PVR5 to be unveiled on the 18th!

[dksuiko]

Anyway, the fact that Unreal had a dedicated PowerVR path means nothing when it comes to overall high-end performance. How long have you been following PowerVR? Do you not know that they never released a high-end performance part?

He just told you, they released a high end part in the early days of 3D graphics cards.

He can speak for himself. Not that he even said that, anyway. Oh well, whatever. In any case, it seems he has come to agree with me - given that he has not replied. Perhaps you can fill us both in on this mysterious high-end part in the "early days of 3D graphic cards"? And no, though it was decent and interesting, the Matrox M3D wasn't high-end.

Seems he knows a lot more about this then you do.

So?! My dad can beat up your dad!

 

[amk]

He just told you, they released a high end part in the early days of 3D graphics cards.

He can speak for himself. Not that he even said that, anyway. Oh well, whatever. In any case, it seems he has come to agree with me - given that he has not replied. Perhaps you can fill us both in on this mysterious high-end part in the "early days of 3D graphic cards"? And no, though it was decent and interesting, the Matrox M3D wasn't high-end.

IIRC:-

Series 1, PCX1 and 2 which included Matrox M3D, lacked a triangle setup engine, and so were highly CPU dependant. If you had a high end CPU then PCX2 was as good a solution as any available at the time. If you didn't have a high end CPU, then it wasn't.

There's a quote somewhere of Carmack saying it was the best card for glQuake.

Whether that fulfills any definition of "high end" is another matter.

 

[Sage]

He can speak for himself. Not that he even said that, anyway. Oh well, whatever. In any case, it seems he has come to agree with me - given that he has not replied.

no i just don't feel it worth my time arguing with someone who doesnt even know that PVR used to be high end.

 

[dksuiko]

He can speak for himself. Not that he even said that, anyway. Oh well, whatever. In any case, it seems he has come to agree with me - given that he has not replied.

no i just don't feel it worth my time arguing with someone who doesnt even know that PVR used to be high end.

It seems you are content in simply chasing the red herring that I'm some sort of 3d-graphics noobie rather than actually discussing the topic. I'm no self-declared 3d graphics sage, but hey... I remember when this site's color scheme was black, orange, and blue (or was it purple?), when perspective correction was an advertisable graphics feature, and when all you needed to play a game were the proper VESA drivers. Not that any of those make my post any more credible, of course. In any case, my position still stands: PowerVR/TBDR has never dominated in high-end performance.

Maybe you forgot or weren't around when those chipsets were actually released and reviewed, so I'll give you a summary. "This is a good card, but..." And there we go, I summarized the whole story. Now don't get me wrong, I'm not bashing ImgTec at all or have anything against TBDR. It's just that the fact of the matter is, they've never dominated the high-end. The reason you bought the card was because you couldn't afford the real high-end cards or for some reason wanted to be different. (Or played Ultimate Race at local comp shop's demo computer w/ a PowerVR card in it and immediately bought it... which I almost did)

You can confirm this yourself, google some reviews.

 

[Sage]

dominating the high end and putting out a high end part are two completely different things. your original post said that they never put out a high end part

 

[dksuiko]

dominating the high end and putting out a high end part are two completely different things. your original post said that they never put out a high end part

And I still say that they never put out a high-end part. A high-end card is a card that has high-end performance. Apocalypse 3Dx was not high-end. Matrox M3D was not high-end. The Kyro series of cards were not high-end. None of them were. I thought it would go without saying here that when one says high-end, they are reffering to performance.

 

[_xxx_]

Sure it was high-end. What else was there back then? Voodoo wasn't really there yet AFAICR. It certainly was VERY high-end compared to ATI's Mach64 or those Trident or S3 cards with 1MB RAM etc. So what's your point? What card was better back then?

 

[Putas]

If its not about price but performance, you can take KyroII too, because GF3 was very unavalaible. PCX2 made it to market half year before Voodoo2, therefore its clearly highend (coupled with highend CPU). I'm curious about rumours of slow performance of m3d, hadn't it slower clocks then normal PCX2?

 

[Simon F]

dominating the high end and putting out a high end part are two completely different things. your original post said that they never put out a high end part

And I still say that they never put out a high-end part. A high-end card is a card that has high-end performance. Apocalypse 3Dx was not high-end.

Not high end compared to what at the time?

IIRC, the voodoo 1 had a peak fill rate of 40+Mpix/sec (please correct me if I'm wrong).

Again, stretching my memory, the PCX2 had a peak texture fill rate of 1 pixel per clock and was running at 60ish Mhz, but sometimes had to do two fetches to the memory, so let's be conservative and say it's achieving 2/3rds of that.

If you then allow for the deferred rendering to remove hidden surface and the fact that the overdraw is at least 2x (there is a background to clear) then I think it's effectibe fill rate was pretty respectable.

For example, Tomb Raider, coded to the native API, ran at 1024x768 @ 30+Hz on my 133. I don't think anything in the PC space matched that at the time.

 

[dksuiko]

And I still say that they never put out a high-end part. A high-end card is a card that has high-end performance. Apocalypse 3Dx was not high-end.

Not high end compared to what at the time?

I'd say compared to Voodoo, but that's only a valid comparison depending on whether or not (I wasn't) you take into consideration Voodoo came a few months later, as putas and xxx pointed out. I suppose for that period, it was the high-end along with Verite, though with Voodoo just around the corner with bilinear filtering and speed.

IIRC, the voodoo 1 had a peak fill rate of 40+Mpix/sec (please correct me if I'm wrong).

Again, stretching my memory, the PCX2 had a peak texture fill rate of 1 pixel per clock and was running at 60ish Mhz, but sometimes had to do two fetches to the memory, so let's be conservative and say it's achieving 2/3rds of that.

If you then allow for the deferred rendering to remove hidden surface and the fact that the overdraw is at least 2x (there is a background to clear) then I think it's effectibe fill rate was pretty respectable.

For example, Tomb Raider, coded to the native API, ran at 1024x768 @ 30+Hz on my 133. I don't anything in the PC space matched that at the time.

I don't remember any benchmark numbers, but like with PCX1 vs Voodoo, wasn't the Voodoo2 considerably faster than PCX2? Given that Vodooo2 couldn't do 1024x768 w/o SLI, I guess you have to give credit PCX2. But I see your point, I concede that it was a bit of an overstatement to declare PCX1/2 as not high-end. It's just that it's been a bit tiresome to hear all these promises about the potential of TBDR only to see minimal ass-kicking. PCX1 had no bilinear filtering, PCX2 still required a fast CPU. Neon250 was late. Kyros lacked... ambition, Gigapixel never materialized anything. Something always came up.

 

[Ailuros]

Something always came up.

I don't think anyone can deny that. Neon250 is probably still holding the record of the biggest delay ever in the PC space, Series3/KYRO wasn't on time too, ST decided to spin the initially planned Series4@150nm on 130nm, left the building a while later, Series4 was shelved entirely and I'm afraid - since it's been too long by now - a similar fate awaits Series5 too for the PC space.

In October 2002 a newsblurb at EEtimes quoted Metcalfe's claim that a 130nm design is going to surface within 2003. It's 2005 by now...

Even if a company aims for high end when starting out development, delays will always shift the final product down to mainstream or even lower according to the analogy of the delay.

If IMG/PowerVR nowadays wouldn't have more than a dozen of separate IP designs under works and would exclusively concentrate on a PC design, I believe that there might have been chances for only minor delays in a worst case scenario. I myself underestimated obviously all the bulk of resources that have been required for the other designs.

Additionaly I still believe that they wouldn't gain any significant profit (if any at all) from a high end PC release either. Au contraire the margins from PowerVR/Ensigma/Metagence IP for other devices seem to be times higher. By now I figure they're already working on next generation IP like "MBX2", "META2" or whatever else they'll be called. They'd win impressions that I can believe and it probably would have helped to sell even more of the other IP designs. Kind of a very expensive advertisement.

Intel should IMHO open it's eyes a bit wider over the 2700G they're selling and consider licensing IP for their next generation integrated graphics sollutions. A slightly cut down version of Series5 in that case would be scary for that market segment at least.

 

[Mariner]

Intel should IMHO open it's eyes a bit wider over the 2700G they're selling and consider licensing IP for their next generation integrated graphics sollutions. A slightly cut down version of Series5 in that case would be scary for that market segment at least.

My thoughts entirely, especially once Longhorn arrives with it's more strenuous requirements on the graphics side of things.

Oh well, there is always hope...

 

[KimB]

It seems obvious to me that PowerVR has decided that the PC space is simply too competitive for them to invest in at the moment. So, they are focusing on side markets in an attempt to hold marketshare where the competition is not so fierce.

In particular, I just do not see the technology licensing business model of PowerVR as being effective in the PC space: this business model certainly lengthens development times, and as such it is unlikely that they could even compete in the PC space, given the rapid release schedules in the market.

 

[_xxx_]

I don't remember any benchmark numbers, but like with PCX1 vs Voodoo, wasn't the Voodoo2 considerably faster than PCX2? Given that Vodooo2 couldn't do 1024x768 w/o SLI, I guess you have to give credit PCX2. But I see your point, I concede that it was a bit of an overstatement to declare PCX1/2 as not high-end. It's just that it's been a bit tiresome to hear all these promises about the potential of TBDR only to see minimal ass-kicking. PCX1 had no bilinear filtering, PCX2 still required a fast CPU. Neon250 was late. Kyros lacked... ambition, Gigapixel never materialized anything. Something always came up.

You also have to consider that PCX2 was a pita to program for and that Glide was an amazing API back then. Coding for Voodoo was child's play in comparison. 3dfx also were very agressive in supporting devs, while PVR didn't try nearly as hard.

If coded properly with PCX2 in mind from the begining (unlike Unreal, where it was half-heartedly done), any game ran just as fast as with Voodoo1. Noting that there was nothing else out there but these two, it clearly was high-end.

Whatever. Who cares?

 

[Simon F]

You also have to consider that PCX2 was a pita to program for and that Glide was an amazing API back then.

Errr... there was very little difference between the low-level native PowerVR library and Glide.

 

[_xxx_]

Well that's what I heard from quite a few people, I programmed for neither of them myself

 

[Megadrive1988]

32 sieres 6 chips together in one machine

naw, I'm waiting for a workstation with 512 PowerVR Series 7s in it

 

[Ailuros]

It seems obvious to me that PowerVR has decided that the PC space is simply too competitive for them to invest in at the moment. So, they are focusing on side markets in an attempt to hold marketshare where the competition is not so fierce.

Not so fierce yet. It's natural that if all of our assumptions are true, that it's of the companys better interest to focus on markets they are successful right now.

In particular, I just do not see the technology licensing business model of PowerVR as being effective in the PC space: this business model certainly lengthens development times, and as such it is unlikely that they could even compete in the PC space, given the rapid release schedules in the market.

Do you have the strategy in mind they had back in the ST Micro days? When an IP company just delivers IP, there's a middleman for transposing it into silicon and the latter then delivers to vendors for board production then chances are extremely high that it'll take a lot longer than usual. I doubt this has been the case though past ST Micro and I would dare to speculate that whatever decision has been taken on it's fate happened around spring 2004.

 

[Lazy8s]

If the hardware became ready for SEGA to start focused game development at the end of last year or early this year, it could still be several months before the arcade games' development cycles are far enough along to show anything.

 

[TEXAN]

02.07.2004
Renamed Next Generation Hardware to Naomi 3 Hardware, just because its using an as yet unknown Imagination Technologies Power VR type setup and it sounds better until it gets a final name, some sort of sword again no doubt
Apparently the first models have already been shipped out to Sega for testing/development purposes.

http://www.system16.com/archive/oldnews04.html

2/7/04[UK] = 7/2/04[US]

So game development begun in June/July of last year.