Predict: The Next Generation Console Tech

[Mobius1aic]

"Internal Main Memory" = eDRAM?

LSI = Large Scale Integration?

"System LSI" = one package to hold all those components or just a single discreet area of the mobo all under the same heatsink? Weird.

The Patent lists the DSP as using the internal main memory, so maybe it's just another general purpose memory pool? Seems more complicated than necessary to me, unless Nintendo is trying to recreate the GC/Wii operating environment for BC which could be done in emulation easily enough if the Wii U does have the rumored specs.

 

[RudeCurve]

"System LSI" looks like Flipper

VRAM = 1MB eDRAM framebuffer
Internal Main Memory = 2MB eDRAM texture cache
DSP = Audio DSP

Could also be the new version of System LSI for WiiU with new GPU and more eDRAM.

 

[AlNom]

No, the internal main memory is described as a general buffer in the document (search 11e). See my post above (it's really more like the Wii with the 24MB 1T-SRAM being the IMM), but I mean, it's all documented there.

 

[Shifty Geezer]

Maybe they just want to protect the controller atm, and they don't want to disclose Wii U hardware yet, so they just used the Wii for patent purposes. Is that feasible/legal? Anyone here well verse don patent law? So many questions,

I just took a look. It's given me a headache just scrolling to the bit that actually matters!!

A patent's meat, the bit that matters, are the claims, specifically the first claim which identifies what exactly is the new idea being patented. The rest of the document is just explanation and examples. Normally the words "one specific embodiment" or similar are used to say that what's described explicitly in the patent is not the only application, and if the patent can be contrived to cover other cases then that's intended (cast your patent as wide as possible to limit infringing/enable patent trolling).

As ever, the actual patented claim is utter crock that isn't worth the words someone's wasted to decribe all this:

What is claimed is:
1. A controller device comprising:
a generally plate-shaped housing;
a display section provided on a front side of the housing;
a first operation section and a second operation section provided respectively on a left side and a right side of the display section above a center of the housing; and
a third operation section and a fourth operation section provided on a back side of the housing so as to generally correspond respectively to the first operation section and the second operation section on the front side of the housing.

That is, they are patenting a controller with a screen and controls on front and back. The rest of the claims are just vaguely expanding that notion with more controls and expanion ports, and a reference to inbuilt game processing in claim 19.

Quite why they feel the need to go into data flow diagrams and console layouts to patent this, who knows. Patent laywers really are an odd breed ('a front side of the housing'? There's more than one front side on this plate-shaped housing??).

Thus we can ignore the claims, and just fish around the rest of the patent for ideas. It could be that they dropped in the mainboard diagram accidentally, but as others say, it's more likely they used the Wii as reference. The specifics of that board have nothing to do with the claims being made so the don't need to be technically accurate. Things like the gun connection show intended application. I can't really see the COD fans going wild over that one...

All in all though, there's very little to see.

 

[lefizz]

Durango already in production?

http://www.tomsguide.com/us/microsoft-xbox-720-durango-xbox-next-gaming,news-15088.html

Strange rumor

 

[DuckThor Evil]

Well, of course he hasn't. I hate to give ammunition the haters, but he did mention his CPU was running hot. 55C and 85C at load I think. Heck I tried overclocking my Q6600 the other day and it shut off at 70C. I didn't know CPU's could go above ~70c.

Anyways I doubt the change in video card would affect things that much, so I'd have to wonder if other X51 CPU's run hot, I assume not since no GAF commenters brought it up.

Thinking seriously, I'm thinking if you dropped a pitcairn/7870 in a console, you would probably have to underclock it to 800.

The 7850 model he has in there leaves majority of the heated air inside the case. I'm surprised it actually ends up working that well as the case is small. The card is very power efficient that's for sure, but a different model that exhausts the hot air out of the case should work significantly better. I assume his CPU gets hot for this very reason, too much hot air running around there.

 

[Kaotik]

Well, of course he hasn't. I hate to give ammunition the haters, but he did mention his CPU was running hot. 55C and 85C at load I think. Heck I tried overclocking my Q6600 the other day and it shut off at 70C. I didn't know CPU's could go above ~70c.

Yes they can, the fact that your shut down at 70C, means your BIOS settings are set at that - for example I had IIRC shutdown temp set to 95C, and my Q6600 (@3.33GHz) went there several times before I realized my stock HSF's mounting had failed. The CPU nor mobo didn't suffer any damage from it.

 

[Acert93]

The 7850 model he has in there leaves majority of the heated air inside the case. I'm surprised it actually ends up working that well as the case is small. The card is very power efficient that's for sure, but a different model that exhausts the hot air out of the case should work significantly better. I assume his CPU gets hot for this very reason, too much hot air running around there.

Something often overlooked is that it isn't the volume of the case or even the number, size, or RPM of the fans but instead the efficiency of drawing cool air over the Heat Sink and expelling the hot air from the case. I built a silent PC in 2005 and it shocked my friends when I showed them the before and after parts. The before parts had some big fast fans but getting a good Heat Sink design (mine was a Thermaltake 90 or the like) and a couple quality fans at lower rpm and a good airflow design with improved cabling.

I mention that because the Alienware is nowhere near a great airflow design, especially compared to what one could do if they were using custom parts designs to be assembled with airflow as a #1 priority.

Adding a GPU that doesn't exhaust doesn't prove anything about future consoles except that the current hardware with HORRIBLE airflow still functions in the 200W range without the benefit of a proper internal layout.

Ps- Dr Evil I was agreeing with you, I just used your post as a jumping point.

 

[DuckThor Evil]

Yep I agree and I almost edited more info about airflow in to my post as well
That X51 case works pretty well with the factory GPU choices, which exhaust the air out, but It's capabilities are soon spent when you install a card with a cooler like in that Sapphire card. Those type of coolers can be very good when a case has good airflow.

 

[babybumb]

http://nukezilla.com/2012/05/02/exc...ave-backwards-compatibility-10-year-lifespan/

New rumour saying Durango will have backwards compatibility only for first three years and this suggesting ARM or x86.

 

[Proelite]

http://nukezilla.com/2012/05/02/exc...ave-backwards-compatibility-10-year-lifespan/

New rumour saying Durango will have backwards compatibility only for first three years and this suggesting ARM or x86.

This is legit. Ive seen the documents first hand in person. It was 4 pages long too without any double spaces.

 

[babybumb]

This is legit. Ive seen the documents first hand in person. It was 4 pages long too without any double spaces.

Sarcasm meter through the roof

 

[Acert93]

LOL They better hope that isn't 64 ALUs! Less than Radeon 7350 here we come! Maybe they meant 64 16-wide Vector SIMD unit as in South Island. In which case that is 1024 ALUs at 1GHz which is... a Radeon HD 7850 at a 7870 frequency and put it just over 2TFLOPs.

EDIT: Oh, and totally unbelievable graphics. The IPC difference between a ARM and x86 is so large that to even throw out the two as possible interchangeable (as indicated by the 6-8 cores) is a good chuckle UNLESS it is indicating a mix of cores.

 

[McHuj]

EDIT: Oh, and totally unbelievable graphics. The IPC difference between a ARM and x86 is so large that to even throw out the two as possible interchangeable (as indicated by the 6-8 cores) is a good chuckle UNLESS it is indicating a mix of cores.

I've thought about an ARM CPU used to run the OS and all back ground tasks. It would certainly help them leverage the Win8 kernal that's being developed for ARM as well as potentially all the Win8/WP8 ARM apps running on the next 360. I think MS realizes that the 360 will have to take on more PC functionality. Who knows how the next AppleTV/GoogleTV's will evolve.

If there's any truth to the always on/always connected/ set-top box functionality that could be handled by a low power ARM processor while rest of the system is powered down.

 

[AlNom]

EDIT: Oh, and totally unbelievable graphics. The IPC difference between a ARM and x86 is so large that to even throw out the two as possible interchangeable (as indicated by the 6-8 cores) is a good chuckle UNLESS it is indicating a mix of cores.

But if you follow the BC graphic, which is one box with the 3x PPC cores, then Application processor must have four of those 6-8x CPUs!

 

[IllusionistK]

Maybe they meant 64CUs? Which would mean 4096 stream processors or maybe a 7990-class GPU in 18 months?

 

[Acert93]

Maybe they meant 64CUs? Which would mean 4096 stream processors or maybe a 7990-class GPU in 18 months?

I like the way you think! Can I come and live in your dream world Seriously, that would be an exciting console. The problem, of course, is 28nm is in mass product in the last 6 months (bleeding edge) and in 18 more months the prospects of a high volume, stable, and relatively affordable 20nm node are... unlikely. So you are looking at a chip bigger than Tahiti on 28nm to hit that target. It isn't impossible, e.g. if they are targeting a fast shrink to 20nm in 2014 and the justification of a long lifespan but the concern of future shrinks, cost, etc all just weight against such. Reality sucks

 

[Ninjaprime]

LOL They better hope that isn't 64 ALUs! Less than Radeon 7350 here we come! Maybe they meant 64 16-wide Vector SIMD unit as in South Island. In which case that is 1024 ALUs at 1GHz which is... a Radeon HD 7850 at a 7870 frequency and put it just over 2TFLOPs.

EDIT: Oh, and totally unbelievable graphics. The IPC difference between a ARM and x86 is so large that to even throw out the two as possible interchangeable (as indicated by the 6-8 cores) is a good chuckle UNLESS it is indicating a mix of cores.

I'm guessing you're right on the vector unit thing. On the ARM vs x86 thing though, depends on which x86 company you're talking about... A15 vs Bulldozer cores are pretty damn close in terms of IPC. I'm sure theres a lot of un-core and cache stuff BD has over ARM in this case, but still, not that far off.

 

[RedVi]

A15 vs Bulldozer cores are pretty damn close in terms of IPC.

No, they are not. I know people really love to exaggerate Bulldozers IPC deficit, but this is taking it too far. The major difference is in the memory hierarchy, as you mentioned with un-core, but besides that, when taking into account the fixed hardware capabilities of Bulldozer that are currently unused (eg FMA4), the difference is staggering.

I take the chart as most likely pointing to a dual core A15/or A9/15 hybrid running the system/OS at all times with x86 cores dedicated to the real performance critical processing.

 

[Ninjaprime]

No, they are not. I know people really love to exaggerate Bulldozers IPC deficit, but this is taking it too far. The major difference is in the memory hierarchy, as you mentioned with un-core, but besides that, when taking into account the fixed hardware capabilities of Bulldozer that are currently unused (eg FMA4), the difference is staggering.

I take the chart as most likely pointing to a dual core A15/or A9/15 hybrid running the system/OS at all times with x86 cores dedicated to the real performance critical processing.

Bulldozer is ~3.78 IPC and A15 is ~3.5 IPC. I didn't say they would perform the same in the real world, just that BD and A15 were fairly close in IPC. BD clocks twice as fast, has a much more robust memory system and better SIMD capabilities, but all those can be worked on and customized for a console. I note that they say the clock is at 2GHz, so at a defined clock speed the BDs advantage diminishes. If they were to add VMX-like units to the A15 and improve the memory/cache system, it could probably compete with the BD.