STM pulls the plug on Kyro

[Johnny Rotten]

On 2002-02-08 19:20, Teasy wrote:
Secondly Kyro III AFAIK is more or less finished and just needs the 0.13 micron process.

Have they even sampled on .13? If not then they arent even 'more or less' finished.

Personally, even assuming STM found a buyer tommorrow, I'd be stunned if a brand new partner brought out the 'Kyro3' in 7 months.

<font size=-1>[ This Message was edited by: Johnny Rotten on 2002-02-08 19:44 ]</font>

 

[Musivarius]

If a fabless company buys STM's GD then how would fab access work for 'Kyro 3'?

STM own a stake in TSMC(?) so they probably have some say over when they can get stuff done but a smaller independent company might not have the same priority or might have to pay to get it.

Would there have to be some kind of guarantees by STM that the new owner would get some of STM's fab time?

 

[MfA]

Question is, can anyone afford to bring KyroIII to market and still compete with GF4-MX on price?

How long does it take for a design to go from tape out to test silicon? If STM had been thinking about this for a while its not unthinkable work was just stopped shy of having the first masks made.


<font size=-1>[ This Message was edited by: MfA on 2002-02-08 19:59 ]</font>

 

[Joe DeFuria]

Hi Rev,

As far as I'm concerned, I only want a PS and VS 1.1 chip that can run as fast as however many shader effects I want plus 4xOGSS in a single level in a game.

That's a decent target to shoot for. My question is, did/does IMG / Whoever plan to have such a PowerVR chip on the market before or at the very latest, at the same time as some competitor does, and a lower price point?

 

[Joe DeFuria]

Question is, can anyone afford to bring KyroIII to market and still compete with GF4-MX on price?

Not just GF-4MX, but Radeon 7x00 and 8x00 series. There are a lot of "very capable" cards right now in the sub $200 price range, and the competition is only going to get more fierce within a year when these value and mainstream parts move to the same 0.13 process that Kryo-III is suppossedly targeted for.

It's really frustrating (speaking as someone who would like to see PowerVR tech be exploited.) Lots of PowerVR "fans" like to say that IMG and its parteners haven't been given their due credit. On the contrary, I don't think IMG and its partners are giving the competition enough credit, as they seem to consistently underestimate the market.

On a related note:

I can see the writing on the wall now:

"Venture Capitalists fund BitBoys Purchase of STM Graphics Division. BitBoys "cans" their current XBA architecture and starts development on a new product integrating XBA and PowerVR tech. First products expected 'soon.'"

 

[Teasy]

I really don't see how a fall release can be all the beneficial for PVR and the "Kyro 3." Perhaps they can hold things over for a while with another stepped-up Kyro 2, but I don't think that it will work for long.

Don't take that as an actual date of release, thats just was a few people around here have guessed at. When I said I don't see why they shouldn't be able to change management and still get Kyro III out by fall I didn't mean that I didn't think tey couldn't get it out earlier, I was just answering the question from someone about if the new company could get Kyro III out by fall or not.

I have hope in the PowerVR technology, but I have never dreamed that IMG would be that ambitious that their chips were that powerful. I was kinda thinking that they would aim for a product to take on nVidia's performance line, not the value line. I wouldn't have a problem paying $150-$200 for a PowerVR card if it would top everything on nVidia's current list. I just don't see that happening if they keep delaying the release of their products. PowerVR development is just too slow

Why do you assume that PowerVR Technologies development team is slow? PowerVR Series 4 (which Kyro III is based on) has been finished for a while now and AFAIK there already working on PowerVR Series 5. Also IMGTEC do not dictate how high end the final chip will be, that decision is made by STM. So if anything it looks like STM are lacking in ambition and not Imagination Technologies.

Have they even sampled on .13? If not then they arent even 'more or less' finished.

I don't know but what's to stop them putting Kyro III on a process they already have access too in order to get all the bugs out and work on drivers (probably running it at a lower clock speed) and then move to 0.13 and bump up the clock speed?.. couldn't this reduce the time between getting 0.13 micron and getting the product out of the door?

Personally, even assuming STM found a buyer tommorrow, I'd be stunned if a brand new partner brought out the 'Kyro3' in 7 months.

Why would that stun you?.. do you know how far a long the Kyro III is right now? How do you know its not almost finished? What if as I said they already had the chip running on a bigger process before 0.13 micron was available and so got a good head start on debuging and drivers. As I said I don't know if STM have had Kyro III on 0.13 yet and neither do you.. its very possible.

Why would it take the same team of technicians under new management (so possibly they'd even add staff to speed this up) 7 months or more to finish the chip if they already have the chip either on a bigger process and ready to move to 0.13 or already on 0.13?

<font size=-1>[ This Message was edited by: Teasy on 2002-02-08 20:30 ]</font>

 

[MfA]

Joe thats a good point, though chips made on an advanced process dont necessarily have to be expensive ... of course you'd have to question the sanity of a company who would target a product which could have been easily produced right now at early production runs of .13u when they desperately needed competetive products on the market.

So either the KyroIII is to far surpass the MX or its a very inefficient design to need .13u ... or STM were incompetent, or what?

Teasy, TSMC has been doing .13u runs for quite a while already even if they arent in full production.

Marco

<font size=-1>[ This Message was edited by: MfA on 2002-02-08 20:17 ]</font>

 

[Teasy]

"Venture Capitalists fund BitBoys Purchase of STM Graphics Division. BitBoys "cans" their current XBA architecture and starts development on a new product integrating XBA and PowerVR tech. First products expected 'soon.'"

LOL, I don't even think thats allowed within the licence agreement (to mix PowerVR tech with other tech).

So either the KyroIII is to far surpass the MX or its a very inefficient design to need .13u

Or it'll far surpass the GF4MX in performance and be quite a bit cheaper (because of the smaller process).

 

[Joe DeFuria]

Why do you assume that PowerVR Technologies development team is slow?

To be clear, he didn't say that PowerVR Technologies development team is slow. He said PowerVR Development is slow.

I think you have some personal connection to IMG-Tech, and try to defend them at all costs...even when they aren't being directly attacked.

I think his point "PowerVR Development is slow" is valid. How many PowerVR PC chips have shipped in the past 3 years, relative to the competition?

PowerVR Series 4 (which Kyro III is based on) has been finished for a while now and AFAIK there already working on PowerVR Series 5. Also IMGTEC do not dictate how high end the final chip will be, that decision is made by STM. So if anything it looks like STM are lacking in ambition and not Imagination Technologies.

You have to learn to separate the difference between IMG-Tech being "done" with a design, and a final "chip" being "finished" by IMG's partner. If Series4 was "done" in the sense that matters, it would be on the shelf. It doesn't matter if IMG-Tech "speced" out deisgns up to PowerVR 10. When the partner takes these specs and puts them into silicon, thats "make it or break it" time.

This all relates back to the licensing model that PowerVR has. The model has its advantages, but IMO, in the PC Space, there seems to be too great a disadvantage to make it viable. I think the "disconnect" between the engineers doing the tech R&amp;D, and the engineers designing the chips, ends up creating too much overhead.

It's also possible that the tech is simply too complicated to be expected to be put on the same sort of 6-9 month cycle of traditional renderers.

 

[Joe DeFuria]

Or it'll far surpass the GF4MX in performance and be quite a bit cheaper (because of the smaller process).

Actually, my original point was, that by the time KyroIII would hit the market at 0.13...what's to stop nVidia from moving their GF-4 MX to 0.13 as well...reducing their own costs (and possibly upping the performance a bit at the same time)? nVidia, does in fact have a history of migrating their parts (Including the GeForce2 MX line) to reap the benefits of economies of scale.

 

[Teasy]

To be clear, he didn't say that PowerVR Technologies development team is slow. He said PowerVR Development is slow.

Ok, I thought by PowerVR Development he meant the development of the actual technology (which is what PowerVR is) and not the chips themselves, which is a easy assumption to make.

I think you have some personal connection to IMG-Tech, and try to defend them at all costs...even when they aren't being directly attacked.

AFAIK (thats knew not know) they were being directly attacked as I explained above, I do no defend anyt attack, I defend against an attack when I think someone is wrong in there attack.

I think his point "PowerVR Development is slow" is valid. How many PowerVR PC chips have shipped in the past 3 years, relative to the competition?

PowerVR is the technology not the chip based on that technology, and its development is no inherently slow. The fact that within the last 3 years only 3 PowerVR graphics cards have been released in the PC space (4 with 4800) does not necasarilly point to slow developement. How many cards are released and when there released rely's on whichever company is using the PowerVR tech to make a product.

You have to learn to separate the difference between IMG-Tech being "done" with a design, and a final "chip" being "finished" by IMG's partner

Actually you have to learn the difference between PowerVR (the technology designed by IMGTEC) and the card made by IMGTEC's partner thats based on that PowerVR tech. PowerVR development to me means the development of PowerVR (which is the design itself). My point in that comment was that IMGTEC are not slow at developing PowerVR tech, the release of products is up to NEC, STM or whichever company is making the product based on PowerVR tech.

If Series4 was "done" in the sense that matters, it would be on the shelf. It doesn't matter if IMG-Tech "speced" out deisgns up to PowerVR 10. When the partner takes these specs and puts them into silicon, thats "make it or break it" time.

Thats all very true Joe but its irrelivent to my comment :smile:

<font size=-1>[ This Message was edited by: Teasy on 2002-02-08 21:03 ]</font>

 

[Johnny Rotten]

I don't know but what's to stop them putting Kyro III on a process they already have access too in order to get all the bugs out and work on drivers (probably running it at a lower clock speed) and then move to 0.13 and bump up the clock speed?.. couldn't this reduce the time between getting 0.13 micron and getting the product out of the door?

Somewhat. But I think you underestimate the impact that a given process has on a given chip designs fabrication.

Why would that stun you?.. do you know how far a long the Kyro III is right now? How do you know its not almost finished? What if as I said they already had the chip running on a bigger process before 0.13 micron was available and so got a good head start on debuging and drivers. As I said I don't know if STM have had Kyro III on 0.13 yet and neither do you.. its very possible.

Why would it take the same team of technicians under new management (so possibly they'd even add staff to speed this up) 7 months or more to finish the chip if they already have the chip either on a bigger process and ready to move to 0.13 or already on 0.13?

Again, I just think you're underestimating the total time investment required for a large scale transition like this.

 

[Teasy]

Again, I just think you're underestimating the total time investment required for a large scale transition like this.

Maybe, but I hope not :smile:

 

[Joe DeFuria]

PowerVR is the technology not the chip based on that technology, and its development is no inherently slow.

The bottom line (what the original poster was getting at), is this:

The development and subsequent release of products based on PowerVR Technology is slow, relative to the rest of the industry.

I hope that satisfies and is agreeable to everyone.

Nobody cares how slow or fast "PowerVR Technology" is developed. What matters is how slow/fast products based on that technology can be brought to market.

How many cards are released and when there released rely's on whichever company is using the PowerVR tech to make a product.

I disagree. It relies on a combination of both IMG-Tech and the chip-maker. If IMG-Tech supplies them with a "design" that effectively requires 0.13 microns to implement, but 0.13 isn't available, then that's not enitrely in the hands of the chip maker.

My point in that comment was that IMGTEC are not slow at developing PowerVR tech, the release of products is up to NEC, STM or whichever company is making the product based on PowerVR tech.

My point is that nobody cares how "slow or fast" IMG are at "developing" IMG tech. What matters is how long it takes to get finished products to market. Now, if STM can't fab the new chips that IMG always has "waiting for them" because IMG is so fast, then what's the problem?

It could be a few things:
1) IMGs "designs" as they are "handed" to STM require a significant feed-back and tweaking and further engineering before they are brought to the level of "production ready design."
2) STM sucks at implementaion.

I'm willing to go with number 1. Basically, I don't think IMG just "hands-over" a design and says "here you go...call us when you want the PowerVR 5 design." It's likely a continuous relationship, so that the "actual" IMG "design" isn't really ever finished until the chip is in production.

Thats all very true Joe but its irrelivent to my comment

Heh...we can go on and on here, because I think your comment was irrelevant to the original point of "PowerVR Development" being slow. I don't think he meant "IMG's role in PowerVR technology development." I'm pretty sure he meant, "how long it takes to get new products out the door", which is what is important / relevant.

(But then, maybe zborgered will qualify what he meant...)

Anyway, there has to be SOME reason why there have only been 3 chips in the last 3 years....IMO, It's either the tech is too difficult to engineer more rapidly, or the business model that IMG-Tech uses doesn't work in the fast-paced PC market. (Or a combination of both.) The only other alternative is that the engineers (be they PowerVR or STM) are sub-par in terms of execution.

 

[deviantchild]

i can't go through all the posts here to see if somebody has already said this

i saw earlier in the thread some ppl saying along the lines of "how come the graphics div. were dropped if STM were doing well? - was it in some way down to IMG/PVR themselves"

STM weren't doing well - being so big, they were hit more than most by the general slump in the chip industry last year:- http://www.pvr-net.com/cgibin/forum/ikonboard.cgi?s=3c642d2d4597ffff;act=ST;f=2;t=71;hl=bbc+gmt

it was a just matter of time before they started restructuring - and PVR/IMG's business was smallfry and got caught up in it

edit: there is a link on that page to the story - but you can only get to it now if you are a cbs marketwatch member

<font size=-1>[ This Message was edited by: deviantchild on 2002-02-08 21:58 ]</font>

 

[Teasy]

The development and subsequent release of products based on PowerVR Technology is slow, relative to the rest of the industry..

If the original poster had said something like that then I wouldn't have argued. But he didn't, he said "PowerVR Development is slow", you took it that he meant "The development and subsequent release of products based on PowerVR Technology is slow, relative to the rest of the industry." And I took it that he was blaming the speed of the development of the technology that is PowerVR for the lack of cards, thats why I asked why he was assuming that any slowness in releasing boards was down to the speed of the design time. If your correct in your reading of what he said, then fine I wouldn't have argued with that.

I disagree. It relies on a combination of both IMG-Tech and the chip-maker. If IMG-Tech supplies them with a "design" that effectively requires 0.13 microns to implement, but 0.13 isn't available, then that's not enitrely in the hands of the chip maker.

STM decide what process the design goes onto, what performance it will have and what cost they want it to end up at. The design is unlikely to be anything over 25 million transistors (assuming it doesn't have DX8 pixel/vertex shaders, which is a pretty safe assumption) so I can't see it needing 0.13 micron to actually make the chip. It may just need 0.13 micron to be at the pricepoint STM want it at.

My point is that nobody cares how "slow or fast" IMG are at "developing" IMG tech. What matters is how long it takes to get finished products to market.

And as I've said I don't disgree with you there but thats irrelivent to the comment your replying to. Because my comment was based on the assumption (a pretty fair assumption considering the wording) that the poster was blaming the speed of design on a lack of cards. If he wasn't saying that then I do not disagree with him.

Now, if STM can't fab the new chips that IMG always has "waiting for them" because IMG is so fast, then what's the problem?

Its not necasarilly that they can't keep up, its more that they don't want too IMO.

1) IMGs "designs" as they are "handed" to STM require a significant feed-back and tweaking and further engineering before they are brought to the level of "production ready design."
2) STM sucks at implementaion.

Add a third: ST don't want to put a high priority and lots of technicians on getting PowerVR designs out ASAP, they just want a nice little card that can sell quite well but without actually having to put much money or effort in (which they would have to do at the highend).

Heh...we can go on and on here, because I think your comment was irrelevant to the original point of "PowerVR Development" being slow.

That depends on what the guy actually meant, if he meant what you think he meant then my comment was irrelivent, but then if I'd known he was saying what you think he was saying I wouldn't have argued with him :smile:

Anyway, there has to be SOME reason why there have only been 3 chips in the last 3 years....IMO, It's either the tech is too difficult to engineer more rapidly, or the business model that IMG-Tech uses doesn't work in the fast-paced PC market. (Or a combination of both.) The only other alternative is that the engineers (be they PowerVR or STM) are sub-par in terms of execution.

Your assuming that everyone wants to hit the highend, and if they don't then either there tech isn't good enough or they can't make it fast enough. IMO STM wanted a nice little product that didn't cost them much to make and could sell well in the high volume market and be a nice little eanrer. If that had happened then they probably would have continued with the same thing, a nice little product thats cheap and makes them some extra cash.

They got exactly that with PowerVR 3 IMO. The problem was STM misjudged the market, they brought out Kyro at 115mhz and it wasn't fast enough. By the time they'd moved to Kyro II they did well but nowhere near as well as if they'd just made the original as fast as Kyro II. In other words instead of taking PowerVR Series 3 and using there own fabs at 0.25 micron and getting a 115mhz Kyro they should have went straight to TSMC and 0.18 and released it at 175mhz back when Kyro was originally released. If they'd done that they would have gotten the success they wanted (more like 3+ million sold rather then 1.2-1.3 million sold).

So IMO this is all more about the ambition of STM and what the company wanted to achieve in the graphics card market rather then the actual ability of the engeneers at either company or the quality of the technology itself.

<font size=-1>[ This Message was edited by: Teasy on 2002-02-08 22:19 ]</font>

 

[Hellbinder]

people can debate the nit picky defenitions of slow development all day long. The Bottom line will ALWAYS be products to market. I can draw up pretty chip designs that smoke everyone all day long for the next 10 years. Guess what. It means exactly ZERO. Nadda, Nothing.

The truth is Since the Neon 250 design IT has offered nothing even close to turning the heads of the majority of the community. Had IT been able to launch the Neon 250 on time the entire Graphics market would look very different today. That card would have been the performance and features leader in a landslide. Setting up IT/PowerVR as the Go-to market leader. Nvidia may have never even gotten off the ground. IT would be a development cycle ahead of everyone else, and the sky is teh limit. Instead the Emphasis was placed on SEGA. A deal that became fruitless in only 24 months.

Now as Far as execution to market is concerned IT is far behind the eightball. On top of that they need a Via, Sony, or other big boy to pick up the ball and score a touchdown. What they need is for someone with oodles and oodles of resources who WANTS to dominate the graphics market ro come in and partner with them.

 

[dksuiko]

As disappointing as this may be, was it really a shock to anybody? It certainly wasn't to me.

STM has not taken any big steps since Early-Mid 2000 with the first Kyro. Unless you've bought into the Kyro III rumors, there was nothing to lead us to believe that it was actually coming. They weren't exactly known for rapid releases. The fact that STM never aimed at the high-end market didn't help matters either. Nor did the Kyro 4800.

We can talk for days about how ImgTec's technology was ready and done (Not that I even believe that one), but none of that matters when they aren't going to come to the market. Other than those who have to place their penis in a hotdog bun, who was really surprised by this announcement?

-dksuiko

 

[Sonic]

What's the possibility of NEC itself retaking its former position and manufacturing the PVR chips again? I'm sure they have the capabilities to do it, IMG just needs to get the ball rolling again.

Sonic

 

[Joe DeFuria]

IMO STM wanted a nice little product that didn't cost them much to make and could sell well in the high volume market and be a nice little eanrer.

If we run with your theory, then the question needs to be asked, why? If the PowerVR technology is in fact such an 'advantage' compared to everyone else's tech why would any company want a "nice little product", then they should be able to dominate the industry, or at least make a large enough dent to reap considerable profits.

I've never known a company to just want to make a "little cash" if the product has inherently that much potential.

They got exactly that with PowerVR 3 IMO. The problem was STM misjudged the market, they brought out Kyro at 115mhz and it wasn't fast enough. By the time they'd moved to Kyro II they did well but nowhere near as well as if they'd just made the original as fast as Kyro II.

Well, you talk as if STM had a choice. That given a price target and timing target, STM could have just "brought out" the KyroI with Kyro II specs.

Personally, I think it was less "underestimation" of the market, and more "lack of execution." It's not that they should have "designed KyroI to have Kyro II specs", they should have been able to get both Kyro I and KyroII out earlier.

So IMO this is all more about the ambition of STM and what the company wanted to achieve in the graphics card market rather then the actual ability of the engeneers at either company or the quality of the technology itself.

Well, IMO you can't separate the two. In other words, IF the quality of the technology was really a cut above, then STM would have more incentive to make more of the product line, and be "more ambitious" with it. So, IMO, the reason for "lack of ambition" is because of something lacking in PowerVR (either the Technology itself, or the business structure around it), relative to the rest of the market.