Original Link: http://www.anandtech.com/show/2512
Futuremark's 3DMark graphics test suite returns after a one year hiatus. What's new and what can we look forward to for the coming ORB competitions?
Another year, another version of Futuremark's 3DMark benchmarking software – except they skipped last year and dropped the year from the name. Whatever. The important thing is, fans of 3D graphics benchmarks everywhere now some new software that can bring your computer to its knees... and then some. 3DMark Vantage officially launched today, and we thought we would do a quick update on what has changed and some general commentary on the newest release.
In terms of changes, the most notable areas are the hardware and software requirements. Like PCMark Vantage, Windows Vista is required in order to run 3DMark Vantage. That's hardly a surprise, but simply installing Vista won't be sufficient; you also need to have a DirectX 10 capable graphics card. If you happen to run a system that's no longer on the cutting edge of hardware and software, you can simply forget about running 3DMark Vantage. We installed it on a system with a Radeon X1950 XTX, just to see what would happen. There were no warning messages during installation, which was surprising, but as soon as we loaded up 3DMark Vantage we were greeted by a software crash. It would probably be more effective if Futuremark alerted users to the fact that their hardware is inadequate rather than simply crashing; most likely they will address this with a future patch.
Besides having appropriate hardware and software, you will probably also want to update your graphics drivers. On the test system, the first time we executed the program, we encountered several error messages and some display corruption. We were running the Catalyst 8.3 drivers, and upgrading to 8.4 at least eliminated the display corruption on the loading screens. (The errors have been reported by others and may simply be a "first run" problem.) While we're on the subject, we might as well get the test hardware out of the way:
|3DMark Vantage Initial Test System|
|Processor||Core 2 Quad Q6600 (2.40GHz 2x4MB
Overclocked to 3.30GHz (1467FSB)
|Memory||2x2048MB OCZ DDR2-800
Running at DDR2-734 4-4-4-12
|Graphics||2 x AMD Radeon HD 3870 (CrossFire)|
|Hard Drive||Samsung F1 750GB (7200RPM 32MB)|
|Operating System||Windows Vista Ultimate 64-bit|
While this is by no means the fastest gaming system on the planet, it just so happens to be my own personal gaming setup and it's more than sufficient for running any current game – though of course Crysis requires a few tweaks in order to run while. This is also not intended as a full benchmarking article, so we'll dispense with things like CPU clock speed scaling, CrossFire performance scaling, etc. This system runs 24/7 fully stable, and has been doing so for several months. (The system will be examined more closely in an upcoming article looking at various midrange gaming systems.)
Like PCMark Vantage, the user interface has undergone an update. The changes here tend to be a bit more far-reaching, however. Instead of the usual 3DMark where you get a standard setting that everyone should use if they want to compare scores, 3DMark Vantage now includes four benchmark scenarios that you can run. The default is "Performance", which is roughly equivalent to the default settings in 3DMark06. It runs at 1280x1024 with a reasonable selection of graphics enhancements. "High" bumps the resolution up to 1680x1050 and increases the various detail settings, while "Extreme" takes things a step further in the detail department and runs at 1920x1200. Don't have what it takes to run at any of these settings? No worries, as Futuremark now includes an "Entry" setting that runs at 1024x768 and disables much of the complexity.
The above four images show the default settings for the four benchmark modes just described. Entry frankly looks quite poor with the disabled effects, but it runs fast. Even on a relatively high-end system, Performance and High modes struggle, and the Extreme benchmark absolutely crawls. Note that we experienced quite a few graphical glitches on Extreme in the Jane Nash test; AMD is likely working to release updated drivers, as it appears to be a GPU issue rather than a CPU/system problem.
Update: We have indeed received confirmation that new drivers are available - and not just from AMD. There's a hotfix driver update from AMD to address performance and graphical corruption issues with 3DMark Vantage and a beta driver from NVIDIA to do the same for their hardware. Since this is merely a first look rather than a review, however, the performance differences aren't a huge concern. We'll save the driver updates for future hardware reviews.
Update #2: Futuremark was kind enough to provide us with a Pro version registration code, and we sort of assumed the normal functionality and limitations that are present in previous 3DMark releases would continue. Not so, good readers! It turns out that you must register in order to even run 3DMark Vantage, and even with your email address you only get to generate one result with your trial - and even that needs to be viewed online. If you want what was normally free, you now have to purchase the Basic version for $6.95, and you still don't get access to all the features (i.e. the four test settings listed above along with the feature tests) unless you upgrade to the Advanced version for $19.95. The Pro version is mostly for business use, priced at a whopping $495. (Ed: Did we mention how thankful we were to receive a Pro code from Futuremark?)
It will be interesting to see how this plays out, because Futuremark may have relegated 3DMark Vantage into irrelevance by this decision. Now, you can't test and retest your system to see how any tweaks may or may not affect your score, and if you have multiple systems you're going to need to generate multiple email addresses. (Ed: I smell a rise in Gmail, Hotmail, Yahoo, etc. email account registrations.) On the other hand, users interested in competing for the top ORB spots will now have to put some money into Futuremark's pockets. But then, Futuremark already receives funding from various sources, like Sapphire obviously, and paying to see ads isn't something most people like.
There are four tests that comprise the main 3DMark benchmark, all of which are brand-new. The first graphics test is called "Jane Nash" and shows a woman escaping a "mad scientist lair". Water effects abound, as do not-so-subtle advertisements for Sapphire. Jane also appears to be a hologram, since she glows in a completely unnatural way - but we would assume that the glowing serves as a good way to stress your GPU. The water effects also look nice, but the bad guys with their "orange camouflage" are even worse than Glow-bug Jane… and they subscribe to the Storm Trooper school of battle tactics, clearly incapable of hitting the broad side of a barn. They were probably too busy checking out Jane's assets….
The second graphics test is "New Calico" and it involves spaceships, asteroids, and an impressive looking planetary bombardment sequence. Perhaps it's intended as a follow-up to the "Proxycon" benchmarks from earlier 3DMark releases; that doesn't matter too much, though, as most users are simply interested in the final performance score. While it doesn't appear to be as stressful as the Jane Nash demo, we actually think the graphics and effects are better in this benchmark.
The two CPU benchmarks use similar looking scenes of differing complexity, showing a bunch of small stunt airplanes flying around a scene. One focuses on AI performance and the other looks at physics processing capabilities. PhysX hardware should improve performance on the latter, although without an appropriate system we were not yet able to verify this. Not that it matters much; improving a benchmark score is one thing, whereas improving gaming performance is a different matter. To date, only a few titles (including Unreal Tournament 3) support PhysX, and nothing we've played actually requires it. Besides the four major benchmarks, Futuremark includes six other "Feature Tests" that look at things like texturing performance, particle simulations, cloth rendering, and ray tracing among other things. These feature tests do not influence the final score, similar to what we've seen with previous 3DMark test suites.
Update #3: For those interested in viewing the first benchmark without going to the trouble of downloading and installing 3DMark Vantage, we've captured the content for a video. The frame rates are a bit choppy from the capturing process, but you can at least get some idea of what the benchmark looks like in action.
I can't drive 55
So how does 3DMark Vantage run on our relatively powerful test system? Without comparison systems, we can't come to any firm conclusion just yet - we'll save that for future reviews. However, here are the scores for the four standard test scenarios. We're sure that the ORB (Online Results Browser) charts are already full of people doubling or tripling our initial meager attempts.
As usual, we don't feel that 3DMark scores are actually better than gaming performance results. Just as performance in Crysis tells you little about how your system will run Flight Simulator X or Half-Life 2: Episode Two, 3DMark Vantage is at best an estimate of how some future DX10 titles will perform on current hardware. We'll include it in some of our reviews as a baseline measurement, but until we can actually play a Futuremark game we'll take the ORB standings with a grain of salt.