If there is a new article, PR, or interviews regarding Sony and the PS3 then post it here. Do not start new threads with just a link to an article and no other reason for discussion. Too many threads are created with PR, articles, hype, and interviews that end up into arguments that do not contribute to the board in a constructive or positive way.
Official Sony PR thread
- Thread starter Sonic
- Start date
koldfuzion1
Newcomer
Sony, Toshiba co-develop 45nm embedded DRAM
Well, I doubt this has jack squat to do with consoles, but it's an article re Sony and Toshiba's DRAM process today, and basically PR. So in case any of the techies here happen to be interested, for a lack of a better place to put it, here is the link:
http://www.eet.com/news/latest/showArticle.jhtml?articleID=164900232
Cheers,
Dave
Well, I doubt this has jack squat to do with consoles, but it's an article re Sony and Toshiba's DRAM process today, and basically PR. So in case any of the techies here happen to be interested, for a lack of a better place to put it, here is the link:
http://www.eet.com/news/latest/showArticle.jhtml?articleID=164900232
Cheers,
Dave
ManuVlad3.0
Regular
Could this technology (45nm eDRAM) be used in PS3?
koldfuzion1
Newcomer
Got me, I figured that's why I'd post it, in case someone here had a clue, I certainly don't. 
Presumably not, as a first-breakthrough is a long way off effecient production runs.
mckmas8808
Legend
Does this mean 65nm could be sooner than later?
BOOMEXPLODE
Regular
Have you guys noticed that this time around Sony didn't give any performance numbers when they announced the PS3. I guess that's because they got so much crap for the "66 million polygons per second" thing with PS2.
The only real performance claims they made (WRT videogames) were 1080p resolution. No polygons per second, no fill-rate, no list of effects, nothing that could come back and bite them later. Contrast that to the announced specs for X360: 500 million polygons/s, 16 gigasamples/s, 48 billion shader operations/s, etc...
The only real performance claims they made (WRT videogames) were 1080p resolution. No polygons per second, no fill-rate, no list of effects, nothing that could come back and bite them later. Contrast that to the announced specs for X360: 500 million polygons/s, 16 gigasamples/s, 48 billion shader operations/s, etc...
From Nikkei Electronics magazine 6/20 issue:
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105803/
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105807/
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105808/
These articles are introductions of full stories.
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105803/
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105807/
http://techon.nikkeibp.co.jp/article/HONSHI_LEAF/20050615/105808/
These articles are introductions of full stories.
BOOMEXPLODE said:
I think this was due to NVidia concerns. I don't think NVidia wanted to reveal anything that might give ATi an advance whiff of what they were up to on the PC side. Which makes you wonder about the things they did talk about - 128-bit HDR etc. - and whether that stuff might be specific to RSX.
So I'm kinda hoping both they and sony will be able to open up about RSX after the announcement of Nvidia's next-gen next week.
ATi, btw, did not have this problem with Xenos since it's on a different track to their imminent PC products. Which is why they may have been more willing to immediately talk about Xenos, without fear of handing "intelligence" to NVidia re. their upcoming PC products.
http://pc.watch.impress.co.jp/docs/2005/0617/kaigai191.htm
Ken Kutaragi interview by Hiroshige Goto part 4
machine translation:
- Weekly overseas news * of Hiroshishige Goto
SCEI Kutaragi president interview(4)
"Reduction in costs strategy of PLAYSTATION 3"
- GDDR3 is adopted from the easiness of development in the graphics.
【Q】 The kind of the memory is also different in CPU and GPU. Cell is XDR DRAM, and RSX is GDDR3. Why is this?
【 Kutaragi 】 Why was not YDRAM (code name of XDR DRAM) used for RSX?Either doesn't change too much in the cost either. In Cell, Rambus combined with the first engineer even if thoroughly going (Mounting on Cell in the XDR interface) and it did. However, there was calling doing this in the graphics another times …… very much, and RSX was made GDDR3. It is easy DDR2, GDDR (To DDR system memory) the graphics side, and to use GDDR3. Because most was the same performance, it was still judged it was good.
【Q】 It was heard that the memory was able to be accessed mutually with Cell and GPU that was CPU.
【 Kutaragi 】 either is also good at the ping-pong. For instance, 256MB of the Cell side can be similarly accessed though even only VRAM of 256MB today is huge when seeing from RSX.
【Q】 The number of DRAM chips has increased from 2 to 8 pieces at a dash.
It cannot be helped that the amount of the memory of 【 Kutaragi 】 total is large. However (The memory chip that can be integrated in the future, it brings together very much. It becomes it very if there are a lot of numbers though there are a problem etc. of the memory grain degree that Mr.. Goto points out, too. The solution is how much if there is a volume of 20 million a year (piece) though it is difficult when the volume is a little.
For instance, it is easy to adjust the memory chip of Rambus from 4 pieces to 2 pieces. It is possible to do without changing the memory band. Because it is with Rambus (The development of XDR DRAM) as soon as. This is developed staring at everything (XDR DRAM). Because we are primary users, YDRAM : though Toshiba, Elpeda, and Samsung, etc. manufacture. Intel (About the next generation main memory standard) says, adopts this, and the same thing.
In addition, there is a method. For instance (The logic chip and DRAM chip are supplied to the one package), SIP(System-in-Package)ing is thought. GPU of Xbox 360 is ..aspect.. ..
- If the format continues for a long term, reduction in costs becomes possible.
【Q】 If the opposite is said, the vender who cannot expect the volume is not good at such reduction in costs.
There is a problem whether it is 【 Kutaragi 】 and it is "Have it" until the cost is downed to begin with.
Only one game machine roughly remains at the end. And, the model that remains has it for several years. It has already had PS1 for ten years. I think that it has already had PS2 for ten years every five years. It continues at long time, and the format won and is a shrink (make minutely)(The case), and can make the stone (chip) slim like PS2. However, Sega that had not heard the story that the stone of Xbox had done the shrink was so. Because the time only of doing the shrink has not been left by them.
【Q】 When the chip is done in the shrink or the cost can be reduced by integrating it if it wins and the format continues for a long term. Will the chip of PS2 also still do the shrink hereafter?Is integration (integration) advanced further?
Naturally, 【 Kutaragi 】 shrink and integration are done. PS2 will surely become the platform of ten years in five years if today. The chip that is manufacturing by a chip manufactured in the process of the first 250nm and another company when thinking about the period of ten years and wants to do to Discon (manufacturing discontinuance) comes out. Everything is integrated when becoming it so. And, PS2 should go because the win with this generation's platform was decided this time.
For instance, a processor different from a different memory doesn't keep the same (single-chip) when usually thinking. However, we do. (The chip is offered. )It talks with each company, a net list (Net-list: circuit chart data) each other is matched, and it achieves it. It can be integrate for us (With the device vender) because it came together though the chip of the chip and AMD of Intel might be brought and it be impossible political combining. It is possible to combine it also with Cell and RSX of PS3 extremely speaking in the long view. Because both (With IBM and NVIDIA) were made from scratch together.
- Shift to 65nm process where it takes time unexpectedly difficultly
【Q】 Both the Cell processor and RSX are scheduled to be manufactured in 90nm process. However, chips such as PS2, PSP, and PSX are manufactured from the current state in all 90nm process. Does the
manufacturing capacity of Fab (manufacturing factory) suffice?When do you about move to 65nm?
【 Kutaragi 】90nm also is (SCEI) also OTSS (Oita Teessemicondacta: joint management Fab of Toshiba and Sony), has Fab2, and has IBM Fab. It is enough though it is not possible to say (The manufacturing capacity. However, I think that it has a hard time taking it to 65nm anywhere unexpectedly difficultly.
One is a problem of the defect density (defect density). The number of garbage that passes the filter (Even if the process is made minute) is the same. Therefore (The transistor), even if it seems to be able to take more chips (To the yield target because the influence of the defect grows), it becomes painful if it reduces it. The problem that the yield doesn't go up simply is larger though in 65nm, there is a problem of Leake (leakage current), too.
The memory is even so OK. Because the redundancy (tedium) is given. (To a defective memory cell. )It only has to bring the memory cell by substitution (tedium).
This time, the architecture that was able to be considered to be a redundancy about SPE was taken for Cell. (The number of SPE made Inabl. )It ..…….. ..(.. ..laughter.. became it even if going by) and 7 pieces because everyone was surely angry that SPE was killed though there was a problem to pieces how many whether made it. I think that the redundancy will terribly become a critical factor in the yield improvement in the future. However, it is crazy, if the chip is not voluntarily manufactured, it is not possible never to do.
Making to 65nm still still has the problem though the redundancy of SPE of Cell contributes to the yield very much. Therefore, it first manufactures it from 90nm now. As for RSX, the idea is the same.
【Q】 In Cell, it is expected that larger-scale Cell like a variety of derivation goods , for instance, 2PE composition version and Cell that reduces the composition oppositely will appear in the future. After it becomes 65nm of making derivation goods
I think that 【 Kutaragi 】 derivation goods come out variously. However, when the derivation goods are made, you should do in a process that is maturer than a state-of-the-art process. It is often said the water fall management in the world of the semiconductor. In a process that manufactures the up-to-date one in an up-to-date process, and is steady ..starting redemption.., you should make general ..comparatively... (In a steady process. )It is possible to do the very cost easily though the chip grows (Because the yield is high). Moreover, the invested capital can be used best.
Because all production lines of 250nm that has invested PlayStation 2 so far or 180nm can be made the best use of, it makes it to cheapness. The stone of PS2 can also return PS2 to an old process again when saying that PSP that uses same 90nm seems to sell terribly and to become insufficient production now though it uses 90nm. This (flexibility of the allocation of manufacturing) is also very important.
- The size of the mass expands the means.
The grain degree of the memory in the game console and the problem of the chip shrink had been written by this column before. If it is an application of a large amount of shipment like the PlayStation family, on the other hand, Kutaragi clarifies that there are various counter measures. Because in a word, general-purpose goods are combined and not made but there can be solution of custom
First of all, there is a possibility of thinking about 2 piece composition of x32 in the future about this though four x16 composition goods of the XDR DRAM memory are expected to be installed in PS3. In that case, the memory number can be decreased without changing the memory band. In this case, it seems that it is reading that it is possible to counterbalance it enough by the number of chips decreasing though Daio Barrhead (cost increase by an area increase of the semiconductor chip) joins in making it to x32 of course. Because PS3 is the maximum user in fact, the SCEI side will be able to lodge the XDR DRAM memory the demand on the specification side to some degree.
It is used SIP, and more radical measures that another one Kutaragi pointed out. For instance, the number of Dai (main body of the semiconductor) and chip can be greatly decreased if the number of DRAM chips is decreased at a stage advanced by making minutely, and it puts it in the RSX chip and the one package of the SIP type. Logic and the memory can be connected at high speed if it makes it to wiring in the package with SIP. Therefore, it ends without decreasing the memory band even if the memory chip is integrated. Moreover, there is a chance that the case can be reduced because wiring on the substrate decreases because it becomes an one package.
SCEI designs the system by thinking about the integration of chips from the beginning. Therefore, Emotion Engine and Graphics Synthesizer were able to be integrated into the single-chip in PS2. When the same thing of PS3 becomes possible, this suggests Kutaragi though this is because of having developed in both chips and SCEI and Toshiba. It is guessed that this shows that the SCEI side has the use right only of it of IP of RSX.
It was clear that one of the problems of Xbox1 of Microsoft was in such respect. Microsoft developed without so considering reduction in costs by making minutely and integration etc. perhaps in Xbox1. The possibility of this not to be able to consider by not only the problem in the technology but also the contract relation to the chip vender perhaps is high. It is sure not to drink easily because sales decrease that much if said that it makes minutely if it makes it to the chip vender side, the cost is downed, and the delivery price is lowered.
Moreover, there is a problem of the scale economy, too. In 10 million number hardware shipped during year, the development cost of the chip is diluted. Therefore, the cancer cancer can reduce the manufacturing cost by making minutely and integrating putting the development cost. It is "Win pattern. " when this is a game machineOn the other hand, if the number of shipments of game machines doesn't reach a critical mass (critical point), reduction in costs on such a semiconductor side is not done easily.
By the way, Microsoft is obviously making hardware by minute making the process in this Xbox 360 on the assumption of reduction in costs. Because of it puts in view including shipment only of it or bet there
The memory of PS3 is asked from the NVIDIA side why it composes two kinds of of XDR DRAM and GDDR3, and it wants to be going to match the RSX architecture one of these days and to introduce it.
More bullish posture it has changed into a steadier remark about 65nm process as for SCEI now before. However, this is not a story of SCEI alone. In the entire industry, as for the start-up of an up-to-date process, not going are becoming common recognition as smoothly as former. The semiconductor manufacturers other than Intel began to come to ally greatly because of the process development because there was a challenge big in each new process. SCEI/Sony also is cooperating with Toshiba and IBM in the process development.
Moreover, the yield improves greatly if the number of SPE made Inabl is decreased because SPE that is a homogeneous module in Cell as pointed out by Kutaragi, and there is in fact tedium. Of course, it is a tolerance though there must strictly be some impacts in the performance when it occasions to kill SPE that is the nearest the memory interface and furthest SPE is killed because the internal ring bus of Cell has latency.
SCEI starts adding to present manufacturing capacity and starting up the line of 65nm. Therefore, it is necessary to manufacture all the product lines from the line of 90nm till then though there is abundant manufacturing capacity when 65nm stands up. However, this company can Maigrat it to Fab of the other companies comparatively easily because it is sharing the base of the process with IBM and Toshiba because of the process development cooperation. Actually, Cell is manufactured from both IBM and SCEI.
Still, when it is possible to correspond by returning the chip manufacturing of the PS2 faction to an old process when capacity becomes insufficient, Kutaragi is pointed out. The degree of freedom of the allocation of such manufacturing is a strong point only of SCEI that manufactures the chip by itself.
Cell is a structure that CPU core is modulated. Therefore, various variations can be made from the chip for a powerful computer to a more low-cost built-in system chip from comparatively few development labors by changing the composition in the core etc.Therefore, the derivation chip of Cell will appear surely first of all.
In the feeling of the interview, it is expected that the derivation goods of Cell appear apparently in 90nm process generation. It is a chip whose one is more powerful and bigger than composition's twice now (Two 2PE=PPE is 16/SPE) that is thought with the derivation goods of Cell. The number of installed SPE is decreased to 4 pieces, and another one be might have taken the place of a built-in system core without VMX etc. , and a smaller chip as for PPE.
Because the former becomes a huge size of 400 square mm if it manufactures it in 90nm process, it is not easy to think. Perhaps, there seems to be a possibility that Cell of the small composition or more of the latter comes out. A present composition of Cell is not matched easily to the application from which low-cost and low power consumption are requested. The performance is also excessive according to the usage. Moreover, there seems to be a meaning that makes the derivation chip of the small composition because it seems to take time to lowering the cost in 65nm. In that case, the application that can apply the Cell architecture will extend further.
Ken Kutaragi interview by Hiroshige Goto part 4
machine translation:
- Weekly overseas news * of Hiroshishige Goto
SCEI Kutaragi president interview(4)
"Reduction in costs strategy of PLAYSTATION 3"
- GDDR3 is adopted from the easiness of development in the graphics.
【Q】 The kind of the memory is also different in CPU and GPU. Cell is XDR DRAM, and RSX is GDDR3. Why is this?
【 Kutaragi 】 Why was not YDRAM (code name of XDR DRAM) used for RSX?Either doesn't change too much in the cost either. In Cell, Rambus combined with the first engineer even if thoroughly going (Mounting on Cell in the XDR interface) and it did. However, there was calling doing this in the graphics another times …… very much, and RSX was made GDDR3. It is easy DDR2, GDDR (To DDR system memory) the graphics side, and to use GDDR3. Because most was the same performance, it was still judged it was good.
【Q】 It was heard that the memory was able to be accessed mutually with Cell and GPU that was CPU.
【 Kutaragi 】 either is also good at the ping-pong. For instance, 256MB of the Cell side can be similarly accessed though even only VRAM of 256MB today is huge when seeing from RSX.
【Q】 The number of DRAM chips has increased from 2 to 8 pieces at a dash.
It cannot be helped that the amount of the memory of 【 Kutaragi 】 total is large. However (The memory chip that can be integrated in the future
, it brings together very much. It becomes it very if there are a lot of numbers though there are a problem etc. of the memory grain degree that Mr.. Goto points out, too. The solution is how much if there is a volume of 20 million a year (piece) though it is difficult when the volume is a little. For instance, it is easy to adjust the memory chip of Rambus from 4 pieces to 2 pieces. It is possible to do without changing the memory band. Because it is with Rambus (The development of XDR DRAM) as soon as. This is developed staring at everything (XDR DRAM). Because we are primary users, YDRAM : though Toshiba, Elpeda, and Samsung, etc. manufacture. Intel (About the next generation main memory standard) says, adopts this, and the same thing.
In addition, there is a method. For instance (The logic chip and DRAM chip are supplied to the one package), SIP(System-in-Package)ing is thought. GPU of Xbox 360 is ..aspect.. ..
- If the format continues for a long term, reduction in costs becomes possible.
【Q】 If the opposite is said, the vender who cannot expect the volume is not good at such reduction in costs.
There is a problem whether it is 【 Kutaragi 】 and it is "Have it" until the cost is downed to begin with.
Only one game machine roughly remains at the end. And, the model that remains has it for several years. It has already had PS1 for ten years. I think that it has already had PS2 for ten years every five years. It continues at long time, and the format won and is a shrink (make minutely)(The case), and can make the stone (chip) slim like PS2. However, Sega that had not heard the story that the stone of Xbox had done the shrink was so. Because the time only of doing the shrink has not been left by them.
【Q】 When the chip is done in the shrink or the cost can be reduced by integrating it if it wins and the format continues for a long term. Will the chip of PS2 also still do the shrink hereafter?Is integration (integration) advanced further?
Naturally, 【 Kutaragi 】 shrink and integration are done. PS2 will surely become the platform of ten years in five years if today. The chip that is manufacturing by a chip manufactured in the process of the first 250nm and another company when thinking about the period of ten years and wants to do to Discon (manufacturing discontinuance) comes out. Everything is integrated when becoming it so. And, PS2 should go because the win with this generation's platform was decided this time.
For instance, a processor different from a different memory doesn't keep the same (single-chip) when usually thinking. However, we do. (The chip is offered. )It talks with each company, a net list (Net-list: circuit chart data) each other is matched, and it achieves it. It can be integrate for us (With the device vender) because it came together though the chip of the chip and AMD of Intel might be brought and it be impossible political combining. It is possible to combine it also with Cell and RSX of PS3 extremely speaking in the long view. Because both (With IBM and NVIDIA) were made from scratch together.
- Shift to 65nm process where it takes time unexpectedly difficultly
【Q】 Both the Cell processor and RSX are scheduled to be manufactured in 90nm process. However, chips such as PS2, PSP, and PSX are manufactured from the current state in all 90nm process. Does the
manufacturing capacity of Fab (manufacturing factory) suffice?When do you about move to 65nm?
【 Kutaragi 】90nm also is (SCEI) also OTSS (Oita Teessemicondacta: joint management Fab of Toshiba and Sony), has Fab2, and has IBM Fab. It is enough though it is not possible to say (The manufacturing capacity
. However, I think that it has a hard time taking it to 65nm anywhere unexpectedly difficultly. One is a problem of the defect density (defect density). The number of garbage that passes the filter (Even if the process is made minute) is the same. Therefore (The transistor), even if it seems to be able to take more chips (To the yield target because the influence of the defect grows), it becomes painful if it reduces it. The problem that the yield doesn't go up simply is larger though in 65nm, there is a problem of Leake (leakage current), too.
The memory is even so OK. Because the redundancy (tedium) is given. (To a defective memory cell. )It only has to bring the memory cell by substitution (tedium).
This time, the architecture that was able to be considered to be a redundancy about SPE was taken for Cell. (The number of SPE made Inabl. )It ..…….. ..(.. ..laughter.. became it even if going by) and 7 pieces because everyone was surely angry that SPE was killed though there was a problem to pieces how many whether made it. I think that the redundancy will terribly become a critical factor in the yield improvement in the future. However, it is crazy, if the chip is not voluntarily manufactured, it is not possible never to do.
Making to 65nm still still has the problem though the redundancy of SPE of Cell contributes to the yield very much. Therefore, it first manufactures it from 90nm now. As for RSX, the idea is the same.
【Q】 In Cell, it is expected that larger-scale Cell like a variety of derivation goods , for instance, 2PE composition version and Cell that reduces the composition oppositely will appear in the future. After it becomes 65nm of making derivation goods
I think that 【 Kutaragi 】 derivation goods come out variously. However, when the derivation goods are made, you should do in a process that is maturer than a state-of-the-art process. It is often said the water fall management in the world of the semiconductor. In a process that manufactures the up-to-date one in an up-to-date process, and is steady ..starting redemption.., you should make general ..comparatively... (In a steady process. )It is possible to do the very cost easily though the chip grows (Because the yield is high). Moreover, the invested capital can be used best.
Because all production lines of 250nm that has invested PlayStation 2 so far or 180nm can be made the best use of, it makes it to cheapness. The stone of PS2 can also return PS2 to an old process again when saying that PSP that uses same 90nm seems to sell terribly and to become insufficient production now though it uses 90nm. This (flexibility of the allocation of manufacturing) is also very important.
- The size of the mass expands the means.
The grain degree of the memory in the game console and the problem of the chip shrink had been written by this column before. If it is an application of a large amount of shipment like the PlayStation family, on the other hand, Kutaragi clarifies that there are various counter measures. Because in a word, general-purpose goods are combined and not made but there can be solution of custom
First of all, there is a possibility of thinking about 2 piece composition of x32 in the future about this though four x16 composition goods of the XDR DRAM memory are expected to be installed in PS3. In that case, the memory number can be decreased without changing the memory band. In this case, it seems that it is reading that it is possible to counterbalance it enough by the number of chips decreasing though Daio Barrhead (cost increase by an area increase of the semiconductor chip) joins in making it to x32 of course. Because PS3 is the maximum user in fact, the SCEI side will be able to lodge the XDR DRAM memory the demand on the specification side to some degree.
It is used SIP, and more radical measures that another one Kutaragi pointed out. For instance, the number of Dai (main body of the semiconductor) and chip can be greatly decreased if the number of DRAM chips is decreased at a stage advanced by making minutely, and it puts it in the RSX chip and the one package of the SIP type. Logic and the memory can be connected at high speed if it makes it to wiring in the package with SIP. Therefore, it ends without decreasing the memory band even if the memory chip is integrated. Moreover, there is a chance that the case can be reduced because wiring on the substrate decreases because it becomes an one package.
SCEI designs the system by thinking about the integration of chips from the beginning. Therefore, Emotion Engine and Graphics Synthesizer were able to be integrated into the single-chip in PS2. When the same thing of PS3 becomes possible, this suggests Kutaragi though this is because of having developed in both chips and SCEI and Toshiba. It is guessed that this shows that the SCEI side has the use right only of it of IP of RSX.
It was clear that one of the problems of Xbox1 of Microsoft was in such respect. Microsoft developed without so considering reduction in costs by making minutely and integration etc. perhaps in Xbox1. The possibility of this not to be able to consider by not only the problem in the technology but also the contract relation to the chip vender perhaps is high. It is sure not to drink easily because sales decrease that much if said that it makes minutely if it makes it to the chip vender side, the cost is downed, and the delivery price is lowered.
Moreover, there is a problem of the scale economy, too. In 10 million number hardware shipped during year, the development cost of the chip is diluted. Therefore, the cancer cancer can reduce the manufacturing cost by making minutely and integrating putting the development cost. It is "Win pattern. " when this is a game machineOn the other hand, if the number of shipments of game machines doesn't reach a critical mass (critical point), reduction in costs on such a semiconductor side is not done easily.
By the way, Microsoft is obviously making hardware by minute making the process in this Xbox 360 on the assumption of reduction in costs. Because of it puts in view including shipment only of it or bet there
The memory of PS3 is asked from the NVIDIA side why it composes two kinds of of XDR DRAM and GDDR3, and it wants to be going to match the RSX architecture one of these days and to introduce it.
More bullish posture it has changed into a steadier remark about 65nm process as for SCEI now before. However, this is not a story of SCEI alone. In the entire industry, as for the start-up of an up-to-date process, not going are becoming common recognition as smoothly as former. The semiconductor manufacturers other than Intel began to come to ally greatly because of the process development because there was a challenge big in each new process. SCEI/Sony also is cooperating with Toshiba and IBM in the process development.
Moreover, the yield improves greatly if the number of SPE made Inabl is decreased because SPE that is a homogeneous module in Cell as pointed out by Kutaragi, and there is in fact tedium. Of course, it is a tolerance though there must strictly be some impacts in the performance when it occasions to kill SPE that is the nearest the memory interface and furthest SPE is killed because the internal ring bus of Cell has latency.
SCEI starts adding to present manufacturing capacity and starting up the line of 65nm. Therefore, it is necessary to manufacture all the product lines from the line of 90nm till then though there is abundant manufacturing capacity when 65nm stands up. However, this company can Maigrat it to Fab of the other companies comparatively easily because it is sharing the base of the process with IBM and Toshiba because of the process development cooperation. Actually, Cell is manufactured from both IBM and SCEI.
Still, when it is possible to correspond by returning the chip manufacturing of the PS2 faction to an old process when capacity becomes insufficient, Kutaragi is pointed out. The degree of freedom of the allocation of such manufacturing is a strong point only of SCEI that manufactures the chip by itself.
Cell is a structure that CPU core is modulated. Therefore, various variations can be made from the chip for a powerful computer to a more low-cost built-in system chip from comparatively few development labors by changing the composition in the core etc.Therefore, the derivation chip of Cell will appear surely first of all.
In the feeling of the interview, it is expected that the derivation goods of Cell appear apparently in 90nm process generation. It is a chip whose one is more powerful and bigger than composition's twice now (Two 2PE=PPE is 16/SPE) that is thought with the derivation goods of Cell. The number of installed SPE is decreased to 4 pieces, and another one be might have taken the place of a built-in system core without VMX etc. , and a smaller chip as for PPE.
Because the former becomes a huge size of 400 square mm if it manufactures it in 90nm process, it is not easy to think. Perhaps, there seems to be a possibility that Cell of the small composition or more of the latter comes out. A present composition of Cell is not matched easily to the application from which low-cost and low power consumption are requested. The performance is also excessive according to the usage. Moreover, there seems to be a meaning that makes the derivation chip of the small composition because it seems to take time to lowering the cost in 65nm. In that case, the application that can apply the Cell architecture will extend further.
OUCH now that was painful to read.
From what i understand, XDR for the RSX wouldn't have helped a lot in terms of performance, and because DDR was easier to implement to begin with, they thought it wasn't worth the trouble to change the memory interface on the GPU to make it work with XDR. Makes sense...
From what i understand, XDR for the RSX wouldn't have helped a lot in terms of performance, and because DDR was easier to implement to begin with, they thought it wasn't worth the trouble to change the memory interface on the GPU to make it work with XDR. Makes sense...
STOP PRESS! Table-tennis playing hardware? Pong next-gen exclusive to PS3?Mikage said:
I have to wait for one or someone else to human-translate these articles. The giggles get me before the headaches ever do
This bit was the best:
Sounds like an old chinese proverb.
Sounds like an old chinese proverb.
And this:
Use cancer to lower manufacturing costs!!! That's just unethical, Ken!! BAD BAD MAN!
Use cancer to lower manufacturing costs!!! That's just unethical, Ken!! BAD BAD MAN! Shifty Geezer said:STOP PRESS! Table-tennis playing hardware? Pong next-gen exclusive to PS3?Mikage said:
I have to wait for one or someone else to human-translate these articles. The giggles get me before the headaches ever do
either is also good at the ping-pong made me laugh...lol...
read that quote with Cosbys voice.... >.>
I'm voicing my displeasure with this whole 'PR-thread' thing. It's at risk of turning what would otherwise be good discussion topics into something that the majority of the usership might just flat-out miss. Afterall how exciting is a thread with the word PR in the title?
That the entire Kutaragi interview thread was locked and essentially banished to here is a sign of what I'm talking about. (banished but never re-posted) That thread could have been quite interesting.
It's no wonder that the PR threads have had such low post activity that the Sony one had already dropped off the front page and the Microsoft one was about to.
That the entire Kutaragi interview thread was locked and essentially banished to here is a sign of what I'm talking about. (banished but never re-posted) That thread could have been quite interesting.
It's no wonder that the PR threads have had such low post activity that the Sony one had already dropped off the front page and the Microsoft one was about to.
JF_Aidan_Pryde
Regular
There's no way people can have a coherent arguement in one jumbo "PR thread". This is plainly not going to work.
Megadrive1988
Veteran
Similar threads
