Announcement

Collapse
No announcement yet.

VigilAnt Rant and Some Observations

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • VigilAnt Rant and Some Observations

    Please discuss the editorial with the above title found on www.murc.ws in this thread.

    Matrox Graphics. How do I take them as an entity now? How should I? At one time they were the best there was. Their legend began with Millenium. At that time, they had about as many employees as they do now (since all the layoffs and resignations have come down). They were focused and they were effective. They concentrated on one thing and they did it like not anyone else could. They weren't trying to make markets where there were no markets. They simply produced the highest performance card at any price and had the most features. It was not novelty card. It was not a made to be a mass-market card. It was meant to kick arse and it did. This product allowed Matrox to grow very large. They won OEM contracts. Everybody who was a geek wanted a Millenium. Some time after all this, Matrox Graphics found they could make a lot of money producing old hat to fit OEMs. Each successive generation of card lost more of that shine that made the Millenium what it was. The focus on the brute performance and the quality that made it so appealing was eroding. Matrox Graphics fell into the rut. The G100 came and flopped, cost big bucks in the numerous revisions they had to make to get it right. G200 came and was a disappointment. It could have been good. But the focus was no longer to put all effort into kicking arse and stomping pcbs. G400 was a huge step over the G200, but was still plagued with problems internally. The card should have been 50% faster than it was. The card should have had better memory performance. The card should have had a few more features. But in the game of trying to be too many things for too many people, it lost some of its shine. Even though it was off target for what the people at Matrox wanted, the public wanted it. But Matrox could not keep up with demand because of poor planning and IMHO poor distribution decisions.

    The powers that be thought that going with another high end card would be a mistake. Even though they already had the technology for an entirely different core ready to go, they instead went with a G450. Dual head for the masses was the idea. Not a bad idea, but they forgot one little thing. There has to be a flagship card to push the mass market card along. Having a high performance, top of the line card is very much important, because when people think of a brand name like Matrox, Intel, Nvidia, they usally think of their best model. That is what allows people to think about the lower end models like Celerons and MX's. Not only does keeping a high end model make a good name for the brand, but it also keeps the technology current and allows it to filter down into the low end cards when the time and need comes to improve them. The same is the story with the G800, oops G550.

    Parhelia, the new hope. I had high expectations for this part because it seemed that perhaps Matrox was getting back in form. They went for the throat when penning the chip. If it hadn't been for the eTV, we probably would have seen the card last year, and it would have been recieved much better. Instead, Matrox was playing catch up. Because they were neglecting their high end needs by focusing their resources on the low end market, they lost out. The end result was a decent card that could have been much more than what it is. Parhelia is one great card. I just bought one, it cost me dearly but I have no regrets. It truly is a good card. It just is a shame that they could not see that the card would not fly in the high end market without reaching the people's expectations first.

    I know it may sound like a bitch session, but I would like to offer some advice to Matrox to dig themselves out of the hole they made with the Parhelia mistakes. :

    1. Ditch the idea of the Triple Head for the mass market idea. You guys at Matrox Graphics know what I am talking about. If you want a Triple Head card for the masses, then use the current Parhelia core for it. This would allow you to fix your mistake with the overstock you have now.

    2. Use the current Parhelia core in a multimedia type card. Your PCB already has foot prints for what a lot of videophiles want. Cater to them. Parhelia has a lot features that aren't out there that would entice some good money your way with such a product.

    3. Release the Parhelia 8X core. It will benefit you right now. Put it in the place of the current Parhelia 128 and price it at perhaps 10 percent lower. The fixes and the added performance alone would allow it to outsell the current Parhelia in short order.

    4. Move forward with Pitou. Get it right. You and I both know it can perform, don't go cheap. Listen to your people who know how. They are not stupid and will make your card a winner.

    5. After all this, if you want to make a mass market card, do it in the form of an easily trimmed core, such as the old Parhelia core or perhaps a trimmed Pitou core. But don't invest precious resources into making a whole new revision just to make a cheap card.


    That about concludes today's steam popper, now it's time to take in an afternoon lager and hope that someone at Matrox will take the time to consider at least the first 4 points.

    VigilAnt
    Last edited by VigilAnt; 16 November 2002, 01:14.
    VigilAnt

  • #2
    I agree with many of your points. In my opinion, it seems like there's some leadership issues with Matrox. I'm not really sure if the marketing team is driving engineering to the extent that has been described or not...I'm not really privy to that type of information.

    This is what I know...Just a couple of observations...

    1. N-Patches: The darn thing supports this feature. Why the heck isn't it enabled? Jeez, you would grab a good dozen or so games that support it. I fail to understand why this isn't enabled.

    2. Anisotropic Filtering: This is just inexusable, IMHO. The original GeForce-256 supported the same level/quality as the Parhelia.

    3. Surround Gaming: I will sit here and swear up and down that it's the greatest thing since sliced bread; however, I'm a realist. I bought > $3,000 in LCD's to be able to enjoy it. There are a lot of people who are lucky to have ONE monitor, much less 3. This should be a feature that you can get ON TOP OF a base solution. IE Parhelia+ model. They've got to figure a way to get their products down in the mainstream market.

    4. Price: Aleady mentioned it...likewise, they need to hit several price points. You just can't focus on the $300-400 market. Look at the success of nVidia. They put all that R&D into the highend product...about 2 months after launch, release the budget variations using a combination of memory/clock/PCB differences. ATI has obviously caught on.

    5. Lack of Evangelizing features: I cannot understand why Matrox has not been proactive in showing the world the strengths of their product...You simply cannot just HOPE that websites are going to do this for you, as evidenced by dickheads like Kyle Bennett. Do some white papers! If nobody will do them, do em yourself. nVidia does it all the time, as well as ATI. Promote your own products!

    Comment


    • #3
      Re: VigilAnt Rant and Some Observations

      1. Ditch the idea of the Triple Head for the mass market idea. You guys at Matrox Graphics know what I am talking about. If you want a Triple Head card for the masses, then use the current Parhelia core for it. This would allow you to fix your mistake with the overstock you have now.
      Necessary?
      Triple Head is one selling point of Matrox, I think it is essential to be on higher end cards.


      2. Use the current Parhelia core in a multimedia type card. Your PCB already has foot prints for what a lot of videophiles want. Cater to them. Parhelia has a lot features that aren't out there that would entice some good money your way with such a product.
      eTV? That would be really cool!


      3. Release the Parhelia 8X core. It will benefit you right now. Put it in the place of the current Parhelia 128 and price it at perhaps 10 percent lower. The fixes and the added performance alone would allow it to outsell the current Parhelia in short order.
      Would that make the overstocked current P even more overstocked?


      4. Move forward with Pitou. Get it right. You and I both know it can perform, don't go cheap. Listen to your people who know how. They are not stupid and will make your card a winner.
      No comment. (since I just bought a P!!!)


      5. After all this, if you want to make a mass market card, do it in the form of an easily trimmed core, such as the old Parhelia core or perhaps a trimmed Pitou core. But don't invest precious resources into making a whole new revision just to make a cheap card.
      Very true indeed!
      That's why GeForce4 MX series is such a good seller.
      P4 Northwood 1.8GHz@2.7GHz 1.65V Albatron PX845PEV Pro
      Running two Dell 2005FPW 20" Widescreen LCD
      And of course, Matrox Parhelia | My Matrox histroy: Mill-I, Mill-II, Mystique, G400, Parhelia

      Comment


      • #4
        I agree with VigilAnt at least on release of AGP 8x version core. It fixes quite few problems on original core and should raise yields as also lower manufacturing costs.

        But we must remember that, at least majority us don't know excat numbers how much parhelia development cost and if I have understood right, Matrox sells cards for living, sooo there is always problem of est. sell amount / price -ratio.

        anyways, who lives will see...
        "Dippadai"

        Comment


        • #5
          now if just matrox would listen to you...
          Main Machine: Intel Q6600@3.33, Abit IP-35 E, 4 x Geil 2048MB PC2-6400-CL4, Asus Geforce 8800GTS 512MB@700/2100, 150GB WD Raptor, Highpoint RR2640, 3x Seagate LP 1.5TB (RAID5), NEC-3500 DVD+/-R(W), Antec SLK3700BQE case, BeQuiet! DarkPower Pro 530W

          Comment


          • #6
            message to Matrox : release Pitou summer 2003



            oh and make sure the memory controllers are the fastest on the world and be sure there is an occlussion unit in it
            Hey! You're talking to me all wrong! It's the wrong tone! Do it again...and I'll stab you in the face with a soldering iron

            Comment


            • #7
              Matrox........

              You have alot of good points VigilAnt, let's hope the right pepole in Matrox lissen to what you say!

              I think if Matrox invests in equipment and in good pepole, and start making right descitions they can become very big!

              The first thing they should to now is to release PArhelia 8x ASAP, and then maybe make an eTV version of the card (Matrox Video Group sholud be responisble for drivers & tools) and get PArhelia 2 out next sommer at the latest (And be shure it can compete with NV30 and the upcomming R350) Then a "value DX9 card fir the masses" in the autum (good time before the chrismas rush)

              And of course, good drivers is very important. Focus 95% of driver development on the Parhelia platform..... (perhaps they do this today.... i don't know)

              I am a Perhalia owner myself, and i realy have to say, it's a very good card, performas well enough in 3D games for me, and the quality is very good! But Anisotropic Filtering is something Matrox should look into ASAP (Is it only a driver problem? What is the problem?)

              Comment


              • #8
                lets just remember matrox have finite resources here, its all very well saying release the 8x core then an etv then develope pitou then improve drivers etc. what they must realise is that the current parhelia will only sell to enthusiasts, they cant have sold very many at the moment and like you say have lots of overstock, so they need a part which will perform favourably in benchmarks with the compitition, doesnt matter what we say about benchmarks on here when joe public buy a pc they want the thing that looks the fastest so they want ther ati's and geforces not an overpriced matrox part which looks like it performs only half the speed of the compitition. if the parhelia 8x can do this then yes put it into action, if not then by the time it hits the shelves then it will be a pile of rubbish so they should save resources and move to the next gen product. what was parhelia really aimed at?doesnt benchmark in games that well, and has serious issues in professional apps performance wise also its way too expensive to be an oem solution. they need to focus on a market and have the best product in that market, faa is great triple head is also great and i cant think that it will add much cost to the card as the way i see it you just send double the resolution to the second head then split it into 2 thats why you can only run it stretched so keep it, but focus on improving consumer oriented figures if thats your market.
                is a flower best picked in it's prime or greater withered away by time?
                Talk about a dream, try to make it real.

                Comment


                • #9
                  if they cant get a competitive high end product out then they must reduce the price of parhelia quickly while it is still a respectable part as although this would dent profits they would shift stock and get some money back from it. at $200 it would sell very well but only for the next few months.
                  is a flower best picked in it's prime or greater withered away by time?
                  Talk about a dream, try to make it real.

                  Comment


                  • #10
                    2. Anisotropic Filtering: This is just inexusable, IMHO. The original GeForce-256 supported the same level/quality as the Parhelia.
                    actually, even the old g400 chip had 4-tap anisotropic filtering, they just didn't include it as an option in the powerdesk gui.

                    I know some murcers had succes in enabling it in Serious sam 1/2 by hacking the registry.
                    it was, of course, horrible slow, but they could have used it as another bullet in their feature-list, like nvidia did with their T&L.

                    who knows it might have been fast enough in the games that were out at that time, like those that were locked in a 640*480 resolution (especially if the g400 had been 50% faster! just imagine were matrox would have been now, if that was the case! a g400max at geforce2-speed! ), just giving the end-user the choice to enable/disable it, would have been nice.
                    Last edited by TdB; 16 November 2002, 08:07.
                    This sig is a shameless atempt to make my post look bigger.

                    Comment


                    • #11
                      thanks for the great news vigil!

                      totally agree with you. matrox should at least get their repetition back by releasing the 8x core logic. IMO matrox should now debug Pitou and make sure they don't make flaws like they did on the P with banding issue/anisotropic (if aniso is hardware problem). IMO Matrox should first make sure Pitou has extreme IQ/features/3d&2d quailty before going into speed. Triple head is a good idea, but would be a log better if it can do 1600x1200 per head with full DVI. Not sure but Intel may be working on 64-bit agp 8x specs. Maybe after the 32-bit 8x AGP on Pitou matrox should make a revision of 64-bit agp bus part. Stick with the traditional June release. According the Vigil and others Pitou is taped out. maybe matrox should just make heavy beta testing/debug on it, and if this is the case, start tweaking the core. Honestly, i think matrox should now foucs on feature sets. good feature sets can get the core sell well (and of course decent speed probably like the 9700 if we are taking about next summer). With decent speed and lots of good features/quality... matrox will certainly be back as a big brand.

                      imo for next summer when pitou is out, make the P 8x for the mass. that way, they can still get some money back, unlike now... i bet P project has lost matrox a lot of money

                      Comment


                      • #12
                        Originally posted by TDB


                        actually, even the old g400 chip had 4-tap anisotropic filtering, they just didn't include it as an option in the powerdesk gui.

                        totally agree. matrox should reveal the feature set on the G-cards. also the TnL + vertex shader on the G550. After all, the vertex shaders on the G550 seems powerful.

                        ***sigh... i still want pixel shader emulated... software rendering all the way***

                        Comment


                        • #13
                          Re: VigilAnt Rant and Some Observations

                          Originally posted by VigilAnt
                          There has to be a flagship card to push the mass market card along. Having a high performance, top of the line card is very much important, because when people think of a brand name like Matrox, Intel, Nvidia, they usally think of their best model. That is what allows people to think about the lower end models like Celerons and MX's. Not only does keeping a high end model make a good name for the brand, but it also keeps the technology current and allows it to filter down into the low end cards when the time and need comes to improve them. The same is the story with the G800, oops G550.
                          VigilAnt
                          Yes you are right!

                          How did AMD get that big in CPU?!? Not by selling cheapest low performance CPUs. They have a best performance/price ratio and some frigging fast CPUs to. That what counts and after that the mass market will buy your, Matrox cheaper, low end card.
                          As simple as that!

                          Comment


                          • #14
                            My 2 cents;

                            1. Sell off the overstocked P cores to OEM's. This reduces inventory, brings more cash flow than letting them sit in a warehouse until they're totally worthless and clears the decks for something productive .

                            2. Get in gear re: the P2. It's been said many times that you have to spend money to make money. It's true. In cases like this cutting development outlays in the short term is comitting hara-kiri in the long term.

                            3. People who do multimedia (video etc.) love Matrox's cards for the DVDMAX feature. The problem is that this currently costs one of the cards monitor ports. Cater to them by putting a totally independent analog output on all new cards so they can use multiple monitors and still preview to a TV.

                            My preference would be component, but S-Video would work and not need a large connector.

                            Dr. Mordrid
                            Last edited by Dr Mordrid; 16 November 2002, 09:25.
                            Dr. Mordrid
                            ----------------------------
                            An elephant is a mouse built to government specifications.

                            I carry a gun because I can't throw a rock 1,250 fps

                            Comment


                            • #15
                              Originally posted by Chrono_Wanderer


                              totally agree. matrox should reveal the feature set on the G-cards. also the TnL + vertex shader on the G550. After all, the vertex shaders on the G550 seems powerful.
                              the problem with the g550 vertex shader is that, even though it has enough registers, it isn´t programmable(afaik), so there is no way they can use it as a dx8 vertex shader, I don´t know if it could "emulate" hardwired dx7 T&L in the g550 vertexshader though, it might give a heavy boost in games like ut2003 in lowres, although it would still be bandwidth-limited.

                              ***sigh... i still want pixel shader emulated... software rendering all the way***
                              you would be looking a single-digit framerates in the best case, some things can´t be emulated effectively in the cpu, the whole point of pixelshading is that the computing takes place in the gpu, where all the needed texture-information is.
                              it would be like comparing the speed of "remote procedure calls" through a network, with the speed of procedures stored in the local cpu-cache.
                              Last edited by TdB; 16 November 2002, 09:18.
                              This sig is a shameless atempt to make my post look bigger.

                              Comment

                              Working...
                              X