Nintendo's hardware philosophy: Always old, outdated tech?

[liolio]

By what means would the various components be identified? How certain could you be of the specifications if you were working on a NES 6 game right now? Maybe they call it a R700 because certain parts of the architecture haven't been exposed yet? Im not sure, maybe someone call enlighten me here...

Actually I'm not sure most site (even the biggest) would think about the difference between the various R7xx products. As I get this rumours is the "core of the GPU is the same as the HD4xxx family", nothing more.
If there is some truth to this rumours I expect the shader core, the RBE, the command processor, etc to be akin to the HD4xxx products line.
That let room for speculation various details (as SIMD width, number of SIMD arrys, number of RBEs, clock speed, etc.) but I can't see N pushing improvment on any of these parts.
the only think I could think that would have make sense for N to implement (especially if they ship a APU) is some power management features. Like they set a peak for power consumption and CPU and GPU frequency vary accordingly.
I don't like the idea of N passing on GDDR5, that would be really troublesome and the chip would be strongly bandwidth constrained. I would actually favor 1GB of GDDR5 to 2GB of DDR3. GDDR5 is power hungry but in case of a single chip you pay the price only "once". I don't remember HD4670 number but I'm confident that part of the increase in power consumption is due memory controler and not the 25 extra MHz vs the 4670. If case the of a single chip the power price is shared between the CPU and the GPU.

 

[Squilliam]

Actually I'm not sure most site (even the biggest) would think about the difference between the various R7xx products. As I get this rumours is the "core of the GPU is the same as the HD4xxx family", nothing more.
If there is some truth to this rumours I expect the shader core, the RBE, the command processor, etc to be akin to the HD4xxx products line.
That let room for speculation various details (as SIMD width, number of SIMD arrys, number of RBEs, clock speed, etc.) but I can't see N pushing improvment on any of these parts.

What I want to know is if what basis the rumours are based on? Do they have final hardware and is it possible the full features of the system haven't been exposed yet? For instance if they are using a custom tessellation API it may not be finished and some people may have concluded no tessellation = R7xx chip as a possible explanation.

I don't like the idea of N passing on GDDR5, that would be really troublesome and the chip would be strongly bandwidth constrained. I would actually favor 1GB of GDDR5 to 2GB of DDR3. GDDR5 is power hungry but in case of a single chip you pay the price only "once". I don't remember HD4670 number but I'm confident that part of the increase in

I would suggest that regular DDR3 with large on die caches would possibly be more power efficient than GDDR5 given better data locality and that would support the data access paterns of the CPU better. Bandwidth like anything is a tradeoff with cost/heat and other design considerations.

 

[Entropy]

Öo

I would suggest that regular DDR3 with large on die caches would possibly be more power efficient than GDDR5 given better data locality and that would support the data access paterns of the CPU better. Bandwidth like anything is a tradeoff with cost/heat and other design considerations.

AMD uses GDDR5 for their mobile products, without (seemingly) either cost nor power draw being a concern anymore. As for average latency for the CPU the traditional solution is to add cache. Assuming the Xenos as the base, simply increasing cache sizes a bit over what was practical on 90nm seems like a nobrainer, particularly as it would help with main memory bus contention, and make locking areas of the cache for guaranteed local processing more practical.

I think your suspiciousness regarding the (lack of) precision in the R700 rumor is wise.

 

[liolio]

What I want to know is if what basis the rumours are based on? Do they have final hardware and is it possible the full features of the system haven't been exposed yet? For instance if they are using a custom tessellation API it may not be finished and some people may have concluded no tessellation = R7xx chip as a possible explanation.

R7xx supported tesselation. But your take is possible. As for the basis for the rumours, well we won't know either somehow that heard stuffs on the distance or some guy that did not want to tell too much or last but not least plain BS.

I would suggest that regular DDR3 with large on die caches would possibly be more power efficient than GDDR5 given better data locality and that would support the data access paterns of the CPU better. Bandwidth like anything is a tradeoff with cost/heat and other design considerations.

Large cache like a a bunch a SRAM? Or EDram? (as it has process implications). If they go with DDR3 they will have to which would imply consequent change to the gpu design.

All this is interesting there is nothing really interesting right now in the gaming realm (at least from my pov) the NES 6 specualtions keeps me busy.

 

[swaaye]

Xenon certainly could be improved upon as a CPU. Just look at how awful some ports play to see how gimpy it is. Look at Oblivion for example. If I recall correctly, it doesn't have enough L2 cache for 3 cores and it is a fairly simple PowerPC (in-order). It has amazing SIMD capabilities but those are meaningless sometimes. I wouldn't be surprised if each core has similar general purpose perf/clock to Atom. I don't really understand why a beefier dual core wouldn't have been better.

 

[AlphaWolf]

Xenon certainly could be improved upon as a CPU. Just look at how awful some ports play to see how gimpy it is. Look at Oblivion for example. If I recall correctly, it doesn't have enough L2 cache for 3 cores and it is a fairly simple PowerPC (in-order). It has amazing SIMD capabilities but those are meaningless sometimes. I wouldn't be surprised if each core has similar general purpose perf/clock to Atom. I don't really understand why a beefier dual core wouldn't have been better.

They might have used Atom if it had been available in 2005.

 

[Deleted member 13524]

They might have used Atom if it had been available in 2005.

It would save them a lot of hassle regarding XBox 1 BC, that's for sure. They might even have had the luxury to enable the X360 to play all XBox 1 games @ 720p or 1080p.

 

[swaaye]

It's too bad Athlon 64 X2s weren't dirt cheap in 2005 like they have been since Core 2 launched.

Yeah I suppose considering the cheap CPU options in 2003-5 that Xenon made much sense.

 

[Squilliam]

AMD uses GDDR5 for their mobile products, without (seemingly) either cost nor power draw being a concern anymore. As for average latency for the CPU the traditional solution is to add cache. Assuming the Xenos as the base, simply increasing cache sizes a bit over what was practical on 90nm seems like a nobrainer, particularly as it would help with main memory bus contention, and make locking areas of the cache for guaranteed local processing more practical.

Large cache like a a bunch a SRAM? Or EDram? (as it has process implications). If they go with DDR3 they will have to which would imply consequent change to the gpu design.

I think AMD uses GDDR5 because thats the only practical alternative aside from simply sticking with DDR3 and having their chips bandwidth starved. In this instance Nintendo can design a system which could have a much larger on board cache. If Microsoft was able to get away with 25GB/S of real bandwidth on the Xbox 360 then surely Nintendo could do the same with 2-3* as much ED-RAM on board and relatively slow DDR3?

 

[swaaye]

Nintendo has really liked low latency RAM since Gamecube. N64 had absolutely awful memory performance and Cube was a design that directly solved that problem (1T-SRAM as main memory!). So it will be interesting to see if they continue to go after low latency memory solutions.

 

[ninzel]

How does the rumoured size of the Wii2 being about the size of the original 360 fit into these other predictions? I assume Nintendo would only make it that big out of necessity.
How much power would make that box size necessary?

 

[AlphaWolf]

How does the rumoured size of the Wii2 being about the size of the original 360 fit into these other predictions? I assume Nintendo would only make it that big out of necessity.
How much power would make that box size necessary?

the 360 shipped at over 200watts, but it also has a brick. I'd expect wii2 to be quite a bit lower power

 

[ninzel]

the 360 shipped at over 200watts, but it also has a brick. I'd expect wii2 to be quite a bit lower power

So why would they use a box as big as the original 360 unless it's necessary? Does that mean they would be using the same older technology as original 360 and have same power,or newer more efficient meaning more power in bigger box?

 

[swaaye]

As long as there isn't also a brick half the size of the console. People forget that when comparing the 360 to Xbox 1 with its internal PSU.

 

[jlippo]

So why would they use a box as big as the original 360 unless it's necessary? Does that mean they would be using the same older technology as original 360 and have same power,or newer more efficient meaning more power in bigger box?

Perhaps it works as a loading station for the controllers.

 

[ninzel]

I was thinking about processing power rather than power draw.

 

[AlphaWolf]

I was thinking about processing power rather than power draw.

They are inseparable. The die size of the of faster processor isn't significant to the size of the box other than the the power draw it takes and the cooling requirements it needs because of that power draw. IF the 360 drew 30 watts at launch, it would have been around the size of the Wii (larger because of hdd, but smaller because of external supply).

 

[ninzel]

They are inseparable. The die size of the of faster processor isn't significant to the size of the box other than the the power draw it takes and the cooling requirements it needs because of that power draw. IF the 360 drew 30 watts at launch, it would have been around the size of the Wii (larger because of hdd, but smaller because of external supply).

Yes of course. So let me try this again.
When the 360 launched it used a certain size chips(90nm?) which gave a certain amount of processing power at approx 200 watts power draw.
Now it uses a smaller chip(45nm?) at around the same processing power but at less power draw thus smaller box.
So now we come to Wii2. Assuming it uses 45nm chips like current 360 but requires a larger box,it would seem to indicate a comparable power draw(200 watts)? I don't know,but I assume Nintendo wouldn't use a larger box unless it needed to(heat caused power draw). So assuming an equal size chip(45nm) but a requirement for a larger box what's causing the extra power draw and heat? Faster more powerful chips?
Can we say that if chips=45nm+ requirement for box large enough to accommodate higher heat,is it possible to accurately predict an/the theoretical processing power?

 

[msia2k75]

Yes of course. So let me try this again.
When the 360 launched it used a certain size chips(90nm?) which gave a certain amount of processing power at approx 200 watts power draw.
Now it uses a smaller chip(45nm?) at around the same processing power but at less power draw thus smaller box.
So now we come to Wii2. Assuming it uses 45nm chips like current 360 but requires a larger box,it would seem to indicate a comparable power draw(200 watts)? I don't know,but I assume Nintendo wouldn't use a larger box unless it needed to(heat caused power draw). So assuming an equal size chip(45nm) but a requirement for a larger box what's causing the extra power draw and heat? Faster more powerful chips?
Can we say that if chips=45nm+ requirement for box large enough to accommodate higher heat,is it possible to accurately predict an/the theoretical processing power?

No, power draw would be much less on the WiiHD.
I read the box for the next Wii was more comparable with the japanese and european version of SNES.

 

[swaaye]

I really expect Nintendo to keep to their tradition of non-furnace like consoles. They have never built anything that was hot. They also build the most reliable hardware and high temps don't lead in that direction. They didn't even have mechanical parts (fan/optical) in the console until Gamecube.