AMD VS INTEL informative read

[Silicon Effect]

Here's My 2cents

now Processor Power = number of calculations that can be done in a set peiod

AMD's pipelines are shorter and fit more instructions per clock cycle into their CPU's

Intel 's Pipelines are longer and have less instructions per clock sycle than AMD's so they may stretch out as many Mhz out of a Core that they can.

Now one thing that AMD definately has over intel other than price is that if the CPU mis-calculates something which hapens too often than not, then the whole entire pipeline needs to be flushed out and start over again which can cause the system to chugg or even crash :S.
Now AMD's pipes are on average ( depending on core ect are 30% shorter and take equally less time to flush if a mistake is made, which is less likely because of the chip's transistors switching at a much lower pace than intel's CPU (looks @ power hungry monster pressy :P ))


AMD's are generally regarded faster then the equivalent intel chip @ Games because of the 3D environment that needs to be drawn (as most games out use Floating Point to render the scene)aswell as AI needs to be taken into account, which really relies heavily on mathematical calculations which is where AMD leaps ahead. But some current and future games will more than likely use SS2 and SS3, in which Intel's Solution is slightly superior in that aspect

Intel's equivalent chip will also generally win in multimedia and bandwidth processes because of its higher bandwidth - Datapath to the Northbridge and Ram


Then there is the Heat and power usage, Yes AMD's processors' use to use much more wattage and put out more heat than the p4 solution but this link should make you a bit skeptical about intel's practices in how reliable their temp monitoring is
http://www.atomicmpc.com.au/forums....;c=6&t=6580
But I have no proof to disprove that AMD is doin' the same thing although i am told by a reliable source i trust that their Diode in the A64 and Barton cored CPU's are 1 silicon layer above the actual Die of the core



oh and note, i did not include A64 to keep both AMD and intel on the same 32bit playing feild,

for those interested> Read on

Because of the on-die Memory controller the A64 makes mince meat of the current P4 solutions in memory bandwidth hungry apps where the P4 out-shines the AXP series,

Another good point is seemless transition from 32-64bit instruction set

But that argument is for another time
here's a screenie or 2 for those still interested in the memory bandwidth performance between some AMD and Intel flagship grade CPU's
http://img.photobucket.com/albums/v...icon/265fsb.jpg
http://img.photobucket.com/albums/v...andra04_MM2.jpg

In Summary

saying "its faster" means sweet :banana::banana::banana::banana:-all its real world performance that matters, thats where game's and benchmarks come in :P





Hope i've covered everything guys, please feel free to post comments and correct me where i have been mis-informed :toast:

[b0bd0le]

actaully the branch prediction that the p4 has is more often then not, correct. Granted it does incur a performance hit, it can also boost it when it's right.

If branch prediction was so bad, why would intel be using it?

[Lithan]

None of your links are working for me.

[Drisler]

for those interested> Read on

Because of the on-die Memory controller the A64 makes mince meat of the current P4 solutions in memory bandwidth hungry apps where the P4 out-shines the AXP series,

Another good point is seemless transition from 32-64bit instruction set



64bit compatibility and the ondie memory controller aren't the only things going on for the A64. They also fit more instructions per clock cycle than both P4s and AXPs.

[Crankster]

Good summary, that in itself is a problem though. Processor architecture is to advanced to begin explain like this. You touch on a lot of important differences but i'm not sure it's enough. You need to study why amd's FPU is the strenght it is for example.
Right now i'm well pleased just saying that amd is roughly 50% faster clock for clock (ROUGHLY).

[slavik]

actually, P4 is better at FP calculations if you optimise correctly ...

P4 - 4FPUs
Athlon (at least AXP AFAIK) - 3FPUs

make this comparison:
integer pipeline is to pixel pipeline is as FP unit is to texture unit

we have AMD with 3x1
we have Intel with 2x2

sort of like when ATI was 8x1 and nVidia was 4x2 ...

2 FPUs on one integer pipeline means u can execute 2 FP operations (if the 4 numbers aren't connected in either way) at the same time on the same pipeline ...

which means 4 FP operations in one clock cycle, for AMD it means only 3 ...

but if u have integer operations:

AMD does 3 in one clock cycle
Intel does 2 in one clock cycle ...

this is all assembler level optimizations and not many programmers optimise manually that low, besides C/C++ compilers are 99% better at optimizing that one human (compiler optimizations have been worked on by teams who know what they are doing ;))