What does it all mean? 🤔 I’ll let you decide.

Unless you found some way of altering the multiplier settings per CPU, you overclocked a Pentium 166MMX to 200MHz. Nothing more or less.

If you want asymmetrical multi-processing, you need to go for a chipset that facilitates it. Take a look at the SparcStation 20 for an extreme and yet easy example. You can run 3 CPUs at two different speeds on those things without the OS (Solaris) even batting an eyelid.

dionb wrote on 2022-01-09, 15:56:

What does it all mean? 🤔 I’ll let you decide.

Unless you found some way of altering the multiplier settings per CPU, you overclocked a Pentium 166MMX to 200MHz. Nothing more or less.

If you want asymmetrical multi-processing, you need to go for a chipset that facilitates it. Take a look at the SparcStation 20 for an extreme and yet easy example. You can run 3 CPUs at two different speeds on those things without the OS (Solaris) even batting an eyelid.

I think the second cpu may have been still running at 166mhz

Even though it said 200 in cpuz the results in super pi suggest a larger difference than if they were both running at 200.

The 166 mmx won’t go over 2.5x multiplier through the bf0,1,2 pins so if it is going at 200 then how did that happen, and if it isn’t that explains the lower super pi score

I've done the same thing with multiplier locked P2s. It is certainly possible to run them at different speeds in dual

Ha, that's interesting.
I could kind of understand why intel wouldn't want people to know about this, since they would lose money with buyers opting for a slower second CPU. But if that were the case, you'd think they try to disable this feature completely.

So what's the deal with matching s-specs and batches? I always remember that being really important for SMP setups...but apparently mismatched CPUs work together just fine?

Honestly, IMO all that matching crap is just a meme. In practice I would keep speeds and cache sizes the same but not worry about anything else

Anonymous Coward wrote on 2022-01-10, 01:22:

Ha, that's interesting.
I could kind of understand why intel wouldn't want people to know about this, since they would lose money with buyers opting for a slower second CPU. But if that were the case, you'd think they try to disable this feature completely.

So what's the deal with matching s-specs and batches? I always remember that being really important for SMP setups...but apparently mismatched CPUs work together just fine?

🤷‍♂️

Maybe so you get consistent performance between cores?

I suppose one reason would be to avoid a p54 and a p55 being in the same machine

Also cpus of different voltages.

But as we all know about any p55c can take 2.8v even if they list 2.2 on the package, just like at 166 I bet a 233 could take 2.2v

I was thinking about trying to do a p54c and a p55c earlier but decided I liked my motherboard.

I bet at 2.8v both chips would probably work at 133 but that’s getting pretty sketch and didn’t want to break anything.

Symmetric vs asymmetric doesn't mean same CPUs but what the CPUs have access to. Asymmetric multiprocessing machine can have one CPU be able to access things that other(s) cannot while symmetric systems have all the CPUs having equal access to all of the resources.

Tiido wrote on 2022-01-10, 04:28:

Symmetric vs asymmetric doesn't mean same CPUs but what the CPUs have access to. Asymmetric multiprocessing machine can have one CPU be able to access things that other(s) cannot while symmetric systems have all the CPUs having equal access to all of the resources.

Asymmetrically clocked? 🤔