Overclock.net › Forums › Intel › Intel CPUs › Best CPU for floating-point calculations
New Posts  All Forums:Forum Nav:

Best CPU for floating-point calculations

post #1 of 8
Thread Starter 
My boss is looking for the CPU that is the best at floating-point calculations. I dont know anything about floating point stuff, so if you could help me out here, it would be great!
post #2 of 8
As this is posted in the Intel section I will give you the best Intel Pentium 4 processors for floating point operations.

Cheaper One: Intel Pentium 4 670 (3.8GHz)
Expensive One: Intel Pentium 4 Extreme Edition (3.73GHz)

These two are Intels top processors for FP processing. As floating point is just really calculations it goes on the speed of the processor, not its features like more L2 cache etc. They come into it, but pure processing power is by far the most important.
post #3 of 8
Thread Starter 
Quote:
Originally Posted by The_Manual
As this is posted in the Intel section I will give you the best Intel Pentium 4 processors for floating point operations.

Cheaper One: Intel Pentium 4 670 (3.8GHz)
Expensive One: Intel Pentium 4 Extreme Edition (3.73GHz)

These two are Intels top processors for FP processing. As floating point is just really calculations it goes on the speed of the processor, not its features like more L2 cache etc. They come into it, but pure processing power is by far the most important.
I posted this in the Intel section only because I thought Intel might be better at it, and also because there is no "general" CPU forum. I am open to AMD chips as well. From what I have read, the app I want to run (G.729) is multi-threaded, so am I corect in thinking that a dual-core chip would be better?
post #4 of 8
Quote:
Originally Posted by SoaDMTGguy
I posted this in the Intel section only because I thought Intel might be better at it, and also because there is no "general" CPU forum. I am open to AMD chips as well. From what I have read, the app I want to run (G.729) is multi-threaded, so am I corect in thinking that a dual-core chip would be better?
Yes, I would agree in the idea that dual core would be better. For doing such a task as floating-point calculations, you would need one of the the top of the line processors. I personally like AMD for the more clocks per cycle, but I think either company would be ok.
My System
(13 items)
 
  
CPUMotherboardOSKeyboard
skt 754 3200 dunno xp pro sp2 wireless 
PowerCaseMouse
neopower 480 cheap wireless 
  hide details  
Reply
My System
(13 items)
 
  
CPUMotherboardOSKeyboard
skt 754 3200 dunno xp pro sp2 wireless 
PowerCaseMouse
neopower 480 cheap wireless 
  hide details  
Reply
post #5 of 8
A dual core CPU would be a fair idea for a multithreaded application, however how efficient is the multithreaded code of the application? if its poor i.e. does not multithread extremely well then I would get a high speed single core processor. As it has some multithreading support I would actually recommend a Pentium 4 processor with HyperThreading Technology like the 670. My reasoning behind this is because the 670 is far faster than the 3.2 840 Dual core (600MHz faster in core clock speeds) so it can process data faster if it is executed on a single core. When heavy multithreading stuff comes into play then there will be a turn in the tides. The 3.8 is a very good processor and because of the clock speed boost I would choose it for an application over the 840 unless it could execute efficiently on 2 physical cores, HT does come into play here as it can help with multithreaded stuff

For the AMD side a 4000+ would be a very good choice as would a 3700+ or an FX series processor, but they can't multi-task as well as Intel processors that process the same amount of data as them.
post #6 of 8
Thread Starter 
Quote:
Originally Posted by The_Manual
A dual core CPU would be a fair idea for a multithreaded application, however how efficient is the multithreaded code of the application? if its poor i.e. does not multithread extremely well then I would get a high speed single core processor. As it has some multithreading support I would actually recommend a Pentium 4 processor with HyperThreading Technology like the 670. My reasoning behind this is because the 670 is far faster than the 3.2 840 Dual core (600MHz faster in core clock speeds) so it can process data faster if it is executed on a single core. When heavy multithreading stuff comes into play then there will be a turn in the tides. The 3.8 is a very good processor and because of the clock speed boost I would choose it for an application over the 840 unless it could execute efficiently on 2 physical cores, HT does come into play here as it can help with multithreaded stuff

For the AMD side a 4000+ would be a very good choice as would a 3700+ or an FX series processor, but they can't multi-task as well as Intel processors that process the same amount of data as them.

Thanks again for your help! I had been thinking about Xeons, as this would be used in a server setting. What are your thoughts about that?
post #7 of 8
Quote:
Thanks again for your help! I had been thinking about Xeons, as this would be used in a server setting. What are your thoughts about that?
A Xeon is a very fast processor, some varients even have 9MB L3 cache incorporated into them. They are built for server work and heavy calculations so they would work well in an environment that suits them. For Floating Point Calculations they will work very well but will still be bested by the 670 because the processing power makes the biggest difference and the 670 has the fastest clock speed. The Xeon's are a very good alternative but I would stick with the Pentium 4 personally unless you want to do extreme number crunching, which in that case the extra features of the Xeon and its ability to process data a bit more effciciently come in handy and outway the cons. The issue with the Xeons are that the other components in the system are also built for intense number crunching and datbase work etc so they are not really built to handle things like games (like the ECC RAM which has poor latencies).
post #8 of 8
Thread Starter 
Quote:
Originally Posted by The_Manual
A Xeon is a very fast processor, some varients even have 9MB L3 cache incorporated into them. They are built for server work and heavy calculations so they would work well in an environment that suits them. For Floating Point Calculations they will work very well but will still be bested by the 670 because the processing power makes the biggest difference and the 670 has the fastest clock speed. The Xeon's are a very good alternative but I would stick with the Pentium 4 personally unless you want to do extreme number crunching, which in that case the extra features of the Xeon and its ability to process data a bit more effciciently come in handy and outway the cons. The issue with the Xeons are that the other components in the system are also built for intense number crunching and datbase work etc so they are not really built to handle things like games (like the ECC RAM which has poor latencies).
Again, thanks for the info! The systems that I would be using the CPUs in would be rack mounts, and would need to support redundent PSUs and HDDs. Can you get that support with the Pentium platform?
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Best CPU for floating-point calculations