I agree when it comes to sheer action gameplay. Metal Gear rising was a much better hack n  slash game than any GOW, but it loses badly in pretty much all other aspects

The issue with Espresso, is that paired singles is something completely different than any other forms of SIMD (it's not even "real" SIMD), and porting heavy FP code from a 128-bit SIMD unit to a 32-bit x 2 Paired Singles isn't an easy operation. It's not like it can't do it, it just won't do it in the same way, nor at the same amount. And the fact that there's only 3 cores is a limiting factor. The amount of stuff it can do is limited, but it's no way a terrible architecture at all. It running several modern games using 32-bit x 2 paired singles is impressive. It's surprising that it was able to do that well in the matmul SIMD tests, something it shouldn't be able to do amazingly.

That core surprises people like this:

http://www.radgametools.com/bnkhist.htm

• Added Wii-U support for Bink 2 - play 30 Hz 1080p or 60 Hz 720p video! We didn't think this would be possible - the little non-SIMD CPU that could!

Something like Espresso, can decode Bink 2 @ 1080p/30 or 720p/60, which is a pure SIMD instruction, even though Espresso doesn't even have real SIMD. I'm pretty sure a more modern CPU with proper SIMD instructions would be able to do this normally, but Espresso being able to do it AT ALL is surprising. Who knows what sorts of things this thing would be able to do in the future.

No, 32-bit x 2 paired singles is not as good as having a true 128-bit SIMD, but that doesn't mean we won't be seeing amazing physics, just not at the same levels as PS4/X1. But calling it "weak" or "terrible" is an understatement. With respect to multi-hundred dollar CPUs, yes, it's a joke, all 8th generation console CPUs  are terribly under-powered compared to even some lower-end CPUs on the market, and that's a shame. Core-for-core, Jaguar isn't going to do stuff that Espresso couldn't handle one way or another (it could do it, just not at the same level), and that won't make downports from next gen impossible, just difficult. Much better than trying to port from Xenon to Broadway though! 

Anyway, Shin'en had better show something good for their next game, because talking "big" like that just has to be supported by actually doing something nice. Nano Assault looks okay, and it's cool that it runs with only the main CPU core, but it's not gonna impress many people. They had better use Tessellation well in their next game, not just for small things. 

Bulldozer/Vishera's IPC is actually lower than the Phenom 2's. Both Stars and Bulldozer are slower in terms of IPC than the Core 2 series, which puts performance into perspective, of course the Phenom 2 will switch into another gear when you pump up the NB clock by roughly 33%, which can yield upwards of 15-20% in IPC increases.

Jaguar has more in common with Brazos than Bulldozer anyway if anything Brazos was an evolutionary/reworking of the K10h architecture.
When comparing Jaguar to Brazos, it's the same 2-wide, L1 cache with the same execution blocks as Brazos it's an evolutionary step, not a revolutionary one, which is standard with PC processors.
AMD went about improving performance by throwing in a loop buffer, improving the cache predicter, improved instruction buffer and providing more instructions than you can poke a stick at.

One take away from it all is that, Brazos had a very poor floating point unit, AMD fixed that with Jaguar by doubling up on essentially all of the FP execution blocks, however what most people don't realise is that, game engines don't only use Floating Point math and CPU's don't just deal with Floating Point, so using FP numbers a way to determine a CPU's performance is completely and utterly pointless.

Jaguar is a 4-core CPU, with 2MB of cache and performance enhancements with new instructions added from Bobcat, that's it. Obviously a Jaguar CPU alone would best a Bobcat CPU since there are more cores, and they have 15% higher IPC, 128-bit SIMD instead of 64-bit, and new instruction sets, but this is NOT going to do much to increase overall performance per-core the way you think. You will not see anything near double the performance out of this because it's only the SIMD width that increased by 2. Real world performance (factoring everything in), will be a bit over 50% more powerful than Bobcat. Again, I'm talking about "per core". 

And an 8-core Jaguar goes at 102.4GFLOPS. 2 cores are reserved for the OS in PS4/X1. 8 FLOPS per cycle, per core. 8 x 1.6ghz = 12.8 GFLOPS per core. 6 usable cores = 76.8 GFLOPS for PS4 CPU, 84 GFLOPS for the X1 CPU (8 x 1.75ghz x 6)

Xenon is based on PPE, which is a PowerPC970, which is a G5. The PPE in Cell is a PowerPC970, which is a G5. G5s were designed to have high clocks, not so high IPC, improved SIMD performances, but they are in-order processors. Here's a link with a good explanation for both:

http://lowendmac.com/ed/bashur/12db/g5-gaming.html

No I can't, because no one has tested Jaguar yet. But if we look at this:

http://semiaccurate.com/assets/uploads/2012/08/slide-1-728.jpg

We can take into account the enhancements from Bobcat, and easily make the calculations from the test blu made in NeoGaf. Those numbers I gave you for Jaguar aren't going to change much (they are probably higher, but likely only about 15-20sh % higher due to the 128-bit SIMD, newer instructions and increased cache). You can also add a bit to the numbers on Espresso as well, considering how there is more cache in it compared to Broadway and it's lower-latency cache eDRAM (I based the Espresso numbers on a 256KB Broadway. Remember that 2 Espresso cores have 512KB of cache, and one has 2MB of cache (as much cache as one Jaguar CPU), that will also add a bit to performance, probably around 5%+/-). But still, the fact that it's that close to Jaguar core-for-core says a lot about how efficient Espresso is. Of course, the advantage the PS4/X1 CPUs will have over Espresso is obviously core count, newer instructions and a 128-bit floating point unit, but core-for-core, it's not goint to blow Espresso away. 

But anyway, this isn't a thread to discuss the CPUs in PS4/X1, this is about Wii U, so I'm gonna stop here. 

He has a point. We can ignore the CPU to some degree since it's part of how the system has been designed (virtually no CPU bottleneck) so it comes down to GPU and RAM.

The GPU alone is above the PS3 and X360, with two thirds of it identified, along with more modern API. The processors and shaders are more advanced, naturally.

The RAM amount is more than double PS3 and X360 when it comes to game development (~~448MB vs. 1024MB). Additionally, if developers use the eDRAM (32MB as opposed to the X360's 10MB) to buffer as much as possible then they can offset plenty of the bandwidth that looks lacking.

Comparing to PS4 and X1 is a different story.

something somethin Red Dead Redemption !!1!!

something something The Last of Us vs. Super Mario Bros. U !!1!11!one !!

I agree such comparisons can be very ignorant since there aren't many multiplats or they simply don't exist that do the WiiU justice in comparison to PS3 and X360. Time will change that, and not to mention inevitable Nintendo 1st party IPs such as Metroid, or even Mario Kart 8.

the reason is  nintendo is embarrassed  to show there specs , its a 160 sp gpu, and nintendo wants to keep it a secret, no other company hides there specs like nintendo, and nintendo only started hiding there specs, with the wii brand, and the reason is, because the specs are embarrasing compared to the competition