THE SQL Server Blog Spot on the Web

Welcome to SQLblog.com - The SQL Server blog spot on the web Sign in | |
in Search

Joe Chang

Early Intel Nehalem Xeon 5500 performance results

SAP numbers were available a few months ago, and now, TPC-C and TPC-E results are available for 2-socket Intel Xeon 5570 systems with the Nehalem quad-core processor. The TPC-C result is on Oracle/Linux, no SQL Server on Windows yet. The TPC-E results are all SQL Server 2008.

I mentioned before that my expectations was that Nehalem would generate substantial performance gain on high call volume applications (ie, the transaction benchmarks), partially attributed to the processor core, but mostly to the return of Hyper-Threading (HT). When HT was first proposed for the Pentium 4 generation, the expectation was that substantial performance gain was possible because certain applications could would many dead cycles waiting for memory access, etc. As with most first generation concepts, the actual performance gains in TPC-C were in the range of 7-10%, and a number of other operations could actually experience performance anomalies, that is, degradation, so HT had to be used with care, or even disabled. I did observe a very large gain (30-50%) with HT in Quest LiteSpeed compression tests on the first Prescott generation NetBurst Xeons. So the expectation was that once the tricky issues with HT in complex code like SQL Server, large gains should also be possible.

System             Configuration                                                    TPC-C

DL370G6           2 Xeon 5570 Quad-core 2.93GHz, 8M L3, 144GB   631,766 (Oracle/Linux)

DL580G5           4 Xeon 7460 Six-core 2.66GHz 16M L3, 256GB     634,825

 

                                                                                                  TPC-E 

Fujitsu RX300    2 Xeon X5570 Quad-core 2.93GHz, 8M L3, 96GB   800.00

x3650M2           2 Xeon X5570 Quad-core 2.93GHz, 8M L3, 96GB   798.00

Dell T610           2 Xeon X5570 Quad-core 2.93GHz, 8M L3, 96GB   766.47

TX300 S4          2 Xeon X5460 Quad-core 3.16GHz, 8M L2, 64GB    317.45

Dell R900          4 Dunnington Six-core 2.66GHz, 16M L3, 64GB      671.35

My expectation is that Xeon 5500 series should show moderate gains in TPC-H style large queries. No results are currently available. It is very dissappointing that vendors only want to show a partial picture of what to expect. SPEC CPU integer results show a reasonable gain of 31.5 for Xeon 5570 versus 25.3 Xeon 5460. The individual components range from a few very large gains, several good gains, a number of modest gains and a slight decrease for the bzip2 component, so expect variations in narrow testing. 

Anyways, I will get a Xeon 5500 system for my own testing as soon as I can. If anyone wants to loan a Opteron Shanghai system, I can do current generation comparison tests. But I do not really have my own budget right now for the Opteron system.

Published Wednesday, April 01, 2009 1:02 AM by jchang
Filed under: ,

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS

Comments

 

Nasser Jasim said:

Hi Joe,

the results you posted show some really interesting performance gains, unfortunately, we bought our servers last month with Xeon 5400 series, I wish I've waited a month!!

anyway, do have any ideas about the performance gains from RDMBs point of view, meaning - will SQL 2K8 benefit more than ORACLE 11g from the new Nehalem architecture?

thanks and regards

Nasser

April 2, 2009 3:54 AM
 

jchang said:

I am strongly inclined to think the performance gains for Nehalem (5500) over Core 2 (5400) is heavily in high call volume applications, ie, small queries, read or write, regardless of RDBMS. Thats why vendors are slow to post TPC-H results. Large queries should still be moderately faster on Nehalem, averaged over a broad spectrum of SQL operations. However, I am expecting parallel execution plans and HT to problematic, ie, should HT be on or off for DW? also, should SQL Server know not to use both logical CPUs on HT?

April 2, 2009 5:02 AM
 

Nasser Jasim said:

thanks Joe for your quick reply,

I also would like to ask another question that was confusing me for a while now, why did MS stop posting TPC-C results for almost two years, they used to be on top the "Top Ten TPC-C by Performance" and over the months they are not even in the top 10 any more, instead they invented the TPC-E which seems like no one is willing to participate in - except MS???

Do you think they gave up on TPC-C because other RDBMs are much better and more advanced than SQL 2008?

Is there any way to test on the same hardware how does SQL 2008 compete against Oracle 11g and IBM DB2 9.5  in TPC-C scenarios?

regards

Nasser

April 2, 2009 6:05 AM
 

jchang said:

TPC-C is an old benchmark and really simplistic. TPC-E is more complex (25+ queries, which is probably reasonable for the number of most frequently called SQL in actual apps)

TPC-E was also supposed to reduce the number of disks required.

Everybody participated in developing TPC-E, read the members list.

I think it was when time came to publish, nobody could beat SQL on common systems, that no one else published. The TPC-E queries contain server-side cursors, which I know in SQL 2000 days would cripple big iron systems. I am certain MS made significant effort first in S2K5 and more so in S2K8 to run TPC-E really well.

which is also why no one can beat MS, hence no other publications,

Since TPC-C is rather simple and has been around a long time, every one runs TPC-C well, especially on servers with less than 8 processor sockets.

I think Oracle and DB2 on unix/linux has special tuning features for better scaling on the HP Superdome, but on Xeon and Opteron, its essentially a wash

April 2, 2009 8:47 AM
 

Daniel Johnson said:

Hi Joe

Thank you so much for your information that you have provided.

I have really obtained a picture of Nehalem's performance.

What fascinates me more is the low power consumption and the higher speed.

April 7, 2009 6:57 AM

Leave a Comment

(required) 
(required) 
Submit

About jchang

Reverse engineering the SQL Server Cost Based Optimizer (Query Optimizer), NUMA System Architecture, performance tools developer - SQL ExecStats, mucking with the data distribution statistics histogram - decoding STATS_STREAM, Parallel Execution plans, microprocessors, SSD, HDD, SAN, storage performance, performance modeling and prediction, database architecture, SQL Server engine

This Blog

Syndication

Powered by Community Server (Commercial Edition), by Telligent Systems
  Privacy Statement