You can increase the gaming performance of your computer by overclocking your video card. Overclocking is a technique that enables a given hardware part to operate at a clock frequency above its standard frequency, thus increasing its performance. In this tutorial, we will explain in detail how to overclock your video card by using several tips and tricks.
If your PC has a video card embedded on the motherboard (i.e., “on-board video”), you won’t be able to overclock it, as your PC doesn’t have a real video card installed – the video is produced by the motherboard chipset. In the following pages, we deal exclusively with real video cards, the ones that are connected to your PC through an expansion slot.
To learn how to overclock your video card, first you need to learn how a video card works. In Figure 1, you see a very basic diagram showing the video card’s main components and how they are interconnected.
The heart of a video card is its graphics chip, also known as GPU (Graphics Processing Unit). It works at a certain clock rate, sometimes referred to as “core clock” or “engine clock.” When we think of overclocking a video card, the first thing that usually comes to mind is to increase the GPU’s core clock.
The GPUs from NVIDIA have two clock signals, one used by its shader engines and another used by the rest of the chip (the core clock we’ve just described), and this second signal (shader clock) is linked to the core clock, so increasing the core clock you automatically increase the shader clock.
The GPU is connected to the video memory (which is physically located on the video card) using a dedicated memory bus (yellow in Figure 1). This bus, known as a “memory clock,” also works at a certain clock rate. We can increase this clock rate in order to increase the performance of your video card, which we will show you how to do.
One important thing to remember is that nowadays the memory bus usually works by transferring two data per clock cycle, a technique known as DDR (Double Data Rate). Because of this technique, the memory clock can sometimes be referred to as the double of its real clock rate, because the transfer rate achieved by the DDR technique is double the transfer rate of a regular memory transferring just one datum per clock cycle. In order to avoid confusion during our tutorial, we will add the letters DDR after clock rates that are “doubled.” For instance, 300 MHz and 600 MHz DDR are the same thing, as this 600 MHz DDR clock rate is really 300 MHz transferring two data per clock cycle.
The memory bus – which can also be referred to as the memory interface – transfers a certain number of bits per time between the GPU and the video memory – 64 bits, 128 bits, 256 bits, etc. This number is fixed and you cannot change it. In other words, there is no way for you to transform your 128-bit video card into one that is 256 bits. This is a physical limitation. Each bit is transferred through an individual wire on the video card printed circuit board, so a video card with a 128-bit memory interface has 128 wires connecting the GPU to the memory. It is impossible to change this number, as you would need to add 128 more wires between the GPU and the memory chips (and also probably add or change the memory chips). The same thing holds true for the video memory size. You cannot transform your 128 MB video card into a 256 MB one simply because you would need to add more memory chips to it.
The GPU is connected to the motherboard through an I/O slot such as PCI Express and AGP. This connection is also performed at a certain clock rate (100 MHz for PCI Express and 66 MHz for AGP). Some motherboards allow you to increase this clock rate, providing a third option to overclock your video card. Notice that this option depends on the motherboard rather than the video card, as it is the motherboard that controls the I/O slot where the video card is installed. Some overclocking-oriented motherboards also afford an option to increase the I/O slot voltage (i.e., the video card voltage), which can make your video card achieve a higher overclocking.
[nextpage title=”Introduction (Cont’d)”]
We can set up three kinds of overclocking: increasing the clock at which the video processor runs, increasing the clock the video processor uses to communicate with the video memory, and increasing the clock the motherboard uses to communicate with the video card. You can even perform these three options simultaneously in order to explore the maximum performance your video card is able to provide. The first two overclockings are accomplished by configuring the video card; you can change these two clocks on any video card. The third one is done on the motherboard setup and will depend upon whether your motherboard provides this configuration option or not.
The first thing you need to do is to discover the core clock and the memory clock that your video card is currently using. The best way to check this is by using a program called PowerStrip. This is also the program we will use to overclock the video card. Depending on the version of the video driver you are using, it can provide the same functionalities as PowerStrip – including overclocking. Since we cannot be certain that your video driver has this feature, we prefer to use PowerStrip.
Upon running this software for the first time, you will immediately see the clocks your video card is using. After the first time, PowerStrip will start minimized on System Tray. You will need to right click on its icon and choose Performance Profiles, Configure.
Be aware that for ATI-based video cards, PowerStrip will report the real memory clock; for NVIDIA-based video cards, it will report the DDR clock (real clock x 2).
Let’s look at two examples. In Figure 2, you can see the clocks used by our Radeon 9800 Pro: 378 MHz for the video processor (“core clock”) and 337 MHz for memory. In Figure 3, you can see the clocks used by our GeForce 6800 GS: 425 MHz core and 1,000 MHz DDR memory.
You can compare the clocks your video card uses with the manufacturer’s default clock for your video card. Click here to take a look at our “ATI Chips Comparison Table” or click here to take a look at our “NVIDIA Chips Comparison Table.” Please notice that the memory clock rates on these two tables are “doubled” (i.e., DDR).
As you can see in our table, Radeon 9800 Pro has default clocks of 380 MHz for core and 680 MHz DDR (340 MHz x 2) for memory. GeForce 6800 GS has default clocks of 425 MHz for core and 1,000 MHz DDR (500 MHz x 2) for memory. Our two video cards were using the default clock rates set by the chip manufacturer. Small differences below 5 MHz are normal, which means that your video card is not running at a “wrong” clock rate.
Sometimes you will find that your video card is factory-overclocked, meaning that the manufacturer has already set it to run at a higher clock rate. Even if this is your case, you can try to overclock your video card even further.
[nextpage title=”Overclocking Your Video Card”]
Let’s now learn the basic procedure of how to overclock your video card.
After installing PowerStrip, it will be launched every time you turn on your computer and will be available as an icon on the task bar, near the system clock (system tray). To overclock your video card, right click on the program’s small icon and select Performance Profiles, Configure, as it is shown in Figure 4.
It is very important to notice that the changes made to your video card using PowerStrip aren’t permanent and are available only when PowerStrip is minimized on the system tray. If you disable PowerStrip (for example, by running Msconfig utility) or uninstall it, your video card won’t be overclocked anymore.
This also means that if your system freezes or something goes wrong while you are overclocking your video card, simply reboot your system and you will have your PC running well again.
On the screen that will be displayed, you can freely set the GPU clock and the memory clock using the available sliders. See Figure 5.
Note that overclocking may or may not work. After adjusting the clock, run a 3D game in its benchmarking mode (Quake 4, for instance), and check whether the computer freezes or restarts on its own. If this occurs, it means that you have set a clock beyond the capability supported by the video card.
The use of a game in its benchmarking mode is also good for you to see how much performance you are gaining with your overclocking. First, run the game with your video card running without any overclocking. Then, you can compare the results with the score achieved under overclocking. If you don’t know how to use a 3D game in benchmarking mode, we’ve posted some tutorials on that: “How to Use Battlefield 2142 to Benchmark Your PC,” “Testing the 3D Performance of Your PC with Quake 4,” and “Testing the 3D Performance of Your PC with Doom 3 and Far Cry.”
Ideally, you should first find the maximum clock frequency supported by the GPU (i. e., the clock you can set without the computer freezing in a 3D game) and then the maximum clock frequency supported by the video memory. If you try to set both simultaneously, if the computer crashes you won’t be able to determine which clock is wrong, the GPU’s or the memory’s.
Overclocking is a boring trial-and-error process. Raise the GPU clock a little and run a 3D game under benchmarking mode. If the system ran without any problems, increase the GPU clock a little more and repeat the process until you find the exact GPU clock under which your system can run without crashing. After determining this, you will need to repeat the same process for the memory clock, followed by the I/O bus if you want to overclock it as well.
After finding the maximum overclocking spot of your video card, we recommend that you run more than one game in its benchmarking mode, at least three times each, to check if your video card overclocking is really stable.
Here are some tips for improving your overclocking chances.
[nextpage title=”Memory Overclocking”]
Sometimes the video card’s memory chips are running at a speed that is lower than their maximum. For example, you have a video card with a memory chip capable of running up to 500 MHz, but the memory is being accessed at 450 MHz.
If you were lucky to get a card like this, you will find that the memory is highly overclockable. This happens because you will be able to put the memory running at a higher clock rate but still under its specs, and then push the memory over its specs.
First you need to know the maximum clock rate of the memory chips of your video card. You can discover this by taking a close look at the chips. The speed grade is marked on the chip’s body after a dash (ex: -40, -50, -5, etc.) as a number. This number is the memory clock, measured in nanoseconds. To find out the maximum clock rate in megahertz, divide one thousand by this number. In the case of two-digit numbers like 40, 45, and 50, place a decimal point between the two digits. For the calculation, you would use 4.0, 4.5 and 5.0, respectively. There is one exception: memory chips from Samsung labeled as “2A” actually are 2.8 ns chips, not 2 ns chips.
Pay attention, because the number you will find is the real memory clock, not its DDR speed, which is the double of the number you will find.
Let’s give you a couple of examples to clarify this. In Figure 6, you see two of the memory chips used on our GeForce 6800 GS. They are labeled as 2 ns, so their maximum labeled clock rate is 500 MHz (1,000 / 2). As we discussed previously, on this video card the memory was accessed at 500 MHz, so in this case the memory was already working at its maximum labeled speed. Of course, we can try to put it to work above its specs.
In Figure 7, there is one of the memory chips from a GeForce 7900 GT. As you can see, it is a 1.4 ns chip, meaning that it can officially run up to 715 MHz. Since on our GeForce 7900 GT the memory was accessed at 660 MHz, we know that we can push the memory clock up to at least 715 MHz and that the video card will still work well, as the memory clock will remain inside the chips’ official specs. We can still try pushing the memory clock even higher for a real overclocking.
On the next page, we will present some tricks that can help you achieve a higher memory clock.
[nextpage title=”Memory Overclocking (Cont’d)”]
One way to increase the video memory overclocking potential is by checking the cooling system used by the memory chips. If you improve the memory chips’ cooling system, you will probably achieve a higher memory overclocking. You can find three situations here:
- The memory chips are cooled down by the same cooler used by the GPU;
- The memory chips use an independent passive heatsink;
- The memory chips don’t use any cooling device at all.
If your memory chips already use passive heatsinks on them, great. You already have a good cooling solution for your memory chips. See an example in Figure 8.
If the memory chips from your video card don’t come with any cooling device, you can buy memory heatsinks and install them (Zalman ZMHRS1 and Thermaltake CL-C0025 are good examples of this kind of product). The installation process is very simple.
If on your video card the cooler used by the GPU is also used to cool down the memory chips, you can improve the video card cooling by installing a better GPU cooler, like the ones from Arctic Cooling. These high-end coolers will help you increase both the GPU and the video memory overclocking potential. Read our first look article on Arctic Cooling Accelero X1 to see the step-by-step installation of this cooler on our GeForce 6800 GS.
[nextpage title=”Memory Overclocking (Cont’d)”]
As you could see in Figure 10, the memory chips from our video card were under the GPU cooler. The problem is that on some video cards the GPU cooler seems to be used to cool down the memory chips; actually, they don’t even touch them. Pay close attention to video cards where the GPU cooler covers the memory chips, checking if it touches them. You can see an example of that in Figures 11 and 12. The GPU cooler appears to be used to cool down the memory chips, but when you take a closer look, it doesn’t touch the chips! In cases like this, the best solution is to replace the GPU cooler with a high-end GPU cooler that touches the memory chips, so you will be improving both GPU and memory chips cooling.
Another case when you might want to replace the GPU cooler that comes with the video card is when the GPU cooler prevents you from installing passive heatsinks on the memory chips. Take a look at Figure 13. On this video card, the GPU cooler covers part of the memory chips without touching them, so the GPU cooler doesn’t cool down the memory chips. At the same time, it prevents you from installing passive heatsinks on them.
Another trick advanced overclockers use is to play with memory timings. Usually, increasing timings reduces the memory performance but allows you to achieve higher clocks. The trick is to check if the higher clock rate you will be able to configure will really deliver a higher 3D performance. Because of the increased timings, you may actually see loss of performance. Memory timing adjustment is performed by editing the video card BIOS. We will discuss later how to edit your video card BIOS; however, we won’t go into detail on how to change memory timings.
[nextpage title=”GPU Overclocking”]
The main method used to increase the GPU overclocking capability is to also improve its cooling. GPUs can come with passive or active heatsinks, i.e., with or without a fan.
If your video card uses a passive heatsink like the one portrayed in Figure 14, consider replacing it with a good VGA cooler. Take a look at our first look articles about the Arctic Cooling NV Silencer 6 and Arctic Cooling Accelero X1 VGA coolers for a better idea of what we are talking about. If you don’t want to spend money, you can at least modify the heatsink so that a fan can be added to it.
Even if your video card already uses an active heatsink, you may want to consider replacing it with a better cooling solution, such as the Arctic Cooling products we’ve mentioned.
Another trick advanced overclockers do is to increase the GPU voltage. Increasing the GPU voltage usually helps the GPU achieve a higher clock rate. One way of increasing the GPU voltage is by editing the video card BIOS. We will later discuss how to edit your video card BIOS; however, we won’t go into detail on how to change the GPU voltage.
[nextpage title=”Tweaking the I/O Bus”]
As we mentioned, another option you have is to overclock the bus where the video card is connected to AGP or PCI Express, depending on your motherboard.
NOTE: Depending on your video card, you may not notice any performance gain by overclocking the I/O bus. This occurs because you may already have enough free bandwidth on the I/O bus, so it is not creating a bottleneck for your 3D performance. It is very important that you run games in their benchmarking modes before and after playing with the I/O bus in order to check whether you had any real performance gain by overclocking it. If you don’t get any performance improvement by tweaking the I/O bus, it makes no sense to keep it overclocked. In this case, leave it back on its default configuration.
We, however, encourage you to at least try playing with the I/O bus if you want to extract the maximum performance from your video card and then, at the end, determine if it was worthwhile.
Here, the options you will have will depend on your motherboard model. We will explore all possible options; however, your motherboard may not have all of them available.
The main problem here is that only overclocking-oriented motherboards will have a separated clock generator for the AGP or PCI Express x16 bus. On simpler motherboards, a single clock generator is used by all the devices found on the motherboard. If you want to increase the AGP or PCI Express x16 clock rate, you will have to increase the master clock generator rate, which will automatically increase the clock rate used by all other devices.
The problem with this setup is that everything will work overclocked as well, not only your video card. Thus, you may face a situation where you won’t be able to pass a certain clock level, not because your video card can’t go over it, but because some other devices on your system that are also overclocked have reached their clock limit.
The following are some examples of how to overclock your AGP or PCI Express x16 bus.
First you will need to enter your motherboard setup utility, which is achieved by pressing the Del key right after you turn on your PC. Inside the setup, you will need to find where the overclocking options are located. The exact location varies according to the motherboard model. Please note that on some motherboards you need to change some configurations from “auto” to “manual” in order to see the overclocking options.
In Figures 15 and 16, there are two motherboards based on the AGP bus. The AGP bus runs at a default clock rate of 66 MHz.
The first motherboard (Figure 15) uses a single clock generator, and to overclock the AGP bus, you need to increase the CPU external bus. By increasing the CPU external bus, you will overclock not only the AGP bus but also the CPU, the PCI bus and all other devices found on the motherboard. On this motherboard, you need to change the “CPU Host Clock Control” option to “Enabled” in order to have access to the external clock rate configuration (“CPU Host Frequency”). Note how you don’t have access to the option “PCI/AGP Frequency.” This option only displays the new PCI/AGP clock rates based on the new external bus configuration you entered.
On the second motherboard (Figure 16), there is a separated clock generator for the AGP bus. We know that because there is an option called “Adjust AGP Frequency,” which defaults to 66 MHz.
[nextpage title=”Tweaking the I/O Bus (Cont’d)”]
The same idea applies to the PCI Express bus. This bus uses a default clock rate of 100 MHz. If your motherboard doesn’t have a separated clock generator for this bus, you will have to increase the motherboard master clock generator, which will increase all clock rates used by all devices on the motherboard.
In Figure 17, you can see a motherboard without a separated clock generator for the PCI Express bus. The only possibility here is to increase the motherboard master clock generator, which will overclock everything connected to the motherboard.
Some motherboards have a separate configuration for the PCI Express bus, but they don’t have a separated clock generator for the PCI Express x16 slot, which is the one used by the video card. Instead, they have a single clock generator for all PCI Express slots and connections, so when you increase the PCI Express clock rate, you will also overclock all devices that use this bus, including the ones connected directly to the motherboard, such as the SATA ports – and hard disk drives are very sensitive to any increase on the SATA clock. This is the case with the motherboard shown in Figure 18. It has a separated clock configuration for the PCI Express bus (“PCIE Clock” option), but this configuration will increase the clock rate of all PCI Express connections, not only the one used by the x16 slot.
The best option is to have a motherboard with a separated clock configuration for the PCI Express x16 slot, like the one shown in Figure 19. On this motherboard, the clock configuration for the main PCI Express x16 slot is called “C51 PCI-Express Frequency.” The clock that will be used by the other PCI Express connections is called “MCP55 PCI-Express Frequency” and should be left at 100 MHz.
So, how do you determine the maximum clock rate your AGP or PCI Express x16 bus will support? Like everything related to overclocking, you do so by trial-and-error. Increase the clock rate a little, save the changes and exit setup, load Windows, run a 3D game like Quake 4 in its benchmarking mode, and see if the system remains stable. If it does, go back and increase the clock rate a little more, and repeat the entire process until you find the maximum clock rate your I/O bus will run without crashing the system.
[nextpage title=”Tweaking the I/O Bus (Cont’d)”]
Besides increasing the I/O bus clock rate, you can also increase its voltage. Usually, by increasing its voltage, you will be able to set it at higher clock rates without making the system crash. However, we recommend that you first find the maximum clock rate the I/O bus will run with stability (i.e., without crashing), and only then increase the I/O bus voltage to see if your system will run without crashing using a higher clock rate. Keep in mind that sometimes you won’t be able to improve the overclocking even by increasing the bus voltage.
Not all motherboards provide this option. But from all the motherboards shown before, only the one portrayed in Figure 17 didn’t provide it. On the motherboard in Figure 15, this option was called “AGP OverVoltage Control,” in Figure 16 it was called “AGP Voltage,” in Figures 18 and 19 you would have to select “Advanced Voltage Control” and then would see this option as “NB Core/PCI-E Voltage” (see Figure 20) or “NB to PCIE VGA Voltage” (see Figure 21).
As you may have noticed, the exact name of each option varies considerably, and we tried to give some examples. However, it is impossible to cover all motherboard models available on the market and tell you what each option will be called on your motherboard in particular. We hope that with our examples you at least got an idea as to how to identify these options.
[nextpage title=”Making Changes Permanent (NVIDIA-Based Cards)”]
As we mentioned before, all changes made to your video card aren’t permanent. This occurs because the video card’s BIOS will instruct the VGA to run on its default configuration every time you reboot your PC. That’s why you need to keep PowerStrip loaded; it will load your personalized configurations and reconfigure your video card every time you load Windows. If you shut down PowerStrip, your video card will use the BIOS default configuration again.
You can edit the video card BIOS and make your overclocking permanent if this is what you want. With this option, you won’t need to load PowerStrip anymore, and the video card will always work under the clocks you programmed at its BIOS.
But don’t worry. In the future, you can reverse this change and make your video card run on its original configuration.
For doing this, you will need two programs: a BIOS editor (used to edit the BIOS contents, i.e., to change your video card clock rates) and a BIOS programmer (used to save the new modified BIOS to the video card). The programs you will need depend on if your video card is based on an NVIDIA or on an ATI chip. Below, we will show you how to modify your NVIDIA-based video card BIOS, and on the next page we will talk about how to do the same thing on an ATI-based video card.
For your NVIDIA-based card, you will need two programs: NiBiTor, which is a BIOS editor, and nvFlash, which is a BIOS programmer.
Run NiBiTor and first backup the BIOS your video card is currently using. If in the future you want to restore your video card’s original BIOS, just update your video card BIOS with this file.
Go to Tools, Read BIOS, Select device and select your video card. Then go to Tools, Read BIOS, Read into File. Give your file a name. This will backup your video card BIOS to a file. Then go to File, Open BIOS and load the BIOS file you’ve just saved. You will see a screen similar to Figure 22.
As you can see, there are two places you will want to change: “Core” and “Memory.” Type in the maximum core and memory clocks that you know your video card can run and remain stable.
After editing these two fields, save the BIOS into a new file by going to File, Save BIOS. Now you have a BIOS file with your overclocked settings.
The next step is to create a bootable floppy disk. Go to https://www.bootdisk.com, select a DOS boot disk, download the .exe file, and run it to create your bootable floppy. Copy nvFlash.exe and your overclocked BIOS file to the floppy disk. Then boot your PC using this floppy (you may need to enter the motherboard setup and change the boot order).
At the command prompt, type:
Restart your computer and your work is done.
[nextpage title=”Making Changes Permanent (ATI-Based Cards)”]
The process for ATI-based cards is similar, but you will need to use different programs: RaBiT, which is a BIOS editor, and ATIWinFlash, which is a BIOS programmer.
Run ATIWinFlash and first backup the BIOS your video card is currently using. If in the future you want to restore your video card’s original BIOS, just update your video card BIOS with this file. Click on Save to save your video card BIOS to a file.
Then run RaBiT and open the BIOS file by going to Open, BIOS file. Click on the Clocking tab to edit the core and memory clocks. After editing these two fields, save the BIOS into a new file by clicking on Save As. Now you have your overclocked BIOS saved to a file.
Run ATIWinFlash again, click on Load Image, and select your overclocked BIOS file. Click on Program, and ATIWinFlash will save your overclocked BIOS to your video card.
Restart your computer and your work is done.