You have to remember that this is Sony's last generation in the game if they can't succeed profitably this coming gen. They will be throwing everything they have at the market this time, including a sharper vision for their platform future from the beginning. They were caught with their pants down in 2006 and the comparisons to XBL, so I'm expecting them to come out all guns blazing. Which is obviously what both they and MS plan to do. Can't wait for the bloody next gen to start.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
PS4 Rumors , APU code named 'Liverpool' Radeon HD 7970 GPU Steamroller CPU 16GB Flash
- Thread starter onQ123
- Start date
- Status
You have to remember that this is Sony's last generation in the game if they can't succeed profitably this coming gen. They will be throwing everything they have at the market this time, including a sharper vision for their platform future from the beginning. They were caught with their pants down in 2006 and the comparisons to XBL, so I'm expecting them to come out all guns blazing. Which is obviously what both they and MS plan to do. Can't wait for the bloody next gen to start.
It just hit me; can gameconsoles be sold with a payment contract (ala smart phones)? For instance, You subscribe to a 2 year contract that, by the end of it, you'd have payed the price of the initial product (what w pay day on off the shelves). I would be VERY attracted to that payment idea.
I don't think Sony are capable of getting their shit together. Unless Hirai really makes some fundamental changes to Sony's corporate culture and direction quickly. It's far too soon to tell, but the Vita can be seen as his baby, so a good thing to judge him on.
And look at the Vita - a wonderful piece of hardware and a gamer's dream handheld, but Sony don't know how to sell or support it. It's floundering in all markets at the moment, with a lack of software on the horizon to see how this could change.
Sony is still very much broken along geographical boundaries. They can't even unify their stores, services and game compatibility. In each case, Japan has the best support, with SCEA and SECC in fierce competition for who is worst. Just look at something seemingly as simple as PS1 games on the Vita. How long it took, and only Japan has full compatibility.
Kazuma Kiryu
Member
official earnings reports.
So you think a next Gen console will launch as a SoC? I also don't see a point in selling the 720 as a cheap cable box when current Gen hardware would do the job a lot better. They can include the same features in the 720 (or ps4) without marketing it as a cheap cable box.
I'm also unsure of 20nm. Even if the process is available, 28nm will still be more mature and easily cheaper.
Signs point to them putting more resources into 1st party titles.
All 3 companies will moneyhat next Gen.
Care to cite specifics? MS's first party practices have not exactly been illustrious.
Proelite
Member
I don't think they have the time necessary to turn around their first party before the next generation starts. I also expect them to push Kinect 2 and casual games from the start this time around, so their first party offering won't match the 360's early years.
James Sawyer Ford
Gold Member
pharmboy044
Member
Microsoft is already doing this.
Kazuma Kiryu
Member
i'm pretty sure microsoft will keep on going down the road of kinect games and keep a handful of studios focused on their main IP's (halo, forza and fable). some of the studios they set up just recently are clearly for more kinect content.
*Cough* Alan Wake sequel
Like that will do any sort of gangbusters. Remedy spent an insidious amount of time crafting the first game to a lukewarm reception and even more anemic sales.
Proelite
Member
Mostly likely Alan Wake 2 will be multiplatform.
grimshawish
Banned
You don't pay the same price though; its a credit option. You end up paying more ultimately.
MS is doing this so they'll know a bit more about how it can work in the general market.
Good for those without the money up front, but remember phones only get it because they have a contract anyway. You'd probably need to chuck something in, such as Live. Otherwise its not something I think anyone in the market will largely react to; customers won't fall for the initial months price tag.
McHuj
Member
Sounds like a great opening for Sony then if Sony have their shit together.
I think Kinect 2 would have to offer significantly more than what the current iteration does for it to matter again. If it's basically the same thing that works just a little better, I don't see it lighting the world on fire.
MS are already doing this right now for the 360 with Live bundled in. I really wouldn't be surprised if it next-gen consoles, most notably any sort of 720 comes out with this sort of plan already in place.
Yes, well, memory cells are worlds different than designing complex processors. *ahem* They're the ideal poster children for touting a new node, which is why a lot of tests are done with them in the first place. Register file arrays are basically SRAM as well.
On a side note, Samsung had 30nm NAND back in 2009. 30nm DRAM trickled out late 2010/H1 2011.
-------
Given the precariousness of fab tech in recent years (especially outside of Intel), it's highly unlikely that Sony/MS would bet on a node that's nowhere near volume in reality. "Expectations" for fabs are nice to hear in PR, but that's what it is.
"Risk production" is also extremely early. Any design basically has to tape out within 6-9 months for volume manufacturing, meaning they're not doing prototype work just to even see if the new process works, let alone starting the design from scratch with newly available design tools.
If we look back to 28nm, TSMC began RP for HKMG in Q3 2010, and 2 years later, they're still ramping up.
Also, if the design is going to incorporate/embed some sort of DRAM, it's going to complicate production significantly. There are extra steps involved which means more time to pump out a chip, so more more lead time for volume. Even IBM has trouble putting out fully enabled Power7+ (32nm) chips - most of the initial batch seemingly have 3-4 cores disabled, and that includes their associated eDRAM.
Anyways.
There's a world of difference between handhelds and consoles in Sony's eyes, 3rd parties' eyes, and consumers'. There's way more investment in consoles from just about everybody. Support for Sony's new console will be there. If you're going to look to the Vita as a sign of Sony's culture, consider the architecture and their design by developer feedback. They were more cost conscious and up until Nintendo lopped off their original retail price, the Vita was considered a decent price among the industry. You'll also notice that they married as much as they could from their lessons with the PS3 into its firmware design too (adding features that were 'missing' from the PS3).
I'd expect the PS4 to follow the same model, and all signs (or rumors) point to that. Easier architecture for devs, price sensitivity, and OS upgrades. I think its folly to assume that lessons won't be learned. Perhaps its fair to say what are those specific lessons they're coming away with; but there will undoubtedly be some. Today their console is basically at feature parity with the 360; only inhibited by whatever decisions they inherited from the original PS3's architectural designs. They go into the next generation knowing what their consumers expect...at the very least all the things you expect in Xbox Live today will be the same for PSN. The landscape changes from this generation to the next, and those services will evolve (tuners, etc), so its likely we may be debating how those new services and features compare.
The driving force behind new consoles is obviously some feature you have over the competition. Nintendo's is the tablet. MS seems to be tuner/smartglass combo. Who knows what Sony's will be. Whatever it is...for any of these consoles...in those early days...I hope its more of a focus on making games better. TV is nice, Movies and music are great. Every electronic device just about gives you another portal into one of them. I'm not hurting for those particular forms of content, I'll get them somewhere...who's going to have the games.
Actually I don't have anything specific and thinking about it more, your earlier post is probably spot on.
Yeah you two are probably right. Especially if they are stupid enough to pack a Kinect with every sku.
Yes, well, memory cells are worlds different than designing complex processors. *ahem* They're the ideal poster children for touting a new node, which is why a lot of tests are done with them in the first place. Register file arrays are basically SRAM as well.
On a side note, Samsung had 30nm NAND back in 2009. 30nm DRAM trickled out late 2010/H1 2011.
-------
Given the precariousness of fab tech in recent years (especially outside of Intel), it's highly unlikely that Sony/MS would bet on a node that's nowhere near volume in reality. "Expectations" for fabs are nice to hear in PR, but that's what it is.
"Risk production" is also extremely early. Any design basically has to tape out within 6-9 months for volume manufacturing, meaning they're not doing prototype work just to even see if the new process works, let alone starting the design from scratch with newly available design tools.
If we look back to 28nm, TSMC began RP for HKMG in Q3 2010, and 2 years later, they're still ramping up.
Also, if the design is going to incorporate/embed some sort of DRAM, it's going to complicate production significantly. There are extra steps involved which means more time to pump out a chip, so more more lead time for volume. Even IBM has trouble putting out fully enabled Power7+ (32nm) chips - most of the initial batch seemingly have 3-4 cores disabled, and that includes their associated eDRAM.
Anyways.
Pretty much this. =D
Oh dear.... =\
You can't sell a family device that's the size of a computer tower or that sounds like the OG 360.
Yes, well, memory cells are worlds different than designing complex processors. *ahem* They're the ideal poster children for touting a new node, which is why a lot of tests are done with them in the first place. Register file arrays are basically SRAM as well.
On a side note, Samsung had 30nm NAND back in 2009. 30nm DRAM trickled out late 2010/H1 2011.
-------
Given the precariousness of fab tech in recent years (especially outside of Intel), it's highly unlikely that Sony/MS would bet on a node that's nowhere near volume in reality. "Expectations" for fabs are nice to hear in PR, but that's what it is.
"Risk production" is also extremely early. Any design basically has to tape out within 6-9 months for volume manufacturing, meaning they're not doing prototype work just to even see if the new process works, let alone starting the design from scratch with newly available design tools.
If we look back to 28nm, TSMC began RP for HKMG in Q3 2010, and 2 years later, they're still ramping up.
Also, if the design is going to incorporate/embed some sort of DRAM, it's going to complicate production significantly. There are extra steps involved which means more time to pump out a chip, so more more lead time for volume. Even IBM has trouble putting out fully enabled Power7+ (32nm) chips - most of the initial batch seemingly have 3-4 cores disabled, and that includes their associated eDRAM.
Anyways.
Thank you for this! lol
Kinect is very important to Microsoft's strategy.
-ImaginaryInsider
Member
*Compiled by somebody on the internet using official earning reports where the Xbox numbers are lumped in with the rest of the Entertainment division, and who probably thinks losses from 2001 still need be paid back "or the investors will get angry".
I think you are partly right and partly wrong. Some shareholders will entirely focus on short term trends, and care little for the overall picture, especially if they weren't a shareholder then. But institutional investors tend to take a much more in depth look ie. was this investment worth it?
It seems like you are contradicting yourself.
Didn't MS do quite well with the 360?
In media mode it'd probably use a very low power mode, possibly a different included SoC if the dual box rumours are still around. So a powerful 720 wouldn't have to be noisy while doing media duties
They know that initial buyers will be enthusiast gamers at the price they launch at. They have nothing to show that kinect is any good for those players. So why add cost to your console and include it?
They've shown they can sell a shit load of them as accessory later in the day. Just including it as a fancy front end or media functions seems crazy to me.
-ImaginaryInsider
Member
The losses are absorbed in that fiscal year, and don't carry over.
I think it can be argued that MS realized the massive investment the original Xbox was, not including the billions it took to setup Xbox Live, and planned on a very long term investment. The 360 followed MS expectations to a tee, and became profitable within the first two years or so.
Why this still comes up is beyond me, its like people rehashing Virtual Boy losses.
jeff_rigby
Banned
Current Consoles have serious design issues that we overlook because that's what we are used to. Think of non-gamer consumers who want instant loading APPs. In handhelds they are always loaded in memory, next generation consoles will have something similar. Next generation video & Skype will be a minimum of 720P which is not possible this generation. Next generation SoCs can be cheaper than even the latest 2010 console SoCs. Next generation will comply with EPA Goldstar power ratings, when viewing a TV program with XTV triggers in them, the consumer can instantly bring up a browser rather than waiting 3+ seconds.
Yup, "The process will deliver twice the density of 28nm and a 60% reduction in static and dynamic power", I expect that IF 20nm is going to be used there will be more "chips" per wafer but yields will be lower than 28nm causing chips to initially be more expensive but I expect that at each 2 month cycle the yields will go up and within months, 20nm will be cheaper.KageMaru said:
It's really hard to get my head around the design lead time AND the needs of a console design going forward for it's "10 year life". You have all sorts of planning designed in for the time it takes to get to the first refresh and what that may bring. You should plan to push technology this year that becomes cost effective at the next refresh.
Memory is already being produced @ 20nm in quantity and when yields increase the cost to make DRAM and Flash is going to drop allowing for more memory at the same costs.
AlStrong said:
Samsung Now Producing Industry's first, Highest Density Mobile LPDDR2 Memory, Using 20nm-class Technology.
SK Hynix Intros 20nm DDR3L-RS DRAM with Reduced Standby Power
20nm DRAM is being released now not two years later, this supports accelerated R&D for the handheld market impacting CPUs as well.
-------
All true but IBM can embed DRAM on SOIAlStrong said:Given the precariousness of fab tech in recent years (especially outside of Intel), it's highly unlikely that Sony/MS would bet on a node that's nowhere near volume in reality. "Expectations" for fabs are nice to hear in PR, but that's what it is.
"Risk production" is also extremely early. Any design basically has to tape out within 6-9 months for volume manufacturing, meaning they're not doing prototype work just to even see if the new process works, let alone starting the design from scratch with newly available design tools.
If we look back to 28nm, TSMC began RP for HKMG in Q3 2010, and 2 years later, they're still ramping up.
Also, if the design is going to incorporate/embed some sort of DRAM, it's going to complicate production significantly. There are extra steps involved which means more time to pump out a chip, so more more lead time for volume. Even IBM has trouble putting out fully enabled Power7+ (32nm) chips - most of the initial batch seemingly have 3-4 cores disabled, and that includes their associated eDRAM.
There is no doubt that 20-22nm would be pushing technology that will be available and with high yields within a year, within 2 years 14nm on 20nm plane will be available. Both TSMC and GloFlo appear to have a multi-year plan for 20nm, it's a long term goal necessary to their survival, it's not a stepping stone to a future node. Handhelds, Servers and embedded which all need low power highly efficient chips appear to have accelerated R&D on 20nm.
Your IBM Power7 @ 32nm and AMD desktop APUs @ 28nm are all high power - high speed and SOI. Are there the same issues with low power lower clock on 20nm? Jaguar CPUs are 1.6 Ghz with a turbo mode to 2.1 I think. AMD CPUs in their desktop APUs run at 3.2Ghz. More CPUs running in parallel at a lower clock speed is more efficient. This is a next generation design not a continuation of "Cell" @ 3.2 Ghz or Power7 which have their own issues that have to be overcome in a Fab design.
Arguments for and against 20nm in next generation uncovers information that makes us all more aware of what's coming each way. Some of AlStrong's arguments are dated some spot on. He does not address what I consider the key, TSMC can not make the core chips for next generation if @ 28nm and gate first which is what AMD is now using for 28nm. GloFlo and IBM would have to make the core chips and package them to be sent to TSMC to be assembled, that process would be more expensive than AMD or TSMC making them 100% in house.SteveP said:
Edit: ARM and Globalfoundries hammer out deal to promote 20nm mobile chips August 2012
AMD Discloses Peculiarities of Next-Generation Jaguar Micro-Architecture. No mention of 20nm
I think what I'm trying to say is that the overall Xbox profit is highly relevant since ultimately MS are judged on whether they spent their money wisely, or could have earned more with less costs investing somewhere else ie. tablets.
It's the type of yardstick management is judged by, which ultimately affects the future of the platform, when people are moved around on the basis of what is considered a success and what isn't.
This is the red herring. If Sony could afford it, they would have went with 8 GB of Ram no doubt. Initial high costs are going to really make people wrack their heads. Here is to hoping that memory costs improve substantially.
Didn't MS do quite well with the 360?
In media mode it'd probably use a very low power mode, possibly a different included SoC if the dual box rumours are still around. So a powerful 720 wouldn't have to be noisy while doing media duties
Size matters as much as sound.
An accessory is still an accessory. In order to reach mass adoption and garner both software support and family support it needs to be included with every console. Read the leaked MS diagram to see that they'd like to be all inclusive with this device, and that will set the priority just as much as gaming did for the OG Xbox.
The very nature of DRAM (capacitors) entails extra steps to the manufacturing process.
The point is that they're having problems with making fully functional chips using eDRAM on the latest node despite all the redundancy built into the chip.
Yes, low power applications for hand-held. That's a far cry from comparing it to mass market, higher performance desktop RAM or actual complex ICs.
Why would I address it in the first place? I have no comment on which particular foundries MS or Sony will use. It could be anyone, and I don't even know how you leapt to sending ICs to TSMC. Designs will have to be adapted if they're outsourcing to various fabs to see who can come up with the best yield (cost). As you should know, even AMD GPUs have been adapted to SOI processes. And who said it has to be 28nm gate-first?
My examples of TSMC and IBM (GloFo isn't in a significantly better position no matter what PR statements exist. There's enough said when AMD is only just getting their 28nm APUs ready for mid-2013 introduction) were to illustrate just how long it can take to ramp up for high end applications, and there is no evidence that moving to newer nodes is any easier than before despite your insistence on using vague PR statements.
There is a world of difference between the mobile market (let alone DRAM) and even the low end of desktop designs. Saying that R&D is accelerated is nice and all, but that is giant leap of faith to think it means 20nm will show up in time for a late 2013, early 2014 release. It makes absolutely no sense for them to be able to have a 20nm design ready right now when the foundries have barely begun risk production this quarter.
Why are you addressing me in the 3rd person?
jeff_rigby
Banned
Overview of memory in processors L1 - L2 - L3 and why eDRAM is atractive
From Deep Trenches to Skyscrapers: - The evolution of embedded memory IBM on SOI is able to drop some of the extra steps needed to produce eDRAM on SOI wafers. Prior to 2011 (Xbox 360S for instance) had the eDRAM connected via interposer not part of the chip. eDRAM designed by IBM is not node agnostic nor process agnostic, last part of the PDF describes IBM's answer to this; 3D stacking of eDRAM under logic (Hybrid Memory Cube used as an example).
Thread on AMD's blog "AMD Fusion, Bobcat, Bulldozer" Of interest is the discussion on the processes and difference in low power and high performance designs. There is more going on than is advertised. IBM is all about high performance and everything is SOI while everyone else is supporting handheld/embedded/server and low power. AMD is getting out of direct competition with Intel on desktop high performance. Remember Sony's comment that they are betting on handheld, this comment could have been miss-translated and it could have been about hand held technology. Edit: IBM & GLoFlo are going bulk silicon at 28nm and 20nm.
RE: AlStrongs comments, he is correct, even on SOI there are extra steps in creating eDRAM on SOI. It is simpler on SOI but creating eDRAM on chip apparently is not going to be accepted by other forges. GloFlo and TSMC want to use bulk silicon wafers (1/4 the cost) with low power (Handheld) designs @ 20nm and gate last HKMG starting in 2013.
If the rumors are correct and both next generation consoles are using Jaguar CPUs and at least one of them has a ARM processor then it will be a low power design. 28nm or 20nm bulk silicon, who knows. 20nm bulk silicon would be incompatible with a PPC (Same for MPA Cell) which currently is designed for SOI.
Are we seeing the issues here? Both AMD and TSMC are stating that Within 6-9 months they will be mass-producing low power 20nm chips and "Stacking". This I assume means 20nm bulk gate last HKMG with the L3 eDRAM TSV 2.5D or 3D connected below logic. This is the advertised 3D stacking node planned since 2008 and should be what is used for next generation and what was planned for since 2010.
6-9 month difference in release schedules between 32-28nm and 20-22nm, first half of 2014 is the Seronx rumor (no die size in his statements) and that timing fits for 20nm and 3D stacking. The AMD Investor conference seems to imply game console win announcements within "quarters" and income from game consoles 3rd quarter 2013 (Start of mass production or contract that delays income till sales start?). Seronx appears to be very connected to the industry but may be outside game console "inside knowledge".
Note: MPA Cell is Multi-Processor-Architecture 1PPU3SPU
From Deep Trenches to Skyscrapers: - The evolution of embedded memory IBM on SOI is able to drop some of the extra steps needed to produce eDRAM on SOI wafers. Prior to 2011 (Xbox 360S for instance) had the eDRAM connected via interposer not part of the chip. eDRAM designed by IBM is not node agnostic nor process agnostic, last part of the PDF describes IBM's answer to this; 3D stacking of eDRAM under logic (Hybrid Memory Cube used as an example).
Thread on AMD's blog "AMD Fusion, Bobcat, Bulldozer" Of interest is the discussion on the processes and difference in low power and high performance designs. There is more going on than is advertised. IBM is all about high performance and everything is SOI while everyone else is supporting handheld/embedded/server and low power. AMD is getting out of direct competition with Intel on desktop high performance. Remember Sony's comment that they are betting on handheld, this comment could have been miss-translated and it could have been about hand held technology. Edit: IBM & GLoFlo are going bulk silicon at 28nm and 20nm.
RE: AlStrongs comments, he is correct, even on SOI there are extra steps in creating eDRAM on SOI. It is simpler on SOI but creating eDRAM on chip apparently is not going to be accepted by other forges. GloFlo and TSMC want to use bulk silicon wafers (1/4 the cost) with low power (Handheld) designs @ 20nm and gate last HKMG starting in 2013.
If the rumors are correct and both next generation consoles are using Jaguar CPUs and at least one of them has a ARM processor then it will be a low power design. 28nm or 20nm bulk silicon, who knows. 20nm bulk silicon would be incompatible with a PPC (Same for MPA Cell) which currently is designed for SOI.
Are we seeing the issues here? Both AMD and TSMC are stating that Within 6-9 months they will be mass-producing low power 20nm chips and "Stacking". This I assume means 20nm bulk gate last HKMG with the L3 eDRAM TSV 2.5D or 3D connected below logic. This is the advertised 3D stacking node planned since 2008 and should be what is used for next generation and what was planned for since 2010.
6-9 month difference in release schedules between 32-28nm and 20-22nm, first half of 2014 is the Seronx rumor (no die size in his statements) and that timing fits for 20nm and 3D stacking. The AMD Investor conference seems to imply game console win announcements within "quarters" and income from game consoles 3rd quarter 2013 (Start of mass production or contract that delays income till sales start?). Seronx appears to be very connected to the industry but may be outside game console "inside knowledge".
Note: MPA Cell is Multi-Processor-Architecture 1PPU3SPU
Yes, that's a nice thing. Again, when I brought up eDRAM complications, it was against the "no-eDRAM" case - ultimately, you're going through and combining two sets of processes (in series) to make such a chip, increasing the chance for defects. The extra time involved in producing the eDRAM simply needs to be accounted for.
In case there's confusion, when they say that they're not "impacting logic Performance, Reliability or Yields", they're talking about the eDRAM process not directly interfering with the logic production. Obviously, they don't say that overall chip yield can be bad or a waste of production time if either part of the fabrication has defects or problems meeting design targets (be it clocks/power envelope, voltage across the arrays etc).
jeff_rigby
Banned
Yes actual complex ICs use several different size transistors with differing doping for different jobs inside complex ICs. This makes for yield issues. Memory is using the same transistors over and over again, this makes for an easier higher yield design.
Edit: AMD's Process optimized building blocks implies multiple less complex and easier to build with higher yield parts but requires 2.5D or 3D stacking.
From one of my May 2012 posts:
AMD's 22nm Southbridge is going to be a separate chip as is L3 Cache (IBM eDRAM possibly) or SSD also rumored to be in next generation. GPU and CPU as different process optimized chips also? There can be a combination of both process and node sizes in this. AMD stating they started 5 years ago is 2008 when AMD and IBM started research on 22nm which was supposed to be ready by 2013.http://www.neogaf.com/forum/showthread.php?t=473780 said:
Game console volumes especially with both Xbox 720 and PS4 using the same forges and Microsoft planning to use the SoC in Cable boxes and Sony planning to use the SoC in medical imaging means many forges should be making the SoC, TSMC is one of the bigger forges and in the past a favorite of AMD.AlStrong said:
28nm gate-first only because the articles I have read have AMD using it NOW and IBM according to SimiAccurate (Oban) is making next generation @32nm on SOI (Dec 2011 10K wafers) along with GloFlo making the same chips three months later. There are multiple problems with the SimiAccurate rumor. Some part of it may be correct, they could have been making fine pitch interposers or MCM/TSV carriers at 32nm or various process optimized CPU packages or other parts of next generation like eDRAM packages.
I've pointed out the reasons and it does make sense to me that they would want to have it ready but everything you've said about what we know of the 20nm node is accurate. I don't know but it seems from everything I have read that the technology target was 2014 HSA with TSV 3D stacking on a 20nm node at volume by multiple forges using low power handheld technology.AlStrong said:
Edit: Game console volumes make it the perfect choice to jump start this technology.
The rumored timing is the issue if we are both correct, 3rd quarter 2013 as the start of mass production makes it possible. Early on it was assumed that 2014 was the start of next generation.....
jeff_rigby
Banned
AMD Ditches SOI: Kaveri goes Bulk at GF, More Details From the New WSA Oct 23, 2012 A few months ago there were statements that GloFlo and AMD would continue to use SOI for high performance APUs, this is aparently in error and my previous comments about 28nm SOI being used in error. It may be, considering the below quotes, that a previous statement as to when AMD goes gate last to be compatible with TSMC @ 20nm is also incorrect, it may also be 28nm. If this is the case, there is not then as much of a reason to push for 20nm. A customized Kabini @28nm could be the next generation game console SoC. But then why the talk of TSVs and stacking?
AMD Fusion Tapeouts Unveiled: 28nm in 2Q 2011, 20nm in 2012 and 14nm in 2014 This was the plan in 4/2011. If they are 6 months behind it's 20nm, one year it's 28nm with grey area between the two especially if looking at prototype or production. If you combine the above and below, 22nm was supposed to be available H1 2014 and 22nm is when stacking starts. 28nm was supposed to be available quarter 1 of 2013.
It's confusing but before the Forge delays, if the target was a 2014 Holiday launch then it was planned to use 20nm; 28nm would have been ready for a holiday 2012 or first quarter 2013 launch. That's only considering the forge not AMD and their design's readiness but that should have been tied to production (forge).
IBM and AMD first to announce 22nm SRAM chips 8/2008
AMD Fusion Tapeouts Unveiled: 28nm in 2Q 2011, 20nm in 2012 and 14nm in 2014 This was the plan in 4/2011. If they are 6 months behind it's 20nm, one year it's 28nm with grey area between the two especially if looking at prototype or production. If you combine the above and below, 22nm was supposed to be available H1 2014 and 22nm is when stacking starts. 28nm was supposed to be available quarter 1 of 2013.
It's confusing but before the Forge delays, if the target was a 2014 Holiday launch then it was planned to use 20nm; 28nm would have been ready for a holiday 2012 or first quarter 2013 launch. That's only considering the forge not AMD and their design's readiness but that should have been tied to production (forge).
EDIT: So this 20nm wafer (IBM-GloFlo) had to be made before January 2012 (test). The 20nm " APUs coming in the 2013 timeframe" will not be Kabini which is at 28nm. So there will be a 28nm APU and 20nm APU out in 2013 but there is no mention from AMD of the 20nm APU name or specs, could it be the "Confidential simi-custom win" read game console?http://www.brightsideofnews.com/news/2011/4/27/amd-fusion-tapeouts-unveiled-28nm-in-2q-20112c-20nm-in-2012-and-14nm-in-2014.aspx said:
IBM and AMD first to announce 22nm SRAM chips 8/2008
systemfehler
Member
Too bad that laser and sillicone don't get along that well. Would solve a lot of problems because all that shrinking is so limited and expensive in the long way. Just prolonging the inevitable...
jeff_rigby
Banned
It appears that 28nm is delayed and 20nm is on the projected track. This creates an overlap that makes it hard to determine anything. My best guess is early 2014 and that a result of Nintendo releasing the WiiU Nov 2012. The target date I think was holiday 2014 and that has been pushed forward depending now on technology that was projected for that time frame being available earlier.
I'm not concerned with when but a target date determined in 2010 would let us determine 32-28nm or 22-20nm and with that 8000 or 9000 series GPU and the extent of 3D stacking in next generation consoles.
The choice to use a Jaguar CPU may point to Low Power Handheld technology and that determines the type of wafer that is used. IF there is any BC, CPUs supporting it will have to be made on SOI or at least a wafer that can support high performance/speed. This means extra expense in the manufacture and attaching it (2.5D or 3D) to the SoC. If there is to be BC for both it makes sense to share this (back to my theories 2 1PPU3SPU CPU packages).
Low Power Handheld technology limits performance/clock speed but on the other hand 20nm and stacking have efficiencies that allow for higher performance. Stacking is only possible if TDP is kept low or within tolerance thus the choice of 20nm to start 3D stacking. Now that we know that AMD will be using Jaguar CPUs and bulk silicon, not high performance SOI silicon (which has it's own efficiencies putting it close to 20nm), the logic supports 20nm LP; that points to 2014. (It can still be 28nm with limited 2.5D stacking.)
XTV - ATSC 2.0 is only going to candidate status 1st quarter 2013 which means the standards will be in place. Implementing them will take time, a Xbox 720 or PS4 designed to fully support AR and XTV would need to be in place sometime late 2013 or early 2014. The PS3 can support XTV with RVU or NASNE in the interim but with lower resolution camera, clumsy interface and with slow loading apps; a Xbox 361 can do the same. This I think was the plan for Sony, the Leaked Xbox720 powerpoint has Microsoft releasing the Xbox 720 holiday 2013 with a Xbox 361 in 2012.
jeff_rigby
Banned
Found this PDF on Driving HPC Performance Efficiency with Heterogeneous Computing posted by Seronx on the AMD blog.
It gives information on 2014 HSA hardware features for the 9000 GPU and APUs . Key is PCIe 3.0 interface (not in later feature lists).
2014 GPU full HSA features: page 10 & 11
GPU compute context switching
GPU graphics pre-emption
Coherent PCIe for discrete GPU >> This was not in later feature lists
Task Parallel Runtime Integration
• Significant and ongoing BW improvements throughout the memory system
• Continue incredible pace of GPU improvements in performance/watt
• Increase sophistication of the GPU so that it becomes a 1st-class citizen of the overall system architecture
• Single unified virtual address space
• Virtual memory support via IOMMU
• Participation in system-level coherency
• Support for context switching
Page 11 has a diagram with the following information. The APU will use a faster memory technology while the second GPU will use GDDR memory, second GPU can be a card or chip.
Memory bandwidth is already known as an issue and a new memory technology is planned for 2014 AMD APUs. 2013 APUs are using DDR3 with two 64 bit channels. This and the IOMMU that interfaces with memory will need to be the same going forward for game consoles. This points to a 2014 design; they can not use the Kabini (2 channel DDR3) and customize it for next generation game consoles, they can't use GDDR5 for the main SoC as it's going to be too hot, requires too much drive current and eventually too expensive so a new memory technology which is probably going to be ultra wide DDR4.
T-RAM signs Globalfoundries to develop high density 32nm and 22nm Thyristor-RAM
It appears that T-RAM can be embedded in bulk silicon, this may be the new memory technology used as cache (L2 and maybe L3) in AMD 2014 APUs. ultra wide DDR4 may be a part of this.
This chart below seems to imply that the GPU on the 2014 discrete card (Volcanic Islands) will be as much a part of the CPU in the APU as the GPU part of the APU is of the CPU is in Sea Islands (2013); multi-apu with hypertransport over pci-express "Coherent PCIe for discrete GPU" mentioned above? Is this being confused with the two APU rumors for the Xbox 720?
Is this being confused with the two APU rumors for the Xbox 720? This is the plan for AMD's 2014 full HSA design.
According to one rumor, Sony is waiting till 2014 for a new memory technology and another, the Xbox will have 2 APUs; they both come with 2014 designs. The PS4 could be a 1 APU design and the Xbox APU plus discrete 2014 GPU. That would have the PS4 design first and the Xbox later which matches another rumor.
Before fanboys attack this, I expect multiple SKUs and the QOS mentioned in the above chart also for 2014 takes care of this. There will probably be a bottom end SKU for both that takes the place of the PS3 and Xbox 360 price point with a top end SKU for both that has a second GPU.
I find it interesting that Sony has a Context switching patent and Microsoft has the QOS patent. Both are needed by both consoles.
There has to be more cooperation and gentleman's agreements between Sony and Microsoft this coming generation than ever before as they are sharing the same or much of the same hardware.
I find statements that one or the other is going to be more powerful or one will be missing a core feature to be less than believable. I find the rumor that only the Xbox 720 will have a ARM A5 processor for "Trustzone" less than believable considering Sony's interest in DRM.
It gives information on 2014 HSA hardware features for the 9000 GPU and APUs . Key is PCIe 3.0 interface (not in later feature lists).
2014 GPU full HSA features: page 10 & 11
GPU compute context switching
GPU graphics pre-emption
Coherent PCIe for discrete GPU >> This was not in later feature lists
Task Parallel Runtime Integration
• Significant and ongoing BW improvements throughout the memory system
• Continue incredible pace of GPU improvements in performance/watt
• Increase sophistication of the GPU so that it becomes a 1st-class citizen of the overall system architecture
• Single unified virtual address space
• Virtual memory support via IOMMU
• Participation in system-level coherency
• Support for context switching
Page 11 has a diagram with the following information. The APU will use a faster memory technology while the second GPU will use GDDR memory, second GPU can be a card or chip.
Memory bandwidth is already known as an issue and a new memory technology is planned for 2014 AMD APUs. 2013 APUs are using DDR3 with two 64 bit channels. This and the IOMMU that interfaces with memory will need to be the same going forward for game consoles. This points to a 2014 design; they can not use the Kabini (2 channel DDR3) and customize it for next generation game consoles, they can't use GDDR5 for the main SoC as it's going to be too hot, requires too much drive current and eventually too expensive so a new memory technology which is probably going to be ultra wide DDR4.
T-RAM signs Globalfoundries to develop high density 32nm and 22nm Thyristor-RAM
It appears that T-RAM can be embedded in bulk silicon, this may be the new memory technology used as cache (L2 and maybe L3) in AMD 2014 APUs. ultra wide DDR4 may be a part of this.
This chart below seems to imply that the GPU on the 2014 discrete card (Volcanic Islands) will be as much a part of the CPU in the APU as the GPU part of the APU is of the CPU is in Sea Islands (2013); multi-apu with hypertransport over pci-express "Coherent PCIe for discrete GPU" mentioned above? Is this being confused with the two APU rumors for the Xbox 720?
Is this being confused with the two APU rumors for the Xbox 720? This is the plan for AMD's 2014 full HSA design.
According to one rumor, Sony is waiting till 2014 for a new memory technology and another, the Xbox will have 2 APUs; they both come with 2014 designs. The PS4 could be a 1 APU design and the Xbox APU plus discrete 2014 GPU. That would have the PS4 design first and the Xbox later which matches another rumor.
Before fanboys attack this, I expect multiple SKUs and the QOS mentioned in the above chart also for 2014 takes care of this. There will probably be a bottom end SKU for both that takes the place of the PS3 and Xbox 360 price point with a top end SKU for both that has a second GPU.
I find it interesting that Sony has a Context switching patent and Microsoft has the QOS patent. Both are needed by both consoles.
There has to be more cooperation and gentleman's agreements between Sony and Microsoft this coming generation than ever before as they are sharing the same or much of the same hardware.
I find statements that one or the other is going to be more powerful or one will be missing a core feature to be less than believable. I find the rumor that only the Xbox 720 will have a ARM A5 processor for "Trustzone" less than believable considering Sony's interest in DRM.
just found a few Sony Patents that may or may not be part of the PS4
Copied from my blog ( not linking my blog just the links to the patents )
_______________________________________________________
is Sony getting on your Nerves with the PlayStation? well that's exactly what they are trying to do (Patent)
A newly published Sony Patent shows how Sony wants to study your nerve responses while using a electronic device so the information can be used to predict your actions when using the device.
CONTROL OF ELECTRONIC DEVICE USING NERVE ANALYSIS
& There is also a Sony Patent for HEADS-UP-DISPLAY SOFTWARE DEVELOPMENT TOOL
& Gaze-Assisted Computer Interface
If I had to guess I would say that all of this will have to do with a newer model of the Sony Heads Up Display that will work with the PS4 or maybe the Clouds or PlayStation Vita.
_____________________________________________________
Copied from my blog ( not linking my blog just the links to the patents )
_______________________________________________________
is Sony getting on your Nerves with the PlayStation? well that's exactly what they are trying to do (Patent)
A newly published Sony Patent shows how Sony wants to study your nerve responses while using a electronic device so the information can be used to predict your actions when using the device.
CONTROL OF ELECTRONIC DEVICE USING NERVE ANALYSIS
& There is also a Sony Patent for HEADS-UP-DISPLAY SOFTWARE DEVELOPMENT TOOL
& Gaze-Assisted Computer Interface
If I had to guess I would say that all of this will have to do with a newer model of the Sony Heads Up Display that will work with the PS4 or maybe the Clouds or PlayStation Vita.
_____________________________________________________
actually there is the potential for MS to consider that they did so well last gen, that this gen they'll cruise. Or take the eye off the ball and push too much for entertainment at the expense of gaming.
If previous generations have told us anything, its that they don't tell us anything.
The software tool bit is 'just' about a HUD on screen while doing graphics work, not HMD related.
The other stuff is more, interesting stuff from SCEA R&D though...they've been patenting a long line of new interface-related stuff over the last 18 months or so. Gaze tracking has popped up in 2 or 3 now. A lot of biometric-y stuff. Will be interesting to see how much actually gets commercialised though.
jeff_rigby
Banned
I believe Sony is trying to find synergies between the medical fields and Game console interfaces. They already have a patent that uses the nano particles from Crystal TV (cheap and reliable high output blue LED light is converted 40% more efficiently into red, green and blue in LCD screens using nano particles rather than colored filters in the LCD screens), the nano particles are combined with chemicals that cause them to bind with cell protein or sugar chains and then when light is shined thru the slide, structures show up better. Super stain compounds and Sony camera technology.
Pushing bright people to think outside the box to capture opportunities presented by cutting edge research in to CE equipment and game consoles. With the baby boomer generation getting older, the medical field is going to be big business.
In the 80's I tried to get my partners interested in the Atari Falcon (1Meg/sec sample rate DSP built in) to write Optical (eye), Audio (ear) and ultrasound medical test routines.
jeff_rigby
Banned
From a SemiAccurate Forum discussion:
From BY3D:
Another vote for next generation game consoles using 2014 AMD designs not a custom version of a Kabini (Jaguar cores as rumored).
From above an edit:
From BY3D:
Just about everyone is expecting a large L3 cache of eDRAM to solve rendering issues but that won't address video stream FPS as I've described them. I'm guessing that the 2013 Fusion APUs with 2 DDR3 channels won't be fast enough thus the New Memory Technology mentioned for 2014 full HSA APUs (page 11) and the reference to needing memory bandwidth in the Sony PDF and here from an author who attended the Sony lecture which used the PDF. With a new IOMMU will come eDRAM support and as you mentioned directly addressing eDRAM and L2 cache by both CPU and GPU with modes that determine by routine size where the routine or data being worked on should reside (L2 CPU or GPU and/or L3) for Zero copy.Gipsel said:
Another vote for next generation game consoles using 2014 AMD designs not a custom version of a Kabini (Jaguar cores as rumored).
From above an edit:
EDIT: So this 20nm wafer (IBM-GloFlo) had to be made before January 2012 (test). The 20nm " APUs coming in the 2013 timeframe" will not be Kabini which is at 28nm. So there will be a 28nm APU and 20nm APU out in 2013 but there is no mention from AMD of the 20nm APU name or specs, could it be the "Confidential simi-custom win" read game console?http://www.brightsideofnews.com/news/2011/4/27/amd-fusion-tapeouts-unveiled-28nm-in-2q-20112c-20nm-in-2012-and-14nm-in-2014.aspx said:
jeff_rigby
Banned
IBM, GloFlo and TSMC are going bulk silicon for 28nm and below but GloFlo is gate first @ 28nm and TSMC is Gate last. At 20nm both GloFlo and TSMC will be gate last.
1) Gate first is a fewer step more reliable manufacturing method but the gate can be destroyed by temperatures in the later steps to place the other transistor junctions down on the Silicon over the gate. IBM developed a technique to allow gate first but it appears to have problems below 28nm.
2) Going Bulk @ 28nm and 20nm The problem with the current partially-depleted SOI (PD-SOI) technology is that the pressure needed for SOI insulation to take place is decreasing yields due to pressure on the already strained silicon. When you pressure the strained silicon transistors, they tend to break.
If next generation consoles are 28nm they will only be manufactured by Common Platform (IBM and GloFlo most likely). If they are 20nm then TSMC can also manufacture next generation.
It appears that T-RAM can be embedded in bulk silicon, this may be the new memory technology used as cache (L2 and maybe L3) in AMD 2014 APUs. ultra wide DDR4 may be a part of this.
T-RAM signs Globalfoundries to develop high density 32nm and 22nm Thyristor-RAM
Any rumors that mention process or die size can now be validated with these known conditions.
1) Gate first is a fewer step more reliable manufacturing method but the gate can be destroyed by temperatures in the later steps to place the other transistor junctions down on the Silicon over the gate. IBM developed a technique to allow gate first but it appears to have problems below 28nm.
2) Going Bulk @ 28nm and 20nm The problem with the current partially-depleted SOI (PD-SOI) technology is that the pressure needed for SOI insulation to take place is decreasing yields due to pressure on the already strained silicon. When you pressure the strained silicon transistors, they tend to break.
If next generation consoles are 28nm they will only be manufactured by Common Platform (IBM and GloFlo most likely). If they are 20nm then TSMC can also manufacture next generation.
It appears that T-RAM can be embedded in bulk silicon, this may be the new memory technology used as cache (L2 and maybe L3) in AMD 2014 APUs. ultra wide DDR4 may be a part of this.
T-RAM signs Globalfoundries to develop high density 32nm and 22nm Thyristor-RAM
http://www.brightsideofnews.com/news/2012/3/14/ibm2c-globalfoundries2c-samsung-to-go-finfet2c-3d-transistor-with-14nm-process.aspx said:
Any rumors that mention process or die size can now be validated with these known conditions.
jeff_rigby
Banned
Good observation from a SemiAccurate forum:
I think this is what the poster above is referencing:
AMD to Set to Make Important Strategy Announcement Next Week. AMD to Further Detail Ambidextrous Strategy on Monday
http://semiaccurate.com/forums/showthread.php?t=6708 said:
I think this is what the poster above is referencing:
AMD to Set to Make Important Strategy Announcement Next Week. AMD to Further Detail Ambidextrous Strategy on Monday
Oct 8 they announce 20nm ready for 2013 and 14nm on a 20nm plane 2014 and on the 29th I suspect they will have details of some of the IP. Is the technology actually ready for 2013-2014 or is this an effort to get the stock prices up? 3D stacking and custom IP in APUs was supposed to be a 2014 design at 20nm.[10/24/2012 07:21 PM]
by Anton Shilov
Advanced Micro Devices has announced that its top executives will host a news conference to provide updates and more detail on the company’s ambidextrous strategy. The news conference will take place in San Francisco, California, on Monday, October 29.
During the press conference, Rory Read, chief executive officer, and Lisa Su, senior vice president and general manager of global business units, as well as an unknown “special guest” will discuss the company’s wide strategy going forward. Ambidextrous strategy involves creation of system-on-chip solutions featuring technologies that were developed in-house as well as those designed by other companies. The strategy is aimed to enable AMD to target markets that require intellectual property AMD simply does not own.
The company naturally provided no details about the forthcoming announcement. However, knowing what the ambidextrous strategy is about, it is logical to expect AMD to announce specific plans or maybe even deals. Earlier this year AMD already said it would integrate ARM Cortex-A5 processing core into its ultra-mobile SoCs to provide advanced security mechanisms developed by ARM. In order to better compete for smartphones and tablets, AMD needs communication, image processing, touch processing and a number of other IPs, which it will need to license from the outside.
Brad Grenz
Member
AMD's dominance of the next generation of console contract does put nVidia in a bad position, but considering how unethical and anti-competitive TWIWMTBP frequently was, they deserve a comeuppance.
jeff_rigby
Banned
onQ123 look to the post 5391 just above yours, I edited it with more information. Also did you see this on BY3D in 2010:
http://forum.beyond3d.com/showpost.php?p=1510165&postcount=3
S3D with Glass 2x Render Performance 2 eye views
Naked-eye S3D 10x Render Performance 5 2 eye views Interesting, rumors are 300FPS (from Sony CTO) 10 eye views (5 S3D) at 30 FPS each and the target of 10X GPU performance. I guess this is where the specs for next generation come from for Sony. They want to support glassless 3-D on 4K TVs with the PS4.
Shifty has very strong views on this subject, he closed the thread, moved the thread and then split the thread editing out the above in the thread moved back to the Console section. I think you are also noticing this in the tone of his posts on 4K after yours.
http://forum.beyond3d.com/showpost.php?p=1510165&postcount=3
Naughty Boy jeff_rigby said:
S3D with Glass 2x Render Performance 2 eye views
Naked-eye S3D 10x Render Performance 5 2 eye views Interesting, rumors are 300FPS (from Sony CTO) 10 eye views (5 S3D) at 30 FPS each and the target of 10X GPU performance. I guess this is where the specs for next generation come from for Sony. They want to support glassless 3-D on 4K TVs with the PS4.
Shifty has very strong views on this subject, he closed the thread, moved the thread and then split the thread editing out the above in the thread moved back to the Console section. I think you are also noticing this in the tone of his posts on 4K after yours.
- Status
Similar threads
- Poll
- Poll