pgn.inertia
06-02-03, 08:53 AM
-- Before I start my rant, let me first state that I once was a true 3Dfx addict (I owned everything from Voodoo Graphics to Voodoo Banshee). That soon changed when they became 3dfx/STB. I have been an NVIDIA follower ever since (NV5/NV10/NV15/NV20/NV25). I hope this NVIDIOT’ism doesn’t reflect too much in the following. Also, in this text I use some harsh words to describe the companies but understand, I don’t hate any one of them, I just try to exaggerate their position to make things more clear. --
The last 2 years (hey, maybe ever since the demise of 3dfx) the world of graphics has become a very interesting situation indeed. From a fairly large group of manufacturers we are down to only two (I’m not counting Matrox and the like, for I think they are manufacturers who focus on niche markets and not on the mainstream or enthusiast markets). Frankly, I thought it would be the death of 3D graphics as a competitive world, for NVIDIA were ruling the world with their offerings. The only competitor left was ATI, which was thought to be a niche market manufacturer too, to a large extent. This situation remained for the longer period of 1 year, up until R200. This core actually was a reasonable one to start with, but it could not rule the market the way NVIDIA products did.
Then something interesting happened. Suddenly ATI bought a console chip developer, ArtX. To me this was a sign of ATI leaving the PC graphics market and fleeing to a safe haven in the form of consoles. This, to me was a sort of conformation to what I feared all the time after the death of 3dfx. ATI however, had very serious plans with this IP company ArtX. They knew the basic ideas behind the Flipper chip were good, and could be used in PC graphics. They use this IP (and the men behind it) and create R300, the most powerful chip to ever come from ATI, and one of the most dominating products in 3D graphics in history. NVIDIA (who dominates at the time with NV25 (GeForce4 Titanium) has nothing to respond with and gets beaten by ATI.
All this above is a nice story about a company that buys another one and does the seemingly impossible, but what really interested me in all this is the parallelism is has with the CPU market. Here, on the one hand we have Intel, which dominated the CPU scene for as long as CPU’s existed and on the other we have the struggling AMD, who largely lives through the sale of second-class CPU’s. Compare this with the graphics situation in 2001, where NVIDIA was absolutely dominating the market with ATI struggling to keep up. This AMD actually releases a very nice core in itself with the K6-II (and eventually Sharptooth/K6-III(+)) with which it can keep up nicely but never surpass Intel. See ATI and it’s R200 against NVIDIA’s NV20 (GeForce3). Then AMD purchases a relatively unknown company (NexGen) and takes IP and employers from that. Again, see ATI who purchases ArtX. AMD uses these assets to create it’s K7 and blows Intel’s offering right off the map (sure, they didn’t surpass Intel on the initial K7 (Slot-A, Argon) offering, but would with their second (Socket-A, Thunderbird)) who still uses the P6 core which was already pretty much on it’s last legs. It’s new architecture, NetBurst (Pentium4), was nearly readying completion, but, on launch, seemed even more pathetic against K7 then the P6 core (in the form of Pentium!!!) was.
AMD fully exploits the power that is in the K7 core and builds upon that (Spitfire, Morgan, Palomino, Thoruoghbred) to become a very serious competitor to Intel. This is very much the same as ATI did with R300, which instantly dominated the whole landscape and is spread over the market fast with Radeon9500 (Pro), Radeon9700 (Pro). ATI also builds upon the core they have to get an even larger performance edge with R350 (Radeon9800 (Pro)). NVIDIA had nothing to put up against this, as their NV2x architecture was on it’s last legs too and their NV3x architecture was also readying completion (but not shipping, as we all know).
Intel, for the first time in it’s recent history, feels threatened by a competitor, but is quick to react and ramps up clockspeeds. They knew their NetBurst idea/architecture was good in itself, but did not have all the necessary means to really show it’s power (0.18 um, Socket-423, 256 kb Level-2 cache). These increases in clockspeed bring Intel closer to AMD but don’t really surpass it. Then Intel releases the Northwood core and gets on par again with AMD. But it never (well, not to this moment) dominates the market as it once did. It has, however, a performance edge on AMD (whose K7 architecture is getting very old indeed). AMD’s new architecture (the well known K8/Hammer) seems to again build upon the basic idea of K7 and is set to reclaim the performance edge from Intel. NVIDIA, who is caught by surprise (maybe not as much as we all thought, but still) has to act, and does. It announces NV30, but to numerous problems (though I don’t think design related are one of them) it never comes to a full release. This architecture (in the form of NV30 (GeForceFX)) would not nearly be enough to reclaim the throne. NVIDIA tries to save a bit of the situation in the form of driver updates (as opposed to ramping up clockspeeds with Intel). But it really takes a design change (256bit DDR-I bus) to get them back in the game (in the form of NV35 (GeForceFX 5900)). They however, do not dominate, but only get roughly on-par with ATI, much as Intel had done with Northwood on AMD.
My whole point to all of this is how much an acquisition of an IP company can mean for a large computer manufacturer. It has saved AMD (not to mention they are respected throughout the entire community) and has made ATI the top manufacturer in 3D graphics at this time. Of course, other examples can be thought of, such as 3dfx buying GigaPixel or even NVIDIA buying 3dfx. These, however were not successful and sometimes even led to the demise of the buyer. I think, that a lot of decisions NVIDIA has made in relation to NV30 were inspired by the 3dfx spirit. I don’t claim 3dfx caused the disaster that was NV30, but I think it didn’t help (just think of the hype that surrounded NV30, the likes of which we had never seen with any other NVIDIA product and compare this with the hype 3dfx created around their VSA-100 (in the form of Voodoo4/5).
The other thing I’d like to mention is how much the graphics market is reminiscent of the CPU market. Jen-Hsung Huang mentioned, not long ago, that the GPU is rapidly becoming a CPU in itself (if I’m correct he even stated to stand up to Intel in on the mid-long term). Now, with ATI mentioning to lengthen the life spans of their cores and NVIDIA not releasing a new GPU architecture every year, the whole market gets another step closer to being a true twin to CPU’s. I think that (with the exception of an incident (like NVIDIA buying IP and building a very powerful and dominating GPU as a result)) the whole graphics market will remind us a lot of the CPU market. With one manufacturer leading (not dominating) but the other never far behind.
Note; this is just my view of things, and I don't expect anyone to agree with me, I just wanted to vent...
Take care...
The last 2 years (hey, maybe ever since the demise of 3dfx) the world of graphics has become a very interesting situation indeed. From a fairly large group of manufacturers we are down to only two (I’m not counting Matrox and the like, for I think they are manufacturers who focus on niche markets and not on the mainstream or enthusiast markets). Frankly, I thought it would be the death of 3D graphics as a competitive world, for NVIDIA were ruling the world with their offerings. The only competitor left was ATI, which was thought to be a niche market manufacturer too, to a large extent. This situation remained for the longer period of 1 year, up until R200. This core actually was a reasonable one to start with, but it could not rule the market the way NVIDIA products did.
Then something interesting happened. Suddenly ATI bought a console chip developer, ArtX. To me this was a sign of ATI leaving the PC graphics market and fleeing to a safe haven in the form of consoles. This, to me was a sort of conformation to what I feared all the time after the death of 3dfx. ATI however, had very serious plans with this IP company ArtX. They knew the basic ideas behind the Flipper chip were good, and could be used in PC graphics. They use this IP (and the men behind it) and create R300, the most powerful chip to ever come from ATI, and one of the most dominating products in 3D graphics in history. NVIDIA (who dominates at the time with NV25 (GeForce4 Titanium) has nothing to respond with and gets beaten by ATI.
All this above is a nice story about a company that buys another one and does the seemingly impossible, but what really interested me in all this is the parallelism is has with the CPU market. Here, on the one hand we have Intel, which dominated the CPU scene for as long as CPU’s existed and on the other we have the struggling AMD, who largely lives through the sale of second-class CPU’s. Compare this with the graphics situation in 2001, where NVIDIA was absolutely dominating the market with ATI struggling to keep up. This AMD actually releases a very nice core in itself with the K6-II (and eventually Sharptooth/K6-III(+)) with which it can keep up nicely but never surpass Intel. See ATI and it’s R200 against NVIDIA’s NV20 (GeForce3). Then AMD purchases a relatively unknown company (NexGen) and takes IP and employers from that. Again, see ATI who purchases ArtX. AMD uses these assets to create it’s K7 and blows Intel’s offering right off the map (sure, they didn’t surpass Intel on the initial K7 (Slot-A, Argon) offering, but would with their second (Socket-A, Thunderbird)) who still uses the P6 core which was already pretty much on it’s last legs. It’s new architecture, NetBurst (Pentium4), was nearly readying completion, but, on launch, seemed even more pathetic against K7 then the P6 core (in the form of Pentium!!!) was.
AMD fully exploits the power that is in the K7 core and builds upon that (Spitfire, Morgan, Palomino, Thoruoghbred) to become a very serious competitor to Intel. This is very much the same as ATI did with R300, which instantly dominated the whole landscape and is spread over the market fast with Radeon9500 (Pro), Radeon9700 (Pro). ATI also builds upon the core they have to get an even larger performance edge with R350 (Radeon9800 (Pro)). NVIDIA had nothing to put up against this, as their NV2x architecture was on it’s last legs too and their NV3x architecture was also readying completion (but not shipping, as we all know).
Intel, for the first time in it’s recent history, feels threatened by a competitor, but is quick to react and ramps up clockspeeds. They knew their NetBurst idea/architecture was good in itself, but did not have all the necessary means to really show it’s power (0.18 um, Socket-423, 256 kb Level-2 cache). These increases in clockspeed bring Intel closer to AMD but don’t really surpass it. Then Intel releases the Northwood core and gets on par again with AMD. But it never (well, not to this moment) dominates the market as it once did. It has, however, a performance edge on AMD (whose K7 architecture is getting very old indeed). AMD’s new architecture (the well known K8/Hammer) seems to again build upon the basic idea of K7 and is set to reclaim the performance edge from Intel. NVIDIA, who is caught by surprise (maybe not as much as we all thought, but still) has to act, and does. It announces NV30, but to numerous problems (though I don’t think design related are one of them) it never comes to a full release. This architecture (in the form of NV30 (GeForceFX)) would not nearly be enough to reclaim the throne. NVIDIA tries to save a bit of the situation in the form of driver updates (as opposed to ramping up clockspeeds with Intel). But it really takes a design change (256bit DDR-I bus) to get them back in the game (in the form of NV35 (GeForceFX 5900)). They however, do not dominate, but only get roughly on-par with ATI, much as Intel had done with Northwood on AMD.
My whole point to all of this is how much an acquisition of an IP company can mean for a large computer manufacturer. It has saved AMD (not to mention they are respected throughout the entire community) and has made ATI the top manufacturer in 3D graphics at this time. Of course, other examples can be thought of, such as 3dfx buying GigaPixel or even NVIDIA buying 3dfx. These, however were not successful and sometimes even led to the demise of the buyer. I think, that a lot of decisions NVIDIA has made in relation to NV30 were inspired by the 3dfx spirit. I don’t claim 3dfx caused the disaster that was NV30, but I think it didn’t help (just think of the hype that surrounded NV30, the likes of which we had never seen with any other NVIDIA product and compare this with the hype 3dfx created around their VSA-100 (in the form of Voodoo4/5).
The other thing I’d like to mention is how much the graphics market is reminiscent of the CPU market. Jen-Hsung Huang mentioned, not long ago, that the GPU is rapidly becoming a CPU in itself (if I’m correct he even stated to stand up to Intel in on the mid-long term). Now, with ATI mentioning to lengthen the life spans of their cores and NVIDIA not releasing a new GPU architecture every year, the whole market gets another step closer to being a true twin to CPU’s. I think that (with the exception of an incident (like NVIDIA buying IP and building a very powerful and dominating GPU as a result)) the whole graphics market will remind us a lot of the CPU market. With one manufacturer leading (not dominating) but the other never far behind.
Note; this is just my view of things, and I don't expect anyone to agree with me, I just wanted to vent...
Take care...