Core2 vs phenom II vs core i

[Chad Boga]

How do you figure? Are you talking about a Phenom II X4 965 clocked at 3.4GHz vs. a Core 2 Quad Q9650 clocked at 3GHz? Of course you then would run these processors using processor intensive applications and not applications which rely mostly on other hardware components (such as games).

If so either overclock the Q9650 to 3.4GHz or down clock the Phenom II X4 965 to 3GHz and compare the two clock for clock.


The question is about the actual architectures not the shipping products.

The shipping clockspeed is also a result of architecture.

 

[enzo matrix]

The shipping clockspeed is also a result of architecture.

So are you saying the Phenom II x4 910 C2 stepping and the Phenom II x4 965 C2 stepping have different architectures? LOL
That's like saying every single CPU has a completely different architecture. I think you fail to grasp the concept of what the architecture is. Unless I misunderstand what you are trying to say.

Oh and I think I made a mistake. Does the core i7 not use the same architecture as the core 2 processors? Intel had netburst before core. Architecture called "core", not "core 2". Netburst was introduced in 2000 and core in 2006. Now with the core i7s, they didn't change the name from core...

 

[ElMoIsEviL]

The shipping clockspeed is also a result of architecture.

Architectures can allow for a certain clock speed range (operating range) within certain specified conditions (voltage, heat etc).

But the actual clock speed with which a product functions is not inherently directly tied to it's architecture because if it were... an overclock on my processor would mean an architectural change.

Now if we look at it all from the perspective of an architecture being engineered with a certain range of clocks speeds in mind we undoubtedly must confess that overclocking yields would therefore be the best indicator of what a given architecture is capable of achieving in relation to a clock speed (and use those yields in order to form a real world statistical analysis via a scientific poll).

In doing so the end result remains no different then that which I portrayed:


Core ix > Core 2 Quad (Yorkfield) > Phenom II X4 > Core 2 Quad (Kentsfield)

 

[Chad Boga]

So are you saying the Phenom II x4 910 C2 stepping and the Phenom II x4 965 C2 stepping have different architectures? LOL

No, and you should have stopped there whilst you were only so far behind.

The rest of your post is just too dumb for me to bother with.

 

[Chad Boga]

Architectures can allow for a certain clock speed range (operating range) within certain specified conditions (voltage, heat etc).

Which was what I was referring to.

Now if we look at it all from the perspective of an architecture being engineered with a certain range of clocks speeds in mind we undoubtedly must confess that overclocking yields would therefore be the best indicator of what a given architecture is capable of achieving in relation to a clock speed (and use those yields in order to form a real world statistical analysis via a scientific poll).

CPU manufacturers have to take into account longevity of product, reliability in worst case scenario's, power and heat consumption, overclocking is not valid as a sign of architectural superiority, as it doesn't take into account these things to the same degree.

How long do these CPU's overclocked to 4+GHz actually run for at that speed, compared to how long they would run at their default speed.

What validation process do these overclocked CPU's go through compared to the validation process they are put through at their rated speeds by manufacturers?

 

[enzo matrix]

No, and you should have stopped there whilst you were only so far behind.

The rest of your post is just too dumb for me to bother with.

There is no need to insult me, sir. It's sad that you think helping others learn about computers is a "bother".

Netburst was an architecture that the pentium 4 CPUs used. Core was Intel's architecture after 2006.
So let me be clear, your point is that an architecture's superiority is based on the performance per clock, times the fastest clock speed the manufacturer has released it's CPU at?

 

[ElMoIsEviL]

Which was what I was referring to.

Now if we look at it all from the perspective of an architecture being engineered with a certain range of clocks speeds in mind we undoubtedly must confess that overclocking yields would therefore be the best indicator of what a given architecture is capable of achieving in relation to a clock speed (and use those yields in order to form a real world statistical analysis via a scientific poll).

CPU manufacturers have to take into account longevity of product, reliability in worst case scenario's, power and heat consumption, overclocking is not valid as a sign of architectural superiority, as it doesn't take into account these things to the same degree.

How long do these CPU's overclocked to 4+GHz actually run for at that speed, compared to how long they would run at their default speed.

What validation process do these overclocked CPU's go through compared to the validation process they are put through at their rated speeds by manufacturers?

What you are doing in that case is that you're assuming that all of the players in question (in this case AMD and Intel being the two major players) operate by the same rules.

If anything, Intel's process technology is superior to that of AMD and this does show when you take retail samples from both and overclock them.

That having been said the way both AMD and Intel calculate the other requirements (Heat, Power Consumption etc) is quite subjective. There is a lot of talk about the variances in how AMD and Intel differentiate when it comes to TDP.

Therefore to assume that a Q9650 is only clocked at 3GHz because of architectural setbacks which don't affect the Phenom II X4 (which currently is clocked up to 3.4GHz with the 965BE) is illogical.

Architecture wise both Penryn/Yorkfield do and can, on average, reach far higher operating clocks using the same or equivalent air cooling method than the Deneb architecture from AMD.

This is even more true, statistically speaking, when we compare Nehalem to Deneb.

Current market incentives are pushing AMD to clock Deneb higher in order to keep up with the competition while these same market incentives are pushing Intel to do the opposite.

This results in lower yields for AMD as well as more defective products but a more competitive product. For Intel this results, in theory, to higher yields, less defective products and higher profits.

 

[cjl]

There is no need to insult me, sir. It's sad that you think helping others learn about computers is a "bother".

Netburst was an architecture that the pentium 4 CPUs used. Core was Intel's architecture after 2006.
So let me be clear, your point is that an architecture's superiority is based on the performance per clock, times the fastest clock speed the manufacturer has released it's CPU at?

Netburst was the architecture used by late P4s. After Netburst came Core, and after Core came Penryn (aka 45nm Core 2, although Penryn was still somewhat under the overall "Core" umbrella). Then Nehalem (45nm i7/i5), and now Westmere (32nm i3-i7, which is still somewhat within the "Nehalem" category).

 

[Chad Boga]

What you are doing in that case is that you're assuming that all of the players in question (in this case AMD and Intel being the two major players) operate by the same rules.

If anything, Intel's process technology is superior to that of AMD and this does show when you take retail samples from both and overclock them.

That having been said the way both AMD and Intel calculate the other requirements (Heat, Power Consumption etc) is quite subjective. There is a lot of talk about the variances in how AMD and Intel differentiate when it comes to TDP.

Therefore to assume that a Q9650 is only clocked at 3GHz because of architectural setbacks which don't affect the Phenom II X4 (which currently is clocked up to 3.4GHz with the 965BE) is illogical.

Architecture wise both Penryn/Yorkfield do and can, on average, reach far higher operating clocks using the same or equivalent air cooling method than the Deneb architecture from AMD.

This is even more true, statistically speaking, when we compare Nehalem to Deneb.

Current market incentives are pushing AMD to clock Deneb higher in order to keep up with the competition while these same market incentives are pushing Intel to do the opposite.

This results in lower yields for AMD as well as more defective products but a more competitive product. For Intel this results, in theory, to higher yields, less defective products and higher profits.

Shipping products at their rated speeds are real, the rest is conjecture.

 

[ElMoIsEviL]

Shipping products at their rated speeds are real, the rest is conjecture.

But this entire topic is not about shipping products.

This entire topic is about the architectures behind current shipping products.

 

[enzo matrix]

Netburst was the architecture used by late P4s. After Netburst came Core, and after Core came Penryn (aka 45nm Core 2, although Penryn was still somewhat under the overall "Core" umbrella). Then Nehalem (45nm i7/i5), and now Westmere (32nm i3-i7, which is still somewhat within the "Nehalem" category).

Thanks.

 

[Chad Boga]

But this entire topic is not about shipping products.

This entire topic is about the architectures behind current shipping products.

Yes, but clockspeeds are still a factor of the architecture.

In the past we have had DEC's Alpha chip squaring off against HP's PA-Risc chip, where they had similar performance but a massive difference in clockspeed due to the intentional design choices the respective architects took.

Sometimes they go the speedrace route and other times they don't, but one shouldn't be ignoring achievable shipping clockspeeds as an aspect of the architecture.

 

[ElMoIsEviL]

Yes, but clockspeeds are still a factor of the architecture.

In the past we have had DEC's Alpha chip squaring off against HP's PA-Risc chip, where they had similar performance but a massive difference in clockspeed due to the intentional design choices the respective architects took.

Sometimes they go the speedrace route and other times they don't, but one shouldn't be ignoring achievable shipping clockspeeds as an aspect of the architecture.

I think you're missing the point and perhaps it might be best for you to sit down and reflect on it some more.

Shipping clock speeds are a factor of an architecture in the sense that the shipping clock speeds are within acceptable parameters (subjectively determined by their maker).

Would AMD prefer to sell a Phenom II X4 965 at a clock speed of 3GHz (assuming it performed better per clk and could compete with the competition at those speeds)? The answer is most definitely yes.

Can a retail sample of a Phenom II X4 scale higher on average in terms of clockspeed (headroom wise when overclocking) than a Yorkfield based Core 2 Quad? The answer is most definitely no.

When Intel launched it's Yorkfield based Core 2 Quad lineup... what did the competition have to counter? The answer is most definitely lower clocked original Phenom architecture based processors (Barcelona).

What incentive did Intel have to release a higher clocked variant of their Yorkfield Core 2 Quad lineup when the competition didn't have anything to compete? None.

Historically speaking, when the competition did have something to compete.. how has Intel behaved? Well if we go back to the Pentium !!! Coppermine, Intel pushed the acceptable limits of their 0.18u architecture in releasing a 1.13GHz variant (which was subsequently recalled) in order to compete with the AMD Athlon (K7) Thunderbird processors.

So if Intel's past behavior is any indication, they will push the envelope when absolutely needed. If we couple that with the fact that selling lower clock variants of processors when your competition has nothing to offer to counter results in better yields (which means less defective units to trash) and as such higher profits and lower rates of defective units after the fact then one could conclude that this is most probably why Yorkfield was clocked as low as it was.

Not because the architecture couldn't scale higher in terms of clock speed but rather because there was no "need" for it (no incentive).

If you can't see the obvious logic behind my entire thought process then yes.. take a breather and think about it for a bit.

 

[Chad Boga]

I think you're missing the point and perhaps it might be best for you to sit down and reflect on it some more.

Shipping clock speeds are a factor of an architecture in the sense that the shipping clock speeds are within acceptable parameters (subjectively determined by their maker).

Would AMD prefer to sell a Phenom II X4 965 at a clock speed of 3GHz (assuming it performed better per clk and could compete with the competition at those speeds)? The answer is most definitely yes.

Can a retail sample of a Phenom II X4 scale higher on average in terms of clockspeed (headroom wise when overclocking) than a Yorkfield based Core 2 Quad? The answer is most definitely no.

When Intel launched it's Yorkfield based Core 2 Quad lineup... what did the competition have to counter? The answer is most definitely lower clocked original Phenom architecture based processors (Barcelona).

What incentive did Intel have to release a higher clocked variant of their Yorkfield Core 2 Quad lineup when the competition didn't have anything to compete? None.

Historically speaking, when the competition did have something to compete.. how has Intel behaved? Well if we go back to the Pentium !!! Coppermine, Intel pushed the acceptable limits of their 0.18u architecture in releasing a 1.13GHz variant (which was subsequently recalled) in order to compete with the AMD Athlon (K7) Thunderbird processors.

So if Intel's past behavior is any indication, they will push the envelope when absolutely needed. If we couple that with the fact that selling lower clock variants of processors when your competition has nothing to offer to counter results in better yields (which means less defective units to trash) and as such higher profits and lower rates of defective units after the fact then one could conclude that this is most probably why Yorkfield was clocked as low as it was.

Not because the architecture couldn't scale higher in terms of clock speed but rather because there was no "need" for it (no incentive).

If you can't see the obvious logic behind my entire thought process then yes.. take a breather and think about it for a bit.

Shipping clockspeeds are real, everything else is conjecture.

 

[outlw6669]

Wow, looks like I missed all the fun!
Stupid sleep

I won't even TRY to post those thousand benchmarks that prove you wrong over and over because I see now that it is pointless. You just continue being irrational and post your own biased benchmarks.

Sorry, I fail to see how I am being irrational
Looking thorough subjective benchmarks and drawing conclusions from imperial data seems pretty rational to me.

Please DO post some of your 'thousands' of benchmarks.
If I find more or better information supports your stance, I would be happy to modify my views.
My views are based on facts and measurements, not any type of bias for any company.

Phenom II is faster than all Core2's. Fact, and an undeniable one. Of course, there are always exceptions, but sometimes PII is even faster than I7 975 and nobody is counting that, so there is no reason to count exceptions on any side.

[:bohleyk:7]

Got proof?
Real benchmarks are preferred.
A small collection of cherry picked Phenom II wins will not work to support your claims.
Remember, whe are comparing all CPU's at the exact same speed...

Shipping products at their rated speeds are real, the rest is conjecture.

Yes, but clockspeeds are still a factor of the architecture.

Shipping clockspeeds are real, everything else is conjecture.

You seem to be missing or distorting the point of this thread.
While it is somewhat true that the CPU speed is tied to the architecture, that is not what is being compared.
We are looking at the efficiency of these architectures, not performance by released modles.
By efficiency we are specifically comparing how the different architectures perform on a clock-for-clock basis.
That is to say, when all the architectures are running under the same conditions (matching the Clock Speeds, RAM, GPU, HD, etc. as closely as possible), which architecture is most efficient (can get the most work done in a set number of cycles).

 

[warmon6]

You seem to be missing or distorting the point of this thread.
While it is somewhat true that the CPU speed is tied to the architecture, that is not what is being compared.
We are looking at the efficiency of these architectures, not performance by released modles.
By efficiency we are specifically comparing how the different architectures perform on a clock-for-clock basis.
That is to say, when all the architectures are running under the same conditions (matching the Clock Speeds, RAM, GPU, HD, etc. as closely as possible), which architecture is most efficient (can get the most work done in a set number of cycles).

+1.

Hey chad,

What where talking about is how efficient is the architecture is.

Example, Lets say i have an core 2 quad and a core i7.

We know the core i7's beat the core 2 quads at the same GHz but we want to know how much more efficient is the "Nehalem" architecture over the "Core" architecture.

http://www.anandtech.com/bench/default.aspx?p=47&p2=51