Cell 4GHz...maybe!?

[Nerve-Damage]

Let’s say a little birdie told me that the Cell processor is actually running at 4.02GHz, and that the reason for the sudden increase was due to the RSX being scaled back to 500MHz. Originally what kept the Cell from performing at the original rated speed of 4GHz was heat issues, mostly heat issues within a closed environment. My source claims that there are Toshiba Cell based workstations (nothing to do with PS3) that have water cooled Cell processors running at 4GHz. The reason for the water cooling setup was to keep the enclosed environment temps from affecting the Cell overall temps (i.e. increasing them).

Which brings us too the enclosed environment of the PS3 system…my source stated the RSX was scaled back on purpose to compensate for some of the heat dissipation in the PS3 unit (see side note). She also states; Sony has came up with a nice heatsink/cooling system for the PS3 system, with noise levels being around 27-29db. Sony felt the RSX 50MHz deficit wouldn’t be that bad, compared to the benefit of increasing the overall performance of the Cell processor.

Personally I think she’s telling the truth…Sony reasons on down clocking the RSX in my opinion is still suspect (if true). Anyhow, I’ve been sitting on this information for quite sometime (about a Month or so), and kind of alluded to MILR awhile back in a private message about it. Anyhow, your thoughts guys????

[xbdestroya]

Are these the same sources that had you speaking of dual-core RSXs with SPE's attached? :smoke:

[Insane Metal]

Well, personally, I don´t believe this.

Anyway, clocking up the cell from 3,2Ghz to 4Ghz would give it an AMAZING performance boost !

[Voidler]

Nope. FCC documents have it at 3.2 Ghz still

http://gullfoss2.fcc.gov/prod/oet/forms/blobs/retrieve.cgi?attachment_id=704267&native_or_pdf=pdf

[Nerve-Damage]

Are these the same sources that had you speaking of dual-core RSXs with SPE's attached? :smoke:

Nope!!

I believe the RSX is more of a Multicore unit... :stirpot:

[Nerve-Damage]

Nope. FCC documents have it at 3.2 Ghz still

http://gullfoss2.fcc.gov/prod/oet/forms/blobs/retrieve.cgi?attachment_id=704267&native_or_pdf=pdf

Well, console specs aren’t usually set in stone (IMO) until they hit the store shelves.

[mario25]

I think is too late for sony to upgrade or downgrade the system, unless this was done some time ago.........

[xbdestroya]

Well, I think the chances of this happening are essentially zero, to put in my take on it.

An RSX at 500MHz wouldn't do much to improve the thermal conditions of the PS3 either, because it's not as if they are choosing to go that route because they wanted to cool it down. Rather, those chips simply hit the target - the majority of the chips running at 550 exceeded it by requirign a voltage bump to reach the intended speed.

[Nerve-Damage]

Can someone please tell me why the RSX clock-specs aren’t present in the FCC report!?

[Sephiroth_VII]

Maybe because they're confidential? Read this (http://gullfoss2.fcc.gov/prod/oet/forms/blobs/retrieve.cgi?attachment_id=704256&native_or_pdf=pdf).

BTW: Welcome back, Nerve! Haven't seen you around for awhile.

[Nerve-Damage]

Well, I think the chances of this happening are essentially zero, to put in my take on it.

An RSX at 500MHz wouldn't do much to improve the thermal conditions of the PS3 either, because it's not as if they are choosing to go that route because they wanted to cool it down. Rather, those chips simply hit the target - the majority of the chips running at 550 exceeded it by requirign a voltage bump to reach the intended speed.

Thats why i said "Sony reasons on down clocking the RSX in my opinion is still suspect (if true)." in the starter topic. I think the reason (if true) has nothing to do with heat issue...more like escape-goating answer on the RSX down clock speed.

[frosty]

If they were going to up the clock speed on cell, wouldn't they have to get the FCC to re-certify it? Doesn't seem likley, though it would be nice.

[xbdestroya]

Thats why i said "Sony reasons on down clocking the RSX in my opinion is still suspect (if true)." in the starter topic. I think the reason (if true) has nothing to do with heat issue...more like escape-goating answer on the RSX down clock speed.

You're not understanding what I'm saying.

It *does* have to do with heat issues; it's just a change out of necessity rather than out of desire. This change will allow no increase in the thermal outputs of other components; it was made because RSX at 550 was going over it's desired wattage envelope.

[RavenFox]

Can someone please tell me why the RSX clock-specs aren’t present in the FCC report!?
good point.

edit: The source is a she and Im curious as to why she would tell you this and make it up? If it is made up..urrr well xb says it is and well Im confused. Does Nerve usually have wrong info or something? Nothing against you Nerve just trying to figure this out.

[Nerve-Damage]

Maybe because they're confidential? Read this (http://gullfoss2.fcc.gov/prod/oet/forms/blobs/retrieve.cgi?attachment_id=704256&native_or_pdf=pdf).



Which brings me to my next question!! :ohno:

If the RSX is just a Nvidia G80 IP as many think...why all the secrecy?


BTW: Welcome back, Nerve! Haven't seen you around for awhile.

Thanks!! :smile:

[frosty]

Actually, G71. And maybe that's why they aren't releasing it's info, because there really isn't that much to be exited about. They probably don't want the RSX to be compared to X360's GPU on paper until after the system's games have had time to speak for themselves. Many will see RSX and Xenos aren't much different power wise while forgetting to factor in the power of cell. This unfair comparison may be the reason for the secrecy.

[Nerve-Damage]

You're not understanding what I'm saying.

It *does* have to do with heat issues; it's just a change out of necessity rather than out of desire. This change will allow no increase in the thermal outputs of other components, it was made because RSX at 550 was going over it's desired wattage envelope.

Sorry for that; I really meant ambient heat temperatures (or contributing too) within a closed environment.

[makeitlookreal]

First of all, I want to say that it's great to have Nerve Damage back in the forums.

Second of all, I want to say that that after all the delays, possible down-clocking of the RSX, and other various issues a nice positive SURPRISE about the PS3 would be very welcome!

Regardless of it's possibility Nerve Damage has sources saying the RSX is a multi-core processor. I also have sources that told me the RSX is going to be much more than what was announced at E3 of 2005, but did not go into much more detail. Also, there are other hints out there on the internet on various websites (many in other langrages) that point to something interesting.

However, we also know from CPI (who has been a great asset and fantastic member on this forum) that the RSX is nothing toooo dramatic or a miracle GPU.

But if these sub-cores are real I doubt CPI could even mention a word about them or even a whisper, and technically they might be considered seperate from the RSX itself. Remember, CPI said the RSX (in itself seperate from the PS3) is nothing that extraordinary. Now, I am NOT in anyway saying that CPI is trying to hint anything whatsoever. I'm just saying there is a small tiny chance that there are sub-cores somewhere in the PS3 working with the RSX. Technically, CPI could have been completely honest with us about the RSX and these cores could exist since they could be considered seperate from the GPU itself.

You see, I'm trying to be positive and hope something surprising (in a good way) will be announced at TGS.

However, realistically I don't think we will get these sub-cores and it's not very likely at all. But I refuse to say it's absolutely impossible.

I expect to hear that the PS3 is pretty much just like we heard at E3 of 2005 with perhaps a few minor differences and with a slight chance of a RAM upgrade (which could explain these great looking games lately).

However, part of me wants to hope that Nerve Damage's sources are telling the truth!

[Nerve-Damage]

Actually, G71. And maybe that's why they aren't releasing it's info, because there really isn't that much to be exited about. They probably don't want the RSX to be compared to X360's GPU on paper until after the system's games have had time to speak for themselves. Many will see RSX and Xenos aren't much different power wise while forgetting to factor in the power of cell. This unfair comparison may be the reason for the secrecy.

I really don’t think that’s the case bud... :drunk:

[xbdestroya]

Which brings me to my next question!! :ohno:

If the RSX is just a Nvidia G80 IP as many think...why all the secrecy?

Who in the world still thinks is based on G80?

[xbdestroya]

gDoes Nerve usually have wrong info or something? Nothing against you Nerve just trying to figure this out.

Yeah, Nerve usually has wrong info. :angelgrin

[makeitlookreal]

I don't think the RSX is based on the G80. Some minor aspects of the G80 like perhaps increased cache sizes might be included, but according to CPI and other developers the RSX is pretty much a stripped down common PC part with a few minor changes.

However, if there were sub-cores in the PS3 they would not absolutely have to be considered as an actual part of the RSX. The RSX itself could still be mostly like the N47 and those sub-cores with a few megabytes each of EDRAM or SRAM could be considered seperate. It would be a huge surprise, but probably is very, very unlikely. If it doesn't happen I'll be fine and I'm still excited about the PS3. But I like to stay optimistic and not rule things out 100%.

[RavenFox]

Yeah, Nerve usually has wrong info. :angelgrin
ahh ok lol:angelgrin

[Nerve-Damage]

Who in the world still thinks is based on G80?

I’m so use to seeing people (on various forums using G80 or G71) I started writing either or. Anyhow, the point really being the RSX isn’t just a current NVIDIA PC dullard. Don’t get me wrong either; the RSX (main-core) will be based around its PC brethren, however the RSX will contain specialized smaller sub-cores (3) to assist with other chores.

[Nerve-Damage]

Yeah, Nerve usually has wrong info. :angelgrin

Be it founded or unfounded…I just report info regardless of what others may think. :whip:

[Siraris]

Nope!!

I believe the RSX is more of a Multicore unit... :stirpot:

I can assure you that the RSX is a 7900GT.

[Nerve-Damage]

I can assure you that the RSX is a 7900GT.

I can assure you that the RSX is not a 7900GT. :smirk:

[makeitlookreal]

Nerve Damage,

Could you please give us a full report on the information you have been told so far? You can easily just mark it as speculation until it is verified. I would love for everyone here to hear the unique and interesting information that was shared with you.

[Siraris]

I can assure you that the RSX is not a 7900GT. :smirk:

...

Making baseless rumors will not endear you to anyone, regardless of the cute face you put after it. The RSX is a 7900GT.

[Insane Metal]

...

Making baseless rumors will not endear you to anyone, regardless of the cute face you put after it. The RSX is a 7900GT.

No. The RSX is not a PC Graphics Card.

[Nerve-Damage]

...

Making baseless rumors will not endear you to anyone, regardless of the cute face you put after it. The RSX is a 7900GT.

Being a trolling a$$-hat doesn’t help your case either!!

[Applefiend]

I have a feeling there's only 5 days of speculation time left....

[Nerve-Damage]

Nerve Damage,

Could you please give us a full report on the information you have been told so far? You can easily just mark it as speculation until it is verified. I would love for everyone here to hear the unique and interesting information that was shared with you.

The only thing I can attest too (personally seeing) was a working mockup GPU that Sony/Toshiba had Ions ago. I firmly believe that technology has survived (to some extent) within the Sony/NVIDIA relationship. To what degree I cant say…

[Sephiroth_VII]

I have a feeling that you're right.

Nerve, could you please compile what you've found out about RSX so far, and what your basis is for suggesting sub-cores.

[Nerve-Damage]

I have a feeling that you're right.

Nerve, could you please compile what you've found out about RSX so far, and what your basis is for suggesting sub-cores.

Give me a bit...I'm working and surfing at the same time. Its good to be the boss. :evil:

[makeitlookreal]

I really want Nerve Damage to give out the information he has been given (even if it's just speculation at this time), but I know other individuals have posted the following online on different forums. I will repeat the very basics of this, but really hope that Nerve Damage will give us the details.

Basically, there is a theory that there will be a ring of three or four seperate sub-processors designed by both Toshiba and Sony. Each one would have a few megabytes of SRAM or EDRAM and basically each of these sub-processors would be sort-of-like a SPE, but different because they will be made specifically for graphical purposes. They also could be similiar to the technology in the PS2 (but upgraded) so they and their fast memory could help with backwards compatibility.

However, this is a BIG rumor and at least I have NO ABSOLUTE confirmation of this. I don't expect this to happen and really think the RSX is just a common PC part, but for the record I don't want to discount this theory totally because I have been told there are unknown features in the PS3 and RSX that are going to surprise us. I don't know if this is true or not. It could be a total lie. But it's something to think about.

[section]

Here we go again :)

[Sephiroth_VII]

Nerve: Lol, you're a boss in some company? Lucky you. Say hello to your for subjects for me.:salute:

EDIT: I can see Cpi on the "current viewers" list. Would youy please care to comment, or even better, clear this up?

[Rubbernek]

There's got to be a name for this: "Anti-FUD".

[acousticvan]

All I can say is "STOP THE SH*T". You woke up the beast inside MILR. The speculation is nonsense. The Cell modules that IBM is selling are configured at 2.8 GHz.

[Applefiend]

I have a feeling that you're right.

Yup, we're at the end of the PS3 development phase, and in production phase, probably for a month or more. No point keeping secrets now. PS3 is a done deal. It's gone from a giant dev kit box and and empty classic case with a boomerang to the final system.

Your stock may go up as well as down.

We can start speculating about PS4 though. :) 7 Speculating years to go. :)

[joe]

http://img42.imagevenue.com/img.php?image=32942_51976_122_589lo.jpg

... 3.2Ghz.

[makeitlookreal]

It's not nonsense. It's speculation and there is nothing wrong with waiting to hear Nerve Damage report on what his sources told him and what he is speculating. He might be totally wrong, but as of right now there are no specs at all on Playstation.com Sony's own Playstation 3 website so until they post the specs speculation is fair game in my opinion.

Do I really think the PS3 is going to have some type of multi-core subprocessor ring? No! But I am willing to hope and find the concept facinating if nothing else simple discussion until the real specs do get revealed. We are not hurting you and if you don't like this thread (although we would love for you to follow it) there are tons of others about various great games that are very, very interesting.

I can't wait to hear Nerve Damage's report.

[cpiasminc]

EDIT: I can see Cpi on the "current viewers" list. Would youy please care to comment, or even better, clear this up?
It's either wishful thinking or stupid confusion.

It's generally very easy to mislead someone by saying something that means nothing. There are things about RSX that sound impressive enough to give MILR 7 heart attacks by themselves, but are actually quite meaningless as they're features pretty much every GPU has these days.

[makeitlookreal]

If they are so meaningless I sure do wish Sony would be honest with their consumers and share the information. I'm not saying their secretly super-powerful, and believe you when they say they are not overly impressive. But I would love to hear Sony tell us the truth about the RSX, and it's because of their lack of information threads like this one where gamers struggle for tiny bits of possible but unlikely information exist.

[Nameless]

I can't wait for these damn PS3 upgrade threads to end...
It just creates confusion and frustration for everyone involved. Peace

[yoshaw]

I imagine he'd be wishing for sony to come clean with RSX info at the PS4 unveiling too. Since that would be their one and only ultimate final chance for revealing RSX specs. So don't expect this to end lol

[LaLiLuLeLo]

Amen to that, nameless.

[Nerve-Damage]

I’m just begging for trouble (flame-bait) on this one, but who cares.:hurt:

I wrote a while back about hypothetically seeing a working mockup of Sony/Toshiba GPU (Visualizer), during an overseas videoconference call about improving industrial waste management (IIRC). Let’s just say my business ties; aren’t just local. I have many friends and associates from all walks of businesses. One being this particular European group; that’s very close to Toshiba and its business affairs. Anyhow, I was hypothetically shown a real-time Matrix (Matrix 1) rendering of the Street scene with the Lady in Red on a computer system that was co-developed by Sony. When I heard Sony…that’s when the questions flowed out. They told me the graphics chips are a revolutionary chipset designed by Toshiba/Sony. And the only thing at that time I knew Sony and Toshiba were working on was the Visualizer and Cell processor. And my friend pretty much confirms that it was the Visualizer. By the way; the Matrix scene was fu**ing unbelievable.

Anyhow, you can call me what you want…my eyes don’t lie. Hypothetically anyway!!

[makeitlookreal]

I for one appreciate everything you have to say.

[yoshaw]

Nerve,
If it was a business sorta meeting, you were probably shown a CG, I'd assume! Because afaict, that's what presentational meetings are for, i.e to really really impress the client/other party or whatever. But maybe you know better.

[Nerve-Damage]

Nerve,
If it was a business sorta meeting, you were probably shown a CG, I'd assume! Because afaict, that's what presentational meetings are for, i.e to really really impress the client/other party or whatever. But maybe you know better.

NOOOooooooo it was real-time!!

Using the keyboard to swing around the agent, neo, morpheus, girl, ect...pretty much laid that to rest.

Edit: By the way guys/gals the meeting wasn’t about computers or anything (improving waste management solutions). The whole seeing Sony/Toshiba system was totally by accident, I made a remark during a brake to my friend about that fugly computer case in the background. That’s when the ball got rolling.....

[jaxmkii]

NOOOooooooo it was real-time!!

Using the keyboard to swing around the agent, neo, morpheus, girl, ect...pretty much laid that to rest. i wish i could trust you:unsure:

still nice to see you back

[xbdestroya]

Nerve Damage, ok you know I mean I'm not saying that didn't happen; Toshiba and Sony were architecting a GPU at one point afterall.

All I'm saying, is that Cell will be at 3.2GHz, and that RSX is based firmly on a tweaked NV47 architecture.

[Nerve-Damage]

Nerve Damage, ok you know I mean I'm not saying that didn't happen; Toshiba and Sony were architecting a GPU at one point afterall.

All I'm saying, is that Cell will be at 3.2GHz, and that RSX is based firmly on a tweaked NV47 architecture.

I knew that bud. :smile:

[Nerve-Damage]

i wish i could trust you:unsure:


Well I really don’t expect you too…my goal isn’t to win a popularity contest around here. My goal is to provide any info (true or false) about the PS3 or any other system, so the readers can make a valid judgment.

still nice to see you back

Thanks

[BruceWayneIII]

We know the 3.2 GHz exist and so does the 4.0 GHz.

My question is: are those two the exact same chip but only controlled by two different crystals or are they two different Cell products?

[Smokey]

Well I really don’t expect you too…my goal isn’t to win a popularity contest around here. My goal is to provide any info (true or false) about the PS3 or any other system, so the readers can make a valid judgment.
youre always a great read mate (true or false) :) keep it up

[xbdestroya]

We know the 3.2 GHz exist and so does the 4.0 GHz.

My question is: are those two the exact same chip but only controlled by two different crystals or are they two different Cell products?

They're the same Cell chip; required voltage to reach certain speeds is the only differentiator between where a chip gets binned speed-wise.

But, if you are asking in order to appraise the plausibility of this rumor, let me reinforce for you that there is no way. :smoke:

[Smokey]

They're the same Cell chip; required voltage to reach certain speeds is the only differentiator between where a chip gets binned speed-wise.

But, if you are asking in order to appraise the plausibility of this rumor, let me reinforce for you that there is no way. :smoke:
so if modders (could) do this to their ps3s it would really make no difference?

[Applefiend]

There's a lot of patents and a lot of tech floating around Sony, most of which won't make it into PS3. What Nerve may have seen is a workstation class GPU card with power requirements not suitable for video game consoles.

Interesting though.

[Epix]

They're the same Cell chip; required voltage to reach certain speeds is the only differentiator between where a chip gets binned speed-wise.

But, if you are asking in order to appraise the plausibility of this rumor, let me reinforce for you that there is no way. :smoke:

Well......chips coming off fab lines are ramped up in speed until they fail. At which point they are dialed back a little and sold at that speed. It's more of a function of when the internal delay's are greater than the clock frequency that determines the final CPU speed.

[cpiasminc]

If they are so meaningless I sure do wish Sony would be honest with their consumers and share the information. I'm not saying their secretly super-powerful, and believe you when they say they are not overly impressive.
Contrary to what you may believe, there are hundreds of thousands of people out there who leap to far crazier, and far more stupid conclusions than you ever would. When that happens, more confusion and more FUD and a more massive flood of nonsensical garbage is bound to ensue.

Misunderstanding circuses are not one of those cases where a crowded hour is worth an age without a name.

[jaxmkii]

so if modders (could) do this to their ps3s it would really make no difference? nope... might see a bump in frame rates if anything.. oh and more crashes

[xbdestroya]

Well......chips coming off fab lines are ramped up in speed until they fail. At which point they are dialed back a little and sold at that speed. It's more of a function of when the internal delay's are greater than the clock frequency that determines the final CPU speed.

We're talking about the same thing, except that sometimes chips can easily reach greater speeds than they are binned at; the binning resulting from demand at a certain price point.

For example, I run an A64 3000+. Normally this chip operates at 1.8GHz@1.4v, but I actually have mine running at 2.2GHz@1.35 volts - so overclocked and undervolted.

I am fairly sure most Cell chips can reached 4GHz if pushed - it's just a matter of what voltage is required.

I present... the schmoo (which everyone has hopefully memorized by now):

http://www.realworldtech.com/includes/images/articles/cell-8.gif

[Epix]

We're talking about the same thing, except that sometimes chips can easily reach greater speeds than they are binned at; the binning resulting from demand at a certain price point.

For example, I run an A64 3000+. Normally this chip operates at 1.8GHz@1.4v, but I actually have mine running at 2.2GHz@1.35 volts - so overclocked and undervolted.

I am fairly sure most Cell chips can reached 4GHz if pushed - it's just a matter of what voltage is required.

There are definitely voltage restrictions set forth before fab begins when chips are designed. That being said, to determine the "shipping" clock frequency the chip is pushed to breaking point (using the predetermined voltage level) and dialed back and shipped.

We're basically agreeing with eachother from different POV's.

[jako]

Reading what Nerve said about this powerfull "PC" that could render Matrix realtime back in the old days..... it is true, I remember reading this, the "PC" was the GSCube:
http://www.answers.com/topic/gscube

So, is Nerve telling the truth about the rest ??

[BruceWayneIII]

They're the same Cell chip; required voltage to reach certain speeds is the only differentiator between where a chip gets binned speed-wise.

But, if you are asking in order to appraise the plausibility of this rumor, let me reinforce for you that there is no way. :smoke:

Nope, just asking a question about the Cell - I'll leave it up to others to speculate. Life is so much simpler when only using official facts.

I was mainly interested in the production of the Cell and this was a perfect thread to ask the question.

[Fazares]

a never ending dream...

[xbdestroya]

There are definitely voltage restrictions set forth before fab begins when chips are designed. That being said, to determine the "shipping" clock frequency the chip is pushed to breaking point (using the predetermined voltage level) and dialed back and shipped.

We're basically agreeing with eachother from different POV's.

Yeah, I think we're both just trying to emphasize different points for the 'audience.'

Moving on from binning itself, I want to point out to people why if 4GHz Cells are a possibility, Sony wouldn't want to go with them.

We take this equation...

Overclocked Watts = Default Watts * (Overclocked Mhz \ Default Mhz) * (Overclocked Vcore \ Default Vcore)²

Now, assuming Cell has to receive 1.2 volts for decent yields at 4GHz, and assuming a 3.2GHz Cell draws an arbitrary 50 watts at 1.0 volts (again the wattage is arbitrary), that means that in comparison to a 3.2GHz Cell (at 50 watts) a 4Ghz Cell would draw 90 watts of power.

This would have a *huge* effect on system thermals and workable cooling solutions within the console. Not to mention, we don't even know if the schmoo chart reflects an accurate average, or if the 'hits' displayed are for a cherry-picked sample to begin with.

[xbdestroya]

Reading what Nerve said about this powerfull "PC" that could render Matrix realtime back in the old days..... it is true, I remember reading this, the "PC" was the GSCube:
http://www.answers.com/topic/gscube

So, is Nerve telling the truth about the rest ??

It could very well have been the GSCube, itself a prototype for future console design decisions. I was thinking about that as well when Nerve Damage mentioned it.

But... why would him having witnessed a GSCube demo make any of the rest of this true? Afterall, this isn't anything to do with what he himself has witnessed, but what some 'source' has told him.

[frosty]

Thank you for sharing this with us nerve, it at the very least gives us something to talk about in this pre-TGS news drought.


=-o You double posted! ^^^ Bad XBD! No Cell for you! (j/k)

[Epix]

Yeah, I think we're both just trying to emphasize different points for the 'audience.'

Moving on from binning itself, I want to point out to people why if 4GHz Cells are a possibility, Sony wouldn't want to go with them.

We take this equation...

Overclocked Watts = Default Watts * (Overclocked Mhz \ Default Mhz) * (Overclocked Vcore \ Default Vcore)²

Now, assuming Cell has to receive 1.2 volts for decent yields at 4GHz, and assuming a 3.2GHz Cell draws an arbitrary 50 watts at 1.0 volts (again the wattage is arbitrary), that means that in comparison to a 3.2GHz Cell (at 50 watts) a 4Ghz Cell would draw 90 watts of power.

This would have a *huge* effect on system thermals and workable cooling solutions within the console. Not to mention, we don't even know if the schmoo chart reflects an accurate average, or if the 'hits' displayed are for a cherry-picked sample to begin with.

"What if" the RSX clock reduction saved 40 watts?

[jako]

But... why would him having witnessed a GSCube demo make any of the rest of this true? Afterall, this isn't anything to do with what he himself has witnessed, but what some 'source' has told him.
He could have made connections with people working on the videogame industry.
That's all we can say.
For me I'll just wait TGS when Kutaragi will give us infos about the PS3 with a lash :)

[xbdestroya]

"What if" the RSX clock reduction saved 40 watts?

LOL! Well, *if* moving RSX from 550 to 500 'saved' Sony 40 watts of draw, then my advice is for everyone to make sure they have fire insurance on their homes and to suspend PS3 in an inert liquid when active. :smoke:

[makeitlookreal]

I goofed up on this post. Nerve Damage corrected me. He does indeed believe that the sub cores are on the same chip as the RSX.

[xbdestroya]

There is only one piece of silicon left unaccounted for in PS3 right now: the EE+GS and/or it's stand-in. Whether this chip can possibly assist in the Playstation 3's playback of Playstation 3 titles - I don't know. Not if it's a normal EE+GS is my feeling.

And I personally will not be entertaining notions of other exotic chip clusters playing a role. My advice is for others t do the same. :smoke:

[Fazares]

wha..?
ee+gs into a ps3..???

[Epix]

wha..?
ee+gs into a ps3..???

For backwards compatability.

[xbdestroya]

wha..?
ee+gs into a ps3..???

Well, exciting piece of news for you I guess! ;)

I'm not saying that it's a full EE+GS, I'm also not saying it's not. All we know is that there will be a hardware consideration made on the part of SCE to ensure backwards compatability.

And on one level or another, that means probably at least the Graphics Synthesizer.

[Fazares]

For backwards compatability.
i never heard they want to include those to give us backward compatibility...

[Nerve-Damage]

Once again, he is NOT saying the RSX is not a tweaked normal GPU. These sub-processors are attached, but not part of the RSX.

You lost me here bud...

The RSX silicon (chip) will contain all the Sony specialized IP sub-cores on one-chip IMO.

[makeitlookreal]

XBD,

I respect your opinion that basically the chance of such additional hardware in the PS3 being impossible. Pesonally, I find it very UNLIKELY for various reasons and especially do to several developers more or less denying such assumptions without actually saying so out loud.

I'm just very eager for information about the PS3's hardware for better or worse. I won't instantly believe anything new I hear about the PS3's hardware, but as long as Sony's not being fourthcoming with the information I feel it's wrong to totally discount anything unless it's truly absolute absurd like 4 GIGS of RAM or QUAD SLI CONFIG or something like that.

At least CPI told us there are a lot of features in the RSX that are common in most other graphics cards. Even if they are pretty common it's going to be great to hear actually what's inside of the beast!

Once I get my PS3 I'm going to be too busy playing it to worry about the specs, but until then I'm facinated by anything someone might hint at. However, I'll also boldly say that in this case I find it not too likely.

NERVE DAMAGE -

I sincerely apologize for saying the wrong thing about your opinion.

[Nerve-Damage]

NERVE DAMAGE -

I sincerely apologize for saying the wrong thing about your opinion.

No biggie my friend... :thumbr:

[makeitlookreal]

Thanks Nerve Damage. By the way, I gave you a +REP for sharing what you saw with the forum. Also, for sharing the information about the RSX. It's not confirmed but you are sharing what you were told.

Everyone, lets give Nerve Damage a big boost of REP for telling us about what he saw and what he has been told. A lot of people would just go about lurking, but he came forward and took the time to tell us!

[mrdarko]

There is only one piece of silicon left unaccounted for in PS3 right now: the EE+GS and/or it's stand-in. Whether this chip can possibly assist in the Playstation 3's playback of Playstation 3 titles - I don't know. Not if it's a normal EE+GS is my feeling.

And I personally will not be entertaining notions of other exotic chip clusters playing a role. My advice is for others t do the same. :smoke:

I totally agree with this.

The only thing i can think of being in the PS3 that's not accounted for is a "version" of toshiba's super companion chip.

And even that is unlikely.............

[woundingchaney]

Thread is kind of long and Im running out of time so I cant look through it, thought this may be helpful.


http://img.photobucket.com/albums/v665/snowcrash007/51968.jpg

http://img.photobucket.com/albums/v665/snowcrash007/51976.jpg

http://img.photobucket.com/albums/v665/snowcrash007/51964.jpg


FCC approval.

[venomv]

The problem with adding the ee-gs and allowing them to accually play a part in the PS3 beyond backwards compatibility is that it takes away Sony's ability to eliminate them if they can get BC running with emulation correctly. That is what they did with the PStwo, correct(?), that is why it's BC isn't quite as good as the PS2's.

[Applefiend]

http://img.photobucket.com/albums/v665/snowcrash007/51976.jpg
FCC approval.


Funny they don't list the frequency of RSX, but yeah, 3.2Ghz folks...

[makeitlookreal]

What do all of you think the chance is that the PS3 in normal function will be able to access the 4MB of EDRAM (or whatever ammount they put in the PS3 for BC)?

Part of what Nerve Damage is talking about is having multiple sub-processors (similiar to the PS2 hardware in some ways) each with a few MB of memory. If you need to put PS2 hardware in the PS3 it would make sense to design it in such a way that the console can utilize it in all games and not just last gen games.

I really think that we are not going to get true emulation of the PS2 because if that's the case then all games would be supported and we have seen disclaimers that not all games might run on the PS3.

[venomv]

Not all PS1 games work with PS2, and as of last I heard not all PS2 games are functional with the PS3, just a vast, vast majority (somewhere in the neighborhood of 98% for PS2-PS3 if I remember right).

[frosty]

^If I remember right, MILR, 4 MB Edram would be a waste in PS3 because it's not enough to hold HD resolutions. I've even heard the 10 MB in 360 isn't near enough.

[makeitlookreal]

Yes, you are right Frosty. The minimum to really utilize antialiasing and HDR is around 12 or so megabytes at the minimum from what I remember. But I was not suggesting the PS3 utilize the memory as a framebuffer. I was thinking more along the lines as an extra quick texture cache or memory pool in the RSX. It would be a place where data could be retrived very, very quickly. I just hate for any ammount of memory to be sitting in there (especially fast EDRAM or SRAM) unused in PS3 games.

[cpiasminc]

The minimum to really utilize antialiasing and HDR is around 12 or so megabytes at the minimum from what I remember.
Assuming 720p, 2xAA and HDR = either FP10 or something akin to NAO32. For 1080p, 4xAA, and FP32 HDR, you're talking 142.4 MB. FP16 would bring it down to 79 MB.

12 MB is the magic number that would have made the most sense for Xbox360 as it means 720p 32bpp framebuffer + compressed Z/Stencil + with 2x AA and NO TILING required. It's also just enough for 1080p framebuffer + compressed Z/Stencil with no AA and no tiling required. MS wasn't willing to push the transistor budget an extra 17 million, so that's all there was to it.

It would be a place where data could be retrived very, very quickly. I just hate for any ammount of memory to be sitting in there (especially fast EDRAM or SRAM) unused in PS3 games.
I do believe the plan is, even if there is an EE+GS in the initial PS3 unit, there won't be one in there forever and the backwards compatibility will eventually be relegated to software emulation. Meaning there will be a point where that eDRAM will cease to exist -- which in turn means no way in hell are we getting access to that eDRAM whether it's sitting there or not.

[woundingchaney]

^If I remember right, MILR, 4 MB Edram would be a waste in PS3 because it's not enough to hold HD resolutions. I've even heard the 10 MB in 360 isn't near enough.
10 mb is enough for tiling but Im under the impresion that 12 mb was the magic number.

[VG Aficionado]

10 mb is enough for tiling but Im under the impresion that 12 mb was the magic number.Almost any amount is "enough for tiling", I think. The question is how convenient a given amount really is. Correct me if I'm wrong anyway :tardbang:

[xbdestroya]

MS would have been much better off (from an ease of implementation standpoint) if they just went with 12MB. I find it surprising actually that they didn't - was the cost differential between 10 and 12 that great?

[woundingchaney]

Almost any amount is "enough for tiling", I think. The question is how convenient a given amount really is. Correct me if I'm wrong anyway :tardbang:
You know Im not really sure what is "necessary for tiling" although the only thing I have ever heard referrence to is that 10mb is enough to tile but 12mb was the preferred number.

Im not sure but I dont think 8 mb would be enough for tiling or a lower number as that, but in all honestly Im not sure.


LOL check Cpis post above.


@XBD I find it hard to believe that they stopped at 10 mb as well, it just doenst make sense to only have to add another couple mb to achieve the "magic number".


Why does this remind me of that fat guys speech "Developers Developers Developers"

[makeitlookreal]

Could they simply theoretically put the 4MB of EDRAM on the die of the RSX and connect it to the PS2 hardware so that when they remove the PS2 hardware the 4MB of EDRAM would be remaining? I know they are probably not going to do it this way, but it would be interesting to determine how feasible it would be.

[cpiasminc]

You know Im not really sure what is "necessary for tiling" although the only thing I have ever heard referrence to is that 10mb is enough to tile but 12mb was the preferred number.

Im not sure but I dont think 8 mb would be enough for tiling or a lower number as that, but in all honestly Im not sure.
All depends how you define a "tile." 8 MB will hold 32bpp 720p framebuffer and uncompressed Z/Stencil without AA and without tiling. 6 will be fine if the Z-Buffer compression is enabled. If you define a tile as being a significant fraction of the display, then you do need a fair bit of room.

Technically, if you wanted, you could to get down to fine granularity tiles and use it more as a framebuffer *cache* while the real framebuffer is off in VRAM, you can still get some net gains, particularly with alpha blended geometry and scenes with high overdraw. Then you can get by with multiple very small tile cache blocks. I do believe I did some basic calculations on this end in an earlier post, but IIRC, a few tens of KB per quad would actually be just fine.

[killermmn]

man the fcc a use not a full hdtv for the test!!

[makeitlookreal]

Too bad I read on these forums that the theory of tile caches in the RSX has been debunked. It would have been neat for the PS3 to have such small sized caches in it's GPU pipelines for such uses.

[cpiasminc]

Too bad I read on these forums that the theory of tile caches in the RSX has been debunked.
Was there ever a theory like that in the first place? AFAIK, any talk of tile caches has been hypothetical "what if somebody did this in a GPU?" kind of talk. Nothing about RSX at any time.

[version]

tilecache will be, and static z-pyramid on the chip
L1,L2 cache , 1 terabytes/s

this is sure , a friend...

4ghz ? hmm i dont think , 2.6...

[woundingchaney]

tilecache will be, and static z-pyramid on the chip
L1,L2 cache , 1 terabytes/s

this is sure , a friend...

4ghz ? hmm i dont think , 2.6...
You just had to do that didnt you.:shrug:






LOL

[makeitlookreal]

nAo once said something about a tile cache that made a few people believe that he was hinting about the RSX. His comments bounced around quite a few forums. I forget how it happened, but either he or someone else debunked the theory saying the RSX was not going to use any type of "defered rendering" using tile caches.

Basically, it was an interesting idea but it turned out not to be used in the RSX. That's all I remember at this moment.

[lips]

I read on wikipedia some stats that are now taken down. There was one claim the LS for each core had to be set at 4.6 ghz (not true, thank god) and that the seventh spu was not actually locked for os and critical functions. This too would be good, but i do not see this listed any more, so it may have just been user submitted knowledge that was indeed just rumors.

[version]

i was buy nitrogen past week :D

[frosty]

Stay off that hippy crack!

[Garfunkel]

@Version: you donut make any sence!

Nerve, where the bloody hell have you been! i know you may be unfortunately wrong but it is something to think about till TGS. dont leave now!

[cliffbo]

Give Nerve some respect everyone! I can't believe the pasting he's received above... its just plain scandalous after all the contributions he's made to these forums. i hope your right Nerve, and something tells me your source is someone who is under NDA.

[Smokey]

Give Nerve some respect everyone! I can't believe the pasting he's received above... its just plain scandalous after all the contributions he's made to these forums. i hope your right Nerve, and something tells me your source is someone who is under NDA.
QFT

[yoshaw]

4ghz ? hmm i dont think , 2.6...

STFU n00b! ;)

And read this, it's officially 3.2ghz
http://img.photobucket.com/albums/v665/snowcrash007/51976.jpg


lol :laugh:

[makeitlookreal]

Version, so you are saying that someone in the know has confirmed to you that there will be a tile caching system in the RSX? Just trying to figure out exactly what you are attempting to report. Thanks for keeping us informed when you hear information.

[version]

tilecache was in the geffx but only L2
rsx has tilecache :)

[woundingchaney]

tilecache was in the geffx but only L2
rsx has tilecache :)
Umm dont all modern gpus have tilecache or am I off base here??

[makeitlookreal]

So according to your sources the RSX has an L2 tile cache. That is interesting.

[cliffbo]

i thought this thread was about the cell at 4.02GHZ?

:)

[cpiasminc]

nAo once said something about a tile cache that made a few people believe that he was hinting about the RSX. His comments bounced around quite a few forums. I forget how it happened, but either he or someone else debunked the theory saying the RSX was not going to use any type of "defered rendering" using tile caches.
Yes, well, people misunderstood that he was talking about something completely hypothetical that would have a lot of the benefits of eDRAM without having to blow through 100 million transistors to get there.

tilecache was in the geffx but only L2
rsx has tilecache
Which RSX are you talking about? The car, perhaps? A friend of mine at Honda mentioned something about the microcontroller in the Acura RSX having multiple "tiles" of cache used as a history stack when keeping track of state of a lot of things.

Umm dont all modern gpus have tilecache or am I off base here??
I think he may be deliberately confusing the idea of a framebuffer tilecache (which some 2nd/3rd tier GPU manufacturers have experimented with in the past -- all to the end result of failure), with caching textures by tiles. The latter has been part of 3d acceleration hardware since before consumers ever saw them.

I seem to recall when companies like 3DLabs and S3 were experimenting with deferred tile rendering and nVidia rather brazenly responded that it was the stupidest idea ever to enter the graphics hardware world.

[makeitlookreal]

If they thought that it was that stupid then probably they would have advised Sony that the RSX should not feature such a technology.

[xbdestroya]

I seem to recall when companies like 3DLabs and S3 were experimenting with deferred tile rendering and nVidia rather brazenly responded that it was the stupidest idea ever to enter the graphics hardware world.

I still own an Evil Kyro II PowerVR based card, so that's a definite artifact there I've got. :)

If they thought that it was that stupid then probably they would have advised Sony that the RSX should not feature such a technology.

Deferred tile rendering is in theory much more efficient than the 'traditional' way it's done now. NVidia's knocking it is more indicative of their propensity to ridicule anything they do not themselves espouse than it is reflective of any failings in the technology/theory itself.

Anyway, but for modern 3D gaming, it's never been implemented in a way that reaches it's potential - games run too 'fast' for the chips and the drivers have never matured to where they need to be.

But this technology is still in widespread use in the less demanding cell phone space, and certain iterations of Intel's integrated graphics are built upon it as well.

Just as there's a cadre out there of SEGA revivalists, a cult of Raytracing, and many various other groups dedicated to fringe technologies, so to is there a group that believes one day deferred rendering will have it's day on the PC.

And actually, speaking of SEGA the Dreamcast used a deferred technique in it's own graphics chip (a PowerVR design), not to mention that until this most recent generation, it was popular in arcade boards as well.

[version]

from nao:

" will use their pixel shader ALUs to perform blending (thus removing blenders from ROPs) it should be easier (but not easy) to support MSAA and alpha blending on FP32 RT.
"
L1 cache on the way , if iam right

[cpiasminc]

Deferred tile rendering is in theory much more efficient than the 'traditional' way it's done now. NVidia's knocking it is more indicative of their propensity to ridicule anything they do not themselves espouse than it is reflective of any failings in the technology/theory itself.
I do remember one company (I forget which offhand) trying to hit the value segment with a deferred tiling scheme back around the time the GeForce3 Ti line was the hot thing, and the GeForce 4 cards were just due to come out.

They were capable of creating GPUs that were ~80% of the performance of the GeForce4s in most cases and at worst about 65% (mainly shader benchmarks), and beating it rather handily in strictly alpha-blending benchmarks. The main thing though is that they were able to do this with a very small chip and a lot less memory bandwidth (and hence, cheaper RAM). So they were able to sell their cards for around $40-$75 a piece.

from nao:

" will use their pixel shader ALUs to perform blending (thus removing blenders from ROPs) it should be easier (but not easy) to support MSAA and alpha blending on FP32 RT.
Where did he say this? Because this sounds more like he's talking about nVidia's long-term roadmap and not specifically about any one product. It is true that nVidia evangelizes the philosophy that pixel shading power alone is the secret of the universe.

They've already been talking about moving blending and AA to the pixel shaders. I think ultimately, they would like to make ROPs part of the pixel shader unit itself -- which I can definitely see as having some niceties about it, but I don't care for the idea of pixel shader power alone in complete abandon of all else.

[makeitlookreal]

I remember nAo saying that too, but I don't remember exactly where. I did a google search for that text and parts of that text but cannot find a match.

[makeitlookreal]

CPI,

I just remembered this, and I meant to ask you before, but I really want to get your take on something.

The official power point presentations for Open GL states that it has PSGL (Playstation Graphic Library) extensions for "queries and conditional rendering." Does this have something to do with defered rendering, because it sounds close but I could be way off here.

What do you think about this CPI? Does this have anything to do with defered rendering or is something completely different?

[overclocked]

I do remember one company (I forget which offhand) trying to hit the value segment with a deferred tiling scheme back around the time the GeForce3 Ti line was the hot thing, and the GeForce 4 cards were just due to come out.

They were capable of creating GPUs that were ~80% of the performance of the GeForce4s in most cases and at worst about 65% (mainly shader benchmarks), and beating it rather handily in strictly alpha-blending benchmarks. The main thing though is that they were able to do this with a very small chip and a lot less memory bandwidth (and hence, cheaper RAM). So they were able to sell their cards for around $40-$75 a piece.


BitBoys eeeey? :smirk:

[Crossbar]

BitBoys eeeey? :smirk:
I would actually bet it was PowervVR with some Kyro card, I remember some similar description concerning them.

The BitBoys never delivered anything more than vapor, sad but true, they had some interesting technoly based on edram with massive bandwidth.

[overclocked]

I would actually bet it was PowervVR with some Kyro card, I remember some similar description concerning them.

The BitBoys never delivered anything more than vapor, sad but true, they had some interesting technoly based on edram with massive bandwidth.


Yes could be them to of course. I remember XGI and Bitboys being bought up namely. ATI or nVidia iirc.

[cpiasminc]

The official power point presentations for Open GL states that it has PSGL (Playstation Graphic Library) extensions for "queries and conditional rendering." Does this have something to do with defered rendering, because it sounds close but I could be way off here.
Sounds more like it has to do with occlusion queries and Z-prepass stuff. It's one of those things we do to avoid drawing objects that can't be seen. We render to a Z-only render target and query whether part of a bounding volume for a given object can be seen -- if it can, we draw the object, otherwise we save some render time by not sending those polys. Not perfect, but if it saves you from some unnecessary vertex processing, it's a net win in the end.

I would actually bet it was PowerVR with some Kyro card, I remember some similar description concerning them.
Might have been. I do recall the card actually coming out. Bitboys, I only saw a prototype demo. Then they gave up on the high-end in the short term (which they should have done, IMO -- learn to crawl before you walk), and went into the mobile sector, and then ATI bought them.

[frosty]

Lol. Learn to crawl before being swallowed up by the big shark.