Thursday , 24 July 2014
Latest Reviews
Home » Reviews & Articles » Hardware » BFG GeForce FX 5900XT OC Review

BFG GeForce FX 5900XT OC Review

BFG Technologies introduces the GeForce FX 5900XT Overclocked. Is this the new king of the mainstrream-performance segment? Read our review to discover if this the card for you!

[review_ad]Introduction


Now that graphics cards have become a staple in the marketplace, more and more end users are finding them within computers at home and at work and are beginning to associate 3d graphics capabilities with ordinary computers. This means that 3d graphics are no longer new (and most of us have known that for years now), and that people expect modern software and games to work on their computers. While the “software-side” of modern computers may have proven less than capable of truly pushing computer hardware to the limit, lately we have begun seeing the introduction of a “new breed” of software (and especially games). This latest evolution of software design and capability has been spurred and kicked into high gear by hardware programmability, feature-rich programming APIs like DirectX (especially the 9.0 generation) and OpenGL, and raw graphics power that enables content developers to push the very limits of virtual reality.

Buyers of recent games like Crytek’s Far Cry and Tom Clancy’s Splinter Cell are finding that although the spectacles witnessed are pure wonders to behold and interact with, the frame rate and playability have diminished rapidly. They will soon be reminded with the introduction of games like i.d. software’s Doom3 and Valve’s Half-Life 2 that this issue will only become dramatically worse and in the worst case a cause to not even try the newest games and entertainment.

What is the culprit in the vast majority of these cases? An inferior and aged graphics card. In a market where computers contain far more power than even the most advanced word processors and web sites, the graphics cards are improving in amazing leaps and bounds. Games and 3d development software have pushed system hardware to the point that developers can no longer rely on a fast CPU; they rely on the graphics card within your computer. Complex graphical calculations that are performed on a modern graphics card are highly specialized and far more demanding than any ordinary Microsoft Word.

If you are in the market for value, performance, and an easy and stable upgrade from your current graphics card, look no further than BFG Technologies’ new GeForce FX 5900XT OC. After many long days of testing, I can confidently and happily pat BFG Tech on the back for a job well done. They have delivered a card that performs excellently, offers excellent capability and compatibility with today’s latest games, and still offers a value that would make one lose sleep at night if it were ignored.

Specifications


nVidia “Performance Card” Specification Comparison
GPU nVidia GeForceFX 5900XT BFG GeForceFX 5900XT OC nVidia GeForceFX 5900 nVidia GeForceFX 5950 Ultra
Bus Type AGP 8x/4x/2x
Memory 128MB DDR 256MB DDR
Core Clock 390MHz 430MHz 400MHz 475MHz
Memory Clock 700MHz 735MHz 850MHz 950MHz
RAMDACs Dual 400MHz
API Support Microsoft DirectX 9.0 and below, OpenGL 1.5 and below
Connectors VGA,DVI,S-Video out,1 Molex Power Connector
Vertices/sec ??? 322 million 300 million 356 million
Memory Bandwidth 22.4GB/s 23.5GB/s 27.2GB/s 30.4GB/s

Packaging/Bundle


When I received the package, I first noticed that it was lighter than it appeared. Upon opening it, I immediately discovered why: BFG opted to not include any extras and went straight for raw utility. All that was included was the GeForce FX 5900XT board, a software install CD, a “Y” power cable splitter, a DVI-to-VGA converter, and a quick install manual – not too much, but enough to get up and running moments after picture-taking was complete.

The card itself has an attractive blue PCB and a sleek heatsink. At first glance, the sink doesn’t seem like it would cool well…but during operation, that first glance is for naught. The heatsink and ram sinks are all located on the same side of the board, as are all 128MB of memory chips. This allows for better cooling, as there doesn’t have to be a separate cooling medium on the other side of the card.

There are a few nice things to be found on the software CD: a copy of nVidia NVDVD 2.0, the usual collection of nVidia product demos, drivers, and an installation manual. I especially liked how BFG included the power “Y” adapter and the DVI-to-VGA converter; these will help many buyers overcome tight situations (literally tight situations…what if you had no more free power connectors within your case, and were forced to purchase a new power supply and/or “Y” adapter?) and use the product’s dual-monitor feature out of the box. Last but certainly not least, BFG included a bunch of stickers. I am still deciding what exactly to do with them, but I’m sure an opportunity will present itself!

Installation/Compatibility


This is an area that many people may either fear or approach doubtfully. Fortunately, both nVidia and 3rd-party developers have you covered! I used DriverHeaven’s Driver Cleaner and Cab Cleaner software to remove all remaining driver/registry residue from my previous Hercules Radeon 8500 128MB installation, and installation of the new BFG card was a dream. I experienced no hardware crashes, game/software compatibility problems or glitches, or any issues whatsoever on my Athlon 2200+/ECS K7S5A/Windows 2000 Professional test system. After testing the card for awhile in my main system for compatibility issues, I installed it in my new benchmarking system with a fresh Windows 2000 Professional install and once again had absolutely no complaints.

Drivers/Overclocking


After installation, it is always a good idea to grab the latest and greatest drivers for your graphics card to achieve maximum performance, fix bugs, and (maybe) introduce new features. This is why I downloaded and installed nVidia’s 56.72 WHQL certified driver set. Read on to see the headline features offered by nVidia’s drivers!

First off, there are several standard options that need to be covered. The Performance & Quality Settings contain image quality options, allow you to modify Antialiasing and Anisotropic filtering settings, and basically give you control over more “gaming specific” things. Read the captions below each image for specific descriptions.

nVidia Control Panel - Image Quality nVidia Control Panel - Performance Options

nVidia includes many options for adjusting/tweaking the color and brightness of your display. Within the “Color Correction” settings you will find nVidia’s Digital Vibrance control which “boosts” colors and can improve color quality. Beyond the more standard Brightness, Contrast, and Gamma settings, you will find nVidia’s Image sharpening slider which helps emphasize individual pixels to allow for a “sharper” picture (if you push the slider all the way over to the maximum, however, things will look a bit strange!).

nVidia Control Panel - Color Settings

Application profiles are one of nVidia’s latest features for their ForceWare driver set. Profiles allow you to set different Antialiasing, Anisotropic filtering, etc. settings per application. For example, you could enable 4x Antialiasing and 8x Anisotropic filtering for Quake3, but have a different profile that will disable them for Far Cry. This allows you to tweak the settings for each game/application so you won’t have to enter the nVidia control panel before each game.

nVidia Control Panel - Application Specific Settings

Overclocking features aren’t enabled by default, but thankfully the easy-to-install Coolbits registry hack grants you access to nVidia’s hidden but powerful overclocking features. The first step to installing Coolbits is of course downloading it at the link in the previous sentence. After installation, you will see a new screen like below in your nVidia control panel, and a “Clock Frequencies” entry in the menu. Simply toggle Manual overclocking, say “yes” to the warning that pops up :-) and begin your overclocking adventures! nVidia helps make sure that you won’t burn out your card by providing newcomers with an “Auto Detect” button that will automatically find and test new clock and memory speed values.

Default Clockspeed Auto Overclock

If you are really feeling adventurous, nVidia makes it easy to get the most out of your graphics card purchase…and BFG Technologies helps you along with their “OC” moniker that offers further encouragement. In my testing, the best overclocking speed I attained using Coolbits was 471MHz for the core clock speed and 795MHz for the memory clock.

Final Overclock after Auto O/C

Overall, I am very pleased with nVidia’s ForceWare driver package and nothing comes to mind that is reason for complaint. Mature drivers are definitely a very big plus for nVidia cards, and this card is no exception.

Testing Platform


Before we go into any performance analysis, it is first critical that the testing platform be presented:

  • AMD Athlon XP 2500+ “Barton” CPU
  • 512MB DDR-RAM running at 333MHz DDR
  • ASRock K7S8XE+ Motherboard
  • SiS 748 Northbridge Chipset
  • SiS 964 Southbridge Chipset
  • WD 310100 5400RPM HD
  • Microsoft Windows 2000 Professional w’ Service Pack 4
  • nVidia ForceWare 56.72 WHQL Drivers
  • Microsoft DirectX 9.0b
  • All Microsoft patches up to 4/15/2004

Look for a motherboard review on the ASRock K7S8XE+ in the future here at Bjorn3D.

Benchmarks


Just as important as the hardware used during the benchmarking process, the software is what we use as an accurate gauge of performance. Below are the software packages I used to review the BFG GeForce FX 5900XT OC:

Real World:

  • Unreal Tournament 2004 Demo
  • Quake 3 v1.32 NV15Demo*
  • AquaMark 3 Commercial
  • Gunmetal Benchmark 2
  • X-2: The Threat Rolling Demo

Synthetic:

  • CodeCreatures Benchmark
  • ChameleonMark
  • 3dMark2001 SE build330

* “This is a high polygon demonstration Level built specificlly for Nvidia.
It has been built completely from scratch, and was built with a 5 week deadline to be ready for GDC 2000 (March 8th, 2000).”

The benchmarks on this page are a arranged for easy viewing and quick gathering of data. The next two pages contain more detailed looks at each graph, and comparisons with the BFG 5900XT OC running at different clock speeds: both the maximum attained clock speed during testing (476MHz core/795MHz memory) and “underclocked” to standard GeForce FX 5900XT speeds (390MHz core/700MHz memory). The in-depth benchmarks are very bandwidth-intensive and a broadband connection is recommended for those with a less patient soul. Don’t forget that if you wish to skip the “In-Depth Benchmarks”, at the bottom of each page is a link that will take you the conclusion of this article.

Just as a quick note on the graphs, the “filled” section of the graph (red or blue) represents the BFG board. The other graphs either represent a “reference clocked” BFG 5900XT OC or the max overclock attained during the review process. A green dashed or dotted line across a graph shows where the stock BFG 5900XT has reached its highest FPS point, and a red dashed or dotted line its lowest FPS point. If the filled graph shows as red, that means that the card’s performance has crossed below the “playability threshold” of 30FPS. Hopefully this will clear things up :-).

First, let’s start off with a few “overview” benchmarks. Let me introduce X2: The Threat:

X2: The Threat is a 3d space shooter simulator, complete with an advanced graphics engine and a huge universe to explore and change. This makes it a perfect candidate for benchmarking. Bjorn3D has a review of it here.

X2: The Threat, Rolling Demo
1024×768, Max Detail
no AA/AF 4X AA/8X AF
53.126 fps 43.693 fps

3dmark2001 SE, a synthetic benchmark, is a standard part of any benchmarking suite.

3dMark2001 SE
Standard Settings
no AA/AF 4X AA/8X AF
11826 3dMarks 9311 3dMarks

Aquamark is another staple benchmark; it is full featured and makes heavy use of Directx 8.x class hardware shaders.

Aquamark 3
Standard Settings
no AA/AF 4X AA/8X AF
38,076 Aquamarks 28,852 Aquamarks

The GunMetal benchmark is another game that makes heavy use of shaders and is an excellent test for modern hardware.

GunMetal Benchmark 2
Standard – no AA/AF
Min 11.99 fps Max 59.29 fps Ave 31.19 fps
Standard – 4X AA/8X AF
Min 11.98 fps Max 57.6 fps Ave 31.17 fps

ChameleonMark is an nVidia-produced synthetic benchmark that tests shader performance.

ChameleonMark
1024×768 – no AA/AF
Glass 178.828 fps Real 230.532 fps Shiny 177.238 fps
1024×768 – 4X AA/8X AF
Glass 123.503 fps Real 108.098 fps Shiny 122.692 fps
1280×1024 – no AA/AF
Glass 126.749 fps Real 165.011 fps Shiny 126.749 fps
1280×1024 – 4X AA/8X AF
Glass 89.0146 fps Real 79.4157 Shiny 88.1746 fps
1600×1200 – no AA/AF
Glass 97.6486 fps Real 127.56 fps Shiny 97.141 fps
1600×1200 – 4X AA/8X AF
Glass 60.7477 Real 56.1186 Shiny 60.3553 fps

Unreal Tournament 2004 is a new game that contains complex geometry and advanced AI.

Unreal Tournament 2004 Benchmarks
Botmatch – 1024×768 – No AA/AF
BR Colossus 84.91 fps AS Convoy 50.98 fps DM Rankin 125.13 fps
Botmatch – 1024×768 – 4xAA/8xAF
BR Colossus 80.34 fps AS Convoy 49.53 fps DM Rankin 104.81 fps

Unreal Tournament 2004 Benchmarks
Botmatch – 1280×1024 – No AA/AF
BR Colossus 89.10 fps AS Convoy 50.70 fps DM Rankin 120.42 fps
Botmatch – 1280×1024 – 4xAA/8xAF
BR Colossus 67.70 fps AS Convoy 47.52 fps DM Rankin 78.90 fps

Unreal Tournament 2004 Benchmarks
Botmatch – 1600×1024 – No AA/AF
BR Colossus 84.04 fps AS Convoy 48.55 fps DM Rankin 107.87 fps
Botmatch – 1600×1024 – 4xAA/8xAF
BR Colossus 53.97 fps AS Convoy 39.13 fps DM Rankin 61.09 fps

CodeCreatures is a synthetic benchmark designed to stress DirectX 8.x class hardware.

CodeCreatures Benchmarks
No AA/AF
1024×768 45.3 fps 1280×1024 36.0 fps 1600×1200 28.5 fps
4xAA/8xAF
1024×768 23.5 fps 1280×1024 15.2 fps 1600×1200 11.7 fps

Quake 3 has been the industry standard since its release in late 1999. What makes this game so long-lived is its OpenGL graphics engine and easy modification capabilities, which have been proven time and again in countless games and modifications.

Quake 3 Benchmarks
NV15Demo – No AA/AF
1024×768 63.1 fps 1280×1024 62.9 fps 1600×1200 62.9 fps
NV15Demo – 4xAA/8xAF
1024×768 62.0 fps 1280×1024 59.7 fps 1600×1200 49.6 fps


In-Depth Benchmarks


After many hours of reviewing, here is the culmination: a complete set of benchmarks that includes a DirectX game, an OpenGL game, and a synthetic benchmark based on DirectX. Hopefully the comments will prove insightful at the very least.

There is a lot of data here to process, so let’s get started!

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1024×768 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
125.13 fps 125.38 fps 124.96 fps

The results here are easily within the range of error. For the most part, it is easy to see that the lower resolutions are more CPU limited than anything else. You can see some very high Unreal Tournament 2004 scores here…definitely a good sign that there is plenty of headroom to be found while playing current games.

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1024×768 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
104.81 fps 95.23 fps 110.88 fps

Things change very fast when we enable antialiasing and anistropic filtering. BFG’s card takes the upper hand over the standard 5900XT board, and the overclock boosts the BFG even farther! Definitely not a bad showing! Keep watching to see what happens as we crank up the resolution.

Unreal Tournament 2004 Demo

Botmatch – AS Convoy
1024×768 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
50.98 fps 51.02 fps 50.93 fps

Without antialiasing, the Assault mode Convoy map also seems to hover at about the same average FPS. Why is this? Same reason as before…more of a CPU limitation. This particular map is very stressful on the CPU, as there is are lot of AI calculations occurring. Fortunately, the game still is playable at this resolution.

Unreal Tournament 2004 Demo
Botmatch – AS Convoy
1024×768 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
49.53 fps 49.91 fps 49.96 fps

The Convoy map hardly takes any performance hit here, even with antialiasing and anistropic filtering enabled. You can tell that graphics card is taking its time, and is hardly being stressed.

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1024×768 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
84.91 fps 85.05 fps 84.98 fps

The Bombing Run demo map manages to maintain a very good FPS across the array of board speeds once again. At these speeds and with scores so close, it is impossible to tell which is truly faster since they are all within the margin of error.

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1024×768 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
80.34 fps 79.85 fps 81.78 fps

The card takes a hit here at this resolution, but if you were playing the game it wouldn’t be noticeable. This board is definitely more than good enough for playing Unreal Tournament 2004 at 1024×768 with all the eye candy turned on. Remember to keep an eye on the graph for big dips! This is a scene of high intensity during the game, and either the graphics or CPU are being slammed.

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1280×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
120.42 fps 113.64 fps 125.49 fps

Once again we find ourselves examining UT2004′s deathmatch map, Rankin. You can finally see a decent difference between each board speed; still, there is plenty of excess power here as the card hasn’t managed to fall even below 80FPS.

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1280×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
78.90 fps 73.00 fps 85.96 fps

Amazingly enough, the board continues to dominate this benchmark at all speeds. Here, the board is doing a lot of processing…fillrate is beginning to suffer, but the game plays smoothly.

Unreal Tournament 2004 Demo
Botmatch – AS Convoy
1280×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
50.70 fps 50.63 fps 50.65 fps

Convoy also has taken a performance hit from the previous resolution, but it isn’t very noticeable. The average FPS still is very similar across all three board speeds, and for the most part (disregarding the start of the benchmark, which could be an error) is still playable.

Unreal Tournament 2004 Demo
Botmatch – AS Convoy
1280×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
47.52 fps 45.55 fps 48.89 fps

The difference between the card speeds increases notably as we turn on antialiasing and anistropic filtering. The game is slowly approaching the 30FPS playability barrier but fortunately still hasn’t crossed it. Will the final resolution finally bring the board across the mark in the Convoy map?

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1280×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
89.10 fps 89.01 fps 89.28 fps

As when last visited, Colossus still manages plenty of FPS across the board and offers excellent playability.

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1280×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
67.70 fps 63.09 fps 73.09 fps

The numbers have plummeted some 10-12FPS on average, but the game is still very playable at this resolution and with antialiasing and anisotropic filtering enabled.

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1600×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
107.87 fps 100.07 fps 116.31 fps

Interestingly, I couldn’t get the benchmark to run at 1600×1200 resolution. I did manage 1600×1024, so the remaining UT2004 benchmarks will be run at that resolution. All three board speeds have an average above 100FPS here, and minimums don’t even drop below 55FPS. There is a significant difference between the speeds here, though. Although they all maintain an excellent FPS, the higher clocked boards are starting to pull away at this resolution.

Unreal Tournament 2004 Demo
Botmatch – DM Rankin
1600×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
61.09 fps 56.35 fps 66.59 fps

As before, the higher clocked boards are markedly faster than the standard reference-clocked 5900XT. The reference speed here dips teeteringly-close to the 30fps line, and barely hovers over it. However, we have reached an important point for this card. The BFG 5900XT has managed to maintain playable frame rates with antialiasing and anistropic filtering enabled, even all the way up to this “max” resolution! We can say with relative surety that this game runs best at 1280×1024 with antialiasing and anistropic filtering enabled, and at 1600×1024 (unfortunately, we couldn’t test at 1600×1200) for those brave souls with either faster CPUs or a will to overclock.

Unreal Tournament 2004 Demo
Botmatch – AS Convoy
1600×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
48.55 fps 48.57 fps 48.65 fps

The convoy map is certainly fascinating in how its average FPS scores have hardly varied across all the resolutions. Like noted before, this is due to CPU limitations. However, there are many extremes to be noted in this case. Towards the beginning and the end of the benchmark, FPS have dropped nearly to our 30FPS barrier!

Unreal Tournament 2004 Demo
Botmatch – AS Convoy
1600×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
39.13 fps 36.63 fps 41.47 fps

With scores already close to 30FPS at the beginning and end of the benchmark, it could have been predicted that the threshold would finally have been crossed with antialiasing and anistropic filtering enabled. Here, we see scores as low as 24FPS for the reference-clocked 5900XT, and only slightly better for the BFG and overclocked models. The game seems to be mostly playable at this resolution (see the DM Rankin results with AA/AF), but so far the most reassuring resolution has been 1280×1024 with antialiasing and anistropic filtering. Of course, a better CPU and motherboard could make things look better here.

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1600×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
84.04 fps 82.27 fps 85.34 fps

Colossus is very playable. Lets see how it holds up with anistropic filtering and antialiasing enabled.

Unreal Tournament 2004 Demo
Botmatch – BR Colossus
1600×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
53.97 fps 49.95 fps 58.59 fps

Great! We still see that the board has managed to keep up excellent FPS, even in such high resolutions and with antialiasing and anistropic filtering enabled. It is at these high resolutions that we see the greatest benefit of BFG’s board over the reference speeds, and of course it is hard to ignore the excellent performance offered by overclocking!


In-Depth Benchmarks (Continued)


Let’s continue…

… there are only a few more benchmarks left!

That means only a few more graphs for you to mull over.

CodeCreatures Benchmark 2002
1024×768 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
45.3 fps 42.5 fps 47.9 fps

CodeCreatures is one of the few synthetic benchmarks here. This can be taken over the game benchmarks with a grain of salt, but it can still be a good indicator of overall board performance. As you can see, this benchmark is more graphics-board limited than the UT2004 runs. There is a noticable gap between each GPU’s speed, and this shows a nice scaling with each speed setting.

CodeCreatures Benchmark 2002
1024×768 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
23.5 fps 21.9 fps 25.4 fps

Wow! This is a totally different picture from the UT2004 benchmarks. Here, you can see that enabling antialiasing and anisotropic filtering have managed to give the card a 50% performance hit at the same resolution. This benchmark is definetly taxing the card quite a bit, with minimum FPS dropping to 11!

CodeCreatures Benchmark 2002
1280×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
36.0 fps 33.7 fps 38.7 fps

Once again, you can see large differences between each speed. Not too much of the graph is in the red, but there is enough to be cautious of the following graph with antialiasing and anistropic filtering enabled.

CodeCreatures Benchmark 2002
1280×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
15.2 fps 14.3 fps 16.3 fps

Ouch. Only the overclocked BFG board has results above the 30FPS boundary.

CodeCreatures Benchmark 2002
1600×1200 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
28.5 fps 26.6 fps 30.8 fps

With scores like this, it can be predicted that 1600×1200 will be a crusher for this card.

CodeCreatures Benchmark 2002
1600×1200 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
11.7 fps 11.0 fps 12.5 fps

Not much to be said here; we saw this coming. Remember that this is a synthetic benchmark, which means that it may not accurately reflect real gaming speeds.

Quake 3 v1.32
NV15Demo – High Quality
1024×768 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
63.1 fps 63.2 fps 63.1 fps

Quake 3 has been a standard OpenGL benchmark for years, and finds itself in service once again here with the NV15Demo map. The NV15Demo map was originally created by nVidia to show off its NV15 board (original GeForce), and it still proves itself to this day as one of the most (if not THE most) stressful Quake 3 maps to this day. This is evident in its very erratic performance that drops below 40FPS at one point!

Quake 3 v1.32
NV15Demo – High Quality
1024×768 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
62.0 fps 61.9 fps 62.1 fps

Once again we see strong but erratic performance. The NV15Demo has been around a long time, but we aren’t seeing the 200+fps you would be witnessing had we been using a standard Quake3 demo benchmark. NV15Demo definetly is stressing the video card.

Quake 3 v1.32
NV15Demo – High Quality
1280×1024 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
62.9 fps 63.0 fps 63.0 fps

Average FPS aren’t dropping below around 63FPS…the game is still running well above our minimum FPS barrier.

Quake 3 v1.32
NV15Demo – High Quality
1280×1024 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
59.7 fps 58.2 fps 60.8 fps

Enabling antialiasing and anistropic filtering pushes it a tad more, but it still is playable. The small “sub-30FPS” drop at the beginning of the benchmark is an interesting phenemonon.

Quake 3 v1.32
NV15Demo – High Quality
1600×1200 NO Antialiasing/Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
62.9 fps 62.8 fps 62.9 fps

Even at 1600×1200 the game’s average FPS hasn’t dropped much below 60FPS…performance is still erratic, but that is due to the stress of the demo run.

Quake 3 v1.32
NV15Demo – High Quality
1600×1200 4X Antialiasing/8X Anisotropic Filtering
BFG 5900XT OC
(430/735)
Reference Standard 5900XT
(390/700)
Max Overclock BFG 5900XT
(471/795)
49.6 fps 47.4 fps 52.6 fps



Conclusion


First of all, I need to say that I have had absolutely no problems with this card. Installation was simple and straightfoward, hardware quality and design is superb, and performance has exceeded expectations. The BFG 5900XT OC has managed to change my opinion of nVidia and has me looking respectively at BFG for a well-executed product. The only real issues I can see is that there is a sore lack of a software bundle, and that the product may be hard to find on the market which could lead to your finding prices closer to the “list price” rather than the street price. Hopefully as the product proliferates throughout the market, we will see the price begin to plummet and settle closer to the other Geforce 5900XT boards (BFG wants this graphics card to be seen as a normal 5900XT but with a free overclock out of the box, and thus at the same price point). This board can be recommended to someone looking to purchase a graphics card in the ~$200 range, and has enough “staying power” to be well-performing in any game for the next 2 years or so. Also, considering overclocking prowess, the BFG 5900XT is perfect for the budding overclocker and experienced tinkerer alike. These reasons plus the impressive compatibility and excellent support offered by nVidia and BFG Technologies merge to give this solid product a rating of 9 and a Bjorn3d Seal of Approval!

List Price*: $229

Street Price**: $215

*Price at BestBuy.com.
**The lowest price I have seen for this board is at Essential Computer for $215.

Pros

  • Excellent Performance
  • Top notch drivers
  • Efficient design
  • Overclocking capabilities
  • Package that inlcudes VGA->DVI Adapter
  • Runs quietly

Cons

  • Price could be lower, and that will happen hopefully as the product spreads throughout the market
  • Bundle doesn’t have many extras
  • Product can be hard to find


(9/10)


Optimization WordPress Plugins & Solutions by W3 EDGE