Core i7 doesn't produce good results in gaming

Status
Not open for further replies.

Hussein19891

Biostar Stole my heart
Messages
2,210
Location
Canada
and if Deneb wins

I doubt Deneb will win. But Deneb will be a Phenom capable of a 4.00ghz over clock on air at a lower cost then the competitor. So thats pretty much why I m on the Deneb band wagon. That and the fact it will have 6MB of L3 cache, which is good news for all the SuperPi (AMD fanboy) addicts. Along with the fact it will be 45nm :D

I just want to see competition in the CPU market as we've all been seeing in the GPU market.
 

veg1992

Golden Master
Messages
19,993
Well SuperPI is more fond of the Core Architecture so its going to get crappy scores no matter what..
 

KSoD

Call me Mak or K, Mod Emeritus
Messages
35,644
Location
C:\
SuperPi is not based toward Intel. It is jsut a Math Processing application. AMD is not good at Math processing but rocks everywhere else. Creating a SuperPi that shows AMD get higher scores and Intel Lower scores will not be just. As it will only show that people are biased against one or the other.

SuperPi does what it is supposed to. It tests the math capabilties of the CPU. It has been long known, even by me who has only 1 AMD CPU ever, that Intel beats AMD in that aspect. AMD was never known for their math. Intel has had a long time to perfect this since they started this back with the 80286 SX and 80286 DX CPU's.

One was jsut a basic run of the mill CPU while the other was the CPU with the math processor. If i remember my CPU history from back tehn the DX had the Math Processor.

Sorry but creating a Super Pi that shows AMD beating Intel will only add fuel to the fire that these tests are biased toward 1 CPU or the other. Then we will never have accurate results cause they will always be brought into question as to which version was used, why that one was used and not the other one, how could the results have varied if the other testing platforms were used and so on.

SuperPi is just fine. It shows what it should. That Intel has perfected their math core but lack overall with their CPU Power against the raw power of the AMD Chip which rocks in raw power but lacks with math.
 

veg1992

Golden Master
Messages
19,993
I remember Apok talking about it a while back..

EDIT:
Me, irrespective of whether Nehalem ends up 10 times faster (it won't though).
I'd say it's most likely the compiler favouring the Intel CPU's.

*edit*
There are compilers that will compile the programs to check if a CPU has a "GenuineIntel" ID, and if so, will enable optimisations on the Intel CPU's
But if it they don't have the "GenuineIntel" ID, those optimisations won't be enabled, even if the CPU is capable of using them.

I'm not sure what compiler PCmark05 used, but there was a test on a Via system - which actually has an unlocked ID - where they used a "GenuineIntel" ID, an "AuthenticAMD" ID, and an unmodified ID.
What they found was that the Via CPU performed highest with the GenuineIntel ID, performed less with an AuthenticAMD ID, and lowest with its own ID.
Nothing else about the system was changed all; just the CPU ID.

AMD Alleges Intel Compilers Create Slower AMD Code
Intel Stinks, Again | Elliot Lee
 

kmote

Seg Fault'n,
Messages
5,801
Location
The dried husk where America came from
This is what i meant.

Make super pi read both compilers equally, and see what happens.

compilers are not "read" by the application at run-time

if there is a claim that all they did is change the cpu id, and then without recompiling the app there was a noticeable performace difference then the cpu id would have to be identified at runtime in which case it will have been written into the code. if it is opensource then we can take a look at the code to see if this is the case. simple
 

ricanflow

Golden Master
Messages
6,317
Location
Euless, Tx
compilers are not "read" by the application at run-time

if there is a claim that all they did is change the cpu id, and then without recompiling the app there was a noticeable performace difference then the cpu id would have to be identified at runtime in which case it will have been written into the code. if it is opensource then we can take a look at the code to see if this is the case. simple


Look at the links posted above and re analyze your statement.
 
Status
Not open for further replies.
Top Bottom