Half Life benchmarks could help us get a life

WELL ISN'T this exciting? I can see the headline now - obscure British case modder gets drunk, offers to write column for leading news site, wakes up next morning with a deadline from Editor. Classic front page stuff.

Some of you guys may recognise my name - the site I co-edit, bit-tech.net, has been featured in Inq hardware roundups for a while - including dire references to our namesake, a diamond drill bit company in the States. But now you guys get the joy of me every fortnight commenting on some of the latest happenings in the hardware and modding world.

So where to start? Well, I've recently been looking at revising the benchmark suite we use on bit-tech. There's been plenty written about the unreliability of 3DMark2003 and other benchmarks, so we figured it was time to re-evaluate. We have been looking at the joys of Gun Metal recently - Nvidia's flagstone DX9 game (well, I say flagstone - it's the only one they have right now). Despite being a prime part of Graphzilla's The Way It's Meant To Be Played regime, it turns out it runs just as well on ATi hardware. Strange no?

Likewise, the Dawn tech demo has caused a lot of wrangling as fans try to get it to run on ATi cards - with success. However, the bizarre thing is that it runs just as well, if not better. That's despite an Nvidia tech telling me it supported shader routines and similar that simply wouldn't work on ATi hardware.

So this leads me to ask, is there really that much difference between the hardware being pushed our way?

Yes, we have different memory bandwidths, clock speeds, shader instruction support and the like. But in the end, does that really give us any different result? There is very little to separate high-end Nvidia cards from high-end ATi cards now in terms of performance, so are we being fed all this marchitecture for nothing?

The worrying thing is that I think the graphics companies are onto this, and are taking more and more drastic measures to try and differentiate their products. I mentioned benchmarking at the beginning - well you may have read the reports that imminent stunner, Half-Life 2, will be coming with a benchmarking tool. 'Heavens!' thinks every hardware reviewer in the world - a game people will want that we can use to benchmark products.

Except for the rumours recently reaching my ears that ATi, following its collaboration to show HL2 at E3 in Las Vegas, has since paid a seven-figure sum to Valve Software to optimise for ATi hardware. Rumour? Possibly, but an unnamed Nvidia source said 'it wouldn't surprise him'. ATi PR refuses to comment, and Valve Software was extremely diplomatic and just fed us the official company line. Read into that what you will. If this is the case, the question is, how much will it make a difference? If Nvidia's TWIMTBP overt regime can only produce a game that runs five per cent faster, how far will, what are now late-stage covert optimizations, get ATi? The proof of the pudding will be in the gaming, and I for one can't wait for the next interesting chapter in this saga. µ