I think many of you remember this article here: http://www.legionhardware.com/document.php?id=869
The article was brought up up during a bitter and vigorous debate here on THG but I figure it deserves mention. Over the course of that debate I had noticed a trend, a trend which was not entirely obvious when looking at the results on their own. So what one of my friends Tactician did was take it upon himself to piece together the data in order to bring to light a clear picture.
His Original post can be seen here: http://tinyurl.com/ycc8wd4
Enjoy
Intel = RED
AMD = ORANGE
What I see here is that with each OC bump there is an increase in FPS for both CPUs. More MHz = more FPS ... okay I get that.
Here I see very minimal FPS increase with more CPU power. Also as far as I can tell these CPUs are on Par.
Lower clocks i7 has more FPS, higher clocks pretty much the same. After 2.6GHz there isn't much gain in FPS for the extra MHz
Here there is quite a noticeable difference up until 3.6GHz. The i7 hit a wall at 2.6GHz and there was no benefit from additional OCing.
Again at the low end of the GHz scale a distinct difference. Both CPUs cease to receive any benefit from MOAR POWAH! at approximately the 3.0GHz mark.
First thing I notice... DAMN really low FPS across all tests. Second thing: another brick wall.
In my opinion this shows no improvement from OCing at all. AMD is a stride ahead at the 2.8GHz mark.
AMD receives a small boost from OCing and steps in front of intel at 3.6GHz
Pretty much starts off infront of the wall here. AMD picks up a few points at the end of the race.
That is a lot of information to process but what you're seeing is that the Core i7 Processor IS the faster gaming CPU (look at it's performance at 2GHz compared to the Phenom II X4). As you increase the clock speed, however, a Graphics bottleneck occurs. The perfect example is the last shot. The GPU becomes the bottleneck even at 2GHz. The Core i7 stops scaling from there.
What many users are quick to point out (particularly those who favor AMD) is that AMD sometimes manager to get as much as ~3FPS more than Intel when a GPU bottleneck is present. These users mistakenly assume this is some sort of proof that "AMD is the better gaming chip offering better performance" as one user suggested. This assertion is patently false. By objectively looking at the data we can come to a rather simple Hypothesis. If one were to alleviate the Graphics Card bottleneck by, let's say, adding another Graphics card... the Core i7 would out-scale the Phenom II X4 (we can deduce this by looking at how many more clocks it takes a Phenom II X4 before it can match a Core i7 running at only 2GHz).
This leaves us with the following question... "Why is the Phenom II X4 out performing Intel by a few frames during instances when a GPU bottleneck occurs?".
One could hypothesize that since this ONLY occurs in instances where we're using a GPU that it has something to do with CPU to GPU communications. There is a bottleneck or a performance inhibitor at that precise area. What could it be?
Well I think I've found the answer and have tried to painstakingly spread the answer in order to further knowledge. Intel holds a patent which suggest (clearly states actually) that there is an inherent incompatibility between the communications protocol the PCI Express bus uses and that of the QPi link. This patent explains how the IOH (X58 chip) acts as a hub converter and "tunnels" the data through (in one cycle). This "cycle" means latency in lay mans terms. Where is this patent you ask? Here: http://www.patentstorm.us/patents/7210000/description.html
And there is your answer.
I would like to thank Tactician for piecing these graphs together as well as the crew over at Legion Hardware for the raw data.
Peace.
The article was brought up up during a bitter and vigorous debate here on THG but I figure it deserves mention. Over the course of that debate I had noticed a trend, a trend which was not entirely obvious when looking at the results on their own. So what one of my friends Tactician did was take it upon himself to piece together the data in order to bring to light a clear picture.
His Original post can be seen here: http://tinyurl.com/ycc8wd4
Enjoy
Intel = RED
AMD = ORANGE
What I see here is that with each OC bump there is an increase in FPS for both CPUs. More MHz = more FPS ... okay I get that.
Here I see very minimal FPS increase with more CPU power. Also as far as I can tell these CPUs are on Par.
Lower clocks i7 has more FPS, higher clocks pretty much the same. After 2.6GHz there isn't much gain in FPS for the extra MHz
Here there is quite a noticeable difference up until 3.6GHz. The i7 hit a wall at 2.6GHz and there was no benefit from additional OCing.
Again at the low end of the GHz scale a distinct difference. Both CPUs cease to receive any benefit from MOAR POWAH! at approximately the 3.0GHz mark.
First thing I notice... DAMN really low FPS across all tests. Second thing: another brick wall.
In my opinion this shows no improvement from OCing at all. AMD is a stride ahead at the 2.8GHz mark.
AMD receives a small boost from OCing and steps in front of intel at 3.6GHz
Pretty much starts off infront of the wall here. AMD picks up a few points at the end of the race.
That is a lot of information to process but what you're seeing is that the Core i7 Processor IS the faster gaming CPU (look at it's performance at 2GHz compared to the Phenom II X4). As you increase the clock speed, however, a Graphics bottleneck occurs. The perfect example is the last shot. The GPU becomes the bottleneck even at 2GHz. The Core i7 stops scaling from there.
What many users are quick to point out (particularly those who favor AMD) is that AMD sometimes manager to get as much as ~3FPS more than Intel when a GPU bottleneck is present. These users mistakenly assume this is some sort of proof that "AMD is the better gaming chip offering better performance" as one user suggested. This assertion is patently false. By objectively looking at the data we can come to a rather simple Hypothesis. If one were to alleviate the Graphics Card bottleneck by, let's say, adding another Graphics card... the Core i7 would out-scale the Phenom II X4 (we can deduce this by looking at how many more clocks it takes a Phenom II X4 before it can match a Core i7 running at only 2GHz).
This leaves us with the following question... "Why is the Phenom II X4 out performing Intel by a few frames during instances when a GPU bottleneck occurs?".
One could hypothesize that since this ONLY occurs in instances where we're using a GPU that it has something to do with CPU to GPU communications. There is a bottleneck or a performance inhibitor at that precise area. What could it be?
Well I think I've found the answer and have tried to painstakingly spread the answer in order to further knowledge. Intel holds a patent which suggest (clearly states actually) that there is an inherent incompatibility between the communications protocol the PCI Express bus uses and that of the QPi link. This patent explains how the IOH (X58 chip) acts as a hub converter and "tunnels" the data through (in one cycle). This "cycle" means latency in lay mans terms. Where is this patent you ask? Here: http://www.patentstorm.us/patents/7210000/description.html
And there is your answer.
I would like to thank Tactician for piecing these graphs together as well as the crew over at Legion Hardware for the raw data.
Peace.