fatbusinessman said:
Xavier said:While the situation on Antialiasing and AF is unlikely to change, it looks like there may be a big turn-around with Half Life 2 performance for NVIDIA...
By now everyone has probably heard about the >$8M OEM and 'Marketing Endorsement' deal between Valve and ATI, and the more observant will probably also be aware of the situation recently with HL2's shaders, where it was shown that ATI themselves provided most of the radeon-friendly pixel and vertex shading code used in the game, while the shading code for the GeForce cards was completed in-house with no apparent input or consultation with NVIDIA... Although they did a good job of diverting all attention from both stories after the new of leaked game code swamped hardware and gaming sites across the net and geeks the world over tried to get their hands on the beta.
We know that at the time of the leak, around 6 days before the original 30th September launch target, the game engine was about 80% 'there' (in valves words), and level-wise only 40% of the maps were populated, with none 100% complete. Valve initially argued that the leaked code was much older but the comments and datestamps both in the source code and leaked executables argue otherwise.
With a new rumoured release date of April '04, that will have given Valve 6 months, not only to get the game environment completed (the creation tools they're using have been referred to as extremely fragile, which may explain in part just how they ended up so far behind by 30/9/03) but also further optimise their shader architecture and render engine to work better with GPUs other than ATI's. Regardless of how big a backhander they recieved, Valve recognise that NVIDIA still hold a pretty large majority in the market and will no doubt do everything they can to make sure that as many potential customers as possible are able to play their game.
We discussed the whole matter with NVIDIA just after their Editors Day, and it was mentioned that not only did Valve reject NVIDIA input in optimisation of shaders for the GeForce chips at the same level to which they'd involved ATI, but also that when the developers spotted the fogging bug in the beta 52-series drivers which Gabe attempted to crucify NVIDIA over, that rather than raise the bug in the usual manner, they dropped all communication - which is pretty damn strange way of working with anybody.
Knowing that the source engine was still incomplete we were pretty surprised just how many sites ran that "exclusive HL2 benchmark"... We along with many others turned it down, unless were offered gold code and corresponding WHQL drivers designated for the time of retail release there's just no point. I made a mistake of leaping for beta code when I wrote the hardware section for PC Zone, entering into a deal with Epic to be one of the first to use UT2003 beta code for graphics benchmarks, and I know only too well how drastically performance changed between that beta and what finally hit shelves, not only in terms of GPU performance but also between the P4 and AMD rigs at that time.
Since the original round of benchmarks, which were run on supplied black-box machines and forbade the use of NVIDIAs 50 series driver, Valve haven't allowed anyone to touch the benchmark again... Positive PR only when it suits them.
Looking at those current HL2 scores as a 'worst case scenario', alongside the performance figures people have touted round the web since the leaked beta it's pretty clear that none of the top-end GeForceFX or Radeon cards are going to have any problem running HL2 or DoomIII. However its worth remembering that by the time both are out we're likely to have news on, if not the launch of the R420 from ATI and NVIDIAs NV40, Pixel and Vertex Shader 3.0, a minor DirectX rev and most significantly PCI Express.
So, unless you've had a significant outage and blown your 3D card sky-high in he process, or simply have pots of cash to waste, I'd wait til May or thereabouts, by which time you should be able to get your mitts on a Prescott based BTX system with PCI express graphics and all other forms of technical wizardry
. Great post. 
However, given a little time, I'll post my thoughts on the 9600XT and 5700Ultra, ATI and nVidia's latest mainstream cards, respectively. I've had the pleasure of testing one of these cards thus far, soon to be both (courtesy of Xavier), so I'll let you know my humble opinion, and hopefully answer any questions you may have.
Really? I thought John Carmack's .plan files were full to bursting with talking about how wonderful ATI's cards were...SawTooTH said:
Since that time (I think?) UltraShadow has been developed and implemented into nVidia's 5900 and 5700 cards, so as to aid lighting techniques in games such as Doom3. When some benchmarks were run a while back (although take into account Xavier's comments, as they're as valid here) nVidia's cards really outperformed ATI's across the board, pretty much.fatbusinessman said:
That's funny, I seem to remember that after ATi fixed the "cheat" (It was actually a bug that showed up only in Quake 3 and had been in the drivers for about 8 releases, naturally nVIDIA waited until ATI had a big product release to tell HardOCP about it - which Kyle had no problem telling the world about. Funny how his tone changed when nVIDIA were caught cheating. I wonder who's payroll he's on.) they didn't actually lose any performance in Quake 3, which considering when it came out I seem to remember it beating the vanilla GF3, means it was still faster than that and the GF3 Ti200.Xavier said:Heh, well all of the initial reviews of the Radeon showed it cheating in Quake3 based games, the 8500 is outclassed by the GeForce3, let alone the Ti500 or any GeForce4Ti, which means the 8500 is slower than every DirectX8 offering NVIDIA have brought to market
Bodhi said:
