Miss-matched CPUs, dual Athlon MP board

[imported_miahallen]

I know most of the dual K7 boards will support the XP, but will they support miss-matched XPs? I have two CPUs, and old XP1600+ (180nm Palamino core, 1.4GHz @ 133MHz FSB), and my current system runs an XP 2500+ (130nm Batron core, 1.83GHz @ 166MHz FSB). I'm planning a new A64 system and want to utilize these for a server.

Will any mobos handle it? If they would, could I run them at their individual speeds/memory timings, or would I have to slow the 2500+ down to the 1600+ speeds? What mobo would you recommend if this would work?

Thanks in advance for any assistance.

 

[Peter]

Theoretically, this would work, since the Athlon-MP architecture has the CPUs on separate busses. However, even if the board's BIOS lets you do this, results in the operating system will be somewhere between surprising and messy - simply because the timekeeping and scheduling in all usual OSes assumes same-speed processors.

 

[Concillian]

If your barton is new-ish, then you won't be able to close the "MP" bridge that allows the processors to work in dualies. Then you have the whole mis-matched cores issue as well as mis-matched voltages that the board probably won't like.

I think your best bet is to try to trade your barton for another Palomino 1600+ (plus a few bucks). You should probably get a few bites on FS/FT. Given your barton is probably superlocked, it would run 1466 anyway (11x133) since the dualie boards don't support 166MHz or 200MHz FSB.

Be aware that if the MP bridge is not already closed on the 1600+ you will need to close it.

 

[aka1nas]

Concillian, the L5 bridges are not super locked on newer Bartons I believe. Mobile Bartons for sure will not have this problem. The above config will not work stably as the two chips do not have the same amount of cache(pretty sure the caches of the chips need to be syncronized).

 

[Peter]

The caches aren't synchronized, they're kept coherent through the front side bus protocol. The two CPU's caches can work in completely different ways, as long as the two cores negotiate properly so that none of the caches contain stale data.

In other words, this wouldn't matter either. Neither would the voltages - regulation for the two CPUs is, and must be, entirely separate.

See above for why things will go wrong anyway.

 

[imported_miahallen]

Peter, do you agree that I may have to "close the MP bridge" on either of these procs? Both processors' multipliers ARE locked.

Given your barton is probably superlocked, it would run 1466 anyway (11x133) since the dualie boards don't support 166MHz or 200MHz FSB.

results in the operating system will be somewhere between surprising and messy - simply because the timekeeping and scheduling in all usual OSes assumes same-speed processors.

So my 1600+ would run at it's stock 1.4GHz and the 2500+ would run at 1.466GHz, is that close enough that it wouldn't create a "mess" in the OS?

 

[Peter]

SAME speed is required.

 

[imported_miahallen]

Alright, thanks for the help