WiiU "Latte" GPU Die Photo - GPU Feature Set And Power Analysis

Status
Not open for further replies.
Anyone find it odd though that the guy took some tidbits from things others have said before? "Punching above it's weight" (referring to the CPU) I know another source said the exact same phrase about the CPU. I feel like the guy read some cliff notes from discussions including from this website to sound legit and help put out that article. Not saying some is true regarding power of the WiiU but something still seems fishy about that guy.
Exactly, the whole things looks like a 'First year history of Wii U, with the 'facts' you don't know filled in by common negative opinions from around the internetz'.
 
Exactly, the whole things looks like a 'First year history of Wii U, with the 'facts' you don't know filled in by common negative opinions from around the internetz'.
It really doesn't. Literally the only thing that wasn't already pretty much known, was the 55nm thing (which is almost certainly a mistake on the anonymous devs part). The fact it all sounds negative is because WiiUs launch and most of the first year were pretty darn negative.
 
From what I understand at 55nm the eDRAM would be much larger than it currently is. I have no idea if it's possible for it to be manufactured on a separate process to the rest of the die or not, but if it is that might explain something.


Yep, when we consider Tron#1's guarantee that the character poly counts in X are significantly higher than 15-20k and fred's assertion that the game is simply not possible on a 160ALU GPU then the Latte being a custom fixed-function/programmable shader hybrid with around half a teraFLOP of processing power and dual geometry engines is all but confirmed.

They are both devs with Monolith right?
Where you get the 500gflops from?. I thought the general consensus was that the GPU had even less gflops then a 360 GPU?
 
It really doesn't. Literally the only thing that wasn't already pretty much known, was the 55nm thing (which is almost certainly a mistake on the anonymous devs part). The fact it all sounds negative is because WiiUs launch and most of the first year were pretty darn negative.
The whole thing being legit sounds entirely plausible.

The whole thing being written by a well-spoken gaf member with an anti-Nintendo bias who has gleaned information over the past 1.5 years is also plausible.

The latter I'd hope is very unlikely, but there's literally nothing in there that could not have been ass-pulled by any well informed forum member here.

And I asked this question in three threads and nobody addressed it... about the Green Hills IDE. The Green Hills licensing deal wasn't until March 2012, yet this guy said it was included in the initial setup of kit/tools that they had... the developers of a LAUNCH GAME didn't get this until potentially less than 8 months before launch???

I'm not saying it's all a made up trolling attempt, but I'm just saying..

Where you get the 500gflops from?. I thought the general consensus was that the GPU had even less gflops then a 360 GPU?
He's being sarcastic. The leading theory is still 176GF. He's making fun of Tron#1 and fred for believing there could be more than meets the eye.
 
It really doesn't. Literally the only thing that wasn't already pretty much known, was the 55nm thing (which is almost certainly a mistake on the anonymous devs part). The fact it all sounds negative is because WiiUs launch and most of the first year were pretty darn negative.
I wish we could have gotten articles like that one for the 360 and PS3 launch. If my friends who worked in the industry aren't blowing smoke up my ass, both of those launches were exponentially worse than the dev environment problems described in the Wii U article.

Also, it seems Nintendo sorted out most of these issues shortly after launch, so what it really boils down to is that Nintendo should have delayed their launch by six months and missed that first holiday. Given what sales have been for most of 2013, they probably would have been better off that way. They could have released with a much better library and game release schedule.
 
Apparently, he's already admitted to breaking a nondisclosure agreement, which is pretty serious already. Career ending, and criminal in the civil sense. So you trying to paint a picture of 'legit' from 'anonymous' just can't and won't be seen that way by many.
Hold on, let me get this straight:

You're not only not addressing what I said, but instead creating an implication that because someone broke NDA, they're either not legit or untrustworthy? So basically, whistleblowers are bad then, correct? Since whistleblowing involves breaking confidentiality agreements, whistleblowing isn't ok?

You're creating an equivalence between breaking an NDA to libel?
 
I wish we could have gotten articles like that one for the 360 and PS3 launch. If my friends who worked in the industry aren't blowing smoke up my ass, both of those launches were exponentially worse than the dev environment problems described in the Wii U article.

Also, it seems Nintendo sorted out most of these issues shortly after launch, so what it really boils down to is that Nintendo should have delayed their launch by six months and missed that first holiday. Given what sales have been for most of 2013, they probably would have been better off that way. They could have released with a much better library and game release schedule.
Indeed. Launches aren't historically that smooth.

It's mostly old news; an insight into early days if Project Cafe. Nothing more, really. Potentially useful for analysing what went wrong at the start, but not for analysing where WiiU development is right now. We have more recent info for that.
 
A

A More Normal Bird

Unconfirmed Member
Where you get the 500gflops from?. I thought the general consensus was that the GPU had even less gflops then a 360 GPU?
Sorry, I was being entirely sarcastic, since the poster I was quoting was calling some random person's "guarantee" that the character polycounts in an unreleased game were significantly above 15-20k an instance of "rational examples coupled with detailed explanations,". Similarly, another poster in this thread was stating that a number of Wii-U games would be impossible on a 160 ALU GPU (presumably they meant one clocked at 550mhz), providing absolutely no substantive reasoning or evidence to support their claim.

Also, Kai was robbed ;)

The whole thing being legit sounds entirely plausible.

The whole thing being written by a well-spoken gaf member with an anti-Nintendo bias who has gleaned information over the past 1.5 years is also plausible.

The latter I'd hope is very unlikely, but there's literally nothing in there that could not have been ass-pulled by any well informed forum member here.

And I asked this question in three threads and nobody addressed it... about the Green Hills IDE. The Green Hills licensing deal wasn't until March 2012, yet this guy said it was included in the initial setup of kit/tools that they had... the developers of a LAUNCH GAME didn't get this until potentially less than 8 months before launch???

I'm not saying it's all a made up trolling attempt, but I'm just saying.
I agree there's a chance that Eurogamer were deceived, but one would imagine they would vet a story like this rather thoroughly. As for the Green Hills thing, rereading the story it seems like the writer is kind of compressing months of time into single paragraphs, talking about general experiences rather than a blow by blow update of what came with each dev-kit. There is also the possibility that Nintendo had the software before the licensing deal was publicly announced.
 
As for the Green Hills thing, rereading the story it seems like the writer is kind of compressing months of time into single paragraphs, talking about general experiences rather than a blow by blow update of what came with each dev-kit. There is also the possibility that Nintendo had the software before the licensing deal was publicly announced.
Reasonable. Certainly could have been rolling it out prior to public announcement.
 
This is all i stated. PEople have said X in its ALPHA stage doesnt look as impressive as open world games of the 7th gen. I said it will when it is RELEASED. i can GUARANTEE when X is released not only will it look better than any open world game of the 7th gen but you will see better DOF, textures, poly count, etc. i also said that i wasnt going to bog down the threads trying to prove this as the game is UNFINISHED and hasnt been released yet. this is something i will happily revisit when the game is released. I dont recall saying anything about believing there is "MORE THAN MEETS THE EYE". NO the Wii U hasnt seen games taking advantage of everything it was to offer so in that sense there is more than what meets the eye.
No worries man, wasn't taking a side. To the extent you're saying here you're probably right. I was just commenting on the guy's broken sarcasm detector.
 
This is all i stated. PEople have said X in its ALPHA stage doesnt look as impressive as open world games of the 7th gen. I said it will when it is RELEASED. i can GUARANTEE when X is released not only will it look better than any open world game of the 7th gen but you will see better DOF, textures, poly count, etc. i also said that i wasnt going to bog down the threads trying to prove this as the game is UNFINISHED and hasnt been released yet. this is something i will happily revisit when the game is released. I dont recall saying anything about believing there is "MORE THAN MEETS THE EYE". NO the Wii U hasnt seen games taking advantage of everything it was to offer so in that sense there is more than what meets the eye.
Being alpha is irrelevant. Last gen games in alpha looked better than X.
 
A third developer has responded to the Eurogamer article calling it out for being exaggerated click bait.

http://nintendoeverything.com/ping-...-wii-u-developer-piece-sticks-up-for-dev-kit/




being an early on developer on any new hardware will pose challenges to those unfamiliar. Programming and coding is most of the time trial-and-error.

Wii U game development takes the same amount of work and attention as a Xbox 360 or PS3 game development. They've done their process of making their API understandable and useful to seasoned programmers. Wii U development = any other game console development time.

That's all I have to say on the subject of Eurogamer's clickbait article. They titled it to stir up the Nintendooom wheel again.
That is three named developers who have weighed in to contradict that piece from Eurogamer.

I find it saddening how so many people choose to praise the word of this no name dev with no credentials while ignoring multiple verfiable devs with confirmed experience. Is this what gaming is about now?
 
I think that at this point, it's pretty clear that the unnamed developer went through the same kind of hurdles every developer goes when developing for a new HW, with all the implications that this has not matter how powerful/weak the HW is.

I'm pretty sure that even Nintendo had it's fair share of problems when working with the Wii U HW for the first time.

EDIT: Just realised this is maybe the wrong thread.
 
A third developer has responded to the Eurogamer article calling it out for being exaggerated click bait.

http://nintendoeverything.com/ping-...-wii-u-developer-piece-sticks-up-for-dev-kit/


That is three named developers who have weighed in to contradict that piece from Eurogamer.

I find it saddening how so many people choose to praise the word of this no name dev with no credentials while ignoring multiple verfiable devs with confirmed experience. Is this what gaming is about now?
Wrong thread? The Eurogamer thread is here.

I think some people were eager to believe the original story because it helped them understand why Wii U ports are shit.
 
Latte doesn't quite have the same raw power, but should be more efficient and have more advanced features. The latter more than likely playing a strong part for the Project CARS team dropping PS360 development but continuing the Wii U version.
I would actually look to the amount of title memory available. Bear in mind that they've already got a DX9 path that's far more limited than the PS360 APIs.

It's one thing to scale the rendering load, but <480MB title memory is a headache enough given they're used to much much more on the PC side.
 
That is three named developers who have weighed in to contradict that piece from Eurogamer.

I find it saddening how so many people choose to praise the word of this no name dev with no credentials while ignoring multiple verfiable devs with confirmed experience. Is this what gaming is about now?
How does anonymous dev = no name with no credentials? The dev has credentials otherwise they wouldn't be getting interviewed anonymously by Eurogamer.

Also, it seems only one developer has weighed in to contradict Eurogamer. Eurogamer's article was about the hassle of developing pre-launch. All but one developer, from what I see, are talking about how developing for the Wii U is right now.
 
Almost certainly off-topic, but I don't quite want to make a new thread, because it would go to shit quickly. Also, I'd need to post a source, and I don't really want to get people in trouble. Anyway, Nintendo apparently started working on a new platform in early 2013 and has already selected a vendor for the SoC after talking to several potential candidates. The SoC might be based on an existing design, but will be changed to fit Nintendo's requirements. I assume it's for their next handheld, though.
 
Almost certainly off-topic, but I don't quite want to make a new thread, because it would go to shit quickly. Also, I'd need to post a source, and I don't really want to get people in trouble. Anyway, Nintendo apparently started working on a new platform in early 2013 and has already selected a vendor for the SoC after talking to several potential candidates. The SoC might be based on an existing design, but will be changed to fit Nintendo's requirements. I assume it's for their next handheld, though.
I would totally be ok with this if it was true. And i hope it's the wii u's successor, the handheld is doing just fine.
 
Well, either way, they selected the vendors for the Wii U back in 2009 and started working on the actual chipset in early 2010, so whatever they're working on could very well be two years off.
If it is Wii U, I see them riding with it at least until 2015 and launching and showing the new console at 2016. I mean, I think they know the Wii U ship has sailed, so I think they will likely ride it out but for 3 years only.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
Almost certainly off-topic, but I don't quite want to make a new thread, because it would go to shit quickly. Also, I'd need to post a source, and I don't really want to get people in trouble. Anyway, Nintendo apparently started working on a new platform in early 2013 and has already selected a vendor for the SoC after talking to several potential candidates. The SoC might be based on an existing design, but will be changed to fit Nintendo's requirements. I assume it's for their next handheld, though.
My guess would be the next handheld as well, just because 3ds came out earlier.
 
Well, either way, they selected the vendors for the Wii U back in 2009 and started working on the actual chipset in early 2010, so whatever they're working on could very well be two years off.
So its sort of a toss up, if its a home console, I would guess its for third party support. While Nintendo would continue to support the WiiU, they're would be minimal support from them. With a game at launch.

Handheld could see a 720p lcd screen, backwards compatibility, etc.
 
My guess would be the next handheld as well, just because 3ds came out earlier.
Same here, but I believe the handhelds are mostly designed in Japan, and the consoles in the US, and this seems to be a US gig (NTD). Or at least that's how it used to be before Nintendo restructured their hardware R&D business.


I'm crossing fingers for Nvidia Tegra K1 variant on next Nintendo handheld.
Doubt it. They reportedly already worked with Nvidia on the 3DS before switching to DMP, and Nvidia is (in)famous for really, really bad deals.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
I'm crossing fingers for Nvidia Tegra K1 variant on next Nintendo handheld.
NV are yet to show they've learned their power efficiency lesson. That said, it could be any one of the established vendors.

Same here, but I believe the handhelds are mostly designed in Japan, and the consoles in the US, and this seems to be a US gig (NTD). Or at least that's how it used to be before Nintendo restructured their hardware R&D business.
Interesting.
 
My guess would be the next handheld as well, just because 3ds came out earlier.
I am of the opposite thinking. When they launch their next system, they are going to want to have a profitable machine already available to offset what will most likely be losses (whether big or small) on the sale of the new machine (be it handheld or home console). The 3ds is profitable and selling well. The Wii U is neither. If they release a new handheld first at a small loss and must rely on Wii U profits and game sales to offset these losses, they won't be in a good position. That is why I think a home console is coming first. They will be able to offset the losses of the new console with profits from 3DS hardware and software sales.
 
Almost certainly off-topic, but I don't quite want to make a new thread, because it would go to shit quickly. Also, I'd need to post a source, and I don't really want to get people in trouble. Anyway, Nintendo apparently started working on a new platform in early 2013 and has already selected a vendor for the SoC after talking to several potential candidates. The SoC might be based on an existing design, but will be changed to fit Nintendo's requirements. I assume it's for their next handheld, though.
I'm still toying with the idea of a die-shrunk Wii U chipset in the next handheld. I know some people noted it was pad limited in terms of memory i/o shrinking, but there's ways around everything. These are already on old 40-45nm fabrication plants, even today we have 28nm available so if it's not starting production for another two years say, 28 will be very mature by then, and non-Intel fabs may already have 14nm by then.

It would benefit from starting with the Wii U install base of games rather than zero, the Wii U would benefit from extended support if a newer console replaces it on a quicker schedule.


Or, going with a balls out Qualcomm or Nvidia SoC would be nice too. If they can provide a sustained bandwidth, the Tegra K1 already beats the Latte in Gflops, but we know that's a superficial measurement, but in a dedicated gaming environment I'm sure they'd engineer out the other bottlenecks.


Correct me if i'm wrong. But isn't Nvidia known to be bad to work with consol wise?
Never tried to program anything on a Nvidia based console. Have no clue.
It's not about the programing, it's about the level of control Nvidia dictates. Same reason no one went with Intel again. When AMD licences out a chip, the console maker is free to alter and shrink it as they see fit. With Intel or Nvidia, they maintain control over the design and shrinks.
 
Same here, but I believe the handhelds are mostly designed in Japan, and the consoles in the US, and this seems to be a US gig (NTD). Or at least that's how it used to be before Nintendo restructured their hardware R&D business.



Doubt it. They reportedly already worked with Nvidia on the 3DS before switching to DMP, and Nvidia is (in)famous for really, really bad deals.
That Nvidia Tegra chip was all hype but didn't perform very very well or even close to spec. Which is why Nintendo had to change suppliers for the 3DS. Nintendo would be better off going with Power VR chips for their handhelds in future. They're more power efficient.
 
These are already on old 40-45nm fabrication plants, even today we have 28nm available so if it's not starting production for another two years say, 28 will be very mature by then, and non-Intel fabs may already have 14nm by then.
The power consumption would have to be reduced to less than 1/10th though (also consider battery/device size). The power reductions aren't really scaling as well lately IIRC.

Hard to say if 14nm will be cheap enough for Nintendo's purposes (or in volume) by 2016. We're only seeing low volume 20nm late this year, mind you.

Nintendo would be better off going with Power VR chips for their handhelds in future. They're more power efficient.
Not an unreasonable choice considering Apple & Sony are using ARM/PVR combinations as well. :)
 
I dont think they will release another home console until at least 2017.
I agree that they will probably release a new handheld before a new home console.
Its hard to say though cause it depends on how sales are 2014 for the 3DS.
 
The power consumption would have to be reduced to less than 1/10th though (also consider battery/device size). The power reductions aren't really scaling as well lately IIRC.
Lets say this would be an average SoC that would draw 4 watts, I don't think it would be a 10x reduction after removing all the console-only parts. The disk drive as a moving part takes a lot of voltage, the fan, the plethora of ports, let alone charging or powering something from that port, that would already reduce the power quite a bit.

We don't really see how much benefit 28nm parts have over the last two generations of 32 and 45 because most chip makers take that chance to add more transistors to gain performance. Without that, they probably draw quite a bit less per transistor. Plus, by 2016 28nm would be as mature as it's getting, look at the power draw shrinks from early 45nm to late 45nm, I can run my 45nm processor on the highest speed step on the voltage for the *lowest* speed step as they originally designed it due to increasing maturity of the silicon. That makes my 35w laptop processor produce about half the heat it normally does, at peak performance and with stability.
 
That eDRAM isn't gonna shrink. Plus, my money is on TSMC as a manufacturer as Renesas seems to be gradually getting out of that biz. So combining the SOI IBM CPU (w/ its own eDRAM) with an AMD/Renesas CMOS part on one SoC would seem to be out of the question.

I wonder how far along AMD's ARM/GCN SoCs are. On 20nm or 28nm, I think they could get 128 alus down to 4-5w after playing with clocks. That is as long as my hunch is correct and they go for a small/medium tablet form factor.

Damn, that's alot of acronyms. :D
 
Almost certainly off-topic, but I don't quite want to make a new thread, because it would go to shit quickly. Also, I'd need to post a source, and I don't really want to get people in trouble. Anyway, Nintendo apparently started working on a new platform in early 2013 and has already selected a vendor for the SoC after talking to several potential candidates. The SoC might be based on an existing design, but will be changed to fit Nintendo's requirements. I assume it's for their next handheld, though.
Cool, thanks for sharing.

Makes sense for 4DS timing. Assuming 2016 release, will be very interesting to see how close to WiiUs specs it comes
(no jokes, you know who you are)

I'm glad I popped my head in here and saw that bit from wsippel. Is it wrong to feel a little excited seeing that now?
Never! Let the 4DS Speculation threads begin! Where's Nibel? We need a hype vehicle collage, stat.
 
I wonder how far along AMD's ARM/GCN SoCs are. On 20nm or 28nm, I think they could get 128 alus down to 4-5w after playing with clocks. That is as long as my hunch is correct and they go for a small/medium tablet form factor.
That's another interesting avenue. We have yet to see what AMDs custom ARM cores will be like.

In any case, I just hope Nintendo doesn't repeat something like the 3DS going with ARM11 CPUs and weird Pica graphics, when chips greater than the Tegra 3 existed at the time in low power draws.

And HD > 3D.
 
Lets say this would be an average SoC that would draw 4 watts, I don't think it would be a 10x reduction after removing all the console-only parts. The disk drive as a moving part takes a lot of voltage, the fan, the plethora of ports, let alone charging or powering something from that port, that would already reduce the power quite a bit.
hm... well, the disc drive will be around 5W for constant spin operation. The fan they use is rated up to 2.5W (I imagine that's high RPM/audible noise level though).

33W -> 25W ? So to hit 5W typical operation, that's still down to 20%.

Of course, there is the matter of fab differences and also the specific node design (bulk/SOI, SHP etc), but that's hard to quantify.

I dunno. Colour me skeptical (45nm Renesas to 28nm elsewhere). :)

Anyways, I suppose there's room to play around depending on what you're thinking for the device size, weight & battery.

iPad Air is roughly 4.6W (Anandtech 32.4WHr/7.07Hr) during a 3D game. Of course, the screen they use also has an effect there, typically the higher % of total power consumption TBH.

----

Consider that the Gamepad's battery was ~5.5WHr (3.7V*1500mAHr) with barely much in the way of processing...

*shrug*


We don't really see how much benefit 28nm parts have over the last two generations of 32 and 45 because most chip makers take that chance to add more transistors to gain performance.
I'm talking more about the actual node transition characteristics. For example, TSMC claims 40% power savings between their 28nm and 40nm processes, but only 25% between 20nm and 28nm. It's slowing down.
 
Status
Not open for further replies.