S3 & PowerVR

Joe,

If all goes according to plans there should be more than one card releases this year (4800 excluded). It would be the absolute track record for PowerVR.

Rev did mention "the one" didn't he? Assuming it would make it be4 the end of this year to shelves it could finally qualify them to give the market some serious competition.

But I know how you stand to "ifs" and "buts" and I can't actually disagree. We'll just have to wait and see I guess.

 

I believe he said, "No, Kyro III would not be the one" (to take the world by storm I presume).

 

I hope there will be something other than KryoII 4800 on the shelves this year from the PowerVR team...just not before September.

 

Joe, would you mind please stopping? The majority of your posts these days seem to be provoking Teasy. I find the banter the two of you thrash through with quote after quote after bloody quote amazingly boring and iritating. He is not much better, and I'm sure many of us can see your point, but enough is enough!

 

I am aware what he meant. Considering the specifications it wouldn't be exaggerated, given the fact that someone takes a 02' release for granted.

 

ROFL

Which year? hehehehe

 

From an ATi presentation

Value 51 %
Mainstream 44 %
HighEnd 3 %
Ultra HighEnd 2 %

I don't know how they define the market segments, or whether the figures are no. of cards sold or by $$$ (more likely), but I think it shows that STM was entirely justified in targetting mainstream first. ST may be a big company, but it's not PC-oriented, so it may not have wanted to divert resources from its core businesses (e.g. hardware MPEG decoders) to make high end gfx cards, when comparatively few would be sold. Nothing to do with the potential of PVR's high-end designs, just resource management.

As to why Creative didn't buy STM GPD: why didn't VIA buy 3DLabs? (Assuming VIA has bought STM GPD). Presumably, Creative want to get into workstations (where TBR is of limited/no benefit - lots of geometry, lots of wire frame models with no need for HSR - and 3DLabs is of course strong), whereas VIA wants integrated chipsets and laptop, where PVR has theoretical advantages (e.g. MBX's low power and size). Hopefully, VIA also want to cover the full spectrum of gfx AIB too.

 

Wondering why STM didn't go after the high end ignores the potentially greater issue. There are many possible reasons STM might not have been interested in the high end, and it makes it easy to side-step any inferences from that to the quality/performance of IMG technology. The more interesting question in considering the potential of IMG technology for the high end is why *no one else*has licensed any IMG designs to go after the mid or high end PC market. STM's got plenty of reasons not to go high end, but lots of other companies would surely love to take a piece of the market, and they have the technological know how and financial ability to do so in a timely fashion, provided they have a good design. And yet, no other licensees of IMG tech for the high to mid level PC space. Why?

- JavaJones

 

I'm guessing that a license for PVR is about $1M US or more, probably plus royalties. It still would take a layout team, plus plenty of assorted engineers to get a product close to market.

The mask set alone for .15u would be close to $1M US dollars.

$2M is not chump change, especially when its only the start!

 

amk,

Actually, I would have thought that it would have been volume sales, not % of total profits. But you're right, we can't tell from the presentation exactly what they mean.

Why is ANYONE making "high end" and "ultra high end" cards if there is no money in it? I'd say the answer is one of two (or both) of the following:

1) There actually IS money in it.
2) There are some less "direct" benefits...like having the "leader" in performance helps push your brand, and increases sales of the lower performing cards via marketing.

While it is certainly true that there's much higher volume in the low to mainstram market, there's also much more competition in terms of the number of competitors, and presumably, the profit margin per unit is lower.

Heh! Well, that can be answered by my original opinion: 3D labs cleraly has a "high performance", not a value part. And Creative apparently wants to go after the high-end mainstream and professional market...at least at first. So if Creative didn't view IMg's tech or the business model as one that is suitable for the high-end, they wouldn't be interested in STM.

JavaJones,

Yes, I said the same exact thing.

And in my opinion, it's a combination of the IMG tech, and the business model that IMG uses to license the technology, that gives companies hesitation. Creative Labs, for example, made it clear in their conference call that they wanted full control over the development of future products (for various reasons). With a 3rd party having the bulk of the responsibility for design with IMG Tech, STM wouldn't fit Creative's needs.

 

If you mean allow ST to target the low end, and another company mid-high - I don't see how that could work. As Joe said, one of the motivations for having a high-end part would be branding, supporting sales of the cheaper parts. Also, high-end cores will tend to migrate towards mainstream over time. I can't imagine how two companies could split PVR tech in that way.

I'm not sure there are 'lots' of companies which could put PVR tech in all market segments. They'd be PC-oriented, heavy-hitting semiconductor firms, but not not so powerful that they demand complete control over future products. VIA perhaps would be IMG's ideal partner - big enough without being an arrogant superpower (e.g. Intel), and upwardly mobile. Who else? AMD is focused on CPUs (even its mobo chipsets exist only to cover late 3rd parties), Intel is IMO likely to want to control everything, SiS and ALi are (perhaps) too small, the memory manufacturers are in financial difficulties...

I don't see any need to question IMG's ability to design a competitive high-end part - IMHO business reasons (including IMG's business model - but if IMG's not big enough to go it alone...) are a more obvious suspect, and adequately explain the continued absence of a high-end PVR part. OTOH if VIA don't produce the goods, I'll get suspicious.

 

One more thing that just hit me is that PowerVR (at least after the KYRO) would have been able to compete in the mid-end (or even high end) market more effectively if all their attempts wouldn't have been haunted by constant delays.

With the specs known so far on the STG5500, would it have been released late fall last year it would have been a much different story than if it launches now somewhere before or within summer this year.

Take the STG4000/KYRO as an example: it was announced in Nov99' as far as I can remember. If it would have been released on time and cards released (even if only a revamped variation) in a 6 month manner, chances are high that the cards presented would be worthy mid-range competitors. Didn't the old roadmap suggest Series5 (STG 6000/6005) for 2002?

I'd rather say that it isn't an inability to design/deliver at all, rather a huge lack in execution whichever's fault that may be.

NVIDIA turned and is remaining the market leader because of it's rapid execution. IMHO the first step to success is to be able to execute. No market "forgives" delays....

 

But do we know if ImgTec managed to design their end on time? I agree with you that 'something(s)' has prevented the timely release of PVR products but I thought that is part of the reason why this thread exists.

It's been said before that execution is a major part of the plan. 3dfx suffered greatly from poor execution as well. Imagine if their Banshee on up had been on time.

 

http://www.digitimes.com/NewsShow/Article.asp?datePublish=2002/03/15&pages=04&seq=24


So Via want the KyroIII, not the KII.

 

If this it the case then I would guess that means that KYROIII is pretty close to being finished - if not VIA could just skip all the hassle and go directly to IMG and buy a Series 4 license...

 

I hope this article is not correct. An arrangement like this will never work.

If correct then IMG would design the chip, ST would manufacture the chips and VIA would sell them. So, an unwilling ST ( otherwise they wouldn't want to sell their division ) would stay in the middle. If true this would be the worst what could have happend !
It seems that VIA needs "something to compete" but is not sure if they even want PVR-technology and so they made an decision based on the smallest common denominator, nothing more. This (if true) looks really really bad. I hope IMG can survive.


Manfred

 

I'm not sure how "close" you are guessing, but that's a pretty big jump there. Series4 might be 2 years away, after Kyro III plus a couple spins of KyroIII variants to up clock speed and tweak features.. I mean, I think KyroIII is close too. (I don't think my guess of "on the shelves in Sept" is far away...it's just one more product cycle in the grand scheme of things.

My attitude is not quite as pessimistic as that, but I generally tend to agree. However, I really have no idea what this new arrangement is. Does this mean that VIA is just another recipient (like Hercules) for STM Kyro chips? Is VIA planning to re-sell the chips? (Will hercules have to buy them from VIA?)

How will the relationship between VIA and STM be different from that of Hercules and STM? Assuming VIA is not buying STM Graphics, is STM still up for sale, or does this new arrangement mean that STM is willing to hold on to the graphics division.

If STM is not bought, then will STM continue to sell chips to 3rd parties like Hercules? Or is this some arrangement to make VIA the exclusive recipient of STM chips?

More questions than answers...

And of course, the big one....what is it about the package that persuaded Via NOT to buy STM? Why doesn't anyone really seem to want to aggressively go for it?

 

Joe, I'm not sure you took that quite correctly; PowerVR Series 4 = KYROIII.

i.e. if ST are still miles away from getting anything into silicon with Series 4 (i.e. KYROIII) then what's the point of VIA buying anything from STM; they could just get a license directly from PowerVR and do it all themselves. There must be some value in going to STM, and as far as I can see the only value that STM has with Series 4 (a.k.a. KYROIII) must be time...

 

Ahhh....forgot about the mis-matched numbers.

To be clear, I think most people (including myself) believe that KyroIII chips are "close" to being ready for production. There's apparently only a disagreement between it being a more "imminent" release (next month or so), or release ready for the fall. I happen to believe the latter.