darkazcura
Member
In terms of features it surely is, but in terms of power, it is not.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Shin'en Interview on WiiU Hardware: GPU "several generations ahead of Current Gen"
- Thread starter The_Lump
- Start date
In terms of features it surely is, but in terms of power, it is not.
JordanN
Banned
Has this been confirmed? We don't even know if Wii U has hull and domain shaders.
Just a tessellator hinted to be in the HD 2000/3000/4000 range. That would put it in the middle (although 360 was close to the HD 2000).
Nice ad hominem.
In some ways it is, in some ways it isn't. Power output isn't one of those things. I'll need to see the posts in question to know what they were talking about. There's a lot of stuff we still don't know.
Architecture wise it certainly is. Raw power wise it is closer to the previous gen consoles. If you don't understand that distinction there are plenty of places to read up on it, even here.
It's open GL 4.x compliant
And he didn't against other posters?
phosphor112
Banned
Some people think the gap between 360-Wii U is greater than the gap between Wii U and Xbone.
So yes, people have held onto that thought.
But I suppose they wouldn't be in "their right mind" anyway.
Isn't what power matters most? Sure it has new features and what not, but it still is only a small leap from current gen consoles in power.
Also, it's extremely low TDP won't all of a sudden put those new architecture features to good work and bring them anywhere near Next-Gen consoles.
darkazcura
Member
True. Info is still pretty scarce. I guess I argue on the side that it's not crazy to believe (or fanboyish) the featureset isn't far off from the One.
Haunted
Member
Even if they're not officially part of the company, I see their impartiality as roughly in the same place as those that are. They've clearly aligned themselves with Nintendo at this point and I treat their statements accordingly.
See, if you say the gap between Wii U and XBO is smaller than the gap between 360 and XBO, meaning that Wii U is more capable than 360, that is correct. That person is not necessarily saying Wii U is more comparable to XBO than it is to 360 in terms of performance.
Reading comprehension.
fallingdove
Member
Let's not pretend that Nintendo fans are the innocent party. I see comments all the time in threads like these where the first comments are - "See, we told you that the Wii U was on the same level as other next gen consoles." These comments are just as guilty of derailing.
H_Prestige
Banned
No. They just refuse to put it on the same level as ps360. Because that's insulting or something. It has to be like 2x current gen.
Oh, I forgot about the "architecture" argument. As if it matters. But hey, anything to avoid speaking of it in the same sentence as current gen I suppose.
PdotMichael
Banned
That's not that important though. The PowerVR SGX based chip of the VITA is more modern than the PS3 GPU, doesn't change the fact that putting the Wii U closer to the Xbox One is absurd.
freezamite
Banned
This affirmation doesn't make any sense. The HD5000 series started development before the HD4000 series was finalized, and the HD4000 series was finalized way before being at sale.M°°nblade;60180481 said:
I don't understand this necessity to downplay whatever facts could benefit Nintendo. EVERY GPU ON THE MARKET STARTS ITS DEVELOPMENT UPON SOME BASE MODEL WHICH OF COURSE IS AN OLDER MODEL, AND THEN EVOLVE FROM IT. AND EVERY GPU HAS A DEVELOPMENT CYCLE OF AT LEAST 2 YEARS (in the case of Nintendo, it may have been a longer development since they go for ultra-customized designs that diverge from standard designs a lot more than what is normal even for console vendors).
I can't understand why the fact that Nintendo started to work with it's GPU in 2008 makes impossible for them to go higher than any 2008 design, when some of the GPU you claim to be superior in terms of architectural advances started their development BEFORE and upon OLDER DESIGNS.
PdotMichael
Banned
It's cool how you didn't even need to read the post to have an opinion.
http://www.neogaf.com/forum/showpost.php?p=60021281&postcount=288
ColdBlooder
Banned
With "miles away" he propably means "Wii vs PS360" away. Wich was a gigantic power gap. The gap is still substantial this time. But not this insanely gigantic.
There's no need to start yelling. If you reread my post you'll see that I'm not downplaying anything since I address people on both sides claiming its either closer to the X360 or closer to the XBone. In my simple mind, the Wii GPU is as many years and generations ahead of Xenos as the Xbone GPU seems to be ahead of the Wii U GPU and it is as much directX feature set levels apart from the X360 as the Xbone.
The real problem is that the (usual) people claiming it is closer to one or another seem to have even less understanding of GPU technology than I do.
Nobody's saying it's impossible.
However, people do think it's weird to assume that, if Nintendo wanted to go higher than an 2008 design, why they have decided to pick up a 2008 GPU and hang some bells and whistles on it in the first place instead of just picking a 2009, 2010 or 2011 GPU which seems a lot easier to achieve the same result.
And like other people just mentioned: does a architecture even matter?
The console is out for more than 6 months and we still haven't seen any releases, screenshots or other footage of 2013 games that put it above PS360 level. Not from multiplatform games, not from exclusive games, not even from Nintendo games.
The proof is in the pudding and all the Wii U pudding we can eat this year tastes like PS360 pudding.
freezamite
Banned
Of course you're downplaying it. You're drawing a picture here where the WiiU GPU is old because its development started in 2007-2008, as if XboxOne and PS4 GPUs were made in less than a year. That's a FALLACY.M°°nblade;60454381 said:
We know that the WiiU GPU was developed over an R700 core for four years. On the WiiU GPU thread, we had informations (from bgassassin) of final silicon not being available until December 2011 - January 2012. And we also know that WiiU had a really premature launch, that the devkits and documentation weren't finalized until post-launch and with an OS that still needed a lot of work, which also means that the GPU is in fact really new.
¿Some "bells and whistles" the design we've seen that differs completely from any R700 design we've seen?M°°nblade said:
If they started to develop the GPU in 2007-2008, why not grab the last finalized model which had a great performance for its time, and then start to work from here?
The R800's development (ATI's first DX11 GPU) started over R600 or even an older prototype, and then reworked it among with the R700 and so on.
Nintendo has had 4 years of technological advancements that of course has used to its favour. Even the CPU which has a really familiar architecture has benefited from this, with a low-latency-eDram as the L2 cache that allowed it to have the largest amount of memory per core of this generation of consoles.
If the launch was premature, and Nintendo has had problems to finalize it as it should have, then its natural than a system with a much modern and different architecture than the 360/PS3 has so many problems to show its strength.
We are at two weeks to see what has done Nintendo over this past year on this regard, and also to show us what this console is capable of. Now will be time to judge.
So, the TDP of the Wii U was mentioned quite a lot in this thread. How come nobody takes a better look at it? Yes, the TDP of the PS360 is a lot higher, which is not a good think, quite the contrary, but that should also give you something to think about, namely the relationship between TDP and power. The Wii U draws a lot less power while still being more powerful than the PS360. In conclusion: The Wii u is a lot stronger, because it has newer hardware that makes this performance possible. I still can't believe that people argue about the power difference between these consoles. Until developers start to optimise their engines for Wii U, you won't see the big graphic leaps you want to see.
Cliff notes version:
0)Dearest freezamite. eDRAM should never appear in the same sentence as "low latency". eDRAM is a denser, slower drop-in for SRAM. Repeat after me: eDRAM is smaller and slower than SRAM.
2)"generations ahead" is a feature set metric. As in, "The Geforce GT610 is several generations ahead of the Geforce 8800 (but this implies absolutely nothing about performance)".
3)Optimizing for caches is both barely possible and hardly necessary. Cache's automatically optimize latency and bandwidth of repeat access patterns. That's the whole idea, and it hasn't changed in 30 years.
The only angle you even have is A)manual prefetch instructions and B)cache-bypassing stores for data you know won't be valuable as a cache entry. Intel's had the relevant instructions since the Pentium 3, AMD since the original Athlon, and of course Xenon and Cell PPE have them as well. Nothing about this is new.
And we can immediately throw away A: manual prefetch, because any modern CPU has automatic prediction-based prefetching, and all you're going to do with manual prefetch instructions is cause interference.
B is a fringe benefit in total edge cases. You gain a complete cache refill if you happen to stream out exactly enough data to overwrite your cache hierarchy once. If you stream less, you don't lose your whole cache, so the penalty diminishes. If you stream more, the penalty stays the same but gets amortized over more work performed.
I've written software "seriously" for 20 years, and I seriously can't tell you how and what to optimize for a CPU cache, much less for a CPU cache of a certain size. My best generic advice is "don't". Completely puzzled by that statement.
0)Dearest freezamite. eDRAM should never appear in the same sentence as "low latency". eDRAM is a denser, slower drop-in for SRAM. Repeat after me: eDRAM is smaller and slower than SRAM.
2)"generations ahead" is a feature set metric. As in, "The Geforce GT610 is several generations ahead of the Geforce 8800 (but this implies absolutely nothing about performance)".
3)Optimizing for caches is both barely possible and hardly necessary. Cache's automatically optimize latency and bandwidth of repeat access patterns. That's the whole idea, and it hasn't changed in 30 years.
The only angle you even have is A)manual prefetch instructions and B)cache-bypassing stores for data you know won't be valuable as a cache entry. Intel's had the relevant instructions since the Pentium 3, AMD since the original Athlon, and of course Xenon and Cell PPE have them as well. Nothing about this is new.
And we can immediately throw away A: manual prefetch, because any modern CPU has automatic prediction-based prefetching, and all you're going to do with manual prefetch instructions is cause interference.
B is a fringe benefit in total edge cases. You gain a complete cache refill if you happen to stream out exactly enough data to overwrite your cache hierarchy once. If you stream less, you don't lose your whole cache, so the penalty diminishes. If you stream more, the penalty stays the same but gets amortized over more work performed.
I've written software "seriously" for 20 years, and I seriously can't tell you how and what to optimize for a CPU cache, much less for a CPU cache of a certain size. My best generic advice is "don't". Completely puzzled by that statement.
In power its closer to 360. In architecture its closer to xbone. Is that so hard to grasp ?
Next gen architecture, current gen power.
lightchris
Member
When I think about optimizing for caches or specific cache sizes things like blocking come to my mind. For example in matrix multiplication it can be useful to split the matrices into several smaller blocks of such a size that each fits into the cache (instead of multiplying the whole matrix at once like you are normally used to).
Just saying though. I have no experience in game development so I'm not sure if such things would often be useful in that context.
PdotMichael
Banned
I don't even know why you quoted me if you just want to write your nonsense.
People were talking about the power of the WiiU and Xbox One.
The talk about architecture is just goalpost moving at this point.
Sorry I thought it was me you were referring to by "people in this thread". And thats all I claimed.
freezamite
Banned
IBM's eDram implementation with the POWER7 series had a latency access times of only 5.7 nano-seconds, compared to 7.5ns of 6MB SRAM on "Tukwila" designs.Rolf NB said:
Fact is that Wii emulation, which had SRAM cache, is done through a single core at the same speed and with the same amount of L2 cache (but this time eDram instead of eSram). If latencies were higher, perfect emulation wouldn't be possible, because having longer latencies in such a low level memory in a CPU affects performance a LOT.
Even the Xbox 360 had problems when it got "mini" because reduced latencies affected how code was executed, and had to introduce some delays to compensate it. And this was to compensate REDUCED latencies!
The fact that Wii emulation is done at the same clock speed confirms that latency to L2 cache is at least as low as it was in terms of clocks.
Of course is a feature set metric, but newer features means better performance in a console. At least first and second parties should take advantage of them.Rolf NB said:
Optimizing for much bigger caches or ones that change specifications is of course possible and much necessary as has been stated by Shin'en.Rolf NB said:
Wii had a more flexible cache implementation than GC, and WiiU is AT LEAST as capable per cycle and per kb as the Wii, but with 2MB of L2 in it's main core and 512KB on the others compared to just 256KB the Broadway had.
That statement had other implications as I read it. While you focus on low level instruction optimization, I think they were focusing on the fact that you have a huge 2MB+1MB L2 lockable cache so you can make use of it in a more flexible way to reduce accesses to main memory.Rolf NB said:
More than to "reprogram" the L2 fetches to perform better, to make code suited for that 2+1MB configuration.
I insist on the latencies, if they were higher, it would be impossible to emulate Wii the way it is done. We're having hardware emulation here.
Mr. Wonderful
Member
These discussions about the Wii U's power are starting to get ridiculous. I think we all have a pretty good idea of what the console is going to be capable of when developed specifically for the Wii U. Take a PS360 game, add high resolution textures due to having 2x the amount of RAM, better draw distances, and some modern lighting/GPU techniques due to having a newer GPU feature-set, and Bam. Done. That's the Wii U.
That's pretty much what I'm expecting, and I'm fine with that.
I honestly think the best looking Wii U games will match the first PS4/Xbone games, albeit at 720p.
Mihael Mello Keehl
Banned
wrong thread lol
I have a degree in Computer Engineering, currently in Electrical Engineering I followed advanced classes on semiconductor production and RISC processor design. I work as a developer on a high-end (non-game) 3D application.
I don't understand most of the discussion in that thread and I doubt many people do. They are just regurgitating stuff they read somewhere without really understanding any of it. Trying to divine the capabilities of a chip set by photographs is a fool's errand.
Why do people keep comparing the WiiU to the X360 & PS3 ?
Also if the WiiU is so powerful where are the games ?
I know the WiiU has eDRAM & its supposed to make all the difference, but why did Nintendo bother with eDRAM as it is so expensive & not just go for more DDR3 or even have GDDR5 ?
I have had a WiiU since launch & i want to know where the games are !?
Also if the WiiU is so powerful where are the games ?
I know the WiiU has eDRAM & its supposed to make all the difference, but why did Nintendo bother with eDRAM as it is so expensive & not just go for more DDR3 or even have GDDR5 ?
I have had a WiiU since launch & i want to know where the games are !?
Cause its competing with them.
Not enough userbase/selling consoles for Third party to bother, and Nintendo is struggling with the shift to HD.
They always have very custom-made hardware, I don't know. They make baffling decisions in this regard.
Store Shelves.
airmangataosenai
Member
ArchangelWest
Member
We already have games announced that crap on PS3 graphics, so so much for that.
Like, any PS3 game's graphics?
shakin_haitian
Banned
Do you honestly believe that? If so, can I buy some pot off you?
ArchangelWest
Member
Well really, 'crap on' has to be filtered through the reality of diminishing returns. But the truth is that Wii U has games already that graphically look as good as or better than the best PS3 games, but also have more modern lighting, draw distance, and still look great when you get in really close up or step out of a corridor type situation, or use a camera that isn't fixed.
airmangataosenai
Member
If you think Nintendo will ever put enough money into a game to get it to look like The Last of Us or GT6, I think I'll come to you first for crack.
Seriously, there's nothing ever shown for the Wii U up to and including tech demos that are anywhere near the level of something like GoW:A. It's like looking at games separated by an entire generation. Go ahead and compare the two.
I dunno why this thread was bumped, but just in case people are wondering what those guys know, and just in case people are still believing they don't understand anything outside of Nintendo hardware: http://www.youtube.com/watch?v=er8CoAAmv2A
Tom Brohanks
Banned
What does this have to do with the Wii U?
Nothing. That's the point. Those guys can exploit hardware in ways probably not even the guys who designed the systems really understand. And that's not limited to Nintendo systems.