Subject: is there any important difference between 21164A-1 and 21164A-2 CPUs?
To: NetBSD/alpha Discussion List <port-alpha@NetBSD.ORG>
From: Greg A. Woods <woods@weird.com>
List: port-alpha
Date: 08/11/2003 18:05:03
I've just upgraded my AlphaServer 4000 to include a second processor and
another 512MB of RAM (for a total of 1.5GB RAM).  Amazingly it's only
drawing an additional 62VA (@120vac) and the system temp has remained
steady at 36 degC!

The second CPU I happened to end up with acquiring is a B3005-AA (21164A
400MHz, 4MB cache, Rev 7.2), and the original in the system is a
B3004-AA (400MHz, 4MB cache, EV5.6 21164A).

Before the upgrade, following a power-up reset, I saw:

	 SROM V1.1 on cpu0
	XSROM V6.0 on cpu0
	BCache testing complete on cpu0
	mem_pair0 - 1024 MB 
	20..21..23..
	please wait 36 seconds for T24 to complete
	24..
	Memory testing complete on cpu0
	starting console on CPU 0
	sizing memory
	  0   1024 MB EDO

After the upgrade I now see:

	 SROM V1.1 on cpu0
	 SROM V3.0 on cpu1
	XSROM V6.0 on cpu0
	XSROM V6.0 on cpu1
	BCache testing complete on cpu0
	BCache testing complete on cpu1
	mem_pair0 - 1024 MB 
	mem_pair1 - 512 MB 
	20..20..21..21..23..
	please wait 36 seconds for T24 to complete
	24..24..
	Memory testing complete on cpu1
	Memory testing complete on cpu0
	starting console on CPU 0
	sizing memory
	  0   1024 MB EDO
	  1    512 MB EDO
	starting console on CPU 1

As you can see there is a different SROM version on each CPU, but the
same XSROM version on both.  If I remember correctly I can't upgrade the
SROM, only the XSROM (which I've already done).  Is that true?

If I understand correctly NetBSD reports them both as identical except
for their minor number:

	cpu0 at mainbus0: ID 0 (primary), 21164A-1
	cpu0: Architecture extensions: 1<BWX>
	cpu1 at mainbus0: ID 1, 21164A-2
	cpu1: Architecture extensions: 1<BWX>

Is there any important difference between these CPUs?  Should I swap
them to put the "newer" one in the cpu0 position?

In any case I've built a kernel now with "make -j 4", while running
GENERIC.MP, and it built in about half the time it took when I built it
with only one CPU running (and without any "-j") (i.e. with a plain
GENERIC kernel running):  17 min. vs. 32 min. with source and
destination on 100baseTX/FDX switched LAN NFS (which is about twice as
fast as the only local drive I'm using on the alpha for the moment).

I.e. I'm very happy with this "new to me" system now!  :-)

(note that "make -j 4 depend" blew up though....)

-- 
						Greg A. Woods

+1 416 218-0098                  VE3TCP            RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>          Secrets of the Weird <woods@weird.com>