Sign in with
Sign up | Sign in
Your question

Xeon vs Dual Core help needed

Tags:
Last response: in CPUs
Share
November 4, 2010 12:53:56 AM

Hi there all

I am on a very limited budget and can only afford used equipment, and even then am limited to what I can purchase.

I require RAW computational power for numeric calculations, something that my existing Single P4 HT 2.8 has much trouble with (some calculations are taking SIX days to complete!!!!). I need to bring these computation times down to less than 24 hours.

I/O is not so important but all architectures support reusing my 146GB 15K SCSI Drives.

My available choices (as part of very similar architectures in other respects - re-using existing SCSI 15000RPM Drives) are:

2 x Dual Core AMD Opteron 285 running at 2.6 GHZ giving 4 cores (2GB RAM)

4 x P4 XEON 3GHZ. (with 8GB RAM)

2 x UltraSPARC IIIi at 1.33GHZ. (with 2GB RAM per CPU)

I will be running the Solaris Operating System.

Many thanks!
Ed

More about : xeon dual core needed

November 4, 2010 1:09:13 AM

I should add the P4 XEONS have 1MB L2 cache on each.
a b à CPUs
November 4, 2010 1:19:44 AM

Lets take a shot at this...
If your numerical calculation software is properly coded to hand many multiple threads up to 8 then the 4 Xeon setup which should be Hyper Threaded might be good.
But I think that overall the Dual Opty 285's would be the best choice.
The 285 is a newer chip and less your application is memory intensive 2gb should be fine.
The 285 will have more instruction sets and also be 64bit which (i dont know much about Solaris) if your OS is 64bit would help.
Also the 285 has 2x1mb cache which is probably better than the Xeon 3.0 since the 1mb cache is shared by the physical and logical HT cores while each core of the 285 has its own cache.
Also the Opty 285 has a Hyper Transport of 1000 mhz which is way better than the 533 FSB of the P4 Xeon 3.0.
I own a dual P4 Xeon 3.2 and I would pick the Opteron 285 definitely.
Related resources
November 4, 2010 1:01:58 PM

many thanks for the reply!

unfortunately my boss mandated we stick with SPARC so I purchased a V210 server with 2 x UltraSPARC III at 1.33 GHZ.

For peoples interest during computation I found that a V100 600MHZ UltraSPARC II was roughly 50% the throughput of a P4 2.4, the UltraSPARC III's should be far faster!

a b à CPUs
November 4, 2010 4:24:52 PM

I don't know much about USPARCs.
I believe their architecture is more sophisticated than Intel and AMD.
How many cores is the US III?
If a v100 is 50 percent a 2.4 than that would be roughly 1200mhz equal.
So the 1.33 should be roughly equal to a 2.66 if the proportion holds true.
I believe that the Sparc has dedicated memory banks?
I ran across these links:
http://en.wikipedia.org/wiki/Multiprocessing
http://en.wikipedia.org/wiki/Symmetric_multiprocessing
http://en.wikipedia.org/wiki/Simultaneous_multithreadin...
This last link refers to Ultra Sparcs about half way down that is interesting since they use a version of HyperThreading though different from Intel's solution.
I think that chip makers have hit a cpu speed barrier for now and that multiprocessors whether they are on one die or multi sockets is a way around that barrier but software developers need to focus on tight multi-threaded apps.
More and more there is more multithreaded software for the home market when traditionally the focus was on business apps (Maya,Photoshop etc).
Even games more and more are using multi threads.
Left 4 Dead 2 is a good example.
It runs great on my system which is pretty old but has 4 logical cores so it can run 4 threads.
a b à CPUs
November 4, 2010 4:28:47 PM

Oh since you are new a little heads up.
If an answer really helps you out than it is good to select it as a best answer.
No I am not fishing for a best answer (they are worth points for you and the person you select) since I really didn't answer your question taking into account the USparcs.
But in the future it is a nice thing to do on this board.
Thank you for an interesting post.
a b à CPUs
November 4, 2010 4:32:46 PM

Also by the way everyone is hot and bothered about cloud computing.
Microsoft has a big focus on it.
In the future we might all have web appliances with minimal hardware and paying
a Computer Processing utility bill.
YUCK!
November 4, 2010 4:45:30 PM

Many thanks

You are correct, each CPU has dedicated memory (these slots are standard memory slots usually adjacent to the processor and part of the CPU module itself).

Nothing special about the memory, bog standard ECC chips.

November 4, 2010 4:52:02 PM

As for performance.

Some UltraSPARC III's have I believe 8MB (!!) of L2 cache.

The Predecessor of the USIII was the USII, this was the one where a 550MHZ (sorry not 600MHZ as I stated) doing complex math/textual manupulation work over 250K records took approximately twice as long as a P4 2.4.

I also believe they are single core.

Sun moved to using Opteron powered stuff in their post low end 2005 servers....
!