Are XBONE graphics closer to Wii U than PS4?

[Pemalite]

ICStats said:

Pemalite said:

ICStats said:

Pemalite said:

dsp333 said:

As weak as it is (not that the PS4 isn't weak as well), no, not even close. The One is still over 4 times the power of the Wii U by its weakest measure.

The Wii U is more in line with the 360 and PS3 than it is the One and the PS4 as pathetic as that is. In fact, by some measures like ram bandwidth and CPU before GPU assistance, the Wii U is actually WEAKER than those machines approaching 8 and 9 years old.

No way. The WiiU's GPU is superior to that of the Xbox 360 and Playstation 3.
The CPU is also superior being an Out-of-Order design.

You're looking at probably 2x the maximum performance of the PS3.
However it's not in the same league as the Next-gen twins.

There is a massive performance disparity between the 7th gen and 8th gen, thanks to the continiously advancing PC technology found in all systems. (And because that generation dragged for so damn long the difference is more pronounced.)

The GPU is superior, but the CPU is inferior.  OoO doesn't make up for the lower clock rate and fewer cores than PS3.  It may be close, but lower.

OoO just makes it easier to get the most out of the cores, but PS3 & XB360 developers already have had 8 years to optimize their code so they are getting the most.  

Kay. Grab one piece of my post, take it out of it's context and disregard everything else I said which also applies.

The PS3 and Xbox 360's CPU's have more in common with an Intel Atom processor.
The Wii U's CPU thanks to the shorter pipelines *and* everything else I stated has more in common with a fully fledge Core based processor.
(Albeit, this point is relatively moot, they have commonalities, but are still very different, but you get the idea.)

The killer in the Wii U's heel is the lack of SIMD's.

Also more cores =/= more performance.
The Cell processor was a poor performer even on release.

A Core i5 4670 which is quad core is faster than an AMD FX 8120 which is an 8 core.
An AMD FX Quad-core running at 4ghz is slower than a Core i5 4670 Quad-Core running at 3ghz.

Kay, I was just commenting on what you said about the CPU.

"The killer in the Wii U's heel is the lack of SIMD's"

OK, well that would do it.  Similar problem to the FX really.  Fewer execution units than an i5.

"Also more cores =/= more performance."

Sure, not without context.

Plus you need to factor in the workload.

At general processing the PS3 was inefficient, and hard to optimize.  For a lot of general game code an i7 OoO, better branch prediction, huge caches, etc. would win easily.  For stream processing the PS3 Cell still rivals Haswell CPUs today thanks to high SIMD throughput, programmable async DMA, 128 register files, better SIMD instruction set, predicated branching, etc.  Code developed specifically for that could beat an i7.  But that is a narrow area like image processing, or doing the GPU's job.

Anyway, it seems like we're kind of arguing the same point.

Nah. The Cell was only good when using iterative refinement in floating point.
Game engines use differing types of math depending on the object, the scene and even the types of A.I, it's never a static environment.

A good analogy would be the Cell travelling down a freeway at 50 miles an hour, on the odd occassion it might hit 150 miles an hour. - A Core i7 however would be sitting at 100 miles an hour during the entire trip.
The i7 would reach the finishing line first simply because it can simply sustain it's speed, regardless of road conditions or weather.

As for Stream processing, the little IGP's that Intel and AMD bundle for "free" with CPU's these days, far outclass the PS3, it's a GPU's forte'.

As for SIMD, nothing Cell has touches AVX and AVX2 (Which also has a 256bit register file) which is found in Haswell and Jaguar, besides not all tasks benefit from SIMD instructions simply because they can't be vectorised to well. (This plays into the highway analogy.)
Intel also has the advantage of macro-op fusion.

As for Predicated branching, well. Intel has such functionality in it's advanced branch tree predictor, Intel hasn't been beaten on this front since the Core 2 days.

All in all, Cell was beaten by the Core 2 CPU's when they were released, the sheer amount of transisters aren't there to look pretty and you would be hard pressed to convince anyone that the Cell could beat something like a 6 Core/12 Threaded Core i7 Sandy/Ivy-Bridge-E based processor. (Which in some cases can be beaten by a Haswell Quad.)
There is a reason why Intel can charge $600-$1000 or more for a CPU and get away with it, there is simply nothing, not even in the server space offered by IBM that can touch Intel in sheer performance.