dragomix

dragomix, on 21 Apr 2014 - 2:53 PM, said:
What about more modern shader design? All the other advantages that came with DX11 GPU's?

Modern shader design is more effecient. thats literally all we got going. We are skimming the bottom of the well until we hit another huge breakthrough.

The wii u (it looks almost certain right now) uses very long instruction word 5, or vliw5. WHat this means, is that each simd 'unti' or 'shader' as it is called, is comprised of 5 alu's.



Well, this turned out to be an effecient use of space, but an ineffecient use of the shaders. Because of things like dependencies (problems that require a variable that hasnt been calculated yet), or cache misses, it would turn out that on average, only 3 or 4 of the 5 units would actually be able work, while the others just sat there, not doing anything. Literally 2-3/5ths of your shaders not even doing anything!

So they switched to arranging their blocks in units of 4 instead. That was called vliw4.

Now, on average, 3-4 of these will work (for the same reasons as stated above), and only the occasional odd man out is left not doing anything. so, now you have 3/4's of your shaders working (with the occasional 2/4 or ugh, 1/4 working). So, even by having the same exact number of shaders, and the same exact resources, you get a 25% increase in performance.

Heres a diagram showing how ineffecient vliw shaders can be if used sloppily. This is vliw4, if it was 5, youd have even more unused shaders. Actually, before that, I need to post a diagram of the wavefronts, and how they are dependent on each others results.



Okay, now we feed the wavefronts into the vliw4.


Now remember, you can add one more unused unit to each of those if this was vliw5... (Which is why they got rid of it, it made no sense to organize that way when almost half the damn things never got used!)

While this is nice, it still literally has no effect whatsoever on what effects these gpu's can do. Its just one has more performance than the other.

Now we have 'GCN' or graphics core next, or the incredibly non descriptive, ridiculously marketing named architecture modern gpu's, as well as ps4 and xbone are using.

Despite its horrible, horrible gimmick name, its a pretty good job.

Using those same wavefronts, we feed it into a gcn architecture.


Well, gcn has a compute unit, that can ORGANIZE wavefronts, and single out ones that would cause a stall because they are dependent on others, and put them on hold and execute other more viable wavefronts instead, that dont need to wait, and then once the needed data comes in, the wavefronts put on hold can go through. There is more to it, but this is the basic gist of the situation.

So now, we have much more effecient use of the shader units, with most of them being used all the time.

Very cool. Less shader units go a lot farther, and more shader units, even farther still.

But again, this does NOTHING WHATSOEVER TO PROVIDE EFFECTS VLIW ARCHITECTURE CAN NOT DO. There is no revolution here. There is no 3d accelerateted hardware vs... Not. THere is no programmable shader vs matrox fixed function. Its just more effecient. They can do the same things, so its not a matter of can it do this effect, its a matter of, how many can it do at once?

Now, I dont know what the heck is REALLY going on in wii u's gpu. All signs point to using a vliw5 architecture with 128 shader units... Except... The games visuals. Mario kart 8 is mindblowing for these specs if they are true. I dont know how they are doing it, but if they are using vliw5 architecture, they must have found a way to increase effeciency. I dont know what it is. Maybe they have AMAZIBALLS prefetchers, maybe its the on the same die cpu and l3 edram cache. I dont know. But its not your run of the mill vliw5 performance. At least not what Nintendo has been showing.

That being said, the 'dx11 range gpu' crap is just that. It is LITERALLY barely anything more than marketing to get people to think their year old gpu is somehow horribly, irrefutably, useless and outdated (Lets face it, thanks to ps4/xbone being WAAAAYYYY underpowered compared to ps360 reveal point in time, several year old computers now have a free ride for another console generation, if you are packing an icore, you are good to go, with minimal upgrades of ram and maybe a mid range gpu in a few years) and buy the 'dx11 compatable' gpu like the good little lemming they are.

• dragomix and NintendoReport like this

Desert Punk, on 13 Apr 2013 - 19:53, said:I remember reading developers comments not that long ago that the PS3 hasn't reached full potential. The PS3 is a absolute nightmare in many ways, the gpu is weaker than 360 and it doesn't have the 10MB of very high speed video memory (although there is graphic memory cache on the gpu). I think the reality is the amount of effort to maximise the performance of the ps3 is just too extreme. Making use of every part of the cell processor and working out how to do so with limited memory bandwidth is just too time consuming and costly. Besides with any console if you have the time and money you can fully optimise the code by hand assembly of the code to maximise performance. When you look at the spec of the cell processor you think the ps3 in theory should easily outperform the 360 but really the only games that show performance above 360 are games that have been written from the ground up exclusively for ps3. Even where the ps3 is the lead platform often performance is only marginally ahead on ps3. When the ps3 was first launched often multiformat games made no use of the additional features of the ps3. So often multiformat games had inferior frame rates and looked worse on ps3 compared to 360. Now programmers know how to maximise performance by offloading some gpu tasks to the cell so it can match or exceed that of the 360 in frame rates. One advantage the ps3 has over 360 is once the 10MB of the 360 is exceeded the 360 has to use far slower memory and this means 1080p is not really an option on 360 unless its more simple games like 2D games but the PS3 has achieved 1080p on more impressive titles like wipeout even though under very high load wipeout will drop resolution.
I think if you compare X to xenoblade chronicles on the wii the game looks to be much the same in game engine but with a visual jump in graphics. It still looks to me like there isn't as much going on in the world as skyrim and the motion and a.i. of the creatures and people looks more simplistic compared to skyrim. In skyrim when people move and run it looks realistic and they attack with some variation in their patterns.


However don't get me wrong 'X' looks brilliant but so far and the important thing is so far the video shows a visually enhanced version of xenoblade and not something of the complexity of skyrim.
I mean no matter how you try and manipulate the information the wii u is operating a tri core arrangement of the wii cpu at 1.25Ghz. The xbox 360 is operating a tri-core cpu at 3.2ghz with 2 threads per core. Both 360 and PS3 are cpu-centric designs and for many games much of this cpu performance goes un-utilised but this is not the case for bethedsa games like skyrim they max out the cpu performance. To me its clear the original wii xenoblade game engine looks much the same albeit with much reduced visual quality.


First and foremost, you have once again COMPLETELY ran away from a HUGE amount of lost arguments and tried to throw together another one as if nothing just happened.

That out of the way.

1. The wii u is running a tricore modified 750 series, NOT a tricore broadway. What you are doing is essentially calling an i3 the same thing as an i7. Broadway was a custom 750cxe, espresso is a tricore custom 750 with two FX's and a GX, which had numerous performance enhancements beyond cache size and clock speed.

2. The ai in skyrim, and all elderscrolls games is crap. Particularly hilarious when compared to the boastings always made and never delivered upon. Nothing in your video showed any ai more sophisticated than a fixed routine that switches to rush at the player head on and attack when a proximity event is triggered.

3. Your poorly worded argument with intangible quantifiers of 'more alive' refers to the difference in combat style, from very poorly done real time in skyrim, to an rpg active time system in xenoblade, not ai.

Before being engaged in a battle, xenoblades ai actually interact with the environment and some with each other, in an ecosystem. More than ive ever seen in wildlife in skyrim, which without fail simply wanders about in idle animations until triggered by the player, then rushes them head on, or, performs the exact same process with npc's deemed 'bad'. Thats the extent of the ai, rushing at crap and beating it. When it works that is, you can see PLENTY of branch misses in that video.

In fact, Xenoblades monsters have more demanding ai than skyrims, they are capable of many, many, many more moves, and the possibility of such planned moves to be erased from the future, and a metric butt ton more branch affecting status effects. Something the 750 series excels at that xenon stutters with. Heavy, heavy branch code. The stock recent 750's (including broadway) series can handle up to 4 misses without penalty, while xenon incurs a 500 cycle (half a GHz) penalty with every miss. The 750 series also has much better branch prediction than xenon, and when it does incur a penalty, its 5 cycles because its not a not goodly deep piped system so as to try and mask its horrible horrible latency. Sucks being an in order processor.

This is why ai pretty much has gone nowhere this gen. The most popular and powerful systems suck at branch code, the foundation of ai. This is why you always hear the same crap and see the same reactions over and over again. Not because they couldnt record more responses, but that the xenon would choke on branch heavy cache misses.

'I used to be an adventurer.... I used to be an adventurer..... i used to be an adventuree.... arrow... knee...'

Ah, love that 'radiant ai'.


The only thing you actually noticed are the difference between combat styles. The 360 also had games with similar combat styles. You dont want to use those though,


And finally, not only would a tricore broadway clocked over a GHz DESTROY xenon in GP branch code like whats actually needed in ai (all the flops in the world arent going to help with branch code).

The wii u utilizes customized fx/gx 750 architecture, NOT cxe (broadway). They have numerous gp performance enhancements over broadway, and im not talking about just cache sizes and clock speeds.

Aside from that, recent benchmarks show nintendos paired singles simd fair surprisingly well, within range of a 1.1/1.2GHz Bobcat core.... at 747MHz.

Any flops the cpu cant muster, can be retreived from the gpu, which has more flops at its disposal than both the 360's cpu and gpu combined.

And on top of all that. X was an ALPHA game, the enemies that were seen were very likely the ONLY enemies that existed at the time, and they probably didnt even have working ai.

This nonsense already has nothing to stand on, which is why you are using intangible unquantifiable nonsense like 'looks more alive'. But lets see how long you can drag it out for. Nintendos going to start showing some games soon. Im guessing around june is when youll give up.

It seems a lot of the people questioning the specs are people not to familiar with PC hardware. These specs are good, but no where near top end in whats available to PC gamer's. Also as to anyone who thinks its not powerful enough, you need to realize that if they attempted to put the most powerful hardware in the machine...we would never see it launch. Technology updates so fast, it would literally be impossible. Any PC gamer can tell you that building a new machine is great and also sucks at the same time, you have many decisions to make and generally even if you blow 3k or so...by the time you build your machine you can bet it will be outdated shortly after.

I, for one didn't buy blops because it's a trash fps game that I have 9001 free or cheaper alternatives that play better on my pc.

http://www.neogaf.co...&postcount=1663

*Comes out of his hole (again)*

I tend to forget some people take this stuff a lot more serious than I do.

First a thanks to Chipworks for going above and beyond for the picture and to blu, Durante, Fourth Storm, Thraktor, and wsippel for the work they did. Shinjohn let me know that the picture had been obtained and sent me a link, but I also checked out the thread. I wanted to come back and help with the confusion and what not.

As some of you know getting info about the hardware was a pain because what Nintendo released essentially boiled down to a features list. And by that I mean general features of a modern GPU that could easily be looked up. Info that dealt with performance apparently was not given out leaving devs to figure have to figure it out on their own. I had two working ideas of the GPU based on a more traditional design (which I was hoping for) and a non-traditional design. I see that some of you actually remembered the non-traditional idea. Wsippel and I would compare notes on whatever info we could come up with. Some of those notes led us to come up with how it may look if Nintendo took the non-traditional route.

http://www.neogaf.com/forum/showpost...ostcount=12053

In this post you’ll see both wsippel’s take and my take. I’m going to address some things in that post because I know some of you will try to take them out of context. First you’ll see wsippel’s baseline ended up being more accurate than mine. When I talked about the potential performance of 1TF or more that was in comparison to the R700 series because new GPUs are more efficient than that line, a higher baseline, and my idea focused on the dedicated silicon handling other performance tasks.

So what was the basis for the non-traditional view? I shared two of those bits of info before.

http://www.neogaf.com/forum/showpost...postcount=6136



Quote:
Well, I can't reveal too much. The performance target is still more or less the same as the last review from around E3. Now it's more balanced and "2012" now that it's nearer to complete and now AMD is providing proper stuff. As far as specs, I don't see any big change for better or worse, other than said cost/performance balance tweaks... It won't make a significant difference to the end user. As far as the kit goes, it's almost like what MS went through. Except more Japanese-ish... If you know what I mean. http://www.neogaf.com/forum/showpost...postcount=6305



Quote:
Anyway, things are shaping up now with the new year. There was some anxiety with some less close third parties about what they were doing with GPU side, whether things were going to be left in the past... but it looks more modern now. You know, there simply wasn't actual U GPU data in third party hands this time last year, just the target range and R700 reference GPU for porting 360 titles to the new cafe control. Maybe now they finally can get to start debugging of the specifics and start showing a difference... Here is one more specific piece that I didn’t fully share.



Quote:
I can't confirm or deny, sorry. The cat is very confidential and I repeat non-final. The target, last checked, is triple core with XX eDram and exclusive Nintendo instructions. 1080/30 capable Radeon HD w/tess. and exclusive Nintendo patented features. On a nice, tight bus that MS wishes they had on 360.
I appreciate the individual for sharing as much as he did. He was a little paranoid though (I can understand) and at one point thought I was leaking info on a messageboard under a different name, but wouldn’t tell me the board or the username, lol.

I’m sure some of you remember me talking about games being 720p. It’s because with this I knew devs would use those resources for 720p development. I’m sure some of you also remember me mentioning the bus. The key thing in this is the “Nintendo patented features”. In the context of things we talked about, it seemed to me these were going to be hardwired features. What is certain for now is that the die shot shows a design that is not traditional, fewer ALUs (in number) from where things supposedly started with the first kit, and GPU logic that is unaccounted for. I’ve seen some saying fixed functions, but that’s too specific to be accurate right now. Dedicated silicon would be a better alternative to use, though I say that as a suggestion. In my opinion I think lighting is a part of this. The Zelda and Bird demos emphasized this. Also in the past it was discussed how Nintendo likes predictability of performance. It would also suggest Nintendo wasn’t ready to embrace a “fully” programmable GPU and kept on the water wings when jumping in the pool.

I did what I could to get as much info on the hardware as possible since Nintendo was giving out so little. From there I gave the best speculation I could based on that info. As of today, I still stand by the evaluations I made about Wii U’s potential performance from all the info I could gather. And until Nintendo’s games show otherwise I’ll continue to stand by them because in the end it’s on Nintendo show what Wii U is capable of.

And if you think I deserve flak for what I’ve said in the past then I’m here, but you’re wasting your time trying because my view hasn’t changed yet.

I made the farewell post to hold myself accountable to avoid posting, but I haven’t done well sticking to that, haha. I wasn’t going to make this post, but since I was one of the primary ones gathering info it’s unfair to you guys to leave things as they were.

You would still need a software layer with wii u when operating in wii mode and its still being stated that the wii u only runs wii games in their original resolutions 480i, 480p or 576p not upscaled.

It's likely the original wii gpu is inside the wii u, the enhanced broadways operate in a wii compatibility mode but other processes may be emulated like sound, any required functionality of the arm star chip or whatever its called and possibly other memory features. The ps3 still had software updates for ps2 compatibility even for the ps3 that had both the ps2 gpu and cpu internally to run ps2 games. I think its highly likely the wii u has the original wii gpu fitted with the built in 2 meg frame buffer and 1 meg texture cache at the very minimum if its to achieve a high level of compatibility.

It seems like the software of the wii u is going to be in a bit of a raw state when you get it out of the box for the first time.

Programmable shader gpu's dont need special hardware. Its far too easy to emulate fixed function gpu's, Its almost like its fixed function or something. You only need working knowledge of the api's. There is no wii gpu in wii u. The update is likely software instructions to emulate wii's fixed function tev.

We know from the iwata asks tear down that special hardware was added in alongside the wii u circuits for wii cpu bc. This was not the case for wii, as wii WAS gc, except overclocked and with more ram. Starlet was simply an i/o that cut off access to resources the gc didnt have, turning the wii into gc at hardware level.

ppc 7xx (broadway line) was discontinued ages ago, never manufactured at 45nm, has no support whatsoever of symmetric multiprocessing, so you cant have a 3core broadway no matter how much you 'enhance' it, and obviously has no support for the kind of on die edram cache like the wii u cpu has.

We know for a fact wii u's cpu isnt broadway. You dont need to add in 7xx compatable circuits if you are using a 7xx series.

'Enhanced broadway'=broadway enhanced, special hardware has been added to the wii u cpu to provide a compatable frontend for bc with wii. There is no broadway cpu.

Numerous devs with HEAVY wii/gc knowledge have gone on record stating their misuse of the wii u cpu SERIOUSLY crippled performance, having the cpu performing as little as 1/6th the capacity as they currently have it now that they have optimised their engines to the wii u's new cpu.

This would not be a thing with broadway cpu's.

There is no broadway in wii u. This is a fact.

I think Dragomix is essentially correct. Next gen will not be as powerful as people expect because of the recession and intrudution of additional hardware.

You might want to edit a word or 2...