Ranked CPU List

[Apokalipse]

I think the Phenom II 940 is between the Q9400 and Q9550 in most cases, anyway.
The Phenom II 955 would be at or slightly above the Q9550 (It can depend if it's using DDR2 or DDR3)

 

[zedman3d]

I don't think a CPU ranking can really work. Performance between two different CPU's depends on a lot of different things.

It can depend on the type of instructions being executed, the compiler used, how sensitive the program is to RAM speed and timings, whether an IMC gives it a benifit, whether large cache helps, predictability of instructions by the prefetching algorithms, what RAM and motherboard it's being paired with, how well it scales with frequency, the chipset drivers and BIOS used, the operating system and what it's compiled with and how it's compiled (especially for GNU/Linux or OpenBSD).
The performance of multi-core CPU's varies depending on how well threaded the application is, and sometimes how fast the intercore communicate is (especially for multi-socket systems).

I disagree, because the points you made can be made towards the graphics cards list.

- Depends what game.
- Depends what driver.
- What Operating System.
- How well does internal SLI/Crossfire scale card to card.
- Etc etc...

All in all, a cpu list is doable and will remove a lot of the clutter in the hardware section.

 

[Apokalipse]

I disagree, because the points you made can be made towards the graphics cards list.

- Depends what game.
- Depends what driver.
- What Operating System.
- How well does internal SLI/Crossfire scale card to card.
- Etc etc...

All in all, a cpu list is doable and will remove a lot of the clutter in the hardware section.

Video cards have a much narrower scope than CPU's do.
Until recently, the only thing they've done is process graphics.

But CPU's have to deal with basically every kind of computation. There are far more variables and there is far bigger variation.

Yes, graphics cards have variation depending on circumstances. But they are much more consistent in general, due to their relatively narrow scope.

 

[zedman3d]

Video cards have a much narrower scope than CPU's do.
Until recently, the only thing they've done is process graphics.

But CPU's have to deal with basically every kind of computation. There are far more variables and there is far bigger variation.

Yes, graphics cards have variation depending on circumstances. But they are much more consistent in general, due to their relatively narrow scope.

This ranking is based off an entire communities knowledge. I understand what you are saying though and ai do agree to an extent.

We all know how one CPU ranks versus another, and if i dont know, Hikari knows, if he doesnt know, you know, if you dont know S0ULphIRE knows. People who actually have had experience with that particular cpu greatly help with determining their general sort of performance.

 

[hikaricloud]

Taste, I don't mind editing the first post, just let me know how you want it done.

Providing the thread stays civil of course.

 

[Apokalipse]

This ranking is based off an entire communities knowledge. I understand what you are saying though and ai do agree to an extent.

We all know how one CPU ranks versus another, and if i dont know, Hikari knows, if he doesnt know, you know, if you dont know S0ULphIRE knows. People who actually have had experience with that particular cpu greatly help with determining their general sort of performance.

I think you're still missing the point.
Sure, you can compare CPU's; When you consider the context in which they're going to be used.

You can't make a system to say that CPU's are ranked absolutely in a certain order; it doesn't work like that.

 

[hikaricloud]

But that's what multiple benchmarks are for, to weed out as many possible instances where the CPU will be performing. There are processors that outperform others at stock in a multitude of benchmarks, hence they are better in general.

Although I do agree that ranking them is a bit hard to do, because it really depends on which benchmarks you go by. For example, the link that Soul posted only goes by one benchmark. In other situations, some of those CPUs would outperform ones higher up on the list. It's just a matter of the amount of benchmarks to judge performance.

 

[Apokalipse]

But that's what multiple benchmarks are for, to weed out as many possible instances where the CPU will be performing. There are processors that outperform others at stock in a multitude of benchmarks, hence they are better in general.

Of course there are instances where a CPU will be clearly better than another. That's not what I'm disputing.
But an absolute measurement based on arbitrary criteria isn't going to be useful in cases where somebody intends to use it in circumstances different to the arbitrary criteria used, and the performance difference is large enough to shift the rankings.

And there will be a lot of those cases. Way more than with video cards.

 

[Luke]

this was suggested a while ago but it did not work out because of these vary reason. Including apok's argument. At stock the performance between tasks on a cpu can vary a lot. Overclocked it can also. The fact is you can't really compare oced cpu's beacause you would need the exact identicle hardware for each oc. But that does not work in cases of scoket type. Then on top of that one cpu may be leaps and bounds better then another with just a different set of ram, or another different motherboard, ect. But if you include that in then that is not fare to other cpu's who don't get the change of hardware. So then if you use the hardware that is ideal for every cpu then the comparison between them becomes irellevent because of the large wake of different hardware being used. You can not compare them.

 

[synergy]

I agree with Apok. The only way to really rank CPUs is to have them in standardized configurations where there are no other variables. Sure GPUs have the same variables with the programs that are giving the instructions, but every GPU is standardized wheras the CPU is so dependent on all the other components. You'll notice whenever a reviewer tests a product they do it on the same mobo, ram, etc.

With this in mind all you need to do to rank these is to look at the clock speeds and probably even price. To me, that really isn't very useful information. While an overclocking ranking would be more controversial I think it would be much more helpful and meaningful in this regard.