Another year, another 3DMarkEver since I started taking hardware performance testing seriously, a version of 3DMark has been an oft-used part of my software arsenal. You need literally dozens of tools to just evaluate a graphics product these days, to get as comprehensive a picture of the hardware as possible in order to make an informed judgement for a discerning readership such as yourselves.
However, and annoyingly so for this technologist, Futuremark have a knack of fairly accurately disseminating information on how technology trends on the PC will progress. That knack results, usually every couple of years, in a new version of 3DMark. And, to date and without fail, each new version of 3DMark has been a great snapshot tool to rank graphics products almost as well as dozens of tools, games, and days of analysis do.
So while it would be folly for us to just rely on 3DMark, despite PD's protesting to the contrary, for our evaluation of graphics products, as a quick indicator of how a certain graphics product is likely to do against its peers, it pretty much rocks. That's always really been the strength of the product from our perspective, with the free download giving you guys the ability to snapshot your own system for rough comparison to others.
It's not all been sweetness and light for Futuremark's flagship benchmarking tool, though. Late 2004 saw the release of 3DMark05 which we introduced and dissected on launch, analysing its ability to do what it has always done best. We felt that it came up short, favouring not-quite-D3D9 ability that only one IHV (at the time) exposed, resulting in too skewed a view to rely on from the point of view of the final score it produces.
Its basic rendering techniques, including perspective shadow mapping with correct self shadowing and its shader-driven special effects, are forward-looking and mostly well implemented, so there's no real complaint there, but for a benchmark that purports to be the overlord of 3D performance testing, we felt it fell short. We explain the reasons why in those referenced articles.
We still use the tool though, since 3DMark05 also bundles a set of basic feature tests that also snapshot GPU subsystems such as texturing rate, vertex shader performance and small batch performance testing. Those tools are entirely useful to us, especially in our evaluation of a reference graphics product when we're looking at the basics of a graphics product's ability to run well and interact with the Direct3D runtime.
So we don't publish the final score in reference product reviews, but we do elsewhere from time to time and the feature tests are a cemented part of our analysis suite. And as a snapshot test, as discussed, we have to give it credit where credit is due.
Hello, 3DMark06Today, the 18th of January 2006, sees the release of the next version of 3DMark, snappily codenamed 3DMark06. And as such, this new 3DMark commands some respect for the grandfathers of modern graphics testing, and a closer look to see whether their new flagship tool is worth a squirt.
And much like the SE update to 01, 06 isn't a complete departure from its predecessor. Let's take a quick look and see what they've done.