PDA

View Full Version : what does this really mean? How does it apply to gaming

gdfo
09-09-2007, 06:28 AM
"Based on early tests, Dell said, a server using Barcelona chips is faster at running floating-point applications than a computer with Intel's Clovertown chips, but Clovertown is faster than Barcelona at running integer instructions."

gdfo
09-09-2007, 06:28 AM
"Based on early tests, Dell said, a server using Barcelona chips is faster at running floating-point applications than a computer with Intel's Clovertown chips, but Clovertown is faster than Barcelona at running integer instructions."

09-09-2007, 07:46 AM
http://forums.ubi.com/groupee_common/emoticons/icon_eek.gif

Skoshi Tiger
09-09-2007, 08:16 AM
Well it's like this.

Computers work in binary. As you can't have a fraction of a bit all 'maths' performed by the computer is in whole numbers (integers)

To do floating point calculations the CPU needs to use algorithms (a sequence of steps to solve a problem) to use integer mathematics to work out decimal points etc. There is actually a section of the CPU that does straight integer maths and another section that does Floating point maths.

Now theres more than one way to skin a cat (not that I've actually skun a cat) and theres more than one way to simulate floating point maths in binary.

Depending on the algorithm used the process is either more or less efficient.

The statement you quoted is just stating that the section of the Barcelona cpu is faster than the other doing floating point maths and the Clovertown is better doing integer maths.

The major advantage the old AMD chips had over their equivelent intel ones was that their FPU (Floating Point Unit) was more efficient. They could have a 'slower' cpu that out performed the intel chip. (sadly no-longer the case)

I'm sure someone will be able to state this in a more tecnical or eloquent manner.

Cheers!

Divine-Wind
09-09-2007, 07:53 PM
It means that Clovertown is a cooler name than Barcelona, so Dell will probably be using the former.

VW-IceFire
09-09-2007, 10:21 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by gdfo:
"Based on early tests, Dell said, a server using Barcelona chips is faster at running floating-point applications than a computer with Intel's Clovertown chips, but Clovertown is faster than Barcelona at running integer instructions." </div></BLOCKQUOTE>
Doesn't apply to gaming...these are server chips.

AMD's Barcelona is the new boy on the block with the newest architecture. What gamers are interested in is the AMD Agena CPU's.

K_Freddie
09-10-2007, 12:31 AM
Floating Point maths can be presented in many ways on a cpu. the two most common are..

- Fixed Point maths which uses integer formats. If you use the 'native' format it's nearly the same as integer calculations which are faster(*). You just have to extract the formats in your software.

- Floating Point presentation (cannot remember the IEEE standards doccy) which has a lot more accuracy for it's speed (*) compared to the rest.

(*) Now the speed of caculations depends on the how the micro-code is implemented on the cpu. Micro-code is the pure logic (transistors units) on the cpu, controlled by a 'Sequencer Controller'. It all works at the 'frequency' of the CPU (Ghz) but some operations take longer than others - FP's being slightly longer due to their complexity. Usually cpu's have all functional blocks (eg Integer Unit/FP Unit) but if some blocks are not implemented on a cpu, don't use it for gaming or general use - you want the whole lot.

Hope that clears things up
http://forums.ubi.com/groupee_common/emoticons/icon_cool.gif

K_Freddie
09-10-2007, 01:28 AM
Nearly forgot..
Another point to consider is that Microsoft have always based their software on Intel products (I heard of a move to AMD, but I think AMD are having problems), so Windoze Kernel code is more likely to be optimised for Intel.

What this means is that it'll run 'faster'. Dell will run into patent problems if they try to copy the micro-code (as did AMD but they got around this) so will have a problem making it faster.

With this in mind you might find that even if Dell's FP unit working faster, it might have unseen problems with Windoze OS, it might have data flow 'bottlenecks' that Intel cpu's don't have because of the Kernel implementation.

This'll mean you'll have to suppliment the Windoze OS with extra drivers..'slowing' it down.

To cut a long story short.. I'll stick to what is 'proven' until such time Dell 'proves' all is well. I've always had Intel motherboards because of all the above - Have very few problems.
http://forums.ubi.com/groupee_common/emoticons/icon_wink.gif