S3 & PowerVR

There’s hardly much of a compatibility worry - PowerVR seem to have fewer troubles then some of the established PC vendors! AFAIK the only game that fails is GP3 (because if its rather shoddy 3D coding) – however, that should probably work if the card is used as an IMR.

I’m talking GigaPixel – the last thing IMG would want, if they were going it alone, is to put NVIDIA’s nose out of join and force them to implement a fully GP based system.

Eh? So you are saying that a chip that is less complex than a GF2MX, but gives a GTS or above a run for its money most of the time doesn’t fit this criteria?

As time goes by the difference will decrease, and possibly (if nothing else is done) even cross over, but we are quite far from that position at the moment IMO.

 

My point is not about how compatible the current chip is, but how much did it "cost" (in terms of time and money) to get there? I wouldn't consider Neon a particularly "compatible chip."

Also, in terms of "compatibility", how much of PowerVR's "advantages" are really realized when all of the PC games are coded to run well on traditional renderers?

Oh...so you're saying IMG and it's partners just don't have the balls and vision to be market leaders in the desktop arena. So why should the case be different with mobile parts and integrated chipsets?

Show me the sales numbers of the GF2 MX boards vs. the KyroII and then you'll have your answer.

 

Joe, I am sorry but your post is very condescending, and I really don't know if your responses to Wavey's post have much relevance.

 

You’ve used one then?

I’ll wager the ‘chip’ was no more or less compatible than any of the rest – the basic design premise of PowerVR from Neon 250 remains in KYRO; the only difference between the two is that its more than just Videologic are marketing Kyro, and more development has gone into the drivers.

Take a look at any comparative benchmark and you have your answer. Remember, the basic level of KYROII in terms of fillrate is no more than that of a TNT2 Ultra…

Joe, I said if they were going it alone. Having a large partner on board gives them a level of abstraction from potential dangers, and it should also afford a greater push in the areas that IMG are just not geared towards (push with board vendors, marketing etc.) because of their size and business model. If KYRO were produced by IMG alone I personally don’t think there would be any hope of Hercules being on board now – they are there because of ST.

Completely different situations for the company’s involved. However, you may like to reference back to Hercules original reason for stepping on the KYRO bandwagon, and their thoughts on the matter.

_________________
'Wavey' Dave
Beyond3D

 

No, but I do remember the (much delayed) launch rather well.

I would wager otherwise.

Neon250's core was much more closely related to PCX1/2 (and Dreamcast) than Kyro. For example, Neon supports PowerSGL if memory serves, and kyro doesn't.

If it's just a matter of drivers, then why doesn't Neon have the same stand-alone GL driver, rather than the mini-GL PowerSGL wrapper that PCX1-2 also uses?

And the KyroII is priced no more than a TNT2 ultra?

So why is Hercules also on the ATI bandwagon? Why doesn't Hercules concentrate exclusively on Kryo tech?

I know the premise: at a given price point, a board manufacturer should have more profit from a PowerVR product than "traditional" products.

Why is ST dropping out then? They should be making a killing by making profits on chips where the competition has such small margins, they should be enjoying large ones.

I keep on going back to the same question. If that premise is a universal truth, then why isn't EVERYONE on board with PowerVR products?

Where are all those 3rd party Taiwaneese, and why aren't they on the PowerVR bandwagon?

 

Look, all I'm trying to get from you (Dave and Teasy), is this:

Assuming that PowerVR tech is in fact, "all that it's cracked up to be", WHY are we talking about the fagile state and possible elimination of PowerVR tech from the PC space, rather than going on about how they are the number 1 graphics technology in PCs?

The only thing I can get from Teasy thus far are vague and unsupportable things like "it's not IMG's fault" and "IMG's partners just weren't interested."

Well, OK, WHY weren't IMG's partners "interested?" The bottom line is, everyone is interested in making money. I would bet that STM WOULD be "interested" in the "high-end" market, if they believed their expectations of the technology lent them to believe they could be profitable in that segment.

<font size=-1>[ This Message was edited by: Joe DeFuria on 2002-03-05 23:29 ]</font>

 

Are you sure the high-end market really makes much money ?

I'm not too sure.

 

Its still based on exactly the same premise though Joe – both KYRO and Neon250 use infinite plane for overdraw detection and removal, the rest is really just aesthetics. The reason KYRO doesn’t support PowerSGL is there’s no reason for it to! Why support another API when you are going to get no developer support for it? (You may actually fine PowerSGL in other newer PowerVR products though…)

No clue how much a TNT2 Ultra cost these days, but that’s not the point. However, if you really want a look at pricewatch you’ll see that the 64MB KYROII stands at just $7 more than a 64MB GF2MX400, and I know which I would take of those!

Because ST/KYRO do not answer all markets!

Who? That lot? :rollseyes:

 

I keep hearing how the next PVR tech is going to take the market by storm. Problem is that it's always the "next". A shame since they were getting closer too. But I guess there's always the next one.

On the other hand maybe they have to take one step back to take two forward.

 

Yep, that's really a shame! Cause if the products would have been on the market in time...

But maybe this time the product is ready in time, cause the main competitor (Nvidia) seems to slow down ( GF4 was delayed 6months, NV30 seems to be delayed 6months too, according to penstar.com ).

Neon250 was late 1,5 years ( shown at Cebit 98; released ~Okt 99 ).

KyroI was delayed at least 9months and slow ( should have been on the market at the end of '99 according to an paper from Videologic; but came to the market only 8-9/00 ).

KyroII was delayed 6months but only due to the delay of the KyroI.

KyroIII / KyroII-Ultra is delayed at least 6months due to ST ( according to ST the KyroIII / KyroII-Ultra should be on the market in 09-10/01 already ).

Don't know the Schedule of the STG 5500 after they cancelled the STG5000 in 2001.

So with the KyroII they got closer, but now they are slowing down again. Hopefully they can reverse this negative trend.


Manfred

 

Joe DeFuria and (oops!) Dave verbally slugged it out with:
Joe:No, but I do remember the (much delayed) launch rather well.

Hmm.. this could get interesting...

Dave:I’ll wager the ‘chip’ was no more or less compatible than any of the rest – the basic design premise of PowerVR from Neon 250 remains in KYRO; the only difference between the two is that its more than just Videologic are marketing Kyro, and more development has gone into the drivers.

Joe:I would wager otherwise.
Neon250's core was much more closely related to PCX1/2 (and Dreamcast) than Kyro. For example, Neon supports PowerSGL if memory serves, and kyro doesn't.

Ok, you are both wrong.

PowerSGL was deemed obsolete for Kyro and so dropped. (BTW the SGL of Neon was quite different to PCX1/2 anyway). Neon has many similarities to Dreamcast but radically different to PCX1/2.

As for compatibility: In D3D, for example, there are a multitude of flags to indicate the various features of a particular piece of hardware and, unfortunately, not all games developers bothered to check them. For example, the "Z-Bufferless rendering", is one such flag, and some games ignored it and insisted on reading a "non-existant" Z buffer on Neon. This is not an issue with Kyro since it can optionally output sections of the Z buffer.

Joe:If it's just a matter of drivers, then why doesn't Neon have the same stand-alone GL driver, rather than the mini-GL PowerSGL wrapper that PCX1-2 also uses?
This is (no doubt) an overly gross simplification, but with GL you have to support everything no matter how obscure or rarely (or never!) used a feature is. If the hardware doesn't have a particular feature then the usual approach is to fall back to use software to render the feature. To do that, you'd at have to be able to access a Z-buffer.....

That's why Neon had a mini-GL (which only supplied functions that were actually used in the GL games), as opposed to a full GL.

Dave:Take a look at any comparative benchmark and you have your answer. Remember, the basic level of KYROII in terms of fillrate is no more than that of a TNT2 Ultra…

Only if you measure one tiny aspect of the performance (eg multiply clock rate by the number of texturing units) and ignore all the other real-world issues such as bandwidth in and out of the chip, and efficiency of the system!

The rest of the debate was business stuff which I'm prudently staying clear of!

_________________
"Your work is both good and original. Unfortunately the part that is good is not original and the part that is original is not good." - Samuel Johnson

<font size=-1>[ This Message was edited by: Simon F on 2002-03-06 10:40 ]</font>

<font size=-1>[ This Message was edited by: Simon F on 2002-03-06 10:42 ]</font>

 

WOW !

SimonF;

You must be lucky when it doesn't matter to you how well Your company does in the business. Or is Your wage so low that it does not matter if You work for IMG or (for example) the post office.


Manfred

 

More on the Via rumour:

http://www.digitimes.com/NewsShow/Article.asp?datePublish=2002/03/06&amp;pages=01&amp;seq=2

 

pffft... but an aquisition of IMG's driver program was never on the cards.

 

Very droll Manfred. FWIW I only discuss technology matters not business.

 

Sorry SimonF;

I found it funny


Manfred

 

They money to be made is at the low end were all of the voulme is. Sure you have a higher profit margin at the high end but the volumes dont come close to the lower end. Ideally you would want to cover all segments of the market but that is not possible with a compnay that has a limit amount of reasouces...

Some one also asked about the power comsumption of the K2, AFAIK I remember reading about 4 watts its what the general card consumed...

 

Yes we do, the PowerVR MBX consumes 60mw of power at 50Mhz (100Mpixels/s) and is 6mm^2 in size. Thats sounds pretty damn good for a mobile chip if u ask me.

http://www.powervr.com/Release.asp?ID=32

Couple that with a superH processer and u have a mobile dreamcast, all u need is the GD-ROM

 

PowerVR MBX does not equal the equivalent of Kyro based lap-top graphics. Completely different market.

Of course the MBX power consumption is good...it better be for handhelds and PDA type devices.

 

Back to the future of PowerVR:

VIA confirmed to tecchannel that they are in negotiations with a graphic manufacturer. They also plan a big press conference on CeBIT.

http://www.tecchannel.de/news/20020307/thema20020307-6902.html