July 25, 2006 at 12:30 am
Hi!
I was trying to evaluate possible performance gain from using 64bit processes with sql 2005 server. The results I got were quite upsetting: the performance gain turned out to be negative. I tested running one, four and eight simalteneous queries on the same hardware with 32bit and 64bit OS and SQL 2005. The table size was about 450MB, CPU time during the tests was 80-100%, memory utilization was 1-2GB during the tests. I intentionally picked a test that would not use more memory as I was testing processor input into the performance results:
One query | 32bit | 64bit | Performance gain |
1st run | 61* | 62 | -2% |
2nd run | 53 | 56 | -5% |
| | | |
Four queries | 32bit | 64bit | |
1st run | 177** | 188 | -6% |
2nd run | 174 | 187 | -7% |
| | | |
Eight queries | 32bit | 64bit | |
1st run | 335 | 368 | -9% |
2nd run | 334 | 353 | -5% |
* Time in seconds ** Average time
Did anyone try similar tests?
Are there standard tests everyone can reproduce to find any 64bit performance advantage? Like test on standard databases? I thought 64bit performance improvement would be obvious and easy to demonstrate.
Thanks.
July 25, 2006 at 4:47 am
do you mean you re-installed 32bit and 64 bit on the same hardware for each test?
Is this an amd or intel box?
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
July 25, 2006 at 4:15 pm
Yes, this is an Intel box: Two way Dell PowerEdge 1850 8GB of RAM, 3200MHz Xeon CPUs.
July 27, 2006 at 12:51 pm
Are you using the same hard drives for each install instance?
July 27, 2006 at 4:41 pm
Yes, the hardware is exactly the same, all I change is OS (Win 2003 Server 32/64bit) and SQL Server (MS SQL 2005 32/64bit).
July 28, 2006 at 8:02 am
OK, the hardware is the same......wondering if somthing needs a new driver??? Just to be sure, I would test the raw machine performance on both machines using a benchmarking tool. I use a program called Sandra.
SiSoftware Sandra is a 32 and 64-bit Windows system analyser that includes benchmarking, testing, and listing modules. It tries to go beyond other utilities to show you more of what is really going on under the hood, so you draw comparisons at both a high- and low-level in a single product.
Do a lookup on Google for the download site. It really helped us track down some hardware stuff.
Hope that this helps...
Also, I remember in the Microsoft SQL Server 2005 road show they had some comparisons between 32 and 64 bit. Might be worth lookin those up and seeing if your results compare.
Eric
July 28, 2006 at 8:16 am
One more thing to look at.....
http://support.microsoft.com/default.aspx?scid=kb;en-us;906892
July 28, 2006 at 9:10 am
Were the stats ever updated on the 64bit?
July 28, 2006 at 10:21 am
at the risk of getting sued by intel I believe intel 64bit chips aren't really too hot .. something to do with the front bus only being 40 bit, so adding more intel procs slows things down whilst adding amd procs speeds things up. So I suspect what you're seeing might be correct. ( This is why intel chips can't adress as much memory as the amd chips )
I do agree about drivers however, make sure you have proper 64bit drivers - colleagues tell me this can be a problem - I have to say all installs I know of using 64bit are AMD and certainly the plans where I am are to go dual core AMD, intel were not even considered.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
Viewing 9 posts - 1 through 8 (of 8 total)
You must be logged in to reply to this topic. Login to reply