Not my favorite benchmark in the world but here goes.
(The closest apples to apples I can find that compares X3 to Opteron.)
TPC.org
IBM x366 w/ X3 chipset and 64GB RAM
3Ghz DC Xeon
IBM DB2 8.2 database 64-bit
784 drive, SAN
TPC-C Throughput 221,017
HP DL585 with 128GB DDR333
2.4Ghz DC Opteron
IBM DB2 8.2 database 64-bit
393 Drive, direct attached
TPC-C Throughput 236,054
-----Original Message-----
From: Alexei_Roudnev [mailto:Alexei_Roudnev@(protected)]
Sent: Wednesday, May 03, 2006 7:32 AM
To: Herman.Maseberg@(protected)
Cc: Anton Dischner; suse-oracle@(protected)
Subject: Re: [suse-oracle] to 64-bit or not to 64-bit that is here the question
Importance: Low
Herman.
I mean _64 bit mode vs 32 bit mode_.
My tests shows that Xeons runs faster in 32 bit mode, and some guys are saying that Opterons runs faster in 64 bit mode.
Which is not surprise, remembering that Intel just borrowed ET64T command set from AMD (so Xeor was designed as 32 bit cpu with 64 extension, and Opteron was designed as 64 bit cpu with 32 bit extension.
I am not comparing CPU's (except 2.5 AMD ==~ 3.2 Xeon).
----- Original Message -----
From: <Herman.Maseberg@(protected)>
To: "Alexei_Roudnev" <Alexei_Roudnev@(protected)>
Cc: "Anton Dischner" <Anton.Dischner@(protected)>;
<suse-oracle@(protected)>
Sent: Wednesday, May 03, 2006 5:47 AM
Subject: Re: [suse-oracle] to 64-bit or not to 64-bit that is here the question
Sorry but I don't agree with this opinion.
Under what situations you can say that ?
The best feature for AMD is local memory for each processor and the low
latency but in real situation like OLTP systems when a data is required for
a processor and this is not in your local memory it query for this data to
a another processor, this mechanism is called snooping but processor cannot
use data until ALL snoops complete, in 4-way systems there are always two
hops for snoops.
Furthermore, if one of these processor hangs you get an automatic reboot
and until these processor is replaced you cannot startup the box.
We test IBM xSeries servers with X3 chipset and there are more performant
in OLTP real situations because they have a memory controller with the
pointer of all data in processor and avoid snoop request.
In summary AMD Opteron is faster than Intel with 8500 chipset (HP, Dell and
many others) and slower than Intel with X3 chipset in OLTP transactions.
Regards,
Herman
"Alexei_Roudnev" <Alexei_Roudnev@(protected)
10:41:14 p.m.:
> 64 bits is faster on AMDbservers and slower on Intel.
>
> I recommend 64 just because 32 became _history_ in a few years.
>
> ----- Original Message -----
> From: "Anton Dischner" <Anton.Dischner@(protected)>
> To: <suse-oracle@(protected)>
> Sent: Tuesday, May 02, 2006 8:02 AM
> Subject: [suse-oracle] to 64-bit or not to 64-bit that is here the
question
>
>
> > Hi,
> >
> > we have 2 new Fujitsu-Siemens Primergy TX200 S2 4GB main memory
> > dual 150 GB SCSI no RAID.
> >
> > I installed SuSE SLES9 and Oracle EE 10.1.0.5 on one and
> > SuSE SLES9-64 on the other with Oracle 10.1.0.5 64 bit version.
> >
> > The CPUs seem to be quite equal in performance.
> >
> > Do you have any reasons/hints to use 64 bit or not?
> >
> > All our other boxes are 32 bit so i tend to use 32-bit.
> >
> > I think with main memory as high as 4 GB there is no reason to use 64-
> > bit Versions.
> >
> > Thanks for all input,
> >
> > Toni
> >
> > PS: I'll run a few benchmarks and report.
> > --
> > +-----------------------------------------------------------------+
> > | Anton Dischner, Unix admin, Oracle-DBA Phone: +49 89 70953202|
> > | Institut fuer Klinische Chemie Fax : +49 89 70958888|
> > | Klinikum Grosshadern Home : +49 89 69708766|
> > | Ludwig Maximilians Universitaet Muenchen Mobil: +49 172 8388880|
> > | 81366 Muenchen Germany |
> > | Marchioninistr. 15 Anton.Dischner"at"med.uni-muenchen.de |
> > +-----------------------------------------------------------------+
--
To unsubscribe, email: suse-oracle-unsubscribe@(protected)
For additional commands, email: suse-oracle-help@(protected)
Please see http://www.suse.com/oracle/ before posting