The last couple of years have seen much hype surrounding graphics cards but the excitement died out soon after. We saw several high-profile launches but not many gamers could actually get their hands on the graphics card owing to the cryptocurrency mining phenomenon. This affected the mid-range cards because they offered the best value for money, not only for gamers but also for miners. So while the mining craze has started dwindling, we are yet to see any relief in the prices. Team Green has been dominating the high-end segment while Team Red is a favourite among the miners. Rumours of new cards have started to arise as NVIDIA, AMD Radeon and Intel prepare new cards. NVIDIA seems to be first in line with a launch slated for Q3-Q4 2018 while AMD Radeon might soon follow. And Intel seems to be working towards producing its own GPUs for a 2019 launch. So the next year seems to be really exciting and will see a lot of you guys buying new graphics cards. Here’s what you need to consider while buying them.
What to consider
- Proprietary technology – Graphics card manufacturers like to tie down their consumers within their ecosystem. Proprietary technology comes in the form of graphics APIs and proprietary hardware. Between NVIDIA and AMD Radeon, NVIDIA has more proprietary technology than Radeon. Radeon, on the other hand, prefers to keep their APIs more open. Which is why G-Sync, which is an NVIDIA hardware, only works with NVIDIA cards and FreeSync which is a Radeon technology works with everything on the market. This comes into play when you’re getting a high-refresh rate monitor with adaptive sync. If you don’t use G-Sync or FreeSync then you will experience frame-tearing and the only recourse then is to switch on V-sync which limits your FPS to 60. Essentially, you just bought a high-end monitor capable of 120-144 Hz and cannot play on it properly. So try to match the graphics card with the monitor technology. There are some issues with using FreeSync on NVIDIA cards but you can trick the NVIDIA cards to use FreeSync, however, we don’t expect the trick to last long.
- Multi-GPU configuration – NVIDIA has SLI while AMD has CrossFireX. The latter has phased out the CrossFireX brand and the former has restricted SLI to only the high-end cards. Both companies have their own reasons and multi-GPU configurations were never simple, to begin with. Some games would lose performance with multi-GPU configuration because the performance profiles weren’t properly defined and you’d have to wait for the manufacturer to fix that. Your best bet in any situation is to ignore multi-GPU configurations and opt for a higher-end card. There used to be scenarios where two mid-range cards would perform just the same as the flagship card. Example: Two GTX 960 would be equal or better than a GTX 980 for half the price. That is no longer possible since mid-range cards no longer support SLI.
- Memory – Everyone’s using either GDDR5 or HBM stacks. That’s not what we wish to point out here. There are graphics cards which have different memory configurations under the same model which offer different performance. This would be the GTX 1060 which comes in 3 GB and 6 GB variants. Ideally, a GTX 1060 wouldn’t have a difference in core configuration otherwise they wouldn’t be called a GTX 1060 as per NVIDIA’s historical nomenclature standards. However, the 3 GB and 6 GB variants indeed had different performance numbers because the CUDA scores varied by 128. This sets a precedent which we might see more of in the coming generations, so keep an eye out for these.
- Build quality – Graphics cards of late don’t bring about a performance difference with clock speeds. There is, technically, a difference but it’s no longer worth spending the extra money. Which brings other add-on features into perspective. One of these is build quality. Manufacturers have very little headroom to improve on their graphics cards against what GPU chip manufacturers provide. So build quality is one aspect where they’re trying to outdo each other. This comes down to the material used in the cooling assembly, the material used for the chassis, the VRM circuitry and the cooling fans. While there have been some gimmicks in the industry over the last year, we’ve only seen the beginning of gimmickry so far. What you should concern yourself with is the material used to build the card. Add too much metal and the card begins to sag, add too much plastic and it becomes a fire hazard. The VRM is best left to experts since there’s no way you can check your graphics card’s VRM before you buy it. The reference design is pretty good with all cards but if you wish to get a custom PCB graphics card, then things may or may not work in your favour. For example, there’s no point in getting the EVGA KingPin edition if you are not going to overclock. Those cards are extremely limited and only professionals can utilise its full potential. Cards with cheap fans tend to give out in 3-4 years which is generally right after the warranty period ends. The only way you can figure this aspect out is by checking out reviews and reading user feedback on e-commerce websites. Or you could always invest in aftermarket custom coolers.
- Usage scenario – Graphics cards, like CPUs, are capable of performing a wide variety of tasks. Gaming graphics cards such as the GeForce and Radeon RX cards excel at providing high FPS for games but with other tasks such as big data number crunching or 3D modelling, they aren’t the very best that you can get. That’s where sub-brands such as Quadro, Tesla, Radeon FE and Radeon Instinct come into the picture. However, these cards are ridiculously expensive and you should only invest in them if you are going to use them for the right conditions extensively. For just dabbling in 3D modelling and starting off with AI, the gaming graphics cards are very good. Especially the high-end ones. NVIDIA’s TITAN V is portrayed as a jack of all trades and so is the Radeon VEGA 56 which has become very popular among video editors. You can use OpenCL to push compute tasks onto graphics cards aside from the proprietary technology that manufacturers have, so that’s another thing that discrete graphics cards support which IGP don’t excel at. For just gaming, it’s best to stick to GeForce and Radeon RX cards.
- Connectivity ports – Most graphics cards have 5-6 ports on the rear but they don’t all support the same number of monitors. This is because the ports are either analogue or digital and when there’s a distinction, you’ll find that those ports are running on a different internal clock. Ideally, you’d want to run the monitors on the ports which have the same clock. There are specialised cards from the manufacturers for more than 4 monitors. The other aspect is if you want to connect a VR headset. You need a free HDMI port and if that port does not share a clock with the port your monitor is connected to, then only one display device can function at any given time. This is why the “VR Ready” moniker is used by brands to show support for a VR HMD + monitor at the same time.
What not to consider
- RGB – Fancy lighting has been touched upon multiple times in this guide so you’re definitely tired of getting the same advice again and it remains the same. Fancy RGB lighting is purely cosmetic and should be the least of your priorities when buying a new graphics card.
- Factory overclock – In the previous generations, just bumping the core clock made a tremendous difference in the performance. However, this has changed and with the recent generations, you are only getting 5-10 FPS difference in most games. So if you’re paying a lot just for the factory overclock, then you’re wasting your money. Unless a graphics card has a really good VRM circuit which can actually aid overclocking, then don’t pay attention to this aspect. Minor overclocks aren’t worth the money.
- Graphics API support – DirectX, OpenGL and Vulkan support. If you are into playing a lot of games then knowing what version of DirectX is important. Then again all cards for the past 2 generations do support DirectX 12 and you don’t need to worry about that. The same goes for OpenGL and Vulkan which are now supported by both GPU manufacturers.
The best place to get it
- Mumbai – Lamington Road
- Kolkata – Chandni Chowk
- Bangalore – SP Road
- Delhi – Nehru Place
- Pan-India – Amazon / Flipkart