooth speed range.

Reply 5 of 12, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

I know this is working on PODP83

Are you sure this works on S4/S5 overdrives?

Last time I checked removing the fan was not changing mutliplier but applying a different sort of throttling. At least when measuring frequency by TSC

Blog|NexGen 586|S4

Reply 6 of 12, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
mpe wrote on 2020-04-14, 21:52:

I know this is working on PODP83

Are you sure this works on S4/S5 overdrives?

Last time I checked removing the fan was not changing mutliplier but applying a different sort of throttling. At least when measuring frequency by TSC

There seem to be a myriad of 'overdrives' out there. I've tested two PODs, on Socket 3 and Socket 5, POD83 and POD200MMX respectively. And they both ran at 1x multiplier for sure.

Reply 7 of 12, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

That would be way too convenient for my tests! Unfortunately, doesn't seem to be the case.

I've just checked this with my PODP5V133, PODP3V150 and PODMMT66x200.

Just as I remember it. Without fan PODP5V133 and PODP3V150 run at full speed and original multiplier (2x and 2.5x). And do not slow down at all.

The MMX overdrive throttles down to ~DX2-50 level (24k Dhrystones), but no change in frequency or multiplier as confirmed by TSC reading. Presumably it is injecting waitstates just as deturbo is doing (neither the turbo button actually changes the frequency).

Only the PODP83 slow downs to 33 MHz with no fan attached.

Last edited by mpe on 2020-04-14, 23:05. Edited 1 time in total.

Blog|NexGen 586|S4

Reply 9 of 12, by mpe

User metadata
Rank Oldbie
Rank
Oldbie
j^aws wrote on 2020-04-14, 23:04:

^^ What are you using to read the TSC? I usually use multiple tools to determine speed. Also, is your POD the same as my POD200MMX? There maybe variations.

By reading TSC counter using the software I've written. Basically you record number of CPU ticks at two intervals and compare with actual time elapsed. Also confirmed by cpuchk.

How do you measure the multiplier? 😀

BTW. MMX overdrive model numbers are PODPMT66X200, PODPMT66X166, PODPMT60X180 and PODPMT60X150. I only have the former two, but I assume they behave the same in this respect.

Blog|NexGen 586|S4

Reply 10 of 12, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
mpe wrote on 2020-04-14, 23:08:
By reading TSC counter using the software I've written. Basically you record number of CPU ticks at two intervals and compare wi […]
Show full quote
j^aws wrote on 2020-04-14, 23:04:

^^ What are you using to read the TSC? I usually use multiple tools to determine speed. Also, is your POD the same as my POD200MMX? There maybe variations.

By reading TSC counter using the software I've written. Basically you record number of CPU ticks at two intervals and compare with actual time elapsed. Also confirmed by cpuchk.

How do you measure the multiplier? 😀

BTW. MMX overdrive model numbers are PODPMT66X200, PODPMT66X166, PODPMT60X180 and PODPMT60X150. I only have the former two, but I assume they behave the same in this respect.

I use neither of those tools. Usually get multiplier confirmed by another DOS tool - 'CPU Identification Utility'.

Also, none of your PODs sound like the one I have. My model is named something like 'POD200MMX' and looks like a POD83. The names of your PODs are different.

Reply 11 of 12, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

That might be actually the problem. Many tools, especially older not specifically written for Pentium (rdtsc is Pentium only instruction) use different mechanism for estimating clock speed. Such as measuring duration of some instruction loops. These algorithms could be fooled by the throttling and report some random numbers. I have a plenty of tools saying my Pentium is 600 MHz 386 or something like that. But it doesn't mean that the CPU is actually running at that speed. With throttling you are facing the same issue the other way around.

Similarly even if you have workin turbo button, it doesn't actually change the frequency to 16 MHz. In this era it was injecting waitstates to simulate that speed. The actual oscilator or clock generator is likely working at fixed frequency and it is not easy to change it at run time. It was only after Pentium era when Intel came with Speedstep technlogy to actually change the frequency at runtime for power management reasons.

I believe there is no such model as "POD200MMX". The 4 models I mentioned above are the only Pentium MMX overdrives produced (although they might also have BPxxxx or BOXxxx prefixes depending on sale channel).

See http://www.cpu-world.com/CPUs/Pentium/TYPE-Pe … 0overdrive.html

Blog|NexGen 586|S4

Reply 12 of 12, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie

^^ 16 MHz was an undocumented FSB setting on the board, and before de-turbo was applied. I distinctly remember the 16 MHz number, and this was found by starting at 66 MHz FSB, then downclocking the FSB incrementally using the boards dipswitches - 66, 50, 40, 33, 25, 16. Then the fan was disconnected and clear multiples were seen.

De-turbo wasn't used to determine clocks and multipliers. Testing was similar to the POD83 as well. I don't have the parts at easy reach to retest, as they're buried in boxes somewhere, but they would be interesting to revisit.