September 8, 2011 at 5:37 pm
as2higpark (9/8/2011)
Just wanted to swing back and close the books on this issue. We determined that there was a default BIOS setting that was causing the performance issues.Supposedly, these new AMD chips have a "green / eco-friendly" setting which is usually disabled on servers, but was enabled on ours. This was causing the chip to not run at full capacity 100% of the time. Since we have disabled this setting, the performance issues have gone away.
Thanks again for all that contributed.
That's 1 for the books :w00t:.
How much gain did you get with the "new" setting?
Thanks for the feedback. Hopefully in the end you got the cake and ate it too ;-).
September 9, 2011 at 6:25 am
The gain was pretty good, it brought it back down to about 10% lower than previous baselines, but that might have more to do with the optimization made prior the change.
September 9, 2011 at 6:40 am
as2higpark (9/9/2011)
The gain was pretty good, it brought it back down to about 10% lower than previous baselines, but that might have more to do with the optimization made prior the change.
Sounds about right. So all in all it wasn't the best hardware choice??
Does anyone know the primary factor? Slower cpus?
September 9, 2011 at 6:44 am
In hindsight, I would say that our workload is so varied that in some cases (our smaller DBs) it is a great choice, but in other cases(our very large clients, < 300GB) it seems to not fit the workload as well. In those cases, the workload benefits more from less, faster CPUs, rather than more slower CPUs.
Sorry for not being exact, but we have seen nice performance with these boxes, just not 100% across the environment.
September 9, 2011 at 6:57 am
as2higpark (9/9/2011)
In hindsight, I would say that our workload is so varied that in some cases (our smaller DBs) it is a great choice, but in other cases(our very large clients, < 300GB) it seems to not fit the workload as well. In those cases, the workload benefits more from less, faster CPUs, rather than more slower CPUs.Sorry for not being exact, but we have seen nice performance with these boxes, just not 100% across the environment.
Sounds like you have a great case to split the workload into 2 different servers.
Fabulous information on the consolidation topic from Brent Ozar in the MCM training : http://technet.microsoft.com/en-us/sqlserver/gg545019.aspx
Viewing 5 posts - 91 through 94 (of 94 total)
You must be logged in to reply to this topic. Login to reply