Is this why AMD has been aggressively promoting Zen?

[Glo.]

You don't understand CPU architectures but you know this is important how? Bigger numbers are better?

What's faster, a car with an 8 cylinder engine or a car with a 4 cylinder engine?

Worst analogy ever, because it depends on a mass of things. Weight of the cars, turbocharging/supercharging the engines, quality of transmission, power output of the engines...

Do you even know what I am implying in my posts? Or are you trying to defend idea that AMD cannot by any way bring good CPU architecture design?

I am just saying that currently... we do not know anything at all. On paper, IPC of Zen should be similar to Broadwell, because of the things that I pointed in that post, IMO. However, as I mentioned, we do not know the final core clocks for Zen, we do not know the bandwidth of the caches. And then there is Blender test, in which Zen edged out Broadwell in the same environment...

Never judge future, just by looking at the past. Nobody predicted that Nvidia was able to achieve massive jump in efficiency with Maxwell GPUs on the same node as Kepler. Same thing can happen with Zen.

 

[IEC]

You don't understand CPU architectures but you know this is important how? Bigger numbers are better?

What's faster, a car with an 8 cylinder engine or a car with a 4 cylinder engine?

My 4-cylinder turbo is quite a bit faster than almost every 8-cylinder out there... That's a terrible analogy.

 

[crashtech]

My 4-cylinder turbo is quite a bit faster than almost every 8-cylinder out there... That's a terrible analogy.

I actually think that was what he was pointing out, but not 100% sure.

 

[Phynaz]

I actually think that was what he was pointing out, but not 100% sure.

Yep.

That was my point. Glo is trying to predict Zen's performance vis a vis Intel CPU's by counting resources. Something like Zen has 95% of the registers therefore it will perform at 95% of the Intel CPU.

It's the same thing as saying a four cylinder engine will perform at 50% of an eight cylinder engine.

I get it, it's easy to compare numbers, but it's apples and oranges.

 

[DrMrLordX]

I like how OP rather explicitly accused AMD of market manipulation--which is fraud--and still hasn't been tagged with a trolling/locked thread

It's just human nature. They have their Chevy and AMD fans have their Ford. Any chance they have to run down the opposition, they'll take it. Victory at any cost.

Hasn't Zen already proven to have better than SB IPC?

Ehhh jury's still out on that one. The picture is starting to crystallize, and AMD has shown that Zen/Summit Ridge can flex its muscles in some fp workload using an unknown selection of SIMD ISAs probably restricted to 128-bit vectors. But the AotS numbers show a different picture, and the Geekbench scores . . . yeah we need more hands-on benching to be sure of things.

What's faster, a car with an 8 cylinder engine or a car with a 4 cylinder engine?

In all fairness to Glo, he is doing more than just counting cylinders. He's counting cylinders and tallying bore, stroke, and compression ratio as well . . . and that gives you a better picture of potential performance.

 

[Einzige]

As a Christian scientist, I find Zen to be blasphemous.

I will not be purchasing.

 

[krumme]

Even at a discount?

 

[Dresdenboy]

You don't understand CPU architectures but you know this is important how? Bigger numbers are better?

What's faster, a car with an 8 cylinder engine or a car with a 4 cylinder engine?

This is not what Intel said. Each of Glo.'s cited numbers correlates with Intel's view, that more is better - as long as we're discussing IPC and these amounts don't need additional pipeline stages in the performance related paths. Other than that we might argue, that uops might be different.

You could have simply stated, that too many of those could either reduce achievable clock frequencies or cause more pipeline stages with a possible performance hit/IPC reduction.