Home 3DMark05 arrives
Reviews

3DMark05 arrives

Scott Wasson
Disclosure
Disclosure
In our content, we occasionally include affiliate links. Should you click on these links, we may earn a commission, though this incurs no additional cost to you. Your use of this website signifies your acceptance of our terms and conditions as well as our privacy policy.

THE LAST REVISION of Futuremark’s 3D graphics benchmark, 3DMark03, brought a storm of controversy on the little Finnish company, largely because they’d incurred the wrath of NVIDIA. Futuremark survived, however, and their benchmark has proven itself over time. Our latest round of tests in our Radeon X700 XT review used a number of state-of-the-art games, including the Doom 3, Half-Life 2’s Source engine, and Far Cry. The performance results we saw in those games tracked very well with the results we got from 3DMark03, somewhat vindicating the benchmark as a decent predictor of performance in the now-current wave of games.

Futuremark intends its new version of 3DMark to give us a glimpse of the future yet again, and if 3DMark05 is any indication, the future of 3D games looks very bright indeed. Read on to see what Futuremark has wrought and how thirteen of the latest graphics cards perform in this new benchmark.

The making of the new 3DMark
The limited time I’ve had with the final version of 3DMark05 has been dedicated almost entirely to benchmarking a wide range of graphics cards with it. As a result, I won’t address the finer points of its technical implementation here. The most important thing to know about the three game tests that make up 3DMark05’s composite score is that they are intended to represent future games. That means Futuremark has attempted to predict and use the same sorts of polygon budgets, dynamic shadowing techniques, and shader programming methods that game developers will use in future games. After reading over Futuremark’s excellent whitepaper on 3DMark05, I’m largely persuaded that they have made a solid set of choices. The app uses “pure” DirectX 9, unlike the previous version, and the floating-point-enabled Shader Model 2.0 is now a baseline. 3DMark05 dynamically generates shader code in Microsoft’s High Level Shader Language and supports a broad range of shader compile targets, so it can take best advantage of the Radeon X700/X800 series via Shader Model 2.0b and the GeForce 6 series via Shader Model 3.0.

Futuremark has also opted to support multiple code paths for dynamic shadowing techniques, so that graphics chips with support for depth stencil textures, like NVIDIA’s, can make use of that capability. The decision to use multiple code paths seems like a reasonable concession to practicality and is very similar to the sorts of methods developers have been using in real-world games. It is, however, something of a new approach for Futuremark, and should be noted. FutureMark even acknowledges that the different shadowing code paths produce some slightly different images.

Let’s have a look at some pictures of the three scenes that together determine 3DMark05’s composite score.

Game test 1: Return to Proxycon
This first scene is a reprise of the space battle sequence from 3DMark03. This new version features much richer detail, including dynamic shadows, some big open spaces, and extremely detailed polygon meshes.

 

Game test 2: Firefly Forest
This test upholds the long-standing 3DMark tradition of amazing-looking outdoor scenes with a big jump in realism over the day’s expectations. The “firefly” is really more of a green blob, but it’s also a dynamic light that casts shadows among the dense vegetation.

 

Game test 3: Canyon Flight
The final test features some of the most amazing water I’ve seen rendered in real time, plus an exquisitely detailed model of an imaginary airship. The water shader is built on the one featured in 3DMark03, but it’s now much improved. The water also features depth fog to make objects further below the surface appear deeper.

Interestingly enough, Futuremark says the rock surface of the canyon is one of the most complex materials in 3DMark05. The shader uses almost all of the 96 instruction slots available in Pixel Shader 2.0.

Now, let’s see how various graphics cards handle the demands of creating images like these.

 

Our testing methods
Both the ATI and NVIDIA cards were left at their driver default settings for image quality, with the exception that we turned off vertical refresh sync on all cards.

Our test system was configured like so:

Processor Pentium 4 3.4’E’GHz Pentium 4 550 3.4GHz
Front-side bus 800MHz (200MHz quad-pumped) 800MHz (200MHz quad-pumped)
Motherboard Abit IC7-G Abit AA8
BIOS revision 2.4 1.4
North bridge 82875P MCH 925X MCH
South bridge ICH5R ICH6R
Chipset drivers INF Update 6.0.1.1002 INF Update 6.0.1.1002
Memory size 1GB (2 DIMMs) 1GB (2 DIMMs)
Memory type OCZ PC3200 EL DDR SDRAM at 400MHz OCZ PC2 5300 DDR2 SDRAM at 533MHz
CAS latency 2 3
Cycle time 5 10
RAS to CAS delay 2 3
RAS precharge 2 3
Hard drive Maxtor MaXLine III 250GB SATA 150
Audio Integrated
Graphics GeForce FX 5700 Ultra 128MB AGP with ForceWare 66.51 beta drivers
GeForce FX 5900 XT 128MB AGP with ForceWare 66.51 beta drivers
GeForce FX 5950 Ultra 256MB AGP with ForceWare 66.51 beta drivers
GeForce 6600 GT 128MB PCI-E with ForceWare 66.51 beta drivers
GeForce 6800 128MB with ForceWare 66.51 beta drivers
GeForce 6800 GT 256MB with ForceWare 66.51 beta drivers
GeForce 6800 Ultra 256MB with ForceWare 66.51 beta drivers
Radeon 9600 XT 128MB AGP with CATALYST 4.10 (8.07) beta drivers
Radeon 9800 Pro 128MB AGP with CATALYST 4.10 (8.07) beta drivers
Radeon 9800 XT 256MB AGP with CATALYST 4.10 (8.07) beta drivers
Radeon X700 XT 128MB PCI-E with CATALYST 4.10 (8.07) beta drivers
Radeon X800 Pro 256MB AGP with CATALYST 4.10 (8.07) beta drivers
Radeon X800 XT PE 256MB AGP with CATALYST 4.10 (8.07) beta drivers
OS Microsoft Windows XP Professional
OS updates Service Pack 2, DirectX 9.0c

The test systems’ Windows desktop was set at 1152×864 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

If you have questions about our methods, hit our forums to talk with us about them.

 

The game tests and overall scores
We ran all tests at 3DMark05’s default resolution (1024×768) and settings, unless otherwise noted. The program automatically chose the most optimal compile target for each type of hardware, so everybody got to put his best foot forward.

Speaking of which, ATI and NVIDIA both managed to get new drivers certified by Futuremark by midday yesterday, and we’ve used those drivers in our testing. Futuremark prefers the use of certified drivers, so they can verify that the drivers don’t include benchmark-specific optimizations.

ATI and NVIDIA should both be happy about these results to some degree. The most obvious result of all is the utter dominance of the Radeon X800 XT Platinum Edition in these tests. ATI’s $500 card beats NVIDIA’s competing GeForce 6800 Ultra by a very healthy margin, probably thanks to its much higher clock speed. Among the $400 cards, we have a tie. At $299, though, the aging Radeon 9800 XT can’t keep pace with the GeForce 6800.

The big prize is the under-$200 market, of course, and in this hotly contested space, we have another virtual tie. The Radeon X700 XT is technically ahead on points, but the difference between it and the GeForce 6600 GT is very slight.

Among the older cards, the Radeon 9800 Pro and 9600 XT both look very good. The same can’t be said for the GeForce FX series, which suffers yet another ignominious humiliation at the hands of a new release of 3DMark. Obviously, 3DMark05 is very polygon and pixel shader intensive, and the GeForce FX architecture just isn’t up to the task, despite the improvements NVIDIA has made to the runtime compiler in its graphics driver over time. Even the GeForce FX 5950 Ultra, with over 30GB/s of memory bandwidth and 3.8 gigatexels/s of fill rate, can’t match the humble Radeon 9600 XT.

As for image quality, we didn’t spot any obvious differences between the cards’ output, as we’ve come to expect in recent months. ATI and NVIDIA have been matching each other pretty closely in their latest hardware and drivers. We’ll have to look further into image quality in 3DMark05 in future graphics reviews.

 

Synthetic feature tests

Futuremark says 3DMark05’s fill rate test should be less dependent on memory bandwidth and more indicative of a GPU’s theoretical peak filling capabilities. That appears to be true in most cases compared to what we saw recently from 3DMark03. However, the new single-texturing test still doesn’t show any difference between the true eight-pipe design of the Radeon X700 XT and the hybrid “8+4×1” design (with fragment crossbar) of the GeForce 6600 GT.

The pixel shader uses the rock face of the cliffs in game test 3 as its subject. Futuremark opted to go with traditional texturing on these cliffs in combination with shader effects, instead of using procedural texturing, because they don’t expect to see wide use of procedural texturing in the next generation of games. As it stands, this test flatters (perhaps justifiably) GeForce 6 series GPUs.

The “simple” vertex shader test has over six million polygons in the scene, so it’s no slouch. These models are affected by a single light, and the vertex shaders are asked only to do static transform and lighting. The “complex” test involves a field of swaying grass blades, each one moving in the wind.

Either way, the ATI cards look relatively strong here. The Radeon 9800 Pro surpasses the GeForce 6800 GT in the simple test, and the Radeon X700 XT outdoes the GeForce 6800 Ultra in the complex test.

 

Shader model performance compared
Now let’s see how using different compile targets affects performance in 3DMark05. We’ll compare the two new “mainstream” cards from NVIDIA and ATI, the GeForce 6600 GT and Radeon X700 XT. Since it’s derived from the Radeon X800 architecture, the X700 XT supports both Shader Model 2.0 and the new 2.0b extension, with more instruction slots and a few other bells and whistles. The GeForce 6600 GT is derived from the GeForce 6800, so it can handle the new Shader Model 3.0, complete with instruction loops, dynamic branching, and 32-bit floating-point datatypes. All Shader Model 2.x incarnations are subsets of Shader Model 3.0, and the GeForce 6600 GT can run them all.

For the tests labeled “SM 2.0b,” I set 3DMark05 to use the 2.0 vertex shader spec and the 2.0b pixel shader spec, because no 2.0b vertex shader choice was available. All the rest used matching pixel and vertex shader revisions.

The Radeon X700 XT doesn’t seem to care which shader model it uses, but the GeForce 6600 GT loses some performance in the vanilla Shader Model 2.0. Most of that performance is regained just by moving to the 2.0a standard intended for the GeForce FX series, though, or even to 2.0b.

The pixel shader test doesn’t seem to care which shader model is used.

The vertex shader test is pretty much the same story. Even the complex test doesn’t benefit from extended shader models.

 
Conclusions
I hate to draw too many conclusions or get too deep into the potential debates over 3DMark05’s usefulness. I, for one, am pleased to have another tool in my benchmarking toolbox, and 3DMark05 offers some unique insights into possible future performance by bringing us graphics workloads much more intensive than even the latest games do.

Plus, hey, bitchin’ graphics demo.

Beyond the three new game tests, the purely synthetic tools in 3DMark05 include some welcome improvements for us hardware reviewer types. I’m glad to see the fill rate test tend toward theoretical peaks; we can test real-world fill rate better in games. However, I’m disappointed the fill rate tool doesn’t offer more granularity, with results for Z/stencil writes distinct from color writes. 3DMark05 does include an interesting new test of performance with different sized batches of triangles. I haven’t had time to play around with it yet, but this test should expose some weaknesses of poorly written graphics drivers. Finally, 3DMark05 packs some very nifty tools to capture statistics from a benchmark run. For instance, it will plot polygon counts over time from any of the game tests. I expect to get some use out of these tools over time, once we’ve further familiarized ourselves with their ins and outs.

You can download 3DMark05 yourself from Fileshack

Latest News

Gold Miner Nilam Resources Shares Surge 22x Amidst Bitcoin Buying Announcement
Crypto News

Gold Miner Nilam Resources Shares Surge 22x Amidst Bitcoin Buying Announcement

BlackRock CEO Goes Bullish on BTC as Spot Bitcoin ETF Crosses $17 Billion
Crypto News

BlackRock CEO Goes Bullish on BTC as Spot Bitcoin ETF Crosses $17 Billion

Spot Bitcoin exchange-traded funds (ETFs) continue to receive massive inflows as investor demand skyrockets, with BlackRock’s IBIT at the forefront. Notably, IBIT has hit a whopping $17 billion in AUM...

Elliott Wave Pattern Indicates Ripple (XRP) Might Surge to $13
Crypto News

Elliott Wave Pattern Indicates Ripple (XRP) Might Surge to $13

Renowned market analyst Tony Severino, also called “The Bull,”  has unveiled a captivating analysis of XRP’s potential price trajectory. His findings suggest that XRP could be poised for an extraordinary...

XRP ETF Premium May Record 100x to $500 Chad Steingraber
Crypto News

XRP ETF Premium May Record 100x Growth Chad Steingraber

NFL
Streaming News & Events

NFL Discloses Moving Two NFL Games Into Streaming in 2024

Crypto News

Unveiling the Most Popular Crypto Presales in March Among Americans

Apple Users Are Being Spammed with Unwanted Password Reset Requests as Part of ‘MFA Bombing'
News

Apple Users Are Being Spammed with Unwanted Password Reset Requests as Part of ‘MFA Bombing’