Crisler
New member
For the last two days I have had the opportunity to play with a Phenom based system. I thought I would post some of my observations here and open a discussion. Let me preface by saying I do not ever, in reviews post benchmark numbers. I feel they are ususally useless. I will reference some benchmarking here but will not quote numbers, let it suffice to say that if I comment on one beating another the difference is over 10% and if I give more descriptive terms the difference is even larger.
To begin we did our typical subjective testing by running the system next to an Intel Core2Quad. We loaded up a number of games and set the graphics to the same level and proceeded to play. I can safely say after a number of hours of play comparision that there was ZERO game play experience difference between the two systems. Every game we threw at each system ran just as smooth. BTW both systems did have identical amounts of RAM and Video cards.
The conclusion from our subjective testing is that in real world applications either chip was more than enough to enjoy an equal computing and gaming experience, in other words there is no loser in real life.
Now at this point I would usually end my testing due to the fact I feel the only truely useful measure of a system had already been done, but with all the info on the Internet I wanted so more data, so I grabbed some synthetic benchmarks designed to push the CPU, FPU and memory system and went to town.
In straight up Integer performance the Intel processor won every time. In fact pushing the Phenom to 3.0 Ghz the Intel chip at a stock speed of 2.4 was still winning but the race was really close. (less than 10%)
Moving to Floating Point however the scale did not just change but almost broke with the Phenom BLASTING the Intel chip by as much as 25% at stock speeds. Even overlcocking the Intel chip did not allow it to take the lead with Floating Point.
When it came to Memory Access the Phenom system also won by a fairly large margin, around 15%. The Intel was running a P35 using DDR2 compared to the Phenom on a 790FX based board.
The last test was one on Sandra that measures Inter-Core Effeicency. I expect the Phenom to rock at this but the Core2 won each time.
The results that count to me are the actual game play but the other results show some things that seem interesting to me.
While the Phenom might lag a bit right now it would seem it might have adge in design as we advance in software. What I mean is right now most software is written for integer math but with the current power of processors and the use of multi core it seems more percise math could now be used with little to no real hit. Now as I am not a programmer I am not sure that my assumptions are full of crap or not but logic would seem to dictate that by switching to floating point for the various calculations AI systems could be made more advance. If this theory is correct then the Phenom could be poised to put some serious smack down. Also the better memory bandwidth is a nice design advanatge the Phenom and the Athlon has shared for some time. Intel needs to learn from this and move in a similiar direction.
The area of concer for me is the fact that dispite being a true on die Quad Core the Sandra test is saying it is not as effecient as the Intel solution. This could be a flaw in the test so I am not sure how accurate this is, but if true is points to a weakness AMD needs to take a serious look at.
Now as I said some of this is just a logical thought process and so I have no basis for this and if I am wrong then please I would be grateful for correction. However in the end I do feel the only true test is the real computing experience with the speedometers turned off. When that is done the chips are to close to call and both deliver great computing experience.
I look forward to some serious discussion on these thoughts.
To begin we did our typical subjective testing by running the system next to an Intel Core2Quad. We loaded up a number of games and set the graphics to the same level and proceeded to play. I can safely say after a number of hours of play comparision that there was ZERO game play experience difference between the two systems. Every game we threw at each system ran just as smooth. BTW both systems did have identical amounts of RAM and Video cards.
The conclusion from our subjective testing is that in real world applications either chip was more than enough to enjoy an equal computing and gaming experience, in other words there is no loser in real life.
Now at this point I would usually end my testing due to the fact I feel the only truely useful measure of a system had already been done, but with all the info on the Internet I wanted so more data, so I grabbed some synthetic benchmarks designed to push the CPU, FPU and memory system and went to town.
In straight up Integer performance the Intel processor won every time. In fact pushing the Phenom to 3.0 Ghz the Intel chip at a stock speed of 2.4 was still winning but the race was really close. (less than 10%)
Moving to Floating Point however the scale did not just change but almost broke with the Phenom BLASTING the Intel chip by as much as 25% at stock speeds. Even overlcocking the Intel chip did not allow it to take the lead with Floating Point.
When it came to Memory Access the Phenom system also won by a fairly large margin, around 15%. The Intel was running a P35 using DDR2 compared to the Phenom on a 790FX based board.
The last test was one on Sandra that measures Inter-Core Effeicency. I expect the Phenom to rock at this but the Core2 won each time.
The results that count to me are the actual game play but the other results show some things that seem interesting to me.
While the Phenom might lag a bit right now it would seem it might have adge in design as we advance in software. What I mean is right now most software is written for integer math but with the current power of processors and the use of multi core it seems more percise math could now be used with little to no real hit. Now as I am not a programmer I am not sure that my assumptions are full of crap or not but logic would seem to dictate that by switching to floating point for the various calculations AI systems could be made more advance. If this theory is correct then the Phenom could be poised to put some serious smack down. Also the better memory bandwidth is a nice design advanatge the Phenom and the Athlon has shared for some time. Intel needs to learn from this and move in a similiar direction.
The area of concer for me is the fact that dispite being a true on die Quad Core the Sandra test is saying it is not as effecient as the Intel solution. This could be a flaw in the test so I am not sure how accurate this is, but if true is points to a weakness AMD needs to take a serious look at.
Now as I said some of this is just a logical thought process and so I have no basis for this and if I am wrong then please I would be grateful for correction. However in the end I do feel the only true test is the real computing experience with the speedometers turned off. When that is done the chips are to close to call and both deliver great computing experience.
I look forward to some serious discussion on these thoughts.