January 29, 2010 at 2:06 am
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...
January 29, 2010 at 2:10 pm
Ritesh Medhe (1/29/2010)
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...
AMD does not hire Kernel mode developers needed to write most of the relevant hardware drivers, while Intel runs a division just writing Kernel mode code so for production it pays to run Intel.
Kind regards,
Gift Peddie
January 29, 2010 at 2:24 pm
Gift Peddie (1/29/2010)
Ritesh Medhe (1/29/2010)
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...AMD does not hire Kernel mode developers needed to write most of the relevant hardware drivers, while Intel runs a division just writing Kernel mode code so for production it pays to run Intel.
Never heard this before, do you have any references?
January 29, 2010 at 2:33 pm
Lynn Pettis (1/29/2010)
Gift Peddie (1/29/2010)
Ritesh Medhe (1/29/2010)
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...AMD does not hire Kernel mode developers needed to write most of the relevant hardware drivers, while Intel runs a division just writing Kernel mode code so for production it pays to run Intel.
Never heard this before, do you have any references?
I actually talked to AMD engineer who told that a while back, Intel now own the most comprehensive C++ tools outside of Microsoft. HP sells AMD servers because HP also hires a lot of Kernel mode coders.
Kind regards,
Gift Peddie
January 31, 2010 at 2:23 am
Ritesh Medhe (1/29/2010)
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...
There are no special settings required to run SQL Server 2008 on an AMD architecture, no.
Most likely there was an obvious difference between the two machines, either in hardware or software configuration, or your test was flawed.
Without knowing a great deal more about the environment and tests, the only answers you will get will be speculative, obvious, or just plain wrong.
Paul White
SQLPerformance.com
SQLkiwi blog
@SQL_Kiwi
February 16, 2010 at 1:42 am
If you are not using intel then you can't use /PAE switch for memory extension.
-Lk
February 16, 2010 at 4:02 am
luckysql.kinda (2/16/2010)
If you are not using intel then you can't use /PAE switch for memory extension.-Lk
Myth.
http://en.wikipedia.org/wiki/Physical_Address_Extension
Busted.
😛
Paul White
SQLPerformance.com
SQLkiwi blog
@SQL_Kiwi
February 16, 2010 at 5:00 am
My bad. Thanks Paul. /PAE SWITCH can be used with AMD as well.
February 16, 2010 at 6:40 am
luckysql.kinda (2/16/2010)
My bad. Thanks Paul. /PAE SWITCH can be used with AMD as well.
No worries. It's a popular misconception for some reason.
Paul White
SQLPerformance.com
SQLkiwi blog
@SQL_Kiwi
February 16, 2010 at 7:14 am
Ritesh Medhe (1/29/2010)
is there any sepcial kind of optimisation setting that needs to be done if SS2008 is running on AMD? I ran same SQL query on Intel and AMD latter was very slow...
Come on now, lets have the full specs of both servers including processor models, memory, disk subsystem etc. as well as the memory allocated to SQL and whether they have the same level of application traffic etc.
Also, check that the actual Execution Plan of the query is the same on both servers.
Do both servers have an identical copy of the database in question, have you got up to date statistics on both?
Pound for pound, AMD seems to lag a little behind Intel on current generation processors of the same clock speed, but the differences are small and only the portion of the query that is CPU bound would be slower.
February 16, 2010 at 9:48 am
Let's assume you're using x64, which AMD developed. Sorry but I really think this amd vs intel cpu stuff is a bit of anon starter.
I do believe when you get beyond 8 sockets then the intel chips may work better, but generally I always sort of understaood that for 4 sockets boxes the amd quads usually outperformed the intel.
However that said being able to measure that difference on the average application server is unlikely.
Things like cpu cache have a big impact, but generally the storage sub system and server memory have a bigger impact. In fact I have a 4 x dual with 12 DAS sas disks which can significantly outperform a 4x quad on a dedicated san ( all 15k raid 10 disks ) with 54 disks.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
February 16, 2010 at 10:00 am
run a showplan on both and also count the data in all the involved tables.
I'm guessing you'll see a difference somewhere.
If not, then I'd be very curious about other differences such as the disk subsystem.
The processors are almost the last place I'd be looking.
Craig Outcalt
Viewing 12 posts - 1 through 11 (of 11 total)
You must be logged in to reply to this topic. Login to reply