gofreak said:
The presumption is it would be only in online games, only leveraging other PS3's also playing that game, and everyone's machine could use others, thus making it "fair".
Ok I'm going to get a few things off my chest. Somethings I'm just not seeing with Cell concerning graphics.
I'm reading things here and there about cell but this is one thing I can look to and roll my eyes a bit. Maybe it is the over confidence forum members have in the tech. Lets say that nice LCD CELL equipped flat panel hi def display has 2 cell processors in it. The set also includes some ultra high bandwidth connection that connects to the PS3 to aid in what ever. That is easier to take, hell at my college they built a node based "super computer" using a bunch of alienware pc's connected by high speed interconnects.
However, I certainly have my doubts about something like this being done over the net. I doubt your cell equipped set combined with your PS3 will help someone's lonely PS3 render more things on screen fast enough to provide two identical experiences. It is obvious certain aspects of the net should already bring doubts to mind but also how do we take the NVIDIA hardware into account? Let me be bold for a moment and say that at the rate that piece of tech (NVIDIAs tech) will churn through pixels will be leaps and bounds beyond over the rate of computed data being brought over the net powered by any number of devices over the internet concerning real time graphics.
Sure enough there are many things that benefit from that type computing. But for the vast amounts of visuals on the screen I don't see it. Perhaps someone can show me the light.
I can picture however some type of MMO network powered by a cell cluster and being aided by every PS3 connected to it in some way. Although rather limited, it would still benefit since the cluster can bank on the spec of every PS3. It knows what it can do with every PS3 at any given time and can work accordingly. (All PS3s will be similarly equipped.) Most of the things concerning the actual player could be computed on the PS3 and sent back to the cluster since the PS3 would technically be an extension of the cell cluster.
Lets say you are seeing some type of cut scene on your PS3 MMO game. The cluster knows whats being used and whats not. It already knows the hardware, software platform, the way the scheduler synchronizes/communicates between processes/threads, and thus the cell cluster can take some processes out of its own queue and queue them right up on your PS3. The process is computed and the results are sent right back to the cluster. Thus one less process for the cluster to keep in queue and the scheduler can give another process its chance in the queue while the other is being computed somewhere else. In the world of multi-processing processes have priorities, thus the high priority processes such as real time computations have the majority of Cells time. When exactly do the lower priority processes get their Cell time? Thus let someone else's PS3 handle it while the Cell cluster churns away at the high priority processes. Making the system more efficient. Of course we are not talking about physics or graphics computations, rather aiding in the scope of running a MMO game, real time tracking, managing weather in the MMO world etc...
Of course this is a ruff idea, but then again I don't expect everyone to be on the same page if I busted out with the tech talk. But I think this is enough for everyone to understand (hopefully) although vague if youre in the industry. (For example. how feedback would be handled between the PS3 and the Cell cluster. What if another high priority process begins computing but requires results of another process sent out to someone's PS3? How long do we want to wait until a process sent out to someone's PS3 is considered lost? How would we estimate time that will give a good balance? I mean, what if we don't wait long enough and cut off too many processes that were on their way back? What if we are waiting too long? The more processes you lose the more work you have to do to get that process computed. When should we decide to just have the process computed locally? The more "misses" the more redundant work that must be done to compute the process. What if this (the process that require the computed data from the process sent out to someone's PS3) is a critical process that must lock a Cell until is done computing which ultimately depends on that process sent out? Sure we can just reload the process and compute it locally but for how many processes do you want to do this? Many of these issues will bring any computing cluster to its knees. Even local thread and process synchronization is one hell of mess with many chances for the system to crash down, Im getting chills thinking of something based across the net.)
Anyways, that is what I think when I hear about distributed computing concerning cell over networks. Maybe someone can explain to me how the graphical side things would be handled. Graphics processing, as much as I love graphics I dont know enough to convince myself someones PS3 with their Cell equipped set will also make someone elses visuals on their PS3 over the net improve.
doncale said:
XBox2's triple core CPU is said to push ~84 Gflops. I cant remember if that was the exact amount but it was 80~90 Gflops. Xbox2 is likely to have one of these sub-100 Gflop triple core CPUs but the newest revealed patents indicate that there could be more CPUs. it is flexable. it's not out of the realm of possibility that MS could have 2 triple core CPUs but thats still under 200 Gflops (80~90 Gflops * 2)
a single Cell Processor Element (1 CPU core plus 8 APUs) is said to push 256 Gflops
the PS3 CPU could be made from 1 to 8 of these 256 Gflop Cell Processor Elements. lets say Sony goes for a 2-PE CPU. thats 512 Gflops. that would outperform even 2 triple core CPUs in Xbox2 (160~180 Gflops) and totally blow away a single triple core Xbox2 CPU (80-90 Gflops).
4.6~4.8 GHz vs 3.5 GHz does in no way indicate what PS3 will be like vs Xbox2, going by clockspeed numbers.
clockspeed matters, yes, but also it's about how much performance you have per clock.
I still not sure what role exactly is NVIDIA's tech playing in the PS3. Do you know? You are directly compairing Xenon's CPU's as they would be doing the exact same thing as to what PS3's Cell chips would be doing. We don't know that, for what Xenon's CPU's may lack, the ATI tech may make up. Of course, when we talk about power we mostly imply graphics. Sure there is AI and physics but we can fake those to different degrees and in some cases don't even have to come close to the real thing as we only need certain aspects of them. I have questions all over the place concerning both these systems. I don't see how many people are coming up with specifics. But we will know soon enough.
But in summary, most seemed to agree that with a refined manufacturing process, and a fan stuck in to it, this thing would be more "normal" in terms of its operating temp.
Freaking multi-thread balancing is a big duh next gen, I love when I read shit like this. It will be RTRM for RTA in the next PS3 release.
