Megahertz Myth

Thx for the reply. I was thinking compared to the whole package, meaning a litany of stress test in gaming, video encoding, etc.

So Intel is faster, or "better" at everything, but is it really worth paying 2, 3, or 4 times the price? for a difference that is impossible to notice without some benchmarking program?
and I'm guessing intel scored better because Intels cpus are "quad-pumped" opposed to amds cpus only being "double-pumped"

Huh, alright, Thank you. Apparently my knowledge was outdated. So your post confused me, what is it currently?

Ok, so currently, Intel is a tiny bit faster?

But really, logically, its not worth paying about 2 or 3 times the price?
(i realize that its almost all personal oppinion on that)

We were just discussing the differences between amd and intel. And thanks.

It is hard to give an exact answer, because it will depend on the app, but in general an i7 is 25%+ faster than a similar clock on an AMD. So, a i7 920 at 2.7 is roughly the same as a an AMD at 3.3. BUT, some apps will exaggerate the difference and make it more like 75% faster, and some will minimize the difference.

So you're saying that the ratio of AMD clock frequencies vs Intel clock frequencies changes depending on what benchmark is used? I'm looking for an overall general easy to remember equation that probably isn't totally accurate but close (i.e. AMD clock frequencies times 1.5 equals Intel's clock frequencies.)

Good point, and I'm sure that my equation that I am looking for is probably not linear either

uhmmmm.....am i invisible? I'll try again. 25%+ is a good base to look at (not counting turbo). So your formula would be i7 is 1.25 faster than phenom II....but understand that will vary greatly (probably +75% /- 10%) depending on the app.

Belial, I did get what you said, I was simply responding to boulard because I wasn't sure he understood what I was asking. Thanks for your input.

I know the OP is asking about Core i7 : Phenom II...
But how about Phenom II : Core 2 Quad? Is it almost 1:1?
I think I've read something like Phenom II 3.0 GHz: roughly the same as Core 2 Quad 2.8 GHz; but I may be wrong here.

Intel still wins with the c2q, but not by as much. c2q also lose to the i7 clock for clock.

I'll make this easy for you:

1.) That thing about AMD processors having more performance per clock? It went away when Intel switched from P4 architecture back to enhanced P3 architecture. Back then it wasn't that AMD's were so good, it was that P4's were so bad.

2.) Core 2 Duos perform a little bit better than Phenom II Dual Cores overall, clock-per-clock. But since they don't outperform the Phenom II at everything, it's hard to quantify.

3.) Core 2 Quads outperform Phenom II X4's by a little bit on average, but again because they don't win at everything its hard to quantify

4.) Core i7's outperform Core 2 Quads in some things, but not in others. They're about 2 performance factors ahead of AMD, since the Core 2 Quad is one performance factor ahead of AMD. But once again, it's hard to quantify.

5.) Intel's current architectures overclock a little bit higher than AMD's current architecture, so you can get both better clock-for-clock and higher clocks with Intel.

6.) Because of all those little "in-general" and "on average" leads Intel has, AMD is practically giving away it's chips. You often get a better value with AMD.

I'll explain the whole Megahertz myth:

The easiest way to put it, is when it comes to speed of execution, there are three factors:
1: How often the component can LOAD/SEND data (Ghz)
2: How much data can be transferred/stored at one time (Bus Width)
3: How much data can be executed per clock cycle (Instructions Per Cycle)

The Pentium4, for instance, went with a high GHz speed in an attempt to overcome a low Bus Width and low ICP. The Athlons went for a lower Ghz speed, and higher ICP in response.

The Core2 line, on the other hand, has a much higher ICP, and eventually reached P4 levels when it came to GHz. Even on one CPU, a duo would beat a P4 due to its higher ICP.

GPU's, on the other hand, went with a far lower GHz speed, but has a much larger data bus (256/512 instead of 32/64) to achieve simmilar speed of execution.

In terms of current performance, I'd argue that late C2Q's have a slightly lower ICP then i7's, although I'd be interested to see head to head comparisions between a i7 920 (2.6GHz) with Hypterthreading off and a Q9400 (also at 2.66 GHz). Keep in mind, the diffrences in cache size/structure, and tri-channel RAM can also effect the results. In apps that don't take advantage of i7's hypterthreading, it appears that late C2Q's are even with i7 most of the time.

Phonom II's (on AM2+ platform) match against later C2Q's (8000-9000 series). Based on subjective analysis, I'd say Phenom II's have a lower ICP then C2Q's, but its hard to quantify, as some apps clearly favor one platform over the other. The difference is neglegable in any case, as a 2.66Ghz Phemon II is a good match against a 2.66 C2Q.

Yes, hyperthreading leads to better performance, but you can't compare 8 to 4. I want to see how a i7 at 2.66GHz running up to 4 threads competes against a 2.6GHz C2Q running up to four threads. This is the only way to compare ICP between the two processors, and thats to use the same exact specs.

(One stick of 4GB of RAM at 1066Mhz would be used, to eliminate dual/tri channel and memory speeds from the comparison as much as possible)

Ive seen lots of PPL disabling HT and have better performance while gaming but encoding and lots of apps are faster if HT is enabled. Remember this is NOT 8 core, but 8 thread.

The example I gave is my first post is about as extreme in difference as it gets.
Core i7 9xx simply doesn't make sense for gamers while pure performance, performance/$ and performance/watt in the area of massively multithreaded environment blow the competitors away.

While generalisation is an easy thing to remember, one must understand this is the worst way to go when making purchasing decisions.
This applies to all other things in life.

*bastage*

Looking at this link
http://www.wepc.com/discussions/vi [...] U_Industry
which addresses much of this, the one place I disagree with on Anand is his speaking to the ease in which to make those decisions, where if anything , I think Intels greatly muddied the waters, be it by pricing, new skt or upgrade path

It's not the bus that makes the difference (though GPUs do have a huge appetite for memory bandwidth), it's the huge number of 'cores' inside the GPU. If you're filling a 100-pixel triangle you can use 100 'cores' to process the rendering for each pixel individually and get it all done 100x faster than if the GPU only processed one pixel at a time.

CPUs tend to be mostly cache with a few processing cores, whereas GPUs tend to be mostly processing cores with a bit of cache (though recent CPUs with integrated memory controllers require less cache for optimum performance and hence are heading in the GPU direction).

Complexity is the current issue, as cpus complexity actually gets in the way.
LRB will be emulating alot of what it does, simply because of this. Having all the extras of a cpu,LRB and its makers/designers, are actually trying to keep them as positives, having all that cache etc, tho, the core count and the latency (core clocks can be made up to account for this) can be solved, its just the drivers that needs work.
Id point out tho, starting with the 200 series, gpus are also heading in the crossover area, having more cache etc, plus adding more cores, and using DP, which is mainly for gpgpu functionality

"Core i7 platform is a fantastic platform for multi-GPU gaming as you'll gain heaps... seriously heaps and heaps more performance opposed to today's regular PCs with a Core 2 Duo processor".....but these guys arguing for their dual cores won't let facts get in their way. They are convinced dual core processors are all they will ever need, while those of us who have stepped up to quad cores know the truth.
So just one more time and then I'll give up. Quad cores will outperform dual cores clock for clock, or even slightly slower clocked, in EVERYTHING including gaming. So there really should be no question about performance. If you want to argue which system you can build cheaper feel free to have at it. But please stop spreading the misinformation that dual cores are "better" at gaming than quad cores.

A good gamer has his system optimized for gaming, that means as little as possible going on. Most games dont benefit from quad core. Dual cores clock higher than quads.


^^^ All the above is true and factual. People have to make up their own minds