First Cell demo (48 MPEG 2 Videos)

[PC-Engine]

[
Well if you have valid reasons why my theory is flawed, then please enlighten me assuming you completely understand the theory.

Yeah, you're so smart we can't hope to understand your theories, you know.
I recall you we're still laughing at those 'real time' madden shots
C'mon genious!

Regarding this discussion you avoided to analyze the bandwith it's needed to decode, to ouput and to downscale 48 dvd streams.
Even if a processor A can play 4 streams in 1/4 of a frame you can't just assume it can play 16 streams in one frame.
If you have 48 SDTV streams you output 48*640*480*4*25 bytes per second. then you have to read them and downscale them to 1920*1080.
So you need (48*640*480*2 + 1920*1080)*4*25 = 3 Gb/s
I don't count compressed streams bandwith cause it's small.
Can a moden P4 based system sustain such bandwith in a real world application?
I don't know but this is not what this discussion is about, right?
Moreover I don't believe such a demo could completely tax a CELL CPU running at 3/4 Ghz as I mentioned before.

I'm waiting for your answer...if you have one. If you were a competent programmer without bias, you'd be able to achieve what was proposed.

I don't need you to tell me if I'm competent or not.
Who do you think you are?

BTW what would be your answer if somebody does get ~ 40 DVD streams running on a dual core HT P4???

I'd answer that somebody has probably forgot to downscale all the streams.

Just remember that this theory is not even close to a 1TFLOPS CELL dream.

So what?

Well yeah Sherlock I already said that the 40 streams DOES NOT INCLUDE DOWNSAMPLING....

48MB/s...understand?

You don't get it do you? The point wasn't to prove a P4 is more powerful than your beloved CELL which it isn't.

Keep hanging on to the belief that CELL is orders of magnitude more powerful than multicore HT Pentiums.

BTW nobody said CELL was being utilized 100%, sh*t we don't even know the clockspeed, but of course you have to rush in and defend it at all costs because you're a competent unbiased programmer.

Where's your dream 1TFLOPS CELL??? PIPEDREAM!!!

Oh and yeah I'm still laughing at YOU regarding that Madden screenshot since you think a realtime cutscene using blur filters couldn't achieve the same look. Of course I'm not surprised with your response since it was for Xenon instead of PS3.

The problem with you PC-Engine is that you have a hyper-sensitivity to appearing to be wrong. I was even nice in my initial post when I referred to the comparison as "silly" even though one of the two programmers you rely on (who definately knows his stuff as well or better than 99% of us) called it a "stupid comparison".

Before that response, however, was Faf who was trying to let you know nicely that you were ignoring OUTPUT bandwidth, which I brought up as well and did NOT chide you on whatsoever.

And now we have nao who is saying the same.

Why you have to be so infantile, I do not know. To think, I used to defend you somewhat in PMs because I felt you could bring interesting thoughts to the table. Now it seems that nearly every topic that is pruned or locked has your hands all over it. Frankly, it is through the behavior of folks such as you, that will cause the Console forum to be shut down. And what a shame that will be.

Read above Sherlock. BTW ERP called it a stupid comparison because I was using a simple bloated PowerDVD with multiple windows example on purpose. It was not intended to be the most efficient understand? It's the same logic aaaaa00 was using to say "if it can do this with all of this overhead then it would surely be able to do more in special cases like the CELL demo" hence the worse case scenario comment. Of course being the great contributor that you are you had to focus all of youre energy on "worse case scenero" and splitting hairs and grasping for straws at the same time for WHAT?? Then you put all of your weight on ERPs comment for WHAT? To make it seem like you're right? It's like idiots that get off on arguing over semantics to shift the focus off of the more important issue to try and sidetrack a thread. That's what you are doing man, and it's old and does NOTHING, NADA, ZILCH. BTW why do you need to point fingers anyway? The thread was fine before YOU came in with one liners because you can't come up with anything worth a damn to discuss. Maybe you should spend more time contributing something with substance instead of following people around to nitpic at their posts.

 

[nAo]

Well yeah Sherlock I already said that the 40 streams DOES NOT INCLUDE DOWNSAMPLING....

Even without downsampling the systems has to support something like 1.4 GB/s worth of bandwith.

Keep hanging on to the belief that CELL is orders of magnitude more powerful than multicore HT Pentiums.

In this kind af applications (stream processing) I have no doubt CELL can do 10x better than a P4. It has the FP power and it has the bandwith.

BTW nobody said CELL was being utilized 100%, sh*t we don't even know the clockspeed, but of course you have to rush in and defend it at all costs becaus you're a competent unbiased programmer.

Frankly speaking I don't think you have the numbers to judge me.

Where's your dream 1TFLOPS CELL???

!?? What are you talking about?!?

Oh and yeah I'm still laughing at YOU regarding that Madden screenshot since you think a realtime cutscene using blur filters couldn't achieve the same look.

I said to you those screenshots couldn't be realtime and well, I was right.

Of course I'm not surprised with your response since it was for Xenon instead of PS3.

Maybe you should use the search button and you could discover that I don't dislike Xenon at all!

I wonder how many times PC-Engine has to nsult someone before moderators start to take care of him.

 

[PC-Engine]

Even without downsampling the systems has to support something like 1.4 GB/s worth of bandwith.

And??? So??? What kind of bandwidth does a modern PC have???

In this kind af applications (stream processing) I have no doubt CELL can do 10x better than a P4. It has the FP power and it has the bandwith.

There's no point in claiming one is 10X more powerful than the other in X application, but of course it didn't stop you from trying did it?

!?? What are you talking about?!?

Exactly that's why it's called a PIPEDREAM!

I said to you those screenshots couldn't be realtime and well, I was right.

It was a guess, and you guessed right so what? Doesn't mean your deduction logic is valid. At least I was smart enough to say it could easily be realtime with blur filters instead of saying it IS IN FACT REALTIME. How in the world would I know for a fact LMAO?

 

[Shifty Geezer]

Firstly, PC-Engine, as has been iterated, this demo wasn't a Cell performance demo but a demo of Toshiba's programming malarkins. It's not officially indicative of Cell's limits.

Secondly, all it would take to prove your theories right is for someone with a P4 to see how many streams they can open. From what's been written so far it seems to me that this doesn't scale linearly, so though you might be doing fine at 10 streams with room to spare, come 15 streams your processor's choked. If people can't agree on whether this theory stands it ought to be taken to the experimental phase, but certainly it's got to the point where thrashing arguments is going nowhere.

Finally - Smilies are intended to portray TRUE meaning to a statement. They're used primarily to convey the more subtle contexts of speech that don't carry over into straight text. If I write

"do that again and I'll rip your arms off!"

it looks pretty offensive, but adding a smiley

"do that again and I'll rip your arms off! "

shows it's meant in jest.

You appear to use smilies defensively. You level criticism and insults at people, then follow them with a couple of smilies as if to say 'no hard feelings'. If you punch a guy on the nose and then smile sweetly, does that make it okay?

I'm still waiting for you to contribute something worthy to this thread other than one liners. You're like a little boy whos trying to get attention while the adults are talking.

That's a derogatory remark, no bones about it. It's obvious what you meant and using a couple of smilies is a thin disguise. The fact you place smilies at the end of seemingly EVERY paragraph you type shows either that don't mean anything you say and your joking all the time (in which case your not really contributing to the 'adult' discussion) or that you don't understand how smilies should be applied.

I don't know which it is, but you ought to think about how you use these new types of punctuation, and also look out how you use them to 'cover up' some rather negative comments you make.

 

[Fafalada]

So you need (48*640*480*2 + 1920*1080)*4*25 = 3 Gb/s

And that's just for pointsampling downscale - add a simple bilinear, or something fancier like cubic and required bandwith goes way higher. Actually we COULD test exact speed requirements for just filtered resizing on a P4, and it'd probably be more meaningfull benchmark then anything else we can do since it could actually be measured clock precise and without overhead of multiple abstraction layers.

Well yeah Sherlock I already said that the 40 streams DOES NOT INCLUDE DOWNSAMPLING....

So your point is that there is no point? You looked at the Cell demo, completely changed what it does, and concluded P4 could run the new demo so and so. Fine, but except for some suggestive smiling, you never explained what you were getting at exactly.
It might help if instead of criticising people for not understanding your point, you would stop dodging and "simply" state what your argument actually is.

 

[PC-Engine]

Uh so now the core issue is how to use emoticons? I need some of that wacky sh*t you've been smoking.

It might help if instead of criticising people for not understanding your point, you would stop dodging and "simply" state what your argument actually is.

Dodging? LMAO. Go back and REREAD the posts if you didn't get it the first time.


BTW here is what ERP said...

OK but all you really need to do to resolve the issue is to change your work pattern.

Say decode 1 second of each stream before swapping so that you get the value of the cache before it ends up being flushed.

Running multiple copies of an application on windows versus a custom written demo application is a stupid comparison anyway. The former is always going to have problems with cache management because the OS is optimised for user experience not running multiple copies of media player.

I wouldn't want to predict how fast a reasonably clocked P4 could do something similar in an similar environment, but it certainly wouldn't be the order of magnitude slower a lot of people seem to be implying.

We don't even have a good idea what the limiting factor was? It could just as easilly been read bandwidth as opposed to CPU cycles.

Here's what aaaaa00 said...

Using a large enough RAM buffer would let you do it, but I suspect that would be a lot of RAM for the 48 buffers.

Not so sure I'd agree. At DVD bitrate, we're talking 1 MB/s per stream.
48 streams = 48 MB/s.

So a GB of RAM is enough to buffer 21 seconds. (I don't think you'll need anywhere near that much buffer though.)

If you read the streams in 1MB chunks from 2 hard disks, you should be able to meet the I/O throughput requirement pretty easily, even with the seeks -- 9ms seek time means you can do about 100 seeks a second... so with 48 streams/1MB chunks you'd spend half your time seeking, and the other half of the time you can be reading data.

Not that this level of HW is practical to ship in a cheap consumer device, but for the purposes of the demo, it seems pretty straightforward to accomplish -- we haven't even gotten into anything semi-exotic (like FibreChannel or RAID) in terms of I/O hardware.

 

[nAo]

And??? So??? What kind of bandwidth does a modern PC have???

Modern PCs have theoretically more bandwith than that..

There's no point in claiming one is 10X more powerful than the other in X application, but of course it didn't stop you from trying did it?

There is a point if the requirements needed to be 10x more powerful are there. You can disprove me with number if you want, it would be nice to see you spitting out some simple math

It was a guess, and you guessed right so what? Doesn't mean your deduction logic is valid.

It means is more valid than yours, that's enough to me

At least I was smart enough to say it could easily be realtime with blur filters instead of saying it IS IN FACT REALTIME. How in the world would I know for a fact LMAO?

Smart guy I already told you why you can't achieve that kind of AA quality
with blur filters.
You can't remove aliasing and keep details at the same time with a blur filter. Those screenshosts had rich small (high frequency) details and superb AA.
Well I really haven't anything to add to this discussion with you, I know you will add some more insults and smileys, go on dude.

 

[PC-Engine]

Modern PCs have theoretically more bandwith than that..

Stop dodging and answer the question...if you dare.

There is a point if the requirements needed to be 10x more powerful are there. You can disprove me with number if you want, it would be nice to see you spitting out some simple math

You mean like the requirements of this demo?
It's kinda ironic that you're arguing over my theory which is based on simple math isn't it?

It means is more valid than yours, that's enough to me

Actually a homeless person living in a cardboard box could have easilly guessed right too by using his own flipping a coin logic. Doesn't mean his logic makes sense.

Smart guy I already told you why you can't achieve that kind of AA quality
with blur filters.
You can't remove aliasing and keep details at the same time with a blur filter. Those screenshosts had rich small (high frequency) details and superb AA.
Well I really haven't anything to add to this discussion with you, I know you will add some more insults and smileys, go on dude.

Those screenshots have rich small high frequency details? Riiiiiight. BTW you seem pretty quiet about ERP and aaaaa00's comments I quoted above, I wonder why...

I guess they're just better programmers than you.

 

[McFly]

Amount of smilies in a PC-Engine post is reciprocal equivalent to what the post is worth.

Fredi

 

[PC-Engine]

Amount of smilies in a PC-Engine post is reciprocal equivalent to what the post is worth.

Fredi

Yeah that's my style so?

How about you contribute something related to this CELL demo thread? Or is that the best you can do Mr. Back to the Future Worthy Poster?

 

[Blazkowicz]

Cable/Satellite TV? Instead of browsing, you have a page with "a lot" of channels all playing properly at once, so u can select what you want to see...
But there are other technology limitations that would make that quite hard anyway.

There's a similar function in my tv to doing that (updates only one channel at a time), but it's much faster to just use the digital tv box to scroll through what's on than watching 48 streams to pick out what I want. Plus, I don't know about splitting a <40" tv into 48 sub pictures. It's kind of... "useless" IMO.

there was some talk about this idea on the other topic, here's the link (I don't want to post the same thing over again ), it's already done here in my country, processed on the TV provider's side
http://www.beyond3d.com/forum/viewtopic.php?t=22250&postdays=0&postorder=asc&start=60

 

[ultimate_end]

blah blah blah..."theory"...blah blah blah

Come on PC, can't you just admit that this demo had nothing to do with Cell performance at all? Quit while you're ahead man, the demo was clearly intended to show Toshiba's software model for this type of application on Cell (and Cell definately needs it).

The one point that I can make out from all your comments was the mention of the SPE that was idle. You may be right that this should not be overlooked. It could indicate that there is not enough bandwidth (or too high a latency) on the Element Interconnect Bus to support this application achieving full SPE utilisation (which would be a concern). But it could easily be something else and Toshiba may simply not have even bothered (what consumer appliance needs to run 48 or 56 simultaneous DVD streams anyway?).

Amount of smilies in a PC-Engine post is reciprocal equivalent to what the post is worth.

Fredi

Yeah that's my style so?

Yes, the smilies are there to be used. But I think that it is how they are used that is at issue here.

 

[Ty]

Well yeah Sherlock I already said that the 40 streams DOES NOT INCLUDE DOWNSAMPLING....

So when I asked:

What is the point? That a P4 could do X amount of streams if we ignore output bandwidth?

Why did you reply with (my bold):

What are you talking about? I'm just removing the downscaling part in my example ie multiple streams at native DVD resolutions with audio being output to multiple windows.

What am I talking about? I was referring to you disregarding the output bandwidth requirement, which is what you were doing by ignoring the downsampling. Either I'm mistaken that the two are related or you were unaware of this relationship.

Furthermore it seems to me that everyone else is saying that downscaling is an important part, partly because of the bandwidth requirement.

Where's your dream 1TFLOPS CELL??? PIPEDREAM!!!

Your strawman argument is woefully transparent. In other words, you seem to be the only one claiming over and over again that Cell was supposed to hit 1TFlop.

I think if you take some time off from these boards and recognize that it's OK to be wrong, you'll grow as a person, in leaps and bounds. Seriously, no one is trying to attack you personally so there is hardly a need to belittle others so immaturely. This forum is a tremendous gathering place for others from all walks of life. . .and I'm afraid your tantrums only serve to dissaude open and honest participation amongst the members here.

 

[randycat99]

Firstly, PC-Engine, as has been iterated, this demo wasn't a Cell performance demo but a demo of Toshiba's programming malarkins. It's not officially indicative of Cell's limits.

Secondly, all it would take to prove your theories right is for someone with a P4 to see how many streams they can open. From what's been written so far it seems to me that this doesn't scale linearly, so though you might be doing fine at 10 streams with room to spare, come 15 streams your processor's choked. If people can't agree on whether this theory stands it ought to be taken to the experimental phase, but certainly it's got to the point where thrashing arguments is going nowhere.

Finally - Smilies are intended to portray TRUE meaning to a statement. They're used primarily to convey the more subtle contexts of speech that don't carry over into straight text. If I write

"do that again and I'll rip your arms off!"

it looks pretty offensive, but adding a smiley

"do that again and I'll rip your arms off! "

shows it's meant in jest.

You appear to use smilies defensively. You level criticism and insults at people, then follow them with a couple of smilies as if to say 'no hard feelings'. If you punch a guy on the nose and then smile sweetly, does that make it okay?

I'm still waiting for you to contribute something worthy to this thread other than one liners. You're like a little boy whos trying to get attention while the adults are talking.

That's a derogatory remark, no bones about it. It's obvious what you meant and using a couple of smilies is a thin disguise. The fact you place smilies at the end of seemingly EVERY paragraph you type shows either that don't mean anything you say and your joking all the time (in which case your not really contributing to the 'adult' discussion) or that you don't understand how smilies should be applied.

I don't know which it is, but you ought to think about how you use these new types of punctuation, and also look out how you use them to 'cover up' some rather negative comments you make.

EXCELLENT POST, RIGHT TO THE POINT, AND ABSOLUTELY ON THE MARK AS TO WHY PCE SHOULD BE GONE BY NOW!!!

 

[PC-Engine]

What am I talking about? I was referring to you disregarding the output bandwidth requirement, which is what you were doing by ignoring the downsampling. Either I'm mistaken that the two are related or you were unaware of this relationship.

Yes what are YOU talking about? How long does it take to go back and REREAD this thread Sherlock? Or are you in serious denial here??? Is there reading comprehension problem? It's amazing that you think I wouldn't know the fact downscaling and output bandwidth is directly connected. It's like showing Einstein how to do math. You should pat yourself on the back for such highly sophisticated reasoning.

Furthermore it seems to me that everyone else is saying that downscaling is an important part, partly because of the bandwidth requirement.

Look man. Did you not get that kid late to class analogy or are you just plain ignorant by default??? Or is it something related to tunnel vision? As usual you go crawling back to that same chronic flawed thought process of what everybody seems to be saying. Get back to me when you've reread this thread enough times to extract a single clue as to what I was saying. Until then maybe you should get to class on time before opening that chronically ignorant trap of yours.

Your strawman argument is woefully transparent. In other words, you seem to be the only one claiming over and over again that Cell was supposed to hit 1TFlop.

That wasn't an argument Mr. English major. Something being a pipedream is a statement not an argument.

I think if you take some time off from these boards and recognize that it's OK to be wrong, you'll grow as a person, in leaps and bounds. Seriously, no one is trying to attack you personally so there is hardly a need to belittle others so immaturely. This forum is a tremendous gathering place for others from all walks of life. . .and I'm afraid your tantrums only serve to dissaude open and honest participation amongst the members here.

You don't even have a single clue man.

I think you're the one with the strawman argument. Nobody claimed the sky is green dude, but that didn't stop you from trying did it?

Maybe you should take my advice and go back and reread this thread at least 10 times since it takes longer for things to click for individuals like yourself. Do that then get back to me...

Or maybe you are the one who should take some time off of this board to do register for a reading and comprehension class, a critical thinking class, and a how to be prompt class.

 

[randycat99]

So let's get you on record here...is it your claim that the Cell that was demonstrated is only equally as powerful as a P4 3 Ghz? Only 2x as powerful? Just how much, per your extrapolation estimation?

 

[ultimate_end]

So let's get you on record here...is it your claim that the Cell that was demonstrated is only equally as powerful as a P4 3 Ghz? Only 2x as powerful? Just how much, per your extrapolation estimation?

Unfortunately, it would be a flawed estimation. The Cell processor in this demo might have only been running at 200MHz for all we know...

 

[PC-Engine]

So let's get you on record here...is it your claim that the Cell that was demonstrated is only equally as powerful as a P4 3 Ghz? Only 2x as powerful? Just how much, per your extrapolation estimation?

My claim is that a single threaded P4 at 3.5GHz can do at least 12 DVD streams minus the downscaling. Then using that as a base, a hyperthreading P4 at 3.5GHz would be able to do 20 streams. Finally based on those two factors, a dual core hyperthreading P4 at 3.5GHz would probably be able to do ~ 40 DVD streams minus the downsampling of course.

I'm not claiming a dual core HT P4 is more powerful than that CELL that was used in that demo. This is plainly obvious because 1. We don't know what clock speed CELL was running at and 2. Even if we knew the clock speed we still don't know the % utilization.

My point was to show that CELL is not orders of magnitude better than a dual core HT P4 which ERP and aaaaa00 also agree. This can be extrapolated with the known information.

 

[randycat99]

So how much then, given the numbers you are working with? Give us a number. Will you go on record to say an Cell is only 20% more powerful than an HT, dual-core 3.5 GHz P4? Is that what your numbers are telling you? Just a yes or no...

(the funny thing is- even if this is absolutely true, I think I can live with that sort of power in a console to entertain me with games... )

 

[PC-Engine]

So how much then, given the numbers you are working with? Give us a number. Will you go on record to say an Cell is only 20% more powerful than an HT, dual-core 3.5 GHz P4? Is that what your numbers are telling you? Just a yes or no...

(the funny thing is- even if this is absolutely true, I think I can live with that sort of power in a console to entertain me with games... )

In order to do that you have to assume a lot of things and I don't think it's appropriate nor do I think we need to know how much more powerful one is over the other. They are both powerful and really I'm only interested in its relative power, besides this demo is only focusing on one type of application an app that favors the CELL architecture more than a P4 architecture ie P4 is a general purpose architecture.