ShaderX 2 Contents - Clues to PowerVR Series 5?

Discussion in 'Beyond3D News' started by Dave Baumann, May 27, 2003.

  1. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,420
    Likes Received:
    179
    Location:
    Chania
    Depends how you count persentages and what they actually represent. I'd agree with the 85% if it would represent a working presenatable sample. I can only guestimate that they're not yet ready with it, otherwise it would be really stupid to not present it, after all they're selling IP and the way I understand it they aim for multiple markets (such as PC graphics, arcade, consoles etc.).


    I'm not so sure if there haven't been hints for Doom3 or more specifically stencil ops already; can't think of many reasons why they sat down and created FableMark, albeit released a bit too early and having only a K2 to run it w/o cubes and dot3.

    If the public wants to hear that all next generation products are most likely going to be at least twice as fast in Doom3 than current accelerators, I personally would like to believe it already, otherwise they can pack up and go home (all of them) :D


    I'm not in a very good shape today, so pardon if I lost you here. I am still under the impression that S5 is going to be on 13nm.


    Don't know. While I do understand your general reasoning about any pre-hype or according marketing strategies I prefer NOT to see ventures as conducted by Trident for it's XP4 (as I said before). I really don't like to see lucky guestimates or empty promises that are endangered to never be fullfilled in the end. Heck I don't digest very well with the term HYPE in the first place to be honest, but that's just me.
     
  2. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
    Of course the number was grabbed out of thin air. I only meant to emphasize that you could let go of preliminary info before you have every detail carved in stone. It's not until you something like a pre-launch that you really have committed yourself and put goodwill and reputation at risk.

    Since they previously have been content to sell their designs, it makes sense (and is probably a contractual necessity) to let the liscensee take care of the marketing through various channels.

    If, however, they are aiming to bring the part to market themselves, the situation is very different, and it's a situation they are not as familiar with.

    Of course, we (or I) only have indications that they may be prepared to produce parts themselves. We have nothing official from ImgTech at all in that respect.

    Well, it was a statement that was forward-looking. If you don't have product this fall, then it follows that the earliest you can get to market would be during the first part of 2004. And that's when the migration from 0.13 to finer geometries is supposed to be starting. Qualifications are already done at all fabs I know, although actual production seems to be Intel only during this year.

    Introducing a new part on a lithographic process that is trailing edge has some benefits of course, but it also ensures that the competition will bring solutions to market that has advantages relative to yours from better processes in terms of working frequencies/power consumption/transistor counts and so on.

    To me, it seems like suboptimal to enter the market in 2004 on 0.13 um if you
    a, want to be competitive from a performance standpoint.
    b, have limited resources to spend redesigning/testing/.... for a new process


    Oh well.
    We'll see.
    Or we won't.
    I really want them to see them come out with something good though, it's frustrating to know there's a talented group of people producing interesting designs that just gather dust in filing cabinets. And it has to be a LOT more frustrating for the engineers involved - sure, the work pays the bills but in the long run....

    Entropy
     
  3. CorwinB

    Regular

    Joined:
    May 27, 2003
    Messages:
    274
    Likes Received:
    0
    Well, I think the R300 proved that you could have an incredibly performing part without being on the most advanced process...

    While PR-lines like "produced on the most advanced process currently available" may sound good, they don't replace a good design (*cough* NV30 *cough*). And, although their parts typically never were targetted for high-end, PowerVR made some excellent designs already (I'm still very fond of the Dreamcast)...

    In the end, what matters to end-users is performance, reliability and IQ. Not manufacturing process, unless the process starts impacting reliability (too much heat, for example). And, here again, being on an advance process is no guarantee of good reliability (dustbuster, anyone ?).
     
  4. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    Entropy,

    I'm not so sure that 90nm is very close at all to tell you the truth.

    http://www.digitimes.com/NewsShow/Article.asp?datePublish=2003/05/30&pages=PR&seq=205



     
  5. Entropy

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,059
    Likes Received:
    1,021
    Hard to say just what timescales they are talking about in that Digitimes quote. It may be that actual production volumes in 0.09um is (significantly) farther off than the fabs have indicated. If so, waiting for it to be viable may indeed make the design go feature-obsolete. Hard to believe that with 3.0 shaders, but still. At this point there is simply too many unknowns for non-insiders like me to make more than rather sweeping hypothesizing.

    Of course, if they've intended to release it on 0.13um the question has to be - why isn't it here? Just what are they waiting for in that case? Both fab capability and DX9 has been around for some time - and there is still no sign either of a licensee or that Img will push to silicon on their own, meaning it will take a fair amount of time from now to get to the market even in a best case scenario.

    Hopefully we will see a Series5 on the market, and we can judge the tradeoffs at that time. But so far, the picture looks bleaker than for the Series4. At least, at that time they had a licensee.

    Damn. And I've so been looking forward to seeing Joe DeFuria eat crow.

    Entropy
     
  6. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,420
    Likes Received:
    179
    Location:
    Chania
    Entropy,

    Wavey saved me some extra time to look up for links concerning 0.09nm.

    That's the reasoning most of us follow; trouble is that their IP/lisencing scheme is not optimal for a cut-throat market like PC graphics. Of course is nothing impossible but it would take a very conscious partner that isn't in a hurry to cash quickly in, but is aware that success in that specific market comes only over time and long lasting and continious efforts. Not to speak all the necessary resources that have to be aligned at all times from all involved parties.

    The market in general is not in it's best shape, meaning even big companies have second thoughts about bigger and long-term investments and while PVR along with ST Micro/Hercules managed to start getting some interest in the budget/mainstream market with the KYROII, I'm afraid most of it has been somewhat lost, since the continuance was broken with the canned/missing Series4 products. I see them starting almost from scratch again, which is as hard as it can be for anyone else plus all the possible milestones of having to channel production/marketing/distribution through third parties.

    I'm not as pessimistic as it may sound, but these are well known considerations to most of us. Assume they really have an extremely capable design this time around, they still need to have a very committed and accurate policy and a buttload of luck too.

    I'm very confident though that when they feel the time is right and products roll out IN TIME, that there isn't much headroom to be very modest with a high end design. The independant tests will hopefully tell the true story as it is, wether it's a good or bad one of course - as you said - remains to be seen.
     
  7. Joe DeFuria

    Legend

    Joined:
    Feb 6, 2002
    Messages:
    5,994
    Likes Received:
    70
    What, seeing Teasy eat (unmentionables) wasn't entertaining enough? ;)

    Seriously, I would like nothing better than to have a third competitor in the mid and especially high-end space.
     
  8. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,420
    Likes Received:
    179
    Location:
    Chania
    *ahem* even when ST Micro abandoned the PC graphics market, there were still two interest parties for Series4. I believe the reason why it finally ended up rotting on a shelve is that it's features where a bit too little for industry standards and it was way too late for it.

    Hey now leave Joe alone; I like the guy. Did he promise to eat crow if S5 get's released? Boy oh boy since advanced 3D discussions have become a rarity at B3D, at least the funny bets some people come up with (either concerning essential body parts or eating yucky animals) at least keeps the interest alive.

    Is that crow going to be served with garlic or without? :shock:
     
  9. Entropy

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,059
    Likes Received:
    1,021
    Heh. No, it's just that his arguments re:tilers has hinged on the absence of a high-performance tiler on the market.
    Actually my statement is a testament to the high regard I hold his (usually) lucid and honest stance on these forums. I'm sure he would recognize the sitiuation and prepare and consume said crow on his own accord. And post links to the DivX. :)

    Entropy
     
  10. Dave B(TotalVR)

    Regular

    Joined:
    Feb 6, 2002
    Messages:
    491
    Likes Received:
    3
    Location:
    Essex, UK (not far from IMGTEC:)
    Im pretty sure Series 4 never made it to market not because they couldn't build it, but STM wouldn't let them beause they owned the IP on many 'to silicon' design techniques used in the chip and any other manufacturer would have to pay royalties. I think.....
     
  11. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    Why do you suspect they need a licensee this time around?
     
  12. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
    Heh, you must have dreamt that Joe, my girlfriend assures me its still there.. its pretty hard to miss infact :lol:
     
  13. Teasy

    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    4,563
    Likes Received:
    14
    Location:
    Newcastle
    Hmm why are guests now allowed in this forum? I forgot to input my username and pass above and instead of asking for my username and pass it just posted as a guest.. how annoying. Who wants annonymous posts here anyway??
     
  14. Joe DeFuria

    Legend

    Joined:
    Feb 6, 2002
    Messages:
    5,994
    Likes Received:
    70
    Aren't we a bit elitist? ;)

    Anonymous posts are allowed in the News forum only. This makes perfect sense to me. I hate OTHER message boards that would force me to register just because I want to make a comment on a news story they are reporting on.
     
  15. Aivansama

    Newcomer

    Joined:
    May 31, 2003
    Messages:
    39
    Likes Received:
    0
    Location:
    Finland
    Regarding 90nm tech, it seems that Xilinx has FPGA chips (Spartan 3) on 90nm up and running. I've heard quoted prices that sound like they're doing ok on that process. Granted, FPGA chips are usually "easy" chips that take the process right after memory chips, but still, 90nm is definitely within possibilities, IMHO. Just my 0.02 EUR.
     
  16. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,420
    Likes Received:
    179
    Location:
    Chania
    Hmmm second hint.....still sounds like a very very limited distribution to me. Ideal would be at least one big vendor with fairly high penetration in the US market (besides the european or asian markets).
     
  17. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
    Yup. I'll quote:
    "The Spartan-3 family is based on IBM and UMC advanced 90nm, 8-layer metal process technology. Xilinx is using 90nm technology to drive pricing down to under $20* for a one-million-gate FPGA"

    Although, as you point out, the chip structure is highly regular.
    I seem to remember that NEC (a dark horse in these matters) is pretty far ahead as well. TSMC is closely followed in the industry of course. IBM is on record stating that they will push East Fishkill into 90nm very agressively.

    On a positive note for a Series5, we are approaching a year since the introduction of the 9700, and so far we have seen only very modest increases in performance. No major performance increases are on the immediate horizon either. Not to mention that the DX9 midrange offerings (5600+9600) are pretty wimpy at their $200 pricepoint.

    If it has 3.0 shaders, the Series5 design should have some longevity in the market, although frankly I think even basic DX9 compliance would suffice. DX10 penetration is a loooong way off.

    And as far as chip fabbing goes, the situation is quite favourable at the moment for the customers. Some remarkably attractive offers available, from the perspective of this armchair.

    So it would appear that there is a window of opportunity.
    Lets hope someone grabs it. And if the part really is strong, being competitive in the high profile gfx market can hardly hurt ImgTech tech/brand recognition.

    Entropy
     
  18. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
     
  19. LeStoffer

    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    1,253
    Likes Received:
    13
    Location:
    Land of the 25% VAT
    Ahh, the good old deferred rendering is coming back and already I have tears in my eyes... :eek:

    I am, however, wondering whether its main virtue will mean much in terms of performance if game engines really start to do the two pass rendering technique: During the first pass the geometry is rendered with outputting only depth values but no color information. During the second rendering pass you should be able to render everything in the front to back order and thus the early pixel rejection of GeForce and Radeon can work fairly efficient.

    Since both ATI and nVidia recommend this technique in apps with a high overdraw, well, PowerVR stands to loose some of it advantage. Just something to consider... :p
     
  20. Ailuros

    Ailuros Epsilon plus three
    Legend Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    9,420
    Likes Received:
    179
    Location:
    Chania
    Yep for what though mainly? DOOM3? If you mean that one then I can only say stencil fillrates.

    IMR: 450MHz * 8 Z/stencil operations per clock = 3.6GPixels stencil fillrate (shared)

    TBDR: (say) 400MHz * (say) 128 Z/stencil operations per clock = :?: :idea: :!: (not shared)
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...