Dual socket 1366 mb able to run two different cpus?

A

acesea

Limp Gawd
Joined
Oct 2, 2004
Messages
211
For example, an e5520 and an l5518? I have supermicro 5500 and 5520 motherboards and am uncertain if I need to have identical cpus.
 
Identical, no. Nearly identical, yes. You can run mismatched CPUs as long as everything but the clock speed is identical in 1366 boards (at which point, the faster CPU will run at the same speed as the slower one). A E5520 and L5518 will not work together.
 
acesea said:
Thanks. I may have spares so I'll do some testing.
Click to expand...

Some boards will let you run CPUs of a different clock rate, in fact some can even run CPUs of different voltages too (Depends on VRM config and I've never seen it officially supported). Quad and Hex core combined works on some boards too.

Also, although Windows is generally OK with different clock rates, some SMP aware software goes a tad awry (My experience with this dates back to running P3s at different clock rates).
 
Stability will almost always be severely affected when using different processors. Hardcore multi-socket folk will often even swear by the fact that you need "matched pair" chips, (processors cast from the same die) in order to maximize stability. You might be able to use different processors, (probably not on those Supermicro boards) but it may not run as well as you'd hoped/at all.
 
alienb21 said:
Stability will almost always be severely affected when using different processors. Hardcore multi-socket folk will often even swear by the fact that you need "matched pair" chips, (processors cast from the same die) in order to maximize stability. You might be able to use different processors, (probably not on those Supermicro boards) but it may not run as well as you'd hoped/at all.
Click to expand...

My Guess is your chances are 50/50, if the L and E series chip stepping match then your chances are actually quite good. If you already have the CPUs and board then it is very unlikely to cause damage so give it a try.L series chip in socket 0, check in the bios immediately after post that core and bus voltages look sensible (L will probably have lower core voltage but 1366 chip voltages vary somewhat). If the board correctly drops the speed of the other chip to match that will help, if it doesnt drop the E5520 multiplier then see if you can change the multiplier from auto to match the L5518.
 
I am intrigued by this issue too. Newer generation CPUs have Turbo modes and what not so the internal on-chip frequency is bound to be different at different times. And these can be different from core to core based on which is going to the Turbo mode.

Further the OS and applications can not (or should not) rely on exact performance since there are so many variables involved in determining performance. Specifically the memory latency which is a function of the context of the program. In one case a CPU may have everything on-chip (L1/L2); in another case everything might be in the RAM with significant performance differences.

What should really matter is the external interface which does not change between many CPUs under the same class.

I think it boils down to the BIOS on the board and what the OEM is willing to support. I see no reason why the frequency to be dropped. For many Xeons of the same series, all that changes is the base clock multiplier which is locked to the CPU. And with the Turbo mode, even this multiplier is constantly changing within different cores. The BIOS should not have to adjust anything for Xeons with the same Base Clocks and external interface.

The only area where the different frequencies can matter is the QPI interface between the two processors. However the QPI interface has its own controller and the current multiplier of the core should not really matter there.
 
Last edited:
You must log in or register to reply here.
Back
Top