it's controlled by the motherboard and cannot be altered

It's down to what the BIOS sets as the boot CPU so you would need a BIOS setup option to select which CPU get's booted.

From your post I assume you have a system with different speed CPUs and I am wondering if my SIV utility correctly deals with this, please will you try it and post the two Menu->File->Save Local files so I can check? It should do OK, but the last time I recall checking was on 22-Feb-2006 on a Gigabyte 7DPXDW-P system with an AMD Athlon @ 2.00 GHz + a Duron @ 1.60 GHz.

Which motherboard allows you to do run 2 CPU's at different speeds?
All motherboards I'm aware of may identity the CPU but will only run at the speed of the slowest CPU

But yes, check the motherboard manual and you'll see the sockets are numbered 1 and 2 with #1 always been the default CPU.
Just as you cant run the system with just a CPU in the 2nd socket and socket 1 empty.

As I remember from my Abit BP6 days, If you have CPUs with different multipliers then they will run at different speeds. I guess some boards might be confused by that and report one speed only.

In a SMP system, usually one CPU is "boot". It is the one that handles legacy interrupts (once APIC is activated) and starts once the computer is power on. The other one gets active only once you boot into SMP-enabled operating system. Usually there is a socket on the motherboard that is marked as "boot" "primary" or "Socket 0" and the secondary socket accordingly.

Thanks for the replies guys.

The board is an S2DG2, manual here https://ibm.retropc.se/manuals/Supermicro/Sup … 20REV%201.3.pdf
- It definately works with a single CPU in either slot with no terminator (Slot 2 or 440GX advantage maybe)
- The CPU slots are marked CPU0 and CPU1, actually the CPU1 seems to be the boot CPU when both are installed
- The CPUs run at the designated different speeds
Quite surprised how it does all of the above without a fuss.

I think that's right about the Boot CPU and no software options. I was reading here https://wiki.osdev.org/Symmetric_Multiprocessing and the Intel Spec specified in the article https://github.com/x86-8/x86-8-docs/blob/mast … ecification.pdf.

The SMP boards build a multiprocessor configuration table during hardware detection, I guess either the last or first of the CPUs in the table is nominated for Boot CPU. Edit, found that there is a CPU flag called BP for Bootstrap processor in the above document

So maybe I could have a backup bios savior device with a switch and somehow edit the bios but it looks like a bit of a challenge to modify the Bootstrap processor flags.
Or perhaps insulate a CPU pin with a switch that prevents it's detection.

red-ray wrote on 2020-10-15, 14:15:

From your post I assume you have a system with different speed CPUs and I am wondering if my SIV utility correctly deals with this, please will you try it and post the two Menu->File->Save Local files so I can check?

Sure I can do this at weekend, does the OS matter between XP/2000/NT?

It definitely works with a single CPU in either slot with no terminator (Slot 2 or 440GX advantage maybe)

you can also run scsi drives w/o a terminator. SOMETIMES. Same thing applies here. A terminator will benefit system stability greatly.

The CPU slots are marked CPU0 and CPU1, actually the CPU1 seems to be the boot CPU when both are installed

It's a quirk of Supermicro boards. at least of that era. In my S2QR6 cpu-4 is primary.

The CPUs run at the designated different speeds

&

I remember from my Abit BP6 days, If you have CPUs with different multipliers then they will run at different speeds. I guess some boards might be confused by that and report one speed only.

NO. No, they don't. If you have mismatched CPU's it will run at the slower speed. Period. End of discussion. It's a fracking miracle it even posts with mismatched processors like that. NT/2K will complain about mismatched processors and might not be stable. XP will NOT boot unless the processors are PERFECTLY matching.

It goes against what many thought about SMP but sorry some of that is wrong.

you can also run scsi drives w/o a terminator. SOMETIMES. Same thing applies here. A terminator will benefit system stability greatly.

It's in the manual: "Question:Do I need to change any settings to use a single processor ona dual processor board?Answer: There are no jumpers or BIOS settings that need to be changed whenrunning a single CPU on a dual processor board. Also, you can use a singleprocessor in either slot."

It's a quirk of Supermicro boards. at least of that era. In my S2QR6 cpu-4 is primary.

Cool good to know and that is a mother of motherboards 😁

NO. No, they don't. If you have mismatched CPU's it will run at the slower speed. Period. End of discussion. It's a fracking miracle it even posts with mismatched processors like that. NT/2K will complain about mismatched processors and might not be stable. XP will NOT boot unless the processors are PERFECTLY matching.

Impossible for the software to influence the multipliers, it's set via jumpers or on the CPU if they are locked.

It did not complain for me but I didn't flex it much with benchmarks so I'll have to prove it with screenshots probably at the weekend. I am sure it has been done with weird combos for Athlon MP systems quite a bit, like the example from red-ray.

Just a note I edited my response earlier, the answer to my opening post is:
The Boostrap processor flag created when populating the multiprocessor table configuration at hardware detection.

Apparently it's process is called MULTIPLE-PROCESSOR (MP) INITIALIZATION PROTOCOL http://datasheets.chipdb.org/Intel/x86/Intel% … re/24319202.pdf Page 266

Hanamichi wrote:

Impossible for the software to influence the multipliers, it's set via jumpers or on the CPU if they are locked.

It did not complain for me but I didn't flex it much with benchmarks so I'll have to prove it with screenshots probably at the weekend. I am sure it has been done with weird combos for Athlon MP systems quite a bit, like the example from red-ray.

True, software can't control the cpu multi. but on the p2/p3 the multiplier might be locked, but it can be set LOWER than spec by the bios. I've done a lot of benchmarking with a top end p2/3 and setting it to lower speeds for comparison testing.

I am surprised if any board or OS can function normally with multiple CPUs at different speeds. To me, it sounds like a car with left and right wheels at different speeds, they would be out of synch. It would involve some sort of overhead to make it work. Not saying it's impossible, just doesn't fit with the logic as I see it.

Oh, I'm 110% curious too. I mean, everything I know about smp comes into question. So i'm anxiously awaiting proof. I'll even go test it on my boards if its possible. Mostly as an academic exercise, as I almost always put a matching set of the FASTEST processors possible. (I run 4x 900's in that previously mentioned S2QR6)

debs3759 wrote on 2020-10-15, 18:38:

I am surprised if any board or OS can function normally with multiple CPUs at different speeds. To me, it sounds like a car with left and right wheels at different speeds, they would be out of synch. It would involve some sort of overhead to make it work. Not saying it's impossible, just doesn't fit with the logic as I see it.

I know it sounds odd but I'm pretty sure modern architecure multi-core CPUs are still SMP technically so I guess you are not web browsing with all cores burning at max frequency.

luckybob wrote on 2020-10-15, 18:44:

Oh, I'm 110% curious too. I mean, everything I know about smp comes into question. So i'm anxiously awaiting proof. I'll even go test it on my boards if its possible. Mostly as an academic exercise, as I almost always put a matching set of the FASTEST processors possible. (I run 4x 900's in that previously mentioned S2QR6)

I had a hunch but I was suprised it didn't really throw wobblies at all but hey I may have been have been not thorough when I had the system assembled last.

I mean if NT/2000/XP can run on a Core 2 or newer with speedstep reducing frequencies on multiple cores why not a mismatched dual/quad socket SMP system just as well? I think it is more the motherboard and chipset limitation not an OS one.

That's true about some downclocking on P2s, I actually wanted setmul to work on this board but it does nothing.

luckybob wrote on 2020-10-15, 18:36:

True, software can't control the cpu multi. but on the p2/p3 the multiplier might be locked, but it can be set LOWER than spec by the bios. I've done a lot of benchmarking with a top end p2/3 and setting it to lower speeds for comparison testing.

Wasn't all p3 locked, and not only limited? Or are we talking engineering samples?

I hope I remember it right as was a long time ago, but I believe at least on my BP6 running mistmatched CPUs was possible. Could anyone with a working BP6 confirm?

CPUs are synchronised using the same front side bus (CPUs shared the same bus in PII era). Their internal speed doesn't have to be the same. I don't think there is any mechanism to negotiate multipliers settings in a SMP system.

Of course it could create issues. The advice back then was that you should be using matching pair. That means matching speed and even the same stepping. However, even running 2 Celerons was a complete hack.

I even remember reading stories about people who were running Slot1 Celeron and PII in one system....

H3nrik V! wrote on 2020-10-15, 19:03:

luckybob wrote on 2020-10-15, 18:36:

True, software can't control the cpu multi. but on the p2/p3 the multiplier might be locked, but it can be set LOWER than spec by the bios. I've done a lot of benchmarking with a top end p2/3 and setting it to lower speeds for comparison testing.

Wasn't all p3 locked, and not only limited? Or are we talking engineering samples?

Mostly I think, some older P2s are unlocked.

Xeon PIIIs 700Mhz and lower can have their multiplier set with jumpers. 900Mhz is locked as it is above the original motherboards specification.

mpe wrote on 2020-10-15, 19:10:

I hope I remember it right as was a long time ago, but I believe at least on my BP6 running mistmatched CPUs was possible. Coul […]
Show full quote

I hope I remember it right as was a long time ago, but I believe at least on my BP6 running mistmatched CPUs was possible. Could anyone with working BP6 confirm?

CPUs are synchronised using the same front side bus (CPUs shared the same bus in PII era). Their internal speed doesn't have to be the same. I don't think there is any mechanism to negotiate multipliers settings in a SMP system.

Of course it could create issues. The advice back then was that you should be using matching pair. That means matching speed and even the same stepping. However, even running 2 Celerons was a complete hack.

I even remember reading stories about people who were running Slot1 Celeron and PII in one system....

Exactly right I think mpe 😁 FSB must be the same
https://arstechnica.com/civis/viewtopic.php?f=8&t=359164 (well it's an ASUS but same)

Actually now I rememeber even my cache is different 1MB and 2MB but somewhere I remember reading a post that said 512KB ones could not be mixed but who knows.

Same kind of thing:
Re: Dual Katmai PIII (Warning: Photos are a bit big)

I think the quick and dirty test here would be running win2k, and multiple threads of something like super pi. if the cpus are running at the same speed, the calculation times will be very similar (if not exactly equal), if a 700 is running with a 900 (for example), then the differences in times should be roughly equivalent to their respective speeds.

None of this will change how I setup a multi-cpu system. I'm still going to run 2x 1.4's in my dual taualatin, 4x 900s in my s2qr6, and 200/1mb ppro's in my ALR setups. I'm just waiting to see if this works.

Hanamichi wrote on 2020-10-15, 16:30:

red-ray wrote on 2020-10-15, 14:15:

From your post I assume you have a system with different speed CPUs and I am wondering if my SIV utility correctly deals with this, please will you try it and post the two Menu->File->Save Local files so I can check?

Sure I can do this at weekend, does the OS matter between XP/2000/NT?

Thank you, the OS does not really matter, but if the system is multi-boot my ordered preference is 2003 -> XP -> 2000 -> NT4. I prefer the newer OSes as they tend to have more recent CPU µCode.

The Menu->Machine->CPU Benchmark panel will tell you how the CPU speeds compare.