Article View: comp.sys.dec
Article #1337Re: Alpha benchmarks
From: jackson
Date: Fri, 14 Aug 1992 19:11
Date: Fri, 14 Aug 1992 19:11
54 lines
2967 bytes
2967 bytes
In article <1992Aug12.160844.2486@nosc.mil>, lem@nosc.mil (Lemuel L. Davis) writes: >>Has anyone run or had DEC run any becnhmarks on Alpha systems? I >>have just seen some..... etc, etc. Before I start, let me say that those who have been given access to Alpha systems (both VMS and OSF/1) in the porting centers and as field test sites have signed agreements regarding releasing performance information. The reason that we have those agreements in place is that those who have access to the systems are told what they can expect from today's systems as well as what they can expect in the future. When information is passed on, those who receive it may not know the condition of the machines, and thus may get the wrong impression (like you have) There are four things that are all in various states on various machines around the world at porting centers. All have serious performance implications at this point. These are: Hardware: Many of the early systems had slower clocks and slower I/O systems due to early prototype chips being used. This will of course slow the systems down OS Software: Both VMS and OSF/1 are early releases of the OS, and thus still have many debugging features "turned on" in the OS. Along with that, some of the OS code has been compiled with compilers that are not final, optomizing versions that generate less than optimal code. Compilers: Once the OS is built, the application has to use those same compilers to generate the application code. These again are less than optimal at this point. PALcode: The low level "microcode" for some of the Priv'd OS internals has yet to be fully optimized. This will also have an impact if many priv'd OS calls are made. Given these, it isn't surprising that any benchmark run today will perform less than expected. This is why one has to sign an agreement saying that they will not publish or distribute any results one may get. I realize that the interest in Alpha performance is great, and there is a desire to "scoop" what Digital will announce in Alpha products. Hang in there, Alpha is coming and it *IS* fast! William Jackson Alpha Workstation Marketing Manager Digital Equipment Corporation 146 Main Street Maynard, MA 01754
Technical Details
Message-ID:
<1992Aug14.191129.12516@engage.pko.dec.com>Path:
arc1-ca.pugleaf.net!sparky!uunet!sun-barr!decwrl!deccrl!news.crl.dec.com!pa.dec.com!nntpd2.cxo.dec.com!nntpd.lkg.dec.com!engage.pko.dec.com!pravda.enet.dec.com!jacksonReferences:
<1992Aug12.160844.2486@nosc.mil> <1992Aug13.093121.4012@rdg.dec.com> <1992Aug13.115706.5652@cl.cam.ac.uk>Show Full Headers
Complete Article Headers
Newsgroups: comp.sys.dec Path: arc1-ca.pugleaf.net!sparky!uunet!sun-barr!decwrl!deccrl!news.crl.dec.com!pa.dec.com!nntpd2.cxo.dec.com!nntpd.lkg.dec.com!engage.pko.dec.com!pravda.enet.dec.com!jackson From: jackson@pravda.enet.dec.com () Subject: Re: Alpha benchmarks Message-ID: <1992Aug14.191129.12516@engage.pko.dec.com> Keywords: alpha, benchmarks Sender: newsdaemon@engage.pko.dec.com (USENET News Daemon) Reply-To: jackson@pravda.enet.dec.com () Organization: Digital Equipment Corporation References: <1992Aug12.160844.2486@nosc.mil> <1992Aug13.093121.4012@rdg.dec.com> <1992Aug13.115706.5652@cl.cam.ac.uk> Date: Fri, 14 Aug 1992 19:11:29 GMT Lines: 57