Some Kyro 4800SE boards has arrived

Tompa:
I don't know if I should take that as an compliment or an insult.
But hey, I'm gearing up with a hardware implementation of a banded scanline renderer with almost free per pixel depth sorting, automatic front to back rendering until destination alpha says stop (controlled by PS), and of course pixel shaders where the PS program is concatenated from all surfaces at that pixel (front to back).

Did that sound like enough mumbo jumbo to keep them away from the vault where they store KyroIII?

 

Just to go into wild optimism mode...

Doom III and John Carmack. We all know that Kyro handles stencil buffers well. In Reverend's interview, JC left open the possibility of a cube map free code path just for Kyro II - and he wouldn't have done that if he didn't think a lowly Kyro II could run Doom at playable framrates. If (hah!) PVR can get a new chip out at around the same time as Doom (preferably just before) they could clean up on the Web's new favourite benchmark. They've got until Q1/Q2 2003, right?

If I were IMG, I'd boast about things going well (e.g. PURE Digital's (formerly VideoLogic Systems) new DAB radios, and Metagence) and not dwell on uncertainties (PowerVR PC).

OTOH... PVR have declared an interest in a next gen console win. This shows ambition (beating either NVidia or ATi - unless there's a fourth platform in the next gen). They intend to achieve this through PC AIBs and integrated chipsets. This shows ambition in the PC space.

 

I'm kind of wondering how much they're going to sell this card for, $5?

I mean this card isn't even competition for the Xabre, let alone GF4 MX, R9000, etc.

At this point they might save money by just dropping the design in the trash.

 

It would be a shame for such a sensible rendering principle to fall by the wayside.
I do not see why PowerVR must always be behind the market in development cycles. Is the design labor intensive? Neither do I see why such an approach so eminently well suited to integrated graphics has not been pushed in that area.

 

While I've been around since the early Dimension3D days, I can't say that I've run a PVR-site. Or any site at all. I think I was more to the PVR side than 3dfx side, but I've generally tried to stay out of "wars".

It seems like some mixup here. I hope this won't demote me all the way down to "slightly above the green goo under the bathtub".

 

Tompa,

Of course, I remember those days!

There were some classic threads with Jezz as well. I wonder where he went off too....Jezz was at least sensible and put forth good and logical arguments. That Sam guy though.... And BTW, I believe it was actually "Jezz's" site that you're thinking of. He did most of the technical writing for it, IIRC, making the case for Power VR tech.

To be clear, I still to this day believe that the "PowerVR Way" is a cool and "elegant" solution. I have always seen it's advantages compared to "traditional" renderers (Bandwidth efficiency related to fill-rate). I got into so many arguments for two reasons:

1) I was never convinced that VideoLogic/IMG Tech/NEC...whoever...was going to make an actual product that would get me excited.

2) "Nothing is for free". In other words, I was always leary of the "flip side" of deferred rendering approaches, (the disadvantages, such as bandwidth efficiency related to polygon count), and was never convinced that the advantages of PowerVR so clearly outweighed the disadvantages, that Deferred rendering would be the next logical step for all 3D graphics architects.

At this point, I almost feel like the opportunity for PowerVR / deferred rendering tech to make a big impact has come and gone. (Similar to embedded D-RAM solutions). I think the really big opportunity would have been to introduce a part around the time of the GeForce1 DDR/GeForce2.

But now going forward, with "traditional" raw bandwidth going beyond 20 gb/sec, polygon counts skyrocketing, and continued evolution of memory interface technology, it's hard to imagine that deferred rendering tech will make a big splash anymore.

I do think that nVidia still has a chance though. If they can apply gigapixel type tech, on a HIGH END part (256 bit DDR/DDRII memory interface), that could be something to behold...

 

Hmm slight exageration there don't you think? The Geforce 2 TI is still $60 in the U.S and £70 in the U.K, so I don't see why the Kyro II SE would need to sell for $5 dollars to be competative. It can sell for around $45-50 and £40-50 in the U.S and U.K and still be a good value solution. After all if you've only got a very limited budget then what are you going to buy?.. a Geforce 2 MX or a Kyro II SE? I suppose we'll just have to wait and see its final price though.

I agree its not a worthwile card as far as enthusiasts go, although I am interested in the EnT&L feature myself, but it can still be a good value card and sell ok at the right price.

Oh and BTW, AFAICS the Xabre 400 is a much better chip then the Geforce 4 MX, so I don't know why you consider the Geforce 4 MX to be the more impresive solution.

In todays games and future games (Unreal 2 ect) its advantages over a IMR architecture do clearly outwiegh any disadvantages it may have compared to a IMR architecture, after all Kyro II would have went out with the TNT2 if it didn't.

Wether that is the case in the distant future with skyrocketing polygon counts is up in the air. I'd say that just as their are solutions to IMR's bandwidth/fillrate problems (specifically Z-bufer compression and early Z, hierarchical Z ect) there are also solutions to any problem a PowerVR design might see in the future with huge polygon counts. I suppose we'll all just have to wait and see, I don't think PowerVR or tilling in general is beaten quite yet.

 

Possibly, yes. But those solutions would require silicon....How much, silicon, and how effective would the solution be? (Please don't speculate and say "shouldn't be any significant silicon" and "it should be very effective"...neither of us really has any idea. My point is, nothing comes "for free." Even if it is a deferred renderer. And all too often my "opponents" (those highly touting the advantages of deferred rendering) simply dismiss the "costs" of deferred rendering as somehow insignificant. To which I always counter, "if it's so insignificant, whay hasn't IMG/PowerVE made a part"....to which my opponents counter "IMG doesn't make parts, it's their partner's fault for not being agressive"...and we get into a useless argument.

 

should I post at all or do you just want to have a conversation with yourself and reply for me?

As for the questions:

A: Yes it may very well require extra transistors, how much I don't know, it depends on the method.

B: How effective it will be again depends on the method.

I'm not going to get into a argument on how pointless and massively oversimplified that little 'overview' was, as that really would be a useless argument.

BTW aren't you getting a little obsessed with our passed discusions on PowerVR? I didn't even reply to your "nothing is for free" comment and you still felt the need to go off on one about the "costs" of PowerVR ect ect.

 

I guess I should use even more smileys in my next reponse.

And BTW, I said "opponents", and did not imply just you when it came to "past discussions" about PowerVR. (This did not start about past discussions with you at all...you jumped in the middle.) And if it seems that I'm "obsessesed", it likely appears that way because you reconize the fact that I tend to repeat the same arguments, to several different people, over and over over the past 5 years.

And there is not a deferred rendering product on the PC market that has have yet to prove me wrong. (To be clear, I'm NOT saying that I have been proven RIGHT....just that apparently, my arguments have some sort of reasonable basis because they have not yet been proven wrong...)

Well, I tried to sum up the entire "Pro PowerVR vs. Con PowerVR" argument for the past 5 (and maybe next 5) years. Because basically (and of course, simplistically), that is all that we ever say to one another.

 

I figured the latter would be more efficient.

 

You used 2 smillies, one at the end of your little 'overview' and one after "Even if it is a deferred renderer", which I didn't even reply to.

You just oversimplified one argument that you used to have allot, about your opinion that PowerVR designs were somehow so complicated that they couldn't bring out a upto date version in the same time frame as companies using IMR. That is hardly the entire Pro/Con PowerVR argument, its just one particular argument.

I'm not suprised that is all we ever talk about, you force it into the conversation whenever you can. As seen in this thread.

It would certainly be more efficient for me anyway. I could then go and do something more useful with my time, like watch paint dry

 

Or possibly do some shopping for sterile surgical instruments, some local anesthetic....

 

No more so then betting that you would eat ..... if you lost a bet.

OMG. Sam, I had forgotten about him. After him though the 'cause' was taken up some other PVR 'fan' who couldn't be reasoned with because he'd always introduce irrelevant arguments. Ah, I remember those days fondly (/banging head on table).

 

Yes brings back memories

Sam and Jezz

 

BTW, ParaKnowYa will be holding a general Q&A session with IMG soon (they don't say when). So if anyone has any questions about PVR...

 

I plan to send ImgTech the questions Monday the 22nd and well be checking this tread so post any question in here.

 

OK Pottsey, I've already emailed in several questions (in two loads), here's some more:

1. On the Beyond3D forums, Teasy famously promised to consume a particularly sensitive part of his own anatomy if Kyro III was not on store shelves by September. As it looks as though he will be unlucky, can you say anything to console him as he tucks in?

2. There has been no high end PowerVR product in the PC marketplace. There has rarely even been a 'mid end' product. Is this because PowerVR's business model is flawed, or because PowerVR's technology isn't good enough?

3. Would the sale of a couple of licenses for integrated chipset cores give IMG enough security to produce its own chips for PC add in boards?

4a. Villagemark has been accused of deliberately sorting back-to-front. I have come to the conclusion that it renders the sky first, then the landscape, then the buildings in random order. Is this correct? *

4b. What nefarious tricks does Fablemark employ to trip up IMRs?

5. In a recent interview for Beyond3D, John Carmack left open the possibility of supporting Kyro in Doom III. Do you think that PowerVR will be able to persuade him to do so?

6. Tim Sweeney said that the Kyro II was "a competent TNT2 class chip" and was "not a viable piece of hardware". We now see the Kyro II performing very well in the latest Unreal Engine benchmarks, always beating a GeForce 2 MX, and sometimes a GeForce 2 Ultra. Has Sweeney apologised for his comments?

7. If PowerVR even did silicon layout for Kyro I/II, what did ST Micro do?

8. How would the stock market react if PowerVR produced its own chips? **

Obviously, the most important question would be along the lines of why the hell won't PowerVR produce its own damn chips?, but probably more polite.


Footnotes:

* Anyone disagree? Kristof?

** My guess for why PVR perseveres with an IP licensing model for PC AIB products.