• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Dell 1850s running slow with 4 cpus than 2cpus.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

el

Senior Member
Joined
Feb 3, 2001
Location
USA Capital
okay basically I have 3 of these servers64x 2x2.8ghz(dual) with 2mbl2 and 4gigs ram.

problem is they are slower than the 2x3.6ghz(single) running in 32bit.

how could this be and what are my options to fix the problem.

basically the tests are running sql server 05 64bit compared to sql server 32bit.

any ideas would be great.

Basically I want to know why 4x2.8ghz servers 64bit is slower than 2x3.6ghz servers running 32bit.
 
Well, if the program is not set for 4 threads, it will definatly see performance decreases because of the clockspeed.

Also, I seem to remember there being some problems with SQL 05 on multithreaded (or maybe hyperthreaded.. not really sure) systems.. seemed to be some inefficiency somewhere.
 
They shouldnt be slower in sql 64 bit. Heck even a 2.8 singlecore in 64 bit should run comparable to a 3.6 in 32. I wonder if licensing is limiting the cores to two?

btw If you don't get an answer here, try dbforums. There are a ton of SQL admins over there.
 
There's a known deficiency in SQL2005 regarding hyperthreading (link). Essentially the thread that frees up cache pages goes into active execution on one logical processor while a worker thread is working on the other, leading to horrid cache thrashing and lock issues. Just more proof us programmers need to learn how to work with the higher degrees of thread parallelism a few months out.
 
Ebola said:
They shouldnt be slower in sql 64 bit. Heck even a 2.8 singlecore in 64 bit should run comparable to a 3.6 in 32.

To date, benchmarks have shown little performance improvements (and in some case, worse performance) in 64-bit mode compared to 32-bit mode.
 
well these are dual core not hyper threading cpus, it could be bandwidth issues, 4 cores all trying to access main memory through a single 800mhz bus could get messy think about how starved for bandwidth one pentium 4 is then mul p4, [#4]! ;)
 
TimP said:
To date, benchmarks have shown little performance improvements (and in some case, worse performance) in 64-bit mode compared to 32-bit mode.


Do you have any links to comparisons? I've only seen the WinHEC demo.
 
yeah that xbitlabs article was really great thanks for the help. Looks like AMD is best for general servers.
 
Untill you factor in cost of entry level, all my servers are Intel just becaues i can build a decent dual xeon setup for $600, you cant do that with AMD :(
 
um if you go just basic with dfi nforce4 and 3800x2 you could make a low end dual with amd. now the opterons are $$$$. If you were going amd which make/model would you buy?
 
I would probably go with a 4200 amd dual core if you can't go with an opty. Since it has more cache and a better core than the 3800.

Also the thing I hate about doing AMD servers is if I build them I have to tech support, them warrenty them, and generally fix any quirks. Its much easier to just call 1800-new-dell and have them build me something thats quiet doesn't crash and works.
 
Back