From owner-freebsd-hardware  Wed Dec 18 10:10:30 1996
Return-Path: <owner-hardware>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.4/8.8.4) id KAA19756
          for hardware-outgoing; Wed, 18 Dec 1996 10:10:30 -0800 (PST)
Received: from pegasus.com (pegasus.com [140.174.243.13])
          by freefall.freebsd.org (8.8.4/8.8.4) with SMTP id KAA19748
          for <freebsd-hardware@freebsd.org>; Wed, 18 Dec 1996 10:10:26 -0800 (PST)
Received: by pegasus.com (8.6.8/PEGASUS-2.2)
	id IAA03525; Wed, 18 Dec 1996 08:10:04 -1000
Date: Wed, 18 Dec 1996 08:10:04 -1000
From: richard@pegasus.com (Richard Foulk)
Message-Id: <199612181810.IAA03525@pegasus.com>
In-Reply-To: "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com>
       "Re: EDO parit RAM (Was Re: Tyan ...)" (Dec 18,  9:34am)
X-Mailer: Mail User's Shell (7.2.5 10/14/92)
To: freebsd-hardware@freebsd.org
Subject: Re: EDO parit RAM (Was Re: Tyan ...)
Sender: owner-hardware@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

} > } > What's wrong with ECC mode only memory?  Where can you find parity
} > } 
} > } 15% loss of performance if I remember well. See the archives of the list.
} > } 
} > 
} > That seems doubtful.  How about 15% on main memory accesses -- most
} > accesses are from cache.
} 
} >From my testing on ASUS PCI/I-P55T2P4 Intel 430HX based boards it was
} infact a 12% degrade in -MEMORY- intensive benchmarks, infact the same
} benchmarks I used to find the 1% EDO speed up showed the 12% ECC slow
} down.  This would tend to indicate that the slow down in the main
} memory system is significant when running in ECC mode, far more than
} one would expect.
} 
} > 
} > This sounds like the EDO gain that turned out to be 1%.
} 
} Except that both where measured using the same set of tests, and thus
} the 15% (12% was my result) is real system performance loss.
} 

So what kind of impact does this have on `make world'?

It can't be *that* bad -- unless ECC disables the cache ...

I think your benchmark is hitting a sour-spot.
 


Richard