How would you react when the PSX3 spec reads 240~480 GFLOPS?

[Paul]

No point being different and ending up being not so good

This is why noone is interested in Xbox2. Because there really are no huge mysteries about it's hardware at all. We *know* by looking at PC hardware time tables what it will use by seeing what's out at what time.

It's just not interesting.

Being "different", custom hardware, is also typically is unmatched for a while more then when compared to typical hardware(aka a new ATI GPU vs a Nvidia GPU).

Why all the talk about PS3? Because it's using new technology that has never been seen before, where on the other hand everyone has seen the technology Xbox2 is going to be using.(Not many surprises)

 

[Citan]

I dont see whats the point here. If its good, its good. No point being different and ending up being not so good

The point is that someone is trying to do something different from the ground up(more or less). We will never know how something different will compete with "normal hardware" if no one ever tries it. The last thing I would want to see in the console market is for it to stagnate into a watered down PC in a box market.

 

[Vince]

I just wanted to compliment MrWibbles on his post. The qualities that he talked of, along with a solid population of educated posters, are what distinguish Beyond3D from the typical forum.

In my opinion, lately the forums, on a whole, have become far too polarized and segregated in their market ideologies. There is a thread in th 3D Forum which is dedicated to the question of "non-traditional" architectures and it struck me how quickly it turned from a more philisophical debate into rhetoric between nVidia & ATI, and ignorant talk of how PS2 is "too traditional" . I'm still trying to find out how discussion of ATI's Anisotropic Filtering has relevence...

At the end of the day do you really think that the end users cares what each architecture can do theoretically on paper and on paper alone?

Think this sums it up quite well; The typical end user might not care, but the typical end user isn't posting here.

 

[notAFanB]

I dont see whats the point here. If its good, its good. No point being different and ending up being not so good

the point is the latter is not known unless you jump in with both feet.

 

[chaphack]

call me a conservative(right sp? ) but i rather go with a more secure route. I dont see whats so un-interesting about PC tech? At the end of the day, they just go about making the best possible 3D graphics per say. So its good, so be good.

 

[chaphack]

ANYWAY, is any one kind enough to provide me with a simple rundown on why Cell is more interesting than conventional PC hardware(which supposedly is about to enter the next phrase of evolution soon)?

Isnt Cell just a bunch of PowerPCs latched together? The distributed computing over the Internet? Dont be thinking that is going to happen any soon. Even over the homes might not be a norm for everyday users soon. REYES rendering? Is it really powerful enough to do it good and pretty? Sorry if i am not keeping tabs on Cell.

Thankie Thankssss.

 

[notAFanB]

ANYWAY, is any one kind enough to provide me with a simple rundown on why Cell is more interesting than conventional PC hardware(which supposedly is about to enter the next phrase of evolution soon)?

from what little I understand of it it's multicore approach is what is gathering interest.

The distributed computing over the Internet? Dont be thinking that is going to happen any soon.

this is happening, tho ur right not anything which will bring the consumer any tangible benefits thus far.

I dont see whats so un-interesting about PC tech? At the end of the day, they just go about making the best possible 3D graphics per say. So its good, so be good.

nothing is un-interesting about it but alot is already known (and proven) which makes a gamble like CELL worth (the ample) discussion/speculation on this board.

 

[akira888]

... and ignorant talk of how PS2 is "too traditional" ...

Argh... ok....
It's pretty clear from my post that what I was referring to as "too traditional" was the GS and only the GS. I believe my point is fairly obvious, seeing as how that chip was far more basic than the CLX2 despite the fact that DC was released 16 months prior to the PS2. I figured my (limited) point would also be obvious especially after my positive comment about VU0 and VU1.

I freely admit that it was a ham-handed attempt at irony that fell short of expectations. But, when discussing topics that have a high degree of subjectivity (especially abstract undefined qualities such as "traditionalness") don't launch barbs at those who happen to disagree with you on the subject at hand.

 

[akira888]

While I might only be expressing my "ignorance" I'll think I'll say what I was going to originally.

In a network latency intolerant environment such as real time gaming, I fail to see how distributed computing would be in any way useable. Think about your "ping" on your network at home. If you can connect with even a latency of around 32ms, that's pretty good on DSL. Yet that represents 2 whole frames to get your data across, and then you need 2 more to retrieve it. While improvements in the internet topology might reduce this, there is a lower bound dictated not only by the processing requirements for network transmission but also by Einstein's Constant! (speed of light).

Another issue is that over the past couple decades at least the amount of bandwidth available to consumers has been at least three orders of magnitude lower (or 1/1000) than the bandwidth of internal I/O inside the machine. So even if latency could be magically eliminated we would still have a tiny fraction of the data carrying capacity available on our network pipe that we would need to really multiprocess.

NB:The past distributed programs all had a quality in common - the amount of data required to execute them was proportionally tiny relative to the processor time they required. In short, they were small "applets" which took massive time to process. The spectral analysis (2048 point FFT) of SETI@Home is the paradigmatic example of this. Game code on the other hand, which relies on massive amounts of high throughput streaming data, certainly is not.

 

[Paul]

In a network latency intolerant environment such as real time gaming, I fail to see how distributed computing would be in any way useable. Think about your "ping" on your network at home. If you can connect with even a latency of around 32ms, that's pretty good on DSL. Yet that represents 2 whole frames to get your data across, and then you need 2 more to retrieve it. While improvements in the internet topology might reduce this, there is a lower bound dictated not only by the processing requirements for network transmission but also by Einstein's Constant! (speed of light).

Current cable/DSL connections don't have the required bandwidth for real time distributed computing over the net.

I don't want to go in depth with this unless someone asks so I'll get to the point.

For real-time distributed gaming over the net you need the under 10ms latency and 10/10mbs full duplex fiber connections. 100mbs would be better.

This is not here, but one day it will. For everyone.

 

[Squeak]

We have online actiongames, with several contestants at a time. Wouldn’t that imply that psychics and AI could be done over a broadband connection?

 

[ChryZ]

At the beginning the PS3 will pretty much standalone, normal broadband (<10mbps) online gaming as we know it. I guess the first crazy CELL stuff will happen at LAN parties, where people hook-up their PS3 with optical fiber cables (>10mbps, low latency), much later when broadband connections reach >10mbps and ISPs are providing CELL nodes, then I guess CELL network clouds will host online games like MMORPG, I doubt CELL WAN networks will ever provide any help at realtime rendering jobs for single CELL clients, but for MMORPG it would be perfect. Instead of burning millions or even billions at server-farms, let the consumers host their own games and in a few year there will be more and more online-gaming-consumers. Sony is smart, they know what could brake their neck. Just my speculation ...

 

[jvd]

At the beginning the PS3 will pretty much standalone, normal broadband (<10mbps) online gaming as we know it. I guess the first crazy CELL stuff will happen at LAN parties, where people hook-up their PS3 with optical fiber cables (>10mbps, low latency), much later when broadband connections reach >10mbps and ISPs are providing CELL nodes, then I guess CELL network clouds will host online games like MMORPG, I doubt CELL WAN networks will ever provide any help at realtime rendering jobs for single CELL clients, but for MMORPG it would be perfect. Instead of burning millions or even billions at server-farms, let the consumers host their own games and in a few year there will be more and more online-gaming-consumers. Sony is smart, they know what could brake their neck. Just my speculation ...

I'm sorry this has been done. I've already hosted a mmorpg on a athlon 1ghz on a cable connection and was able to have 68 people online at once. Added another 1.5ghz athlon and another cable line i was able to have over 200+ people at once. The game was ultima online. Its that bandwidth that was important not the cpu. So unless someone wants to pay for the same connections i had (which my guild payed for through banners on our clan site.) i doubt anyone would want to pay that for cell gaming. I was running a 1000 dollar bill.

Mabye a lan would be better. But the wiring and hubs would be very expensive for this. I also don't know if a 100mbit full duplex would be enough. I know we start to get problems with a 16 player neverwinternights lan party and we have full duplex 100mbit connections. We are going to upgrade the hubs for the next one . One of those gigalans.

 

[Megadrive1988]

Current cable/DSL connections don't have the required bandwidth for real time distributed computing over the net.

I'll bet that even VDSL does not have the required bandwidth. VDSL is sort of like 'HD-DSL', the king of DSL/Broadband in the U.S. which offers upto 54 Mbps, and only availble in a few communities.

VDSL is said to over a similar leap over current Cable/DSL that Cable/DSL offered over 56K dialup.

 

[ChryZ]

I'm sorry this has been done. I've already hosted a mmorpg on a athlon 1ghz on a cable connection and was able to have 68 people online at once. Added another 1.5ghz athlon and another cable line i was able to have over 200+ people at once. The game was ultima online. Its that bandwidth that was important not the cpu. So unless someone wants to pay for the same connections i had (which my guild payed for through banners on our clan site.) i doubt anyone would want to pay that for cell gaming. I was running a 1000 dollar bill.

My point was not "who has done it first", but you illustrated wonderfully the trouble of hosting, now scale it up to a potential massive userbase and CELL doing it all natively, there wouldn't be the need to host excessively many other people, if everyone host a few or as much as feasable ....

Again pure speculation of a possible CELL application.

 

[jvd]

I'm sorry this has been done. I've already hosted a mmorpg on a athlon 1ghz on a cable connection and was able to have 68 people online at once. Added another 1.5ghz athlon and another cable line i was able to have over 200+ people at once. The game was ultima online. Its that bandwidth that was important not the cpu. So unless someone wants to pay for the same connections i had (which my guild payed for through banners on our clan site.) i doubt anyone would want to pay that for cell gaming. I was running a 1000 dollar bill.

My point was not "who has done it first", but you illustrated wonderfully the trouble of hosting, now scale it up to a potential massive userbase and CELL doing it all natively, there wouldn't be the need to host excessively many other people, if everyone host a few or as much as feasable ....

Again pure speculation of a possible CELL application.

Sorry are you saying instead of the giant servers that they use now they will be replaced by giant servers of cell chips ? There really is no diffrence . If anything i think it will be slowed down. the current servers have tons of ram and pretty cheap for what they do. Also most software is writen on those servers. It will all go down to bandwidth and so why use cell chips. Not only that but the servers don't need to display graphics. So what really is the question is how well the cell chip stacks up in interger performance.

 

[Chris123234]

Current cable/DSL connections don't have the required bandwidth for real time distributed computing over the net.

I'll bet that even VDSL does not have the required bandwidth. VDSL is sort of like 'HD-DSL', the king of DSL/Broadband in the U.S. which offers upto 54 Mbps, and only availble in a few communities.

VDSL is said to over a similar leap over current Cable/DSL that Cable/DSL offered over 56K dialup.

VDSL AKAIK only offers top speed within ~1 mile of the station. After that the speed drops dramatically.

 

[ChryZ]

Sorry are you saying instead of the giant servers that they use now they will be replaced by giant servers of cell chips ? There really is no diffrence . If anything i think it will be slowed down. the current servers have tons of ram and pretty cheap for what they do. Also most software is writen on those servers. It will all go down to bandwidth and so why use cell chips. Not only that but the servers don't need to display graphics. So what really is the question is how well the cell chip stacks up in interger performance.

If you host a game (giant server, small server, server-farm), you'll pay the bandwidth and all equipment. That's central.

CELL is central or peripheral, peripheral would mean the game is hosted by all connected CELL nodes (end-user PS3, ISP nodes, Sony's CELL servers). Distributed computing = distributed/shared bandwidth/load = no focal point of costs. Sorry last post, I am off ...

 

[jvd]

Sorry are you saying instead of the giant servers that they use now they will be replaced by giant servers of cell chips ? There really is no diffrence . If anything i think it will be slowed down. the current servers have tons of ram and pretty cheap for what they do. Also most software is writen on those servers. It will all go down to bandwidth and so why use cell chips. Not only that but the servers don't need to display graphics. So what really is the question is how well the cell chip stacks up in interger performance.

If you host a game (giant server, small server, server-farm), you'll pay the bandwidth and all equipment. That's central.

CELL is central or peripheral, peripheral would mean the game is hosted by all connected CELL nodes (end-user PS3, ISP nodes, Sony's CELL servers). Distributed computing = distributed/shared bandwidth/load = no focal point of costs. Sorry last post, I am off ...

I don't see how this will work. Most of the nodes will be taxed by displaying grpahics or whatever thier jobs are. Then you'd have thousdands of systems shutting off and on with out notice. This will not work for a few more generations of internet. Not only that but there would need to be a ton of un used ps3s sitting online. Not many people would like this . Larger farms for mmorpgs are needed. That is not going to change.

 

[Paul]

VDSL AKAIK only offers top speed within ~1 mile of the station. After that the speed drops dramatically.

Now yes. However in time vendors will make better chipsets offering that 52mbs top speeds at farther distances.