Predict: The Next Generation Console Tech

[TheChefO]

I'm expecting Cell to be dead and non-existent next gen.

BC should keep Cell around for a while yet. Modified likely, but fully BC.

 

[bgassassin]

If there is backwards compatibility I'm guessing it would be through software emulation then. I see Sony going for easier development this gen and in essence acting like Cell never happened.

 

[Brad Grenz]

It seems to me (as a PS3-exclusive console owner) that the hardware next generation is likely to be less interesting than the software. There are only so many CPUs and GPUs to pull out of the hat now, but whether or not Sony can come up with an operating system / network environment rich enough to be competitive next generation is a much more open question to me.

With the 4.0 firmware, the 'About..' item in the XMB now has a lot more disclosure about what software components Sony is using. Apparently, they're using eCOS (http://en.wikipedia.org/wiki/ECos) for at least pieces of their operating system, along with a great number of other open source componentry.

eCos is apparently a non-multitasking operating system, which fits with what I've seen on the PS3. I wonder if they'll have anything meaningfully better next time around.

I think the Vita OS gives us a promising look at the direction Sony is moving.

 

[Crossbar]

Ok, I want to start a serious line of discussion. Based on the AMD "future GPU" hints let's discuss:

(1) Stacked Memory. When will it be available in volume? What are the specs in regards to Bandwidth, Power, and Cost. What constraints are we looking at in terms of # of modules and bus? What are the providers? What are the pro/con and what are the competing options worth considering?
.

I read a really good article on this subject just some month ago, but I can´t find it right now. But here is some stuff on the subject, basically there are two directions:

1. Low power, small foot print for mobile devices (WideIO)
This is basically already available.

http://www.ecnmag.com/News/Feeds/20...ung-wide-io-memory-for-mobile-products-a-dee/
http://www.invensas.com/expertise/Documents/IWLPC.pdf
http://www.edn.com/article/518874-T...demand_drives_new_trends_in_mobile_memory.php

JEDEC expects WideIO to provide performance, energy efficiency, and compactness for smartphones, tablets, handheld gaming consoles, and other high-performance mobile devices. WideIO mobile DRAM uses chip-level 3-D stacking with TSV (through-silicon-via) interconnects and memory chips on SOCs (systems on chips). WideIO will suit use in systems that require memory bandwidth as high as 12.8 Gbytes, including 3-D gaming, 1080p high-definition video, and similar applications.

2. High performance, The Hybrid Memory Cube (HMC)

http://www.i-micronews.com/news/Mic...ed-memory-collaboration-closer-look,7766.html
http://www.eetimes.com/electronics-news/4229220/Micron--Samsung-seek-hybrid-memory-spec-
http://www.eetimes.com/electronics-news/4231051/IBM--Micron-to-build-hybrid-memory-with-TSVs
http://www.electronics-eetimes.com/...-with-3d-cmos.html?cmp_id=7&news_id=222910313

HMC technology uses advanced TSVs—vertical conduits that electrically connect a stack of individual chips—to combine high-performance logic with Micron’s state-of-the-art DRAM. HMC delivers bandwidth and efficiencies a leap beyond current device capabilities. HMC prototypes, for example, clock in with bandwidth of 128 gigabytes per second (GB/s).

Micron Technology has announced the production of a new memory device built using the first commercial CMOS manufacturing technology to employ through-silicon vias (TSVs) on IBM's advanced TSV chip-making process.

 

[TheChefO]

It is claimed that the technology provides 15X the performance of a DDR3 module, uses 70% less energy per bit than DDR3 and uses 90% less space than todays RDIMMs. Current DRAM burns a huge amount of the power in laptops and phones. HMC draws less power because of the wider I/O capabilities and greater I/O bandwidth significantly cut the amount of energy needed per bit - ~ 10% of the energy per bit of a DDR3 memory module.

Yes please.

Only problem I see with these is there is no clear roadmap for production.

Spec won't get finalized until next year sometime (Haven't seen a more refined date than 2012) and only after a spec is nailed down will we see the product commercially.

Do you think either MS or Sony will be the first to use this tech nextgen?

The bandwidth is great, but I'm thinking there are other methods to reach suitable bandwidth without resorting to prototype technology that hasn't been commercially viable prior to launching their nextgen consoles.

OF course if I'm completely wrong and 2014 is the intended date for nextgen, then this memory technology may have time to come to fruition and be a viable alternative to using GDDR5 (or a large stack of DDR3 and EDRAM).

http://www.i-micronews.com/news/Mic...ed-memory-collaboration-closer-look,7766.html

 

[hoho]

Biggest problem with stacked memory is that you can't really stack 8+ of them ontop of each other to get anywhere near the amount of total RAM current machines have.

 

[TheChefO]

Biggest problem with stacked memory is that you can't really stack 8+ of them ontop of each other to get anywhere near the amount of total RAM current machines have.

Well this is more a problem of cost than anything which leads back to the question of commercial viability prior to use in a console.

If they are 10x as "dense" as a DRAM module, then it would take 1/10 the number of modules to make up the same total memory. But if it costs 10x as much per module, then nothing is gained/saved .... aside from PCB traces ...

XDR2 should be able to help WRT board complexity, but royalties would play a role there.

Smart thing Rambus did with XDR2 is the memory controllers can use DDR as well as XDR chips. It should help XDR2 get a foothold which should bring costs down.

 

[pipo]

This talk about two different xbox models got my imagination running...

How about a cheap model with kinect that streams onlive style; xbox live platinum. Works for a lot of casual stuff.

A high end extender / local server that can be placed where you like for high-end / core gamers.

I'm not sure about the lag for kinect games with streaming, but things are moving fast and virtualization has a lot of future. Basically sell a HD streamer now and you're done for quite some time...

 

[hoho]

aside from PCB traces.

You do realize that the whole point of stacked RAM is to not have them connected to the CPU/GPU over a long distance but to have them literally sitting ontop of them instead? If you are going to connect them over PCB with traces you are better off using regular RAM modules instead.

 

[TheChefO]

You do realize that the whole point of stacked RAM is to not have them connected to the CPU/GPU over a long distance but to have them literally sitting ontop of them instead? If you are going to connect them over PCB with traces you are better off using regular RAM modules instead.

...

Yes, that is the intent.

However.

As you said.

The amount of ram will be limited to the die size if used in that way.

Hence my statement.

 

[TheChefO]

This talk about two different xbox models got my imagination running...

How about a cheap model with kinect that streams onlive style; xbox live platinum. Works for a lot of casual stuff.

A high end extender / local server that can be placed where you like for high-end / core gamers.

I'm not sure about the lag for kinect games with streaming, but things are moving fast and virtualization has a lot of future. Basically sell a HD streamer now and you're done for quite some time...

Here's the interesting thing pipo:

xb360 could serve this purpose of a kinect/casual/streambox ... today.

The problem is, the network still isn't ready for it (bandwidth/latency).

Years from now it may be, and by this time, the casuals would likely start yearning for a greater experience than xb360+kinect can give them.

At this time, xb720 should be near break-even for cost.

So then MS has two choices: pitch an onlive type service for any xb360 or xb720 console, or pitch xb720 as the upgrade path if the onlive solution isn't viable.

I'd be willing to bet that MS has plans to introduce their own streaming game service through live as it is a good business model. Just need the tech to catch up.



Speaking of the xb360+kinect_forcasualfamiliesconsole, considering that MS was bundling the kinect and arcade unit pretty much everywhere for $200 (or less) this Black Friday, I think it's safe to assume it costs them less than this for both a kinect and xb360.

I don't think it will be long (next year) before this bundle is in the $150 retail region. After this, there isn't much more to be gained from a price demographic perspective. Essentially they'd have another year of strong sales (2013) when they hit the next price bracket ($99) and at that point they've hit saturation.

If they have nothing new on the shelf by then (2013), they have nowhere to go but down.

 

[Madcutter]

I'm new to this forum, I enjoy this thread the most on the whole message board and have been following it for a while now. Quick question; what would come after 28nm and in which year would it arrive?

 

[TheChefO]

I'm new to this forum, I enjoy this thread the most on the whole message board and have been following it for a while now. Quick question; what would come after 28nm and in which year would it arrive?

TSMC states production of 20nm will be rolling out q4/2012.

http://www.electronicsweekly.com/blogs/david-manners-semiconductor-blog/2011/11/arm-tsmc-moving-fast-to-20nm.html

 

[homerdog]

22nm would be the obvious step after 28nm, but TSMC has been skipping to the half nodes lately, so they may skip 22nm and go for its optical shrink instead. I'm not sure if there is any official information on this yet.

Oh there you go. Thanks Chef I wasn't sure about that.

 

[Heinrich4]

Interesting is TSMC confident with yields rate 28nm process considering better than 40nm* at the same level of development,maybe superior to 20/30%.

Perhaps in the middle or end of 2012 yields rate on 28nm rise to levels consistently higher than 40% to allow for the chips for next gen console even end this year(Wii U and "X720"?).

* http://www.behardware.com/news/10521/tsmc-40-yield-for-the-40nm-process.html

 

[jonabbey]

If there is backwards compatibility I'm guessing it would be through software emulation then. I see Sony going for easier development this gen and in essence acting like Cell never happened.

The consensus on B3D seems to be that BC is unlikely to be accomplished without some kind of Cell chip in PS4. If IBM was able to pull a clock-rate doubling or tripling out of its pocket for the PS4 CPU, maybe they could get away without having actual SPU element hardware, but otherwise, it's hard to see how it could be done.

Maybe IBM will be able to come out with a 8 core Power architecture chip with processor cores that can act as SPUs for backwards compatibility and as more PPE-like elements for general purpose computing, but that seems as big a development task as the original Cell was.

 

[hoho]

Interesting is TSMC confident with yields rate 28nm process considering better than 40nm* at the same level of development,maybe superior to 20/30%

For how long have they been shipping 28nm stuff? If it's just a couple of months (if that) it's not really all that important.

 

[TheChefO]

Perhaps in the middle or end of 2012 yields rate on 28nm rise to levels consistently higher than 40% to allow for the chips for next gen console even end this year(Wii U and "X720"?).

Indeed.

It will be interesting to see how the new GCN AMD (and Nvidia) GPUs look in q1/2012.

If we see a quick breakdown of the product lineup (salvaged GCN parts), limited availability or missing clock targets along with where the new chips sit on the TDP, transistor, and die-size scales.

They will tell a lot about how 28nm is coming along.

 

[bgassassin]

The consensus on B3D seems to be that BC is unlikely to be accomplished without some kind of Cell chip in PS4. If IBM was able to pull a clock-rate doubling or tripling out of its pocket for the PS4 CPU, maybe they could get away without having actual SPU element hardware, but otherwise, it's hard to see how it could be done.

Maybe IBM will be able to come out with a 8 core Power architecture chip with processor cores that can act as SPUs for backwards compatibility and as more PPE-like elements for general purpose computing, but that seems as big a development task as the original Cell was.

Thanks for the info. Well I guess there won't be any emulation then because I fully expect Cell to go away completely. That includes PPEs for Xbox3. Keep your PS360 ready people.

I'm predicting all three consoles have a CPU that is OoO.

 

[AlphaWolf]

Thanks for the info. Well I guess there won't be any emulation then because I fully expect Cell to go away completely. That includes PPEs for Xbox3. Keep your PS360 ready people.

I'm predicting all three consoles have a CPU that is OoO.

I also do not see Cell making a return.

The 360 managed with very limited bc (how many sales would none even have cost it?), no reason Sony couldn't manage similar.