Support NeoGAF

Marlenus · Jan 27, 2016

Rodin said:
Crysis 3 runs like ass on last gen consoles and GTA V, while being quite an accomplishment for such outdated hardware, runs possibly even worse. Nintendo games are usually locked 60fps, use advanced shader and we've seen some display of radiosity global illumination (which i don't recall was ever used in any last gen game) in combination with hd textures and a fair amount of polygons (nothing super impressive, but not underwhelming either all things considered). The only two 30fps first party games are Xenoblade X, which features maybe the biggest non-procedurally generated world ever, with zero loading screens and locked 30fps despite its giant areas full of elements on screen and despite it was made with a mid tier budget, and Zelda U, which isn't out yet but it showed things never seen on last gen hardware, especially in an open world game.

All these games are not without problems, image quality is usually poor and lack of AA in so many games is quite baffling, not to mention the generally poor (or lack of) anisotropic filtering. I think that for a console that came out in 2012 they should've aimed at least for hardware that allowed this level of fidelity at 900p-1080p, but that doesn't make Nintendo games worse than last gen best accomplishments by any mean. They just use resources differently.

Or

AmyS said:
We know Xbox 360 GPU was 240 gflops and PS3's GPU was 192 glfops, It's really hard to believe the Wii U GPU is that much higher, with 352 glops.

176 glops seems a lot more believable.

While not specific on specs, Gamesindustry.biz and Eurogamer article from early 2012 citing developer sources would be right in line with the lower number.

http://www.eurogamer.net/articles/2012-04-03-wii-u-not-as-capable-as-ps3-xbox-360-report
http://www.gamesindustry.biz/articl...ess-powerful-than-ps3-xbox-360-developers-say

This is the RV770 (radeon 4870) die. Latte is based on RV770 (although very cut down) but the basic VLIW5 uarch is the same.

This is the latte GPU.

The Rv770 has 180 lots of 5 shaders giving a total of 900 shaders of which only 800 were enabled on the full fat die as the other 100 were spares to increase yields.

Looking at the latte die it appears to have 32 lots of 5 shaders which is 160 cores meaning 176 GLOPS is the correct answer.

The 4670 was the closest thing to the 3870 which was just a lower power respin of the 2900 XT which is the basic uarch of the Xbox 360. Generally the 4670 was slightly behind the 3870 which itself was around 10% faster than the 2900XT meaning the 4670 was likely on par with the 2900XT (despite fewer ROPS and a much narrow memory bus). Bottom line is that despite the Wii U having a more updated architecture it was no where near enough to overcome a 64Gflop deficit.

Feel_The_Pain · Jan 27, 2016

Who flopping cares.

AmyS · Jan 27, 2016

Marlenus said:
This is the RV770 (radeon 4870) die. Latte is based on RV770 (although very cut down) but the basic VLIW5 uarch is the same.

This is the latte GPU.

The Rv770 has 180 lots of 5 shaders giving a total of 900 shaders of which only 800 were enabled on the full fat die as the other 100 were spares to increase yields.

Looking at the latte die it appears to have 32 lots of 5 shaders which is 160 cores meaning 176 GLOPS is the correct answer.

Thanks. I haven't seen a breakdown like this before, at least I don't remember.

I do remember that Rv770 had 800 active shaders, which provided 1.2 and 1.0 Tflops peak performance in the Radeon HD 4870 and 4850, respectively. Both released in 2nd half of 2008.

AlStrong · Jan 27, 2016

Marlenus said:
The Rv770 has 180 lots of 5 shaders giving a total of 900 shaders of which only 800 were enabled on the full fat die as the other 100 were spares to increase yields.

hum... I have never heard of a 900 shader assertion or theory. Dave Baumann has ever maintained 80*10. The Anandtech article was quite illuminating in the sense that AMD added shaders because of padding issues.

Azalean · Jan 27, 2016

BennyBlanco said:
it only flopped once afaik

Amazing.

Fourth Storm · Jan 27, 2016

AlStrong said:
hum... I have never heard of a 900 shader assertion or theory. Dave Baumann has ever maintained 80*10. The Anandtech article was quite illuminating in the sense that AMD added shaders because of padding issues.

I think what people mistook for the "half size" shader groups to the left of the SIMD cores were actually the local data shares.

Marlenus · Jan 27, 2016

AlStrong said:
hum... I have never heard of a 900 shader assertion or theory. Dave Baumann has ever maintained 80*10. The Anandtech article was quite illuminating in the sense that AMD added shaders because of padding issues.

On larger dies they always have extra shaders. The PS4 and Xbox One both have extra shader blocks that are disabled to enhance yields. It is a very common practice.

pulsemyne · Jan 27, 2016

Just to add some horribly long forgot fuel to this topic we have to remember the possibility that latte is on 40nm and RV700 chips were on 55nm.

ChronotriggerJM · Jan 27, 2016

dragonzdogma said:
WTH! really....so why aren't people buying more Vitas than?
I'm sold, which model do you recommend/version/bundle ect do you recommend?

Whichever is the cheapest if all you're looking for is the Remote play function lol. The Borderlands bundle was the best bet for a while because it came with an 8gb memory card, which is more than enough if your good at management.

Rodin · Jan 27, 2016

AmyS said:
Thanks. I haven't seen a breakdown like this before, at least I don't remember.

I do remember that Rv770 had 800 active shaders, which provided 1.2 and 1.0 Tflops peak performance in the Radeon HD 4870 and 4850, respectively. Both released in 2nd half of 2008.

Which should tell you a lot about comparing flops from different architectures.

pulsemyne said:
Just to add some horribly long forgot fuel to this topic we have to remember the possibility that latte is on 40nm and RV700 chips were on 55nm.

The wii u mcm should be on 45nm.

Marlenus · Jan 27, 2016

Fourth Storm said:
I think what people mistook for the "half size" shader groups to the left of the SIMD cores were actually the local data shares.

Or maybe that, the bottom one of the pairs does look slightly different to all the other clusters.

RV770 was not that large a die but I would not be surprised if the top one in each block is an extra shader cluster to improve yields.

Either way the Wii U GPU only has 160 cores which is pathetic, even for when it launched.

Locuza · Jan 27, 2016

If I see it right, AMD got 40 ALUs per SIMD in Llano vs. the 20 in earlier designs:
http://images.anandtech.com/doci/7677/DieShot%20-%20Llano.jpg

If the same is true for the Wiiu it could be 320 ALUs, if it's the classic 20 ALU-Design, then 176 GFLOPs are right.

Domaje · Jan 27, 2016

Relaxed Muscle said:
You should care, because you could have all these beautiful/creative game without jaggies that could cut your eyes.

Nah I also don't care. If a game is good enough thank god I can easily forget that I'm "cutting my eyes" with jaggies.

Marlenus · Jan 27, 2016

Locuza said:
If I see it right, AMD got 40 ALUs per SIMD in Llano vs. the 20 in earlier designs:
http://images.anandtech.com/doci/7677/DieShot%20-%20Llano.jpg

If the same is true for the Wiiu it could be 320 ALUs, if it's the classic 20 ALU-Design, then 176 GFLOPs are right.

The Wii U GPU is based on RV770 so I doubt it has any of the enhancements made after RV870 was released. Heck the original RV870 in the 5870 was a 20 ALU-Design. This only got enhanced with llano in the Sumo GPU.

Salbug · Jan 27, 2016

BennyBlanco said:
it only flopped once afaik

Oh! ...Oh...

Netherscourge · Jan 27, 2016

BennyBlanco said:
it only flopped once afaik

Overside · Jan 27, 2016

There is a solution to make everyone feel better.

Depending on what architecture you want to use to normalize the flops to, you could have a wide variety of *OFFICIAL* Wii U Gflop numbers.

ninjablade · Jan 27, 2016

Overside said:
There is a solution to make everyone feel better.

Depending on what architecture you want to use to normalize the flops to, you could have a wide variety of *OFFICIAL* Wii U Gflop numbers.

Not really. almost every tech head is in agreement that it's most likely 176 gflops, then you have leaked documents from bigassain,saying 176 gflops along with devs article confirming those numbers before wiiu release. you can also look at the games nothing suggests wiiu is has almost double the gpu power of 360/ps3 unless there delusional.

Locuza · Jan 27, 2016

Marlenus said:
The Wii U GPU is based on RV770 so I doubt it has any of the enhancements made after RV870 was released. Heck the original RV870 in the 5870 was a 20 ALU-Design. This only got enhanced with llano in the Sumo GPU.

If we specify chips then it's probably based on the 40nm IP-Blocks from the RV740.
Well sadly it's a speculation.
Although the Wii U emulator guys probably should know.
Maybe there are some documents.

FStubbs · Jan 27, 2016

Kapow said:
They're also way more expensive than a WII U...

Also, let's see how long that battery lasts pushing 588 Gflops.

leehom · Jan 27, 2016

It's not 176 or 352, it's just a flop.

AmyS · Jan 27, 2016

I clearly remember an official Nintendo interview, statement about Wii U's GPGPU (Latte). Nintendo stated that Wii U's GPU is 20 times more powerful than the GPU in the original Wii (Hollywood).

Knowing that Hollywood was just a 1.5x (50%) overclocked version of Flipper,

GameCube Flipper GPU - 9.4 gflops
Wii Hollywood GPU (Flipper + 50% (1.5x) - 14.1 gflops
:
If Wii U GPU is 20 times more powerful than Hollywood.

14.1 gflops x 20 = 282 gflops. That is likely the performance of Wii U's GP/GPU.

Wii U Latte GPU = 282 gflops (?)

So, higher than the often quoted 176 gflops, but not as high as the equally often quoted = 352flops

Overside · Jan 27, 2016

ninjablade said:
Not really. almost every tech head is in agreement that it's most likely 176 gflops, then you have leaked documents from bigassain,saying 176 gflops along with devs article confirming those numbers before wiiu release. you can also look at the games nothing suggests wiiu is has almost double the gpu power of 360/ps3 unless there delusional.

Holy shit dude.

You are the gift that keeps giving.

Flops don't convert directly across architectures.

What might take one architecture 1 flop to complete could take another architecture 8 flops to complete. In order to compare floating point computation power between architectures you need to normalize the flops, depending on what architecture you decided to normalize to, you could make the flops number increase or decrease, the amount of work completed, would not change.

It was a joke you hopeless goof.

Dicer · Jan 27, 2016

It's over 9000 jiggawatts of fun, and that's all that matters to me....

PK Gaming · Jan 27, 2016

Feel_The_Pain said:
Who flopping cares.

The person who posted this thread

The people posting in this thread

Plot twist

you

dmr87 · Jan 27, 2016

Sub Boss said:
A Wii U tech thread?
We needed one more of these, i missed them

What year is it?

ninjablade · Jan 27, 2016

Overside said:
Holy shit dude.

You are the gift that keeps giving.

Flops don't convert directly across architectures.

What might take one architecture 1 flop to complete could take another architecture 8 flops to complete. In order to compare floating point computation power between architectures you need to normalize the flops, depending on what architecture you decided to normalize to, you could make the flops number increase or decrease, the amount of work completed, would not change.

It was a joke you hopeless goof.

Except there both amd gpu's, and pretty similar based on everything we know, wiiu gpu is only slightly more efficient.

AmyS said:
I clearly remember an official Nintendo interview, statement about Wii U's GPGPU (Latte). Nintendo stated that Wii U's GPU is 20 times more powerful than the GPU in the original Wii (Hollywood).

Knowing that Hollywood was just a 1.5x (50%) overclocked version of Flipper,

GameCube Flipper GPU - 9.4 gflops
Wii Hollywood GPU (Flipper + 50% (1.5x) - 14.1 gflops
:
If Wii U GPU is 20 times more powerful than Hollywood.

14.1 gflops x 20 = 282 gflops. That is likely the performance of Wii U's GP/GPU.

Wii U Latte GPU = 282 gflops (?)

So, higher than the often quoted 176 gflops, but not as high as the equally often quoted = 352flops

It's can only be 352flops or 176, based on the gpu configuration, it's all but confirmed to 176 gflops.

nkarafo · Jan 27, 2016

Is there a chart that compares computational power between consoles, PCs, handhelds and smartphones?

I still don't have a clear image of the gaps between them.

Overside · Jan 27, 2016

ninjablade said:
Except there both amd gpu, and pretty similar, based on everything we know wiiu gpu is only slightly more efficient.

That is completely irrelevant to what I said.

Your inability to comprehend what is going on while pretending you do is the stuff of legends.

AmyS · Jan 27, 2016

ninjablade said:
Except there both amd gpu's, and pretty similar based on everything we know, wiiu gpu is only slightly more efficient.

It's can only be 352flops or 176, based on the gpu configuration, it's all but confirmed to 176 gflops.

I believe it now. Before I wasn't sure.

Marlenus · Jan 27, 2016

Overside said:
Holy shit dude.

You are the gift that keeps giving.

Flops don't convert directly across architectures.

What might take one architecture 1 flop to complete could take another architecture 8 flops to complete. In order to compare floating point computation power between architectures you need to normalize the flops, depending on what architecture you decided to normalize to, you could make the flops number increase or decrease, the amount of work completed, would not change.

It was a joke you hopeless goof.

Comparing fixed function to unified is nebulous but the 4670 which is a 320 shader part is not much faster at all than the 2900xt which is basically the architecture in the Xbox 360. That shows the flops in those cards is comparable.

Overside · Jan 27, 2016

Marlenus said:
Comparing fixed function to unified is nebulous but the 4670 which is a 320 shader part is not much faster at all than the 2900xt which is basically the architecture in the Xbox 360. That shows the flops in those cards is comparable.

That is also irrelevant.

*Hint* The original statement, as well as what you quoted, has has nothing to do with the 360.

Read what is being said in what you quoted and think about it.
If what you end up with is seriously trying to look at specs of any sysem, you missed by a mile.

Flops don't convert directly across architectures. They have to be normalized before they can really be compared. That means, if you choose an architecture that takes a lot of operations to do what today's processors, like say the Wii u's GPU, only need one or two flops to do, and normalize to that architectures flops, you can inflate the arbitrary flops number for the Wii u making people who rely on what is actually an arbitrary number as if it were actually something concrete, feel like it is more powerful because the number is higher. The actual amount of work done, remains unchanged

.

neurosisxeno · Jan 27, 2016

tkscz said:
Architecture usually plays a larger roll than pure numbers. FLOPS are like the new BITS honestly. Just because something has more, doesn't mean it's better (not saying it isn't better for having more, but I think you get what I'm saying).

As apparent by the Iris Pro numbers.

Alchemy · Jan 27, 2016

kabel said:
It doesn't matter, 176 gflops or 352 gflops.
It's both pretty low.

Even smartphones are way more powerful. (Snapdragon 820 -> 588 gflops)

There is a hell of a lot of things you can't get performance wise just from raw calculation estimates. Smartphones are still teetering around in that bullshit "console quality" "ehh we're almost 360/PS3 level" space. Is there anything in the mobile space that even comes close to looking as good as Mario Kart 8?

anothertech · Jan 27, 2016

I looked at the thread title and seriously did a double take at the date. Thought this was a necro bump or something! And look at the chip charts and hardware discussions !

The Abominable Snowman · Jan 27, 2016

Alchemy said:
There is a hell of a lot of things you can't get performance wise just from raw calculation estimates. Smartphones are still teetering around in that bullshit "console quality" "ehh we're almost 360/PS3 level" space. Is there anything in the mobile space that even comes close to looking as good as Mario Kart 8?

Hm, I mean there aren't any mobile games that look better than the top Vita games, IMO, but Vita is based on the same designs. And Vita does come close (I'd say 70%) to approximating PS3 at roughly 1/4th the resolution.

Now we have mobile GPUs 10-15x faster in sub-$400 handsets.

Feel_The_Pain · Jan 27, 2016

PK Gaming said:
The person who posted this thread

The people posting in this thread

Plot twist

you

Plot twist

you too

GhostTrick · Jan 27, 2016

The Abominable Snowman said:
Hm, I mean there aren't any mobile games that look better than the top Vita games, IMO, but Vita is based on the same designs. And Vita does come close (I'd say 70%) to approximating PS3 at roughly 1/4th the resolution.

Now we have mobile GPUs 10-15x faster in sub-$400 handsets.

Its all about resolution, heat and budget.
Top Vita games are often subnative res. As for Vita, its a 28gflops part. But what makes the difference is the lower res. When your game is 360p or 400p, its easier to target higher grade graphics than on a 1080p or 1440p screen. Especially when mobile games are targeting 720p to avoid the thing to look to blurry and messy on 1080p screen.

Heat and budget are the reason why we wont see mobile games using the hardware power available.

LeleSocho · Jan 27, 2016

nkarafo said:
Is there a chart that compares computational power between consoles, PCs, handhelds and smartphones?

I still don't have a clear image of the gaps between them.

Roughly:
Today's smartphones and tablets have finally arrived at level of PS360 but they are limited by batteries (aka while they are able to push it you'll never see that quality because otherwise the batteries would last nothing and the devices becomes lava in your hand)
Ps4 is like a 7850 with lots of vram and super high end netbook cpu... let's say 2009 high end pcs (???)
Vita is an high end 2012 tablet, think of in between iPad 3 and 4 (more like iPad 3 on steroids)
Today's high end PCs are several times more powerful than todays consoles

The Abominable Snowman · Jan 27, 2016

GhostTrick said:
Its all about resolution, heat and budget.
Top Vita games are often subnative res. As for Vita, its a 28gflops part. But what makes the difference is the lower res. When your game is 360p or 400p, its easier to target higher grade graphics than on a 1080p or 1440p screen. Especially when mobile games are targeting 720p to avoid the thing to look to blurry and messy on 1080p screen.

Heat and budget are the reason why we wont see mobile games using the hardware power available.

Yeah, and also the Vita has an efficient large heat spreader and better heat dissipation so it's able to consistently use more of the hardware potential without overheating. Most phones rely on contact with the back casing, while many tablets (Surface) are using ionic (air) cooling methods

In phones, that's usually done away with for thinness, weight, better signal, space and other things that just won't matter as much in a dedicated handheld.

And of course the power schedulers in Android are overbearing and tend to kill performance and benchmarks while maintaining multiple tasks and services. The chips are still operating 5W at full load or under, but we are conditioned into thinking that somehow battery life has gotten substantially worse when it has gotten better. A Note 5 with its 'measly' 3000mAh battery manages over 5-6 hours in VR Oculus gaming, no power saving enabled.

But if Nintendo or Sony wanted a handheld with a Snapdragon 820 or similar SoC, a Surface-like ionic cooling system, and a 3-4000mAh battery for 6+ hours of game time, there's nothing stopping them, and we would then see what the chips are really capable of.

ninjablade · Jan 27, 2016

Overside said:
That is also irrelevant.

*Hint* The original statement, as well as what you quoted, has has nothing to do with the 360.

Read what is being said in what you quoted and think about it.
If what you end up with is seriously trying to look at specs of any sysem, you missed by a mile.

Flops don't convert directly across architectures. They have to be normalized before they can really be compared. That means, if you choose an architecture that takes a lot of operations to do what today's processors, like say the Wii u's GPU, only need one or two flops to do, and normalize to that architectures flops, you can inflate the arbitrary flops number for the Wii u making people who rely on what is actually an arbitrary number as if it were actually something concrete, feel like it is more powerful because the number is higher. The actual amount of work done, remains unchanged

.

I presume you're talking about 360 architecture here? Seriously makes no sense. we know what both architecture are, cpu,ram, and bandwidth, both have advantages and disadvantages.

L.O.R.D · Jan 27, 2016

It doesn't matter, because most of the devs on Wii u can't get the advantage of that power.

DonMigs85 · Jan 27, 2016

It's really a shame how low Wii U's floating point power is, especially considering it may have had 320 stream processors originally only to be cut down due to heat issues in the tiny case. Heck, even its raw texel fillrate is barely more than half that of the PS3 and 360 since it only has 1 TMU per ROP. What really saves it is the relatively large eDRAM and double the amount of usable system memory versus the older consoles. At the very least some games like Need for Speed Most Wanted and Sonic Racing Transformed managed to keep pace with or slightly exceed the other versions.

Marlenus · Jan 27, 2016

Overside said:
That is also irrelevant.

*Hint* The original statement, as well as what you quoted, has has nothing to do with the 360.

Read what is being said in what you quoted and think about it.
If what you end up with is seriously trying to look at specs of any sysem, you missed by a mile.

Flops don't convert directly across architectures. They have to be normalized before they can really be compared. That means, if you choose an architecture that takes a lot of operations to do what today's processors, like say the Wii u's GPU, only need one or two flops to do, and normalize to that architectures flops, you can inflate the arbitrary flops number for the Wii u making people who rely on what is actually an arbitrary number as if it were actually something concrete, feel like it is more powerful because the number is higher. The actual amount of work done, remains unchanged

.

For all AMD architectures from the Xbox 360 to the latest GCN and probably Polaris flops is calculated by the simple equation shaders * clock * 2.

Flops are not arbitrary but they are far from a complete measure of performance nor are they an accurate predictor of real world performance across architectures.

I get what you are attempting to say and I agree but what you actually said is categorically incorrect.

nkarafo · Jan 27, 2016

LeleSocho said:
Roughly:
Today's smartphones and tablets have finally arrived at level of PS360 but they are limited by batteries (aka while they are able to push it you'll never see that quality because otherwise the batteries would last nothing and the devices becomes lava in your hand)
Ps4 is like a 7850 with lots of vram and super high end netbook cpu... let's say 2009 high end pcs (???)
Vita is an high end 2012 tablet, think of in between iPad 3 and 4 (more like iPad 3 on steroids)
Today's high end PCs are several times more powerful than todays consoles

Thanks, puts some things to perspective.

Hermii · Jan 27, 2016

I wonder if there will be as much discussion about the NX gpu flop count.t

lherre · Jan 27, 2016

I have a deja vû

GhostTrick · Jan 27, 2016

LeleSocho said:
Roughly:
Today's smartphones and tablets have finally arrived at level of PS360 but they are limited by batteries (aka while they are able to push it you'll never see that quality because otherwise the batteries would last nothing and the devices becomes lava in your hand)
Ps4 is like a 7850 with lots of vram and super high end netbook cpu... let's say 2009 high end pcs (???)
Vita is an high end 2012 tablet, think of in between iPad 3 and 4 (more like iPad 3 on steroids)
Today's high end PCs are several times more powerful than todays consoles

More like iPad 3 and maybe even slightly less. Vita is dramatically underclocked. The CPU runs at 333mhz and can reach 444mhz. The GPU is at 222mhz.

LeleSocho · Jan 27, 2016

GhostTrick said:
More like iPad 3 and maybe even slightly less. Vita is dramatically underclocked. The CPU runs at 333mhz and can reach 444mhz. The GPU is at 222mhz.

Yeah i actually forgot all about the clocks and only counted the cores... my bad.

GhostTrick · Jan 27, 2016

LeleSocho said:
Yeah i actually forgot all about the clocks and only counted the cores... my bad.

Eh, cant blame you, no one expected that.

Support NeoGAF

Is the Wii U gpu 176 gflops or 352 gigaflops.

Member

Member

Member

Member

Banned

Member

Member

Member

Member

Member

Member

Member

Member

Member

Member

Banned

Banned

Banned

Member

Member

Member

Member

Banned

Banned

Member

Member

Banned

Member

Banned

Member

Member

Banned

Member

Member

Member

Member

Member

Banned

Banned

Member

Banned

Member

Member

Member

Member

Member

Accurate

Banned

Banned

Banned

Similar threads