PowerVR Series 5 to debut in 2003 using 0.13u

[Simon F]

There's a patent explaining how PVR works when it runs out of binning space, I haven't got a link handy, Simon could probably point you at it

No need. Ailuros gave a (long) link to it on the first page of this topic.

 

[Dave Baumann]

No need. Ailuros gave a (long) link to it on the first page of this topic.

Patents make my head hurt - any chance of explaining it in laymens terms Simon?

 

[Simon F]

IIRC, the abstract is fairly simple to read.
[EDIT] On a second inspection, it does contain a reasonable quantity of lawyer-ish, doesn't it? [/EDIT]

 

[Dave Baumann]

On a second inspection, it does contains a reasonable quantity of lawyer-ish, doesn't it?

Yes.

I started to read it before and my basic thoughts were that this was similar to the talked about principal whereby when the Bin space is filled it would render whats in the bin, refill the bin and then blend the results.

However, I really didn't try to read too much!

 

[Ailuros]

Dave,

Kristof had posted a thread about it some time ago with a brief explantion on it I'll see if I can find it through the boards search function.

(Warning layman's explanation): basically from what I understood it splits up in macro and later on into micro tiles. I called it on another board "hierarchical tiling" (varying size of tiles), but I'm really not certain if the term does make any sense here. Simon can always correct me.

edit:

Here you go

http://www.beyond3d.com/forum/viewtopic.php?t=758&highlight=patent

 

[rubank]

[0006] A block diagram of the type of processor used is shown in FIG. 1. This comprises a tile accelerator unit 2 which performs the tiling operations referred to above and supplies the tile information to a display list memory 4. This in turn supplies tile object data for processing to the image synthesis processor (ISP) 6 which performs the ray/ surface intersection tests referred to above. This derives image data which comprises object identification and depth data. After this, the thus derived image data is supplied to a texturing and shading processor (TSP) 8 which applies texturing and shading data to surfaces which have been determined as visible and outputs image shading data to a frame store.

[0009] Preferred embodiments of the present invention provide a method of managing the display list in such a manner as to impose an upper bound on the memory consumption, whilst at the same time attempting to minimise the memory bandwidth consumed by the system. This may be achieved by storing the state of the system (ISP and TSP) to memory before rendering of a tile is complete, and reloading this state at a later time in order to finish the rendering. We refer to this technique as "z/frame buffer load and store".

[0025] In the system of FIG. 1, polygons which are to be displayed are broken down into planar triangles. The data representing these is transformed by either hardware or software into a screen space representation and is then stored in local memory. The tiling process then performed by the tile accelerator 2 creates a list of pointers to the transformed triangle data for each tile in the scene. This is then stored in the display list along with parameters pertaining to the object of which the triangle it a portion.. This is shown schematically in FIG. 2 in which the region headers 10 correspond to the identities of the tiles 1-8 shown in the Figure. As can be seen, the tiles cover two triangles named object A and object B. object A is visible in tiles 1, 2, 5, 6 and 7 and object B is visible in tiles 3 and 7.

This sounds interesting, pointing to two processors - or will they be inside a single die?
Or am I just stupid?

 

[Teasy]

Don't flatter yourself, Teasy. I'm not following you around. I make more posts on this board than just "Teasy and PowerVR related" you know.

Joe you post this sort of stuff in all sorts of threads I post in.. I don't think I'm flattering myself when I say that you seem to follow me around with this stuff whenever I post in this forum.

Lighten up, dude.

Tell you what, we'll make a deal, you grow up and I'll lighten up? That's not an insult either, you have really been acting like a spoilt brat with me for a while now.

So......Care to wager?

Yeah absolutely, I'm not one to shy away from a bet, even if I did not say this was a certainty. So what are we betting for then? Remember this is a serious bet this time so lets make it something serious. Oh and as I said, I did say "most likely" not deffinately so I'm not betting £50 or something on this ok

In any case (wager or not), what are you basing your speculation on? I don't see much in the way of PowerVR product history or market drive to suggest that they ever release new hardware based on the latest brand new API. (Which is what 'higher than DX9' would be next year.)

I'm basing my opinion partly on things said in the article at the start of this thread and partly on.. well other things that I'm not about to talk about in this thread.

 

[Joe DeFuria]

Tell you what, we'll make a deal, you grow up and I'll lighten up? That's not an insult either, you have really been acting like a spoilt brat with me for a while now.

"Acting like a brat" appears to be in the eye of the beholder.

Oh and as I said, I did say "most likely" not deffinately so I'm not betting £50 or something on this ok

Fine. You propose the amount (obviously less than £50 , and I'll agree to it.).

Then we'll have to formalize the bet with more precise terms. (For example, "what we mean by beyond DX9?" What happens if "Series 5" is delayed beyond 2003, etc.

Here's by proposed terms. You win the bet if:

1) Series5 based PC Graphics bords are on the shelf in 2003, and
2) They include hardware support for pixel and vertex shaders version 3.0 or higher.

RE number 2: That is an acceptable "Beyond the first version of DX9" in my definition for graphics hardware...even though the first version of DX9 API might actually ship with version 3.0 shaders in the API...I'm cutting you a little slack here. )

I win the bet otherwise.

I'm basing my opinion partly on things said in the article at the start of this thread and partly on.. well other things that I'm not about to talk about in this thread.

And I'm basing my opinion mostly on past history and what my common sense tells me. Seems like we based our opinions on our last bet on the same premise.

 

[Teasy]

Here's by proposed terms. You win the bet if:

No I most deffinately do not agree to those terms. AFAIR I did not make any comment on betting on when the product will be released. I only mentioned DX9+. Now I do believe it will be out there in 2003 obviously, but I am not close enough to launch to be THAT sure it will all work out in time. So no I won't bet one a timeframe at this time.

How about if they are not on the shelf in 2003 the bet is null and void? As it would be unfair on you to keep the bet past 2003 incase they release Series 6 as Series 5 or something.

So the bet would be:

If Series 5 is released in 2003 and has hardware support for pixel and vertex shaders version 3.0 or higher, I win

If Series 5 is released in 2003 and does not have hardware support for pixel and vertex shaders version 3.0 or higher, you win

If Series 5 is not released in 2003 the bet is void.

Is that ok?

And I'm basing my opinion mostly on past history and what my common sense tells me. Seems like we based our opinions on our last bet on the same premise.

Not really no, bettinh on release dates and betting on rough specs are very different things.

 

[Joe DeFuria]

If Series 5 is released in 2003 and has hardware support for pixel and vertex shaders version 3.0 or higher, I win

If Series 5 is released in 2003 and does not have hardware support for pixel and vertex shaders version 3.0 or higher, you win

If Series 5 is not released in 2003 the bet is void.

Agreed.

However, I originally said "product on the shelf". ("Released" is a bit ambiguous). If you are saying that you would win if IMG "announces" a product in 2003, but we can't buy one until 2004 (or ever), then I wouldn't take the bet.

So if we agree on "product on the shelf" (as in IN STOCK) during 2003, then we have a deal. Just speicify the amount.

The shipping date is crucial to the bet. Simply because the longer it takes to deliver the product, the more likely it will fall into the "not competitive with high end."

My opinion is based upon the premise that we won't see a PowerVR part that is competitive with the best of the bleeding edge competition at the time (as that is historically what has happened.) And I see PS 3.0 as "bleeding edge" only until the end of 2003. Once PS 3.0 is no longer bleeding edge, it wouldn't surprise me to see a PowerVR part that supports it.

Not really no, bettinh on release dates and betting on rough specs are very different things.

You misunderstood. By "same premise" I meant your position is again based on some public info about future products and some inside info that you won't publically disclose, and my position is again based on what I believe is common sense applied to past historical facts.

And to be clear on all of this: much like BitBoys, I really am pulling for IMG Tech. I would LOVE to see more competition in the industry, especially in the high end. I will quite happy to lose the bet if that means more competition!

 

[LeStoffer]

So the bet would be:

If Series 5 is released in 2003 and has hardware support for pixel and vertex shaders version 3.0 or higher, I win

If Series 5 is released in 2003 and does not have hardware support for pixel and vertex shaders version 3.0 or higher, you win

Is anybody gonna bet that Joe will win this one flat out?

 

[Teasy]

Joe, I agree with the product on the shelf bit. For the purpose of this bet released will mean available to buy and in stock.

However on second thoughts I am having a problem with the bet. Why does both pixel and vertex shaders have to be above DX9 for the spec to be above DX9?

If a card is released and has DX9 spec and a big feature like either VS or PS is above the DX9 spec then surely that is above DX9? Also I was thinking, well PS and VS aren't the only things that could make a chip higher then DX9 spec. A chip with DX8.1 PS and VS that has something like displacement mapping is still above DX8.1 spec don't you agree.

I suppose a bet has to be nailed down to a degree as to not cause arguments later, but I'm not sure it has to be THIS simple. I'm not deffinately saying I won't take the bet we have sorted out so far. I'm just not going to rush into this bet so I think we should try to iron it out or at least discuss it a little more before I commit to it.

Also I think we should discuss the payment/forfit for the bet. I'm currently un-employed and basically skint, well I just bought a new graphics card and a few GameCube games, but relitavely skint So perhaps money would not be nearly as much of a forfit to you as it is to me.

Anyway we have plenty of time to iron out this bet.

 

[Joe DeFuria]

Why does both pixel and vertex shaders have to be above DX9 for the spec to be above DX9?

Because that's my opinion. Some chip makers are not even putting vertex shaders at all in hardware. That doesn't impress me in terms of engineering achievement.

To be clear, I'm not saying ABOVE PS and VS 3.0. I'm saying PS/VS 3.0 or higher. That is, I consider PS/VS 2.0 as "DX 9.0", and 3.0 as "DX 9.1".

If Series 5 has PS and VS 3.0 (and nothing beyond that), you would win the bet.

If a card is released and has DX9 spec and a big feature like either VS or PS is above the DX9 spec then surely that is above DX9?

IMO....it "all depends." We're starting to "muddle the waters" in terms of beyond DX9 spec again. Part of the problem is that DX9 isn't released yet, so it's understandable. That's why I'm trying to make this as clear-cut as possible.

Rumor has it that DX9 will ship with both PS/VS 2.0 and 3.0. however, no hardware yet supports 3.0 or higher. (And 3.0 may actually be "disabled" in the API for end user purposes.) When 3.0 becomes "enabled", we don't know if DX9 will still be called DX9, or if it will be "DX 9.1, DX10" or whatever.

I'm basically saying that the "first version" of DX9 means that the hardware supports PS and VS 2.0. IMO, the "big boys" (ATI and nVidia) will support both PS and VS 3.0 next year or early 2004.

I don't think IMG TECH will do the same...both PS and VS 3.0 in hardware at the same time I expect ATI / nVidia to do it. That is what my opinion (my side of the bet) is based on.

A chip with DX8.1 PS and VS that has something like displacement mapping is still above DX8.1 spec don't you agree.

Yes. But displacement mapping is a significant feature. Whether or not some other part is "beyond DX9" depends on what said feature is. Displacement mapping? Probably. Some extra texture format? Probably not. And there's no way to do a meaningful bet about "extra features" IMO. We can't forsee such things now and determine which "features" would qualify or not.

So, in my mind, "Beyond DX9" means at least PS and VS 3.0 hardware support. That is unambiguous, there's no way that I can see for either of us to say "but...but...."

I suppose a bet has to be nailed down to a degree as to not cause arguments later...

Exactly.

I'm just not going to rush into this bet so I think we should try to iron it out or at least discuss it a little more before I commit to it.

Actually, the real purpose of making this bet is not even the bet itself, but to go exactly what we're going though. Because it's a formal way to clear up exactly what our two positions are...and it may turn out they are not too different.

When you said "most likely DX 9.1" you can see now that than can mean a WIDE variety of things, depending on your interpretation. If your definition of DX 9.1 is vastly different than mine (mine being VS and PS 3.0 or higher), then we may not actually disagree.

So perhaps money would not be nearly as much of a forfit to you as it is to me.

Well, we should probably rule out sacrificing body parts too. They may not nearly be as important to you as to me. We'll think of something!

 

[Kristof]

This sounds interesting, pointing to two processors - or will they be inside a single die?
Or am I just stupid?

ISP and TSP terms have been used for a while and KYRO also has an ISP and a TSP, basically the ISP is the hidden surface removal unit (Z Check) and TSP is the Texturing Pipeline (So the Pipes and TMUs). All on the same chip, just names for the blocks. Check some of the presentations on www.pvrdev.com for more details.

Obviously I am not making any statements about the future, just that ISP and TSP are not really new concepts.

Oh and Teasy and Joe... can you make the bet a bit interesting... some extra motivation for the teams here is always a nice bonus

K-

 

[Joe DeFuria]

Oh and Teasy and Joe... can you make the bet a bit interesting... some extra motivation for the teams here is always a nice bonus

Hmmmm....how about this. If Teasy wins, then I won't post a naked picture of myself.

(Me now expects to see Series5, in all its PS/VS 5.0 glory, hitting the shelves next week...)

 

[demalion]

For the love of God, Kristof!!!!!


Please...?

 

[Ailuros]

Dumb question is either of you two or both even aware what supposedly PS/VS3.0 even includes?

I'd like to hear at least educated guestimates if you don't mind.

Hopefully that was the kind of motivation Kristof was asking for. I have little to no interest to read about bets concerning critical body parts again.

 

[Randell]

I have little to no interest to read about bets concerning critical body parts again.

Oh I dunno you are Greek after all

 

[rubank]

Kristof,
thanks for clearing things up.

 

[LittlePenny]

If Teasy no longer has a penis, what do you call his testicles? External ovaries perhaps? j/k

I do have an idea for a bet though. Teasy would have to eat a duck dinner if he lost( I want pictures ). And Joe would have to stop being an asshole to Teasy.