Why the Wii U is Probably More Capable Than You Think It Is
Manfred Linzner of Shin’en had been kind enough to supply us with quite a bit of insight into the Wii U’s capability, and has done so once again.You can read our past interviews here before reading on, if you like:
Here is the latest. My personal commentary will follow below after the divide.
“The Wii U eDRAM has a similar function as the eDRAM in the XBOX360. You put your GPU buffers there for fast access. On Wii U it is just much more available than on XBOX360, which means you can render faster because all of your buffers can reside in this very fast RAM. On Wii U the eDRAM is available to the GPU and CPU. So you can also use it very efficiently to speed up your application.
The 1GB application RAM is used for all the games resources. Audio, textures, geometry, etc.
Theoretical RAM bandwidth in a system doesn’t tell you too much because GPU caching will hide a lot of this latency. Bandwidth is mostly an issue for the GPU if you make scattered reads around the memory. This is never a good idea for good performance.
I can’t detail the Wii U GPU but remember it’s a GPGPU. So you are lifted from most limits you had on previous consoles. I think that if you have problems making a great looking game on Wii U then it’s not a problem of the hardware.”
So, let’s break down these comments for explanatory purposes. If you like, you may want to use the earlier links for further insight.
“The Wii U eDRAM has a similar function as the eDRAM in the XBOX360. You put your GPU buffers there for fast access. On Wii U it is just much more available than on XBOX360, which means you can render faster because all of your buffers can reside in this very fast RAM. On Wii U the eDRAM is available to the GPU and CPU. So you can also use it very efficiently to speed up your application.”
The Xbox 360 has 10 MB of eDRAM. With that eDRAM, the console was able to use it for several rendering targets. Yet, the amount was slightly insufficient for even 720p resolution at times once Multi Sampling Anti Aliasing was used. This was why at times you would see games that had sub 720p resolution, or the Console had to chug extra hard with Predicated Tiling in order to pick up the slack. As Mr. Linzner explained to us in an earlier interview, about 16 MB of eDRAM is required for 1080p rendering with a double buffer, even without Anti Aliasing.
So what is meant by: “On Wii U it is much more available”?
Two main things. For one, all in all, the Wii U has around 38 MB total of eDRAM if you count the total of 35 MB on the GPU die, and the 3 that exist as Cache for the CPU. That gives the Wii U in total, nearly 4 times the amount of very fast eDRAM that the Xbox 360 (which had enough to use it for 720p frame buffers) has, with more than double the amount needed to render in 1080p. Hence the statement that ALL of your buffers can reside in this very fast RAM.
Mr. Linzner then makes it a point to say once again, that on Wii U, the eDRAM is available to the GPU and CPU. So effectively, outside of the CPU Cache eDRAM, this RAM exists as a unified pool that can be used as needed for a variety of tasks. I.E. “Very efficiently, to speed up your application.”
At this point, you may be wondering, how does the speed of this RAM make such a difference? The Wii U has 2 Gigabytes of RAM, with one currently available for gaming, along with the eDRAM described above. How can 38 MB of RAM make that much of a difference?
The easiest way that I can explain this is that when you take each unit of time that the Wii U eDRAM can do work with separate tasks as compared to the 1 Gigabyte of slower RAM, the amount of actual Megabytes of RAM that exist during the same time frame is superior with the eDRAM, regardless of the fact that the size and number applied makes the 1 Gigabyte of DDR3 RAM seem larger. These are units of both time and space. Fast eDRAM that can be used at a speed more useful to the CPU and GPU have certain advantages, that when exploited, give the console great gains in performance.
The eDRAM of the Wii U is embedded right onto the chip logic, which for most intent and purposes negates the classic In/Out bottleneck that developers have faced in the past as well. Reading and writing directly in regard to all of the chips on the Multi Chip Module as instructed.
So what then is the 1 Gigabyte of slower RAM for? As stated:
“The 1GB application RAM is used for all the games resources. Audio, textures, geometry, etc.”
Think of it in terms of the quote/unquote larger, slower bit of RAM holding the foundations of what you see on screen. While the quote/unquote smaller, faster bit of RAM is responsible for allowing the CPU and GPU to express much of the detail that makes a High Definition game, a High Definition game in a way that doesn’t call for insane amounts of heat producing power and Developer annoying latency.
In a single second, a single Megabyte of eDRAM, embedded on the GPU, can receive and send amounts of data, that so greatly surpass the number on paper, it’s scary. Being so close in proximity to the shader units, and the CPU, and embedded directly into the same logic wafer that connects them all, every piece becomes so much more than they would be when separated and connected in the more classic way. Not to mention far less wasteful, and more capable than most people realize or acknowledge.
“Theoretical RAM bandwidth in a system doesn’t tell you too much because GPU caching will hide a lot of this latency. Bandwidth is mostly an issue for the GPU if you make scattered reads around the memory. This is never a good idea for good performance.”
The picture appears to become more and more clear.
Finally, when I asked Mr. Linzner if there were any features of the latest version of Direct X that Wii U couldn’t perform the equivalent of, he had this to say:
“I can’t detail the Wii U GPU but remember it’s a GPGPU. So you are lifted from most limits you had on previous consoles. I think that if you have problems making a great looking game on Wii U then it’s not a problem of the hardware.”
GPGPU = General Purpose Graphics Processing Unit. In the case of the Wii U, this chip is connected to the CPU, with eDRAM as the ‘magic’ facilitator. A chip that as a whole, can do whatever it’s programmed to, and do it fast and efficiently. Even the Xbox 360 and PS3 could be configured to perform effects and mimick features beyond what their chips were actually designed to do. The Wii U Multichip module in addition to its standard feature set, is a canvass upon which a talented artist can achieve whatever it is that he can imagine through code, in a way that will not cripple the hardware.
How can a machine that has such a (relatively speaking), low wattage be so capable? In short, Nintendo has created a tight focal point where the most important pieces are all integrated in a way where speed is great but has a low cost, while travel distance is very short and interwoven. A lot of power is not necessary, so in terms of power draw, the machine, and in fact, all 3 new consoles cannot be judged in precisely the same ways as before.
Nintendo, having had the Wii U painted for the most part as less capable than it is by the gaming public at large, continues to have to push the boundaries of their comfort zone as far as pricing their hardware at a financial loss to them goes. However, as a result, they, by far have the lowest price point and a machine of true quality, which now that the library of games begins to become rather stout, Gamers will have to strongly consider Nintendo’s Wii U once they take their wallets out to buy a new console.
Speaking of quality, and games, keep your eyes open and ears tuned for future announcements from Shin’en Multimedia regarding the two new games that they’ve been working on, and if you haven’t yet bought the absolutely sweet Nano Assault Neo (which is only $10 on the Wii U eShop), then you are missing out big time!