| Overclocking the NVIDIA GeForce Video Card |
| Articles - Featured Guides | |
| Written by Olin Coles | |
| Saturday, 06 September 2008 | |
Overclocking NVIDIAIt has been almost two years since I sat down and wrote my guide on How To Overclock the NVIDIA GeForce 8800 Series. As the title implies, the material covers what is now considered an obsolete product... so it's a little outdated, but certainly not irrelevant. When I originally wrote that article, I was an avid enthusiast who was just beginning to write product reviews. Back then I had only one video card worth playing video games with, and I massaged it in every way possible looking for that last bit of hidden performance. But now that things are different for me and my collection of products fills the entire Benchmark Reviews Video Cards section, I find that overclocking is a hobby of the past. Because I receive many different products for review testing, there are two things that have happened as a result: 1) I don't have time to play video games anymore, and 2) I don't need to overclock any of these graphics products because already arrive that way. However, occasionally I receive a top-end product that already tips the scales, and the enthusiast in me comes back out to play with some overclocking experiments. That's why I've returned to this article, and decided to update my previous guide with a fresh and current approach to the topic. Overclocking can take on many forms, and experiments can range from minor product improvements to a total re-engineering project that completely alters the hardware. Because of their long history of successful product launches, NVIDIA is the most recognized name in the graphics industry. Their GeForce series of desktop graphics accelerators has become an enthusiast favorite, primarily because of excellent overclocking potential. For gamers, overclocking unlocks a hidden potential to change the graphics quality settings from medium to high without a hardware upgrade. The results are often times worth the risk, and in this article Benchmark Reviews will focus on achieving the most gain possible from any NVIDIA GeForce video card through overclocking, while using the least amount of effort.
Some industry voices have called overclocking a hobby, while others have compared it to product misuse. However, I believe that if you are reading this article, you are probably one of the many computer enthusiasts who believe it is perfectly acceptable to get something more out of a product without it costing you more money. When I think about it, everyone enjoys getting something for nothing; it's human nature. Additionally, it is also human nature to blame someone else if something goes wrong. This is where I present a warning to you, the reader of this article, that this article is not a recommendation to overclock your video card. This article explains how I conducted these experiments on my own personal property. Benchmark Reviews and the author of this article will not be responsible for damages or injury resulting from experiments you choose to conduct on your own property. If you read beyond this point, you are accepting responsibility for your own actions and hold Olin Coles and the staff of benchmarkreviews.com harmless. So with the legal disclaimer out of the way, here's what you can expect. This article covers several key areas, most of which you will likely skip so that you can jump right into overclocking. The first section will cover the end-goal of overclocking and discusses when it should be temporary or made permanent. Then the discussion moves on to the preparation and actual overclocking procedure. After that, I review a few different tools that could help make the tasks much easier. At the end of this article, Benchmark Reviews offers up our results on the performance difference between a reference stock and overclocked video card. So if you're ready, let's begin... Planning the ProjectLike any project, there are a number of different items that need consideration before you begin. Not everyone wants their hardware modifications to become permanent, and others are forced to make them temporary because of manufacturer warranty or store return policy concerns. In this section, I will cover the different directions that can be taken with this video card overclocking project. Before I begin though, there are a few common understandings that should be committed to memory. The first is that SLI sets should not be overclocked or tested together as a team. It is best to remove one video card from the set while overclocking and testing on the other, and then vice-versa. This allows each video card to be stress-tested for faults and instability before they are joined into a mulit-card array. The next thing to consider is the limitation of your product. Most manufacturers are already "factory-overclocking" their graphic cards, so often times these products have very little head-room for additional performance. Lastly, consider the side-effects of overclocking on the rest of your computer system. Overclocked video cards generate much more heat than normal, and adjacent components suffer a reduced stability without additional cooling. To begin my project, I decided to exeriment with both temporary and permanent overclocking. There's a short list of free software needed to complete these tasks:
Once I have all of the tool's needed for this project, I am ready to begin work... almost. Before I begin to overclock any video card, I ensure that my results are not somehow skewed by an outside source. One particular concern is drivers, because having the latest developers alpha or testers beta driver can often cause stability problems that make lead you to think there's something wrong with the hardware when it's really a software (driver) issue. It's a smart idea to use only the most commonly accepted stable version of the graphics driver available. Driver Cleaner Pro is also a good tool for making sure you have no previous driver remnants attached to the Operating System.
Next up is cooling. Heat is never friendly towards electronics, and there are times when you'll want to use this to your advantage. Keeping in mind that the cooling fan(s) found on most modern video cards can have their speed modified, I usually stress test with a reduced fan output (thus increasing test temperatures) and later increase the full-time fan output. Doing this allows me to test stability in my video card at the most grueling temperature acceptable, and once my experiments are complete the added fan power ensures that heat never ruins my game. Testing the stability of a video card with the fan turned up on full power only removes the opportunity to discover overclock instability. RivaTuner is a great tool for adjusting fan speeds from within the system settings (both standard and low-level). With the fan set to low, I am now ready to begin the overclocking. The next section covers a temporary overclocking solution, which may be as far as some people are comfortable in going. I usually dislike the extra software running in the background, as it only leaches away system resources from the real focus of my attention. Nevertheless, the steps I will cover in the next section are critical and required before any permanent changes can safely be made. ATITool OverclockingThe idea of permanently overclocking the video card can be a frightening thought for many enthusiasts. Most people aren't overclocking as part of a deep-rooted passion or hobby; they are simply gamers who want more. So it's very convenient (and safe) for the overclocking to be conducted on a reversible software level. I have used two different tools for many years now: ATITool and RivaTuner. Using these tools together offers me the ability to control fan output, change clock speeds, and create load for testing the video cards ability to function without producing artifacts. ATITool works extremely well with 8800 series and older GeForce products, while RivaTuner works well with everything new and old. Although W1zzard has named his application ATITool, this great tool has worked for both ATI and NVIDIA products for many years now. After a straightforward installation and system reboot (to complete the installation of the driver-level service), I open ATITool and see a lot of options confronting me. Try not to be overwhelmed, since all of these options could strike fear in the hearts of the inexperienced. Of the two tools mentioned this is the easier of the two in my opinion, but it's limited to older GeForce models until development resumes. The first screen displayed will be the only screen that is really needed. While the memory clock speed copies its adjustments across all three performance levels (2D, 3D low power, and 3D performance), the software allows the user to make individual speed settings to the GPU core clock speed. Initially, the default values are saved in the profile named "default", but as you make changes you may save and delete profiles as needed. For this project, I kept raising my speeds and saving them into a profile named "MAX".
To begin the overclocking process, I start by raising the temperature of the GPU core by using the "Show 3D View" button to display a rotating fuzzy cube. It is critical that the video card attain the highest temperature possible prior to overclocking, because the results of a cold overclock may prove unstable during gaming conditions. After reaching the loaded operating temperature, I change to the "Scan for Artifacts" view by pressing the button. Although I could have used the "Find Max Core & Mem" buttons to have ATITool automatically work out the best settings, I choose to manually test each incremental improvement on my own.
Experience has taught me that overclocking the memory is the best starting point, since it has a very small impact on operating temperatures when compared to overclocking the GPU. I have also learned that clock speed improvements can be made in larger steps on RAM (10MHz steps), than they can be made on the GPU (3MHz steps). It is recommended that each clock be adjusted in small increments one after the other. Do not find the maximum RAM clock and then set out to find the best GPU clock; this will give skewed results at either end. I tested at different points in my experiment using the "Scan for Artifacts" function for ten-minutes. After stability was successfully tested, I played some of my favorite video games for an hour to confirm real-world stability. The settings can be saved to a custom-named profile at any time, and the program can load with Windows or be triggered when a 3D application is opened. RivaTuner OverclockingAlexey Nicolaychuk, also known as Unwinder, is the brilliant mastermind behind the excellent tool named RivaTuner. Much more than just an overclocking tool for video cards, RivaTuner can manage all aspects pertaining to the computers display: from multimedia and video game settings, to monitor refresh rates and positioning. Once installed, the application opens to a complex assortment of tabbed menus. Even though I have used this tool for many years, the ergonomics of the program can leave the inexperienced enthusiast a little confused. For this guide, I will focus on the tools required to complete temporary overclocking on a graphics card.
Although ATITool is a past favorite because of BIOS level features, it doesn't offer many of the settings needed for basic temporary changes like RivaTuner does. Low-level system setting changes can be completed after the hardware is manually detected or after a complete system reboot. The fan controls (illustrated in the last section) and overclocking can be completed from the system settings section, which is opened by clicking the "Customize" arrow on the far right side of the Driver settings bar.
For this article I used the ZOTAC GeForce GTX 280 AMP! Edition Video Card, which takes the NVIDIA reference values (602 MHz GPU, 1296 MHz shader, and 1107 MHz RAM) and factory overclocks them (to 700 MHz GPU, 1400 shader, and 1150 MHz RAM). Most people would leave well-enough alone, but I have faith in NVIDIA's chip-binning process and believe that there's plenty more performance to be squeezed from this graphics card. So after several very small incremental increases to the core clock (which I kept linked to the shader clock), I would return to ATITool for and FurMark for artifact scanning and stress testing. Because this is a low-level process that modifies settings through direct access to the hardware driver, whenever my overclocking was too zealous and resulted in a crash all I had to do was restart my computer. This would happen several times, but eventually my effort paid off. After several trial-and-error experiments to find the optimal combination of GPU, shader, and memory speeds, I settled at a stable configuration. Using both the synthetic tests and real-world game play, the final overclock was: 715 MHz GPU (113 MHz over stock and 15 MHz over factory OC), 1430 MHz shader (134 MHz over stock and 30 MHz over factory), and finally 1385 MHz RAM (278 MHz over stock and 235 MHz over ZOTAC's factory overclock). Was it all worth it? Wait until you see the results in the following section! Once the maximum stable speeds for the GPU, shader, and RAM have been found and tested, it's time to make a big decision: do I keep using the ATITool or RivaTuner software to overclock my video card, or should I program the new settings into the video card BIOS and make the changes permanent? If I ever want to use this video card as part of an SLI set, there's only one choice: I will need to flash the settings to the video BIOS of each card. But if I only use a single video card, then the choice is purely personal. OC Performance ResultsI'm sure that people often wonder what kind of difference overclocking can have on a video card, which is exactly why I've put these charts together. Obviously every product is different and each model will produce its own unique results, but this should confirm the obvious: overclocking gives you more performance for free! Stepan Hrbek is the mastermind behind Lightmark 2007, a program that allows you to benchmark real-time global illumination. Typical workloads in real-time rendering will shift, and Lightsmark simulates it. Global Illumination renders often take hours, so before Lightsmark real-time global illumination was limited to small scenes, small resolutions, low speeds, and specially crafted scenes with handmade optimizations. Lightsmark breaks all limits at once, running in reasonably sized scene (220000 triangles) in high resolutions at excellent speed. Lightsmark is comparable to lower-demand OpenGL video games, such as: Prey, Quake 4, and Doom 3. This scene places medium-high demands on a graphics card processor and tests the maximum speed with which the scene can be properly displayed at each resolution. At the 1920x1200 resolution tested, the reference NVIDIA GeForce GTX 280 video card produced a 455 FPS for the scene, which the ZOTAC GeForce GTX 280 AMP! Edition rendered at 510 FPS. After an hour of trail-and-error overclocking, that same ZOTAC graphics card was pushed to produce 521 FPS. This results in a 15% improvement over stock, and 2% increase over the factory overclock. 
Call of Duty 4: Modern Warfare runs on a proprietary game engine that Infinity Ward based off of the tried-and-true Q3 structure. This engine offers features such as true world-dynamic lighting, HDR lighting effects, dynamic shadows and depth of field. "Bullet Penetration" is calculated by the Infinity Ward COD4 game engine, taking into account things such as surface type and entity thickness. Certain objects, such as cars, and some buildings are destructible. This makes distinguishing cover from concealment important, as the meager protection provided by things such as wooden fences and thin walls does not fully shield players from harm as it does in many other games released during the same time period. Bullet speed and stopping power are decreased after penetrating an object, and this decrease is calculated realistically depending on the thickness and surface of the object penetrated. This version of the game also makes use of a dynamic physics engine, a feature which was not implemented in previous Call of Duty titles for Windows PC's. The new in-game death animations are a combination of pre-set static animations combined with ragdoll physics. Infinity Ward's use of the well-debugged Quake 3 engine along with new dynamic physics implementation allows Call of Duty 4 to be playable by a wide range of computer hardware systems. The performance may be scaled for low-end graphic cards up to 4x Anti-Aliasing and 16x Tri-linear anisotropic texture filtering. The coup d'etat scene when Al Asad takes over control offers medium-high stress on the video card. At 1920x1200 the reference NVIDIA GeForce GTX 280 video card produced a 92 FPS for this intro scene, which the ZOTAC AMP! Edition rendered at 98 FPS. Once I further-overclocked the ZOTAC graphics card, my performance improvements produced 104 FPS. This equals a 13% improvement over stock reference, and 6% / 6 FPS increase over the factory overclock.
Crysis uses a new graphics engine: the CryENGINE2, which is the successor to Far Cry's CryENGINE. CryENGINE2 is among the first engines to use the Direct3D 10 (DirectX10) framework of Windows Vista, but can also run using DirectX9, both on Vista and Windows XP. Roy Taylor, Vice President of Content Relations at NVIDIA, has spoken on the subject of the engine's complexity, stating that Crysis has over a million lines of code, 1GB of texture data, and 85,000 shaders. To get the most out of modern multicore processor architectures, CPU intensive subsystems of CryENGINE 2 such as physics, networking and sound, have been re-written to support multi-threading. Benchmark Reviews uses the Crysis Benchmark Tool by Mad Boris to test frame rates in batches, which allows the results of many tests to be averaged. At 1920x1200 with no post-processing AA enabled, the reference NVIDIA GeForce GTX 280 was able to produce 43.9 FPS after three test loops. Given a healthy overclock at the factory, ZOTAC's GTX 280 AMP! Edition gave a 48.3 FPS but later pushed performance up to 49.8 FPS after an additional overclock. Crysis places the most stress possible from any currently available video game, and our overclocked GTX 280 produced a 13% improvement over stock and 3% over ZOTAC's factory overclocked performance.
So what does it all mean? Without getting into a complex answer, this data means that overclocking can yield a very notable improvement in video game performance. I tool an already maxed-out ZOTAC GeForce GTX 280 AMP! Edition video card, and pushed it ever farther. But could the same thing be done with a stock GeForce GTX 280? No, probably not to the same degree. NVIDIA has adopted a technique called "binning" where the best GPU's are sorted apart at their Taiwan Semiconductor Manufacturing Company (TSMC) fabrication plant. The very best graphics processors are then later resold to NVIDIA's add-in card partners (AIC's) who build their factory overclocked products with them. Regardless, overclocking your video card clearly offers a benefit to the video frame rates in every game we tested. So now that you're armed with this information, would you make the changes permanent? In the next section, I outline the steps necessary to flash your video card BIOS with settings I've tested to be absolutely stable. Building a custom BIOSIn the first few steps of this guide I discovered the maximum potential of my already factory-overclocked video card. Since I have completed several weeks of testing on my overclocked product with both artificial and real-world usage, it is now safe enough for me to modify the video card BIOS file with my custom settings and make this video operate with enhanced performance without additional software running in the background. Flashing the BIOS is a risky endeavor, and although most often times there is nothing abnormal in the event the risk involved is still very high. If the processes fails, or an error is encountered during the operation, the graphics card will be rendered useless hardware. For these reasons alone it is imperative that precautions are taken to ensure the process of flashing the BIOS does not become a disaster. I have been referring to the process of flashing a custom BIOS to the video card as a permanent event, which isn't entirely true. If a backup of the original BIOS is made and preserved, the overclocked video card can always be returned to its original factory state by re-flashing the BIOS from the backup. But the backup isn't just about reverting back to the factory configuration; by exporting the BIOS we will have a good starting point for adding any custom settings. The biggest hurdle for most people will be getting the computer system to boot into MS-DOS, which is made difficult because 3.5" floppy disk drives are considered obsolete legacy hardware. I personally still use floppy diskettes, but only with the help of a more modern USB-based floppy disk drive. As an alternative, any USB flash drive or recordable optical media can also offer alternatives for creating a boot disk. Windows can create a boot system disk when you choose to format a floppy disk, but there are several other solutions available with the assistance of an online search. I recommend using whatever is most comfortable and convenient.
Save a backupI should only continue after creating a backup of the original video card BIOS, and it's highly recommended that this backup BIOS file be copied and renamed so there will be both a working copy and the original backup file available. I named the original BIOS file "BACKUP.ROM" (although *.ROM and *.BIN work exactly the same), and then created a copy of it which I named "PROJECT.BIN" so I wouldn't get the files confused. There are several ways to backup the video card BIOS, none of which require MS-DOS. Using GPU-Z, you can use the shortcut beside the BIOS version to save the file (or submit to an online database). Another method is with ATITool, which will also save a copy of your video BIOS file to a local drive. However, since I will be using NiBiTor to re-configure the BIOS and also use it to demonstrate the backup process:
Now that a backup of the video BIOS has been copied and stored away for safe keeping, the next step for reading the BIOS file and making changes begins with: File → Open BIOS (choose the renamed copy "PROJECT.ROM" of the video BIOS here). Once again, the novice could become very concerned about the many tabs and options available in NiBiTor, but for my purposes I will only use the Clockrates tab to change the values for 3D speeds. Using the values discovered and tested to be stable in my previous steps, I apply these values into the appropriate fields in the "Extra" row (which is standard-power 3D mode), replacing the original values. Once I have typed in the new Core and Memory values, the modified video BIOS file is saved by choosing: File → Save BIOS (save this modified file onto a formatted floppy disk and name the file something simple with less then eight characters such as "PROJECT.BIN").
NiBiTor supports changes to the low-power 3D mode, and 2D mode, however they offer no performance improvements. Occasionally I will reduce the 2D settings on older NVIDIA video cards in an effort to conserve energy, however the newest GeForce video cards already come with a configuration so low that reducing it further will create screen-draw delays. Alternatively, I could also increase the low-power 3D settings to match those in the standard-power (dubbed 'Extra') field. In the next section, I demonstrate how to flash the modified (overclocked) BIOS back to the NVIDIA GeForce video card. Programming the BIOSUp to this point, I have saved my modified video BIOS named "PROJECT.BIN" onto a floppy disk. I will then also save the nvFlashprogram (which is available for free download) onto this same floppy disk. On a second floppy disk I will use Windows XP (or other tool) to format and create an MS-DOS startup disk. It may not necessary to split the project files and nvFlash from the MS-DOS startup disk because of file sizes and available space, but this is a safe practice which also decreases the chance for possible media problems. Flashing a video BIOS is a very simple process; yet extra care and precaution must be taken or the hardware being flashed may be rendered non-operational. I have taken steps to ensure my computer systems stability will not be compromised by removing any system component overclocking (CPU, RAM, front-side bus speeds) I have previously made, and I have also placed my system on a 1500VA backup battery UPS. With stability ensured, I am ready to move forward and flash the video BIOS using the following steps:
That's it! The hard part is done. Once the computer system reboots after successfully completing the BIOS flash, the video card will reflect the enhanced new performance settings. But am I finished? Probably not. Overclocking quickly becomes an addiction, with a thirst that is difficult to quench. Overclocking Conclusion
|
|
Comments
Thanks.