If you’re a PC gamer, you’ll likely be familiar with the concepts of updating your drivers, DirectX, game patches and so on. Much of this is now automated, with applications like Nvidia’s GeForce Experience gently nudging you “hey, time to update me!”. Steam of course takes care of itself, auto updating games in the background so you’re always running the latest patch. But for those gamer’s who wish to get their hands dirty and take a look at how their PC’s are performing, then there are a few essential tools.
Monitoring your graphics card might seem a little overkill to some, but in reality there’s a myriad of different uses and scenarios where it’s an invaluable tool to improve performance or bug test. For example, due to cards using ‘boost’, which actively raises or lowers clocks based on power targets and temperatures, the ability to monitor those (along with the average and max clocks) is extremely important if you’re keen to maximize the performance from your system. You can also get a read on what your RAM usage is, to help you understand if your card is hitting a VRAM wall during gameplay, which can help account for stuttering.
TechPowerUp’s GPU-Z is a utility many will be familiar with, allowing you to monitor pretty much all of your Graphic Card’s statistics. With modern day graphics cards relying on ‘boost’ like technology to increase its clock speed providing the card has the spare power and heat to do so, monitoring temps is fairly essential. So how do we go about measuring your temperature and GPU statistics then?
First things first, you’ll need to download the latest and greatest release of TechPowerUp’s GPU-Z, it’s priced very reasonably (it’s totally free) and is pretty damn powerful. I’d also recommend downloading benchmarking software (such as 3d mark) or alternatively, you could use a games built in benchmark. Titles like Tomb Raider, Sleeping Dogs, Metro (both Last Light and 2033) are among just a few which features their own built in benchmarks.
Temperatures of a Graphics Card are affected by several conditions, including the voltage being fed into the GPU (which you may adjust if you’re overclocking), ambient heat in the case along with airflow, and the usage of the card. Quite simply put, while the card is idling on desktop, it’s not being strained, and many modern cards will automatically ‘downclock’ to a slower speed to save power, noise and temps. Monitoring your cards temperatures can be essential to ensure that the card is maintaining a safe operating temperature, and just to get an idea of how your system is doing.
First things first, launch GPU-Z, and and head to the ‘Sensors’ Tab. From there, you’ll want to left click the arrow to the right of ‘GPU Temperature’ and then select ‘Show Highest Reading’. Now, open a second GPU-Z (just double click on the icon again). Head to sensors on second instance of GPU-Z (we’ll call the second instance GPU-Z2) and have it show the ‘Average Reading’. I’d recommend now running a game benchmark, that’ll strain the card to it’s max. Something like 3d Mark, Heaven or possibly playing a game such as Crysis 3.
If you notice that your temps are above the safe limits for you card, or generally pretty damn high, then it’s a good idea to think about lowering them. Setting a more aggressive fan profile, improving the air flow in your case, or possibly cleaning out your PC / GPU are a few examples.
For the purposes of GPU Core Clock, it’s pretty much the same scenario. You’ll want to set the average and max, and generally will want to look to see what the difference is between them. If you notice your card doesn’t boost as high as you might like, then it could be temperature or power limit related. This is more useful when you’re manually setting the boost levels on a card while overclocking than a reference card, but can still be useful to find what your cards average clocks
Logging to file (clicking Log to File at the bottom of GPU-Z) can be a useful tool if you’re making changes to your system and you’re wishing to compare the GPU states before and after. For example, SLI can drastically affect the clocks for cards, so disabling SLI and testing each card in turn let’s you see if one is naturally a weak link. If you’re changing things in your case (such as a new PSU, or perhaps different cooling) you can also have a written record of the changes.
Monitoring is often done not as a solo activity, but in groups. For example, we might decide to monitor the GPU Core Clock, temps, load and fan speeds. We might also decide to test out different fan speeds (manually setting them) and take a look at our temps. We could then use this to decide what fan speeds provide enough cooling to not throttle performance, and yet aren’t so noisy to drive you crazy. Of course, this all comes down to the GPU you’re using.
AMD’s recent Radeon R9 290 and R9 290X are two cards which run very hot, on their stock coolers in particular. Both cards actually throttle when they hit 95 degrees. AMD have improved the fan profiles for the cards to help take into account the heat of the cards. But if you’re using a reference card, using GPU-Z to monitor the average clocks can be a great way to find out if temps are killing your performance.
RAM usage monitoring is one of the most important uses for GPU-Z, particularly if you’ve just bought a new (larger) screen, or you’re trying out a new game and getting microstuttering. As I discussed in my EVGA GeForce GTX 780 Ti SuperClocked ACX review, RAM usage on cards varies greatly. During my testing on Crysis 3, I’d said ““Welcome to the Jungle” level of Crysis 3, on Windows 8, which of course already loads a few hundred MB of data into the GPU before you load up any games. Crysis 3 running at the highest graphics settings, using FXAA and 1440P hits a max of 2.1GB of memory, leaving 900MB spare.”
In the above image, you can clearly see the difference in memory usage. If this test had been conducted on a card with 2GB or less RAM, I’d clearly have hit the memory limit. For what it’s worth, I manage to hit the 3GB limit using MSAAx8 at 1440P on my GTX 780 Ti on Crysis 3. Metro Last Light, with SSAA + FXAA running at 1440P, with everything maxed (including Nvidia’s Hardware Physx) was less memory intensive. GPU Z showing that just over 1.5GB of memory was used, significantly smaller than Crysis 3.
If you’re overclocking, keeping a close eye on the Power Consumption is imperative. Higher clocks use greater amounts of energy, and of course raising the GPU’s Power Target / Voltages are one way we achieve higher clocks. All in all, there is a lot of different uses of GPU-Z and it’s monitoring potential. This is far from a comprehensive guide, but it should give you the basic ideas on how to setup your own monitoring. It’ll also come invaluable when we start delving into Overclocking and DownSampling guides within the next few days.