Buying a new GPU? Just watch this…
If you want to buy a graphics card but don't know where to start, it's almost impossible to decide which card to buy. In this video, you'll learn everything you need to know about graphics cards. To suit everyone, I'll cover both the basics and more advanced details. So, without wasting any more time, let's dive in. Firstly, let me explain the terms you need to know so you won't get confused. It might seem a bit
complicated, but don't worry, you'll know everything when this video ends. What are DLSS, FSR, and XESS? These three technologies found in modern graphics cards essentially do the same job. They upscale lowresolution images to highresolution ones using artificial intelligence or algorithms. For example, you might be unable to play a game smoothly in 4K resolution because your graphics card isn't powerful enough. These three technologies upscale the resolution to 4K by processing the game graphics in 1080p and then filling in the gaps using AI. Depending on your
settings, it might cause minor artifacts or slight distortions in the image, but it makes games run smoother. The pioneer of this upscaling technology is Nvidia. Nvidia's DLSS is more successful compared to its competitors in the case of quality and performance thanks to artificial intelligence. DLSS can only
be used in Nvidia cards. On the other hand, FSR is the upscaling technology developed by AMD. While DLSS can only be used with Nvidia cards, as I've just mentioned, things are a bit complicated in AMD's FSR, it varies between versions. For example, while FSR 3.0 also supports Nvidia graphics cards, FSR 4.0 only supports RX9000 series graphics
cards. XESS is the upscaling technology developed by Intel. It's like a mixture of DLSS and FSR. Alongside Intel
graphics cards, XESS also supports many Nvidia and AMD graphics cards, but it performs better on Intel graphics cards. I think you understand upscaling technologies. Next, let's talk about frame generation. What is frame generation? Frame generation, just like upscaling technologies, is used for improving gaming performance. But the way this works is different. While
upscaling technologies use AI to increase the resolution, frame generation analyzes the frames in the image as you're gaming to add in new frames using AI. This way, games become smoother. The impact it has on performance is way higher compared to upscaling technologies. In fact, it
increases the performance so much that Nvidia's multiframe generation technology can sometimes lead to three to four times higher FPS. Of course, as it's doing this, it can also cause distortions. Furthermore, the worst part is that it greatly increases input lag.
Because of this, it becomes impractical for competitive games. AMD's frame generation technology is called AFMF, and Intel's frame generation technology is called XESS frame generation. They may not be as effective as Nvidia's multiframe generation, but they might cause less input lag. Now, let's talk
about ray tracing. What is ray tracing? I can say that it's the reason the DLSS and the frame generation technologies I've just mentioned exist. Ray tracing creates more realistic images and reflections by calculating the interaction of light and the objects in the scene. As you can probably guess, as this technology requires a lot of calculations to be made, the raw power of graphics cards alone is not enough. Therefore, we need to make use of technologies such as upscaling and frame generation to get a good performance out of ray tracing. Ray tracing has been
used for years in the animation and the movie industries. In the case of games, Nvidia has done a great ray tracing ad campaign with their RTX cards in 2018. Nvidia cards have better ray tracing performance compared to their competitors, but AMD increases on their ray tracing performance with each new generation. Intel is also trying to compete, but generally speaking, their entry to mid-tier graphics cards fall flat on ray tracing performance and compatibility. Now, let's dive into CUDA cores. But before we dive in, there's
something you should definitely check out. Whether you're working alone or with a team, choosing the right AI model for your project and keeping track of them all can be a real challenge. That's where Chat LLM Teams comes in. Chat LLM Teams simplifies getting the most out of various models with Root LLM, which automatically selects the best large language model for your prompt. But
that's just the beginning. You can also generate images and videos using a wide variety of models. On top of that, Chat LLM can handle your documents like a pro with features like chatting directly with PDFs, interacting with PowerPoint files, and even creating Word documents with just a few clicks. If you're a developer, Chat LLM has something special for you, too. It lets you ask questions about your codebase, and it includes Code LLM, a fullfeatured code editor based on VS Code, completely free. With Chat LLM, you can instantly
generate and run code or use the built-in AI engineer tool to create your own chatbot from scratch. And even if you don't know how to code, no worries. You can still build your own app using App LLM. It's designed to let you turn your ideas into working applications with an intuitive code-free interface.
Worried about AI detection? The humaniz feature rewrites your generated content so it sounds more natural and can bypass most AI detectors. And here's the best part. For just $10 per month per user, you get access to multiple powerful LLMs, often for less than the cost of a single model elsewhere. Visit
chatlm.abocus.ai or click the link in the description to get started today. Now, let's move on to CUDA corores. What are CUDA corores? I will explain it in simple terms to not confuse you. It's a type of core found in the graphics card. These cores developed by Nvidia allow the GPU to offer high performance in tasks requiring high processing power such as graphical operations and artificial intelligence. The faster and
more these cores are, the higher the performance is. CUDA is available on Nvidia GPUs. The equivalent of CUDA cores on AMD cards are the stream cores and the equivalent on Intel cards are the XE cores. Among these three, the performance of Nvidia graphics cards is generally better than the other two manufacturers. So as you can understand from here, Nvidia is significantly better than other brands in terms of graphics card technologies. All right,
now you know everything you need to know about graphics card technologies. Now let me tell you how to pick a graphics card. Which graphics card manufacturer should I pick? Nvidia, AMD, and Intel. These three manufacturers are the most popular GPU manufacturers. There are many differences among the cards of these three manufacturers. When buying a graphics card, it is very important to pick the right manufacturer. Let me put
it this way. Nvidia is the most popular GPU manufacturer. It is known for its technologies such as ray tracing, DLSS, frame generation, and CUDA. Nvidia's drivers are usually more stable compared to other manufacturers. Therefore, they are considered more reliable. Its power consumption is very good as well. Of
course, with all of its good qualities, it is also very expensive. Alongside this, lately, there have also been many stock shortages with the RTX 5000 series's launch. This is actually one of the reasons why it's more expensive than MSRP in many countries. Meanwhile, AMD is another manufacturer that is considered a competitor to Nvidia. It has alternatives to DLSS and frame generation technologies, though they aren't as successful as Nvidia. The prices vary between countries, but it can compete with Nvidia thanks to generally being more affordable. In
terms of raw gaming performance, some AMD cards offer significantly better performance compared to similarly priced Nvidia cards. I want to emphasize that it is a competitor only in the area of gaming performance. Nvidia performs much better than AMD in areas that require calculations such as ray tracing, content creation, streaming, and AI thanks to CUDA. Besides this, there's a lot of bias against AMD graphics cards due to their high power consumption and past driver issues like the Wattman error. Nowadays, next to none of these
issues are present, so we can say that these biases are meaningless. Besides these two, there is also a new player in the market, Intel. It manufactures entry and mid-tier graphics cards that focus on the price toerformance ratio. Intel also has image enhancement technologies that can be alternatives to DLSS and they perform rather well. Its power consumption is at a reasonable level too, but there can be some firmware related issues as they are new in the market, but they are hard at work. so much that we can see significant increases in GPU performance within even just 1 month thanks to driver updates add-in board manufacturers. So
here's the thing despite the fact that the GPUs and graphics cards are manufactured by Nvidia, AMD or Intel, the graphics card you'll buy from the stores will usually be cards made by thirdparty brands such as ASUS, MSI and Sapphire. These brands source the GPU from these three manufacturers and then sell it after making customizations in areas such as the cooling, clock speeds, and design. In fact, there are different cards like entry tier, mid-tier, and high tier, even within the same brand.
This means that, for example, not all ASUS cards are the same. I have two videos where I explain this more in depth. You should definitely check it out if you're curious. Now, you know the brands. Now, let's move on to maybe one of the most confusing, but at the same time, one of the most important topics, graphics card naming schemes. Nvidia,
AMD, and Intel have various different GPUs. To cater to more users, they separate their cards into entry-level, mid-tier, and high-end cards. This is where graphics card namings come in. You need to learn about these namings in order to pick the most suitable card according to your budget and use case. Firstly, let's start with Nvidia. What
does the GTX or the RTX on the left mean? The GTX or the RTX phrase actually signifies the technological base. At the present time, cards that support advanced technologies such as ray tracing and DLSS are counted as parts of the RTX series. What do the first two digits on the left mean? These two digits indicate the series the card belongs to. There are series such as the
RTX 20, 30, 40, and even 50 series. Now, the higher this number is, the newer the tech in the card is. For example, you get lower power consumption, higher performance, and extra software features. In fact, there are
technologies such as multiframe generation, which is exclusive to the RTX50 series. What do the last two digits mean? The last two digits indicate the performance the card can put out. Things are a bit more clear here. It goes like 50, 60, 70, 80, 90. The higher the number is, the higher end it is. This means that the RTX 4070 is
more powerful than the 4060. Within the same series, the performance and the price increases as the number gets better. But be careful, the newest card doesn't always necessarily mean the most powerful card. For example, the RTX 3070 is a little more powerful than the RTX 4060 Ti or a GTX 1080 Ti and Old Legend perform similarly to the RTX 3060. This
means that it isn't enough to just look at the naming scheme. You must also check real world performance tests. We will talk about this in detail in a bit. What about super or TI? The terms TI or super at the end of GPU model names are typically used for cards that perform between two other models. For example,
there's a gap between the RTX 4070 and the RTX 4080. So, to fill this gap, cards such as the RTX 4070Ti are released to be more powerful than the 4070, but cheaper than the 4080. You might see Super in some card names. These super models are upgraded, more powerful versions of the standard versions of cards in the same series. For example, there's a performance difference of around 20% between the RTX 4070 and the RTX 4070 Super. In short,
the Ti and Super suffixes are added to provide more options and performance ranks within the same series. Next up is AMD. To be frank, the way AMD names cards is a total mess, especially in older series. But don't worry, they moved onto a more organized naming scheme, similar to the one for Nvidia cards, like the 70 tier of the AMD RX9000 series XT version. You may think of it as the same thing I explained on the Nvidia side, just with the numbers changed. Now, let's move on
to the more confusing part. How does the naming scheme work on AMD cards? RX. This phrase has been in use for a long time. It's the general name of AMD graphics cards. The first digit indicates the series the card is in, like the RX 5000, 6000, or the 7000 series. The second digit indicates the segment of the card. For example,
RX6700 means the 700 tier of the 6000 series. If the last two digits are 50, this card is a different version of the normal card. For example, you may think of the RX6750 as a higher performance version of the RX6700, but generally there aren't any big performance differences between these cards. Their performances are similar. They are usually cards that
aren't worth the price difference. XT and XTX. You might think of these as the Ti in Nvidia cards, meaning it's the more powerful version of the same card. For example, the RX6700 XT is more powerful than the RX6700. Let's move on to the Intel
cards. Intel is very new on the graphics card market. Due to this, the naming side of things are a bit weird. For
example, the most powerful card in Intel's previous generation graphics cards is called the ARK A770, while the most powerful graphics card in the current generation is called the ARK B580. As you can see, there is little to no correlation between the naming of their generations. Because of this, it isn't possible to say anything about the naming scheme of Intel cards just yet.
Their user base is not that big anyways, but it is still important to mention that Intel has been improving itself quite rapidly. Now, you know what the expressions on graphics card names mean. Let's take a look at other important stuff you need to know when picking a graphics card. Resolution. Resolution,
as you know, refers to the number of pixels on the screen. The most popular resolutions today are 1080p, 1440p, and 2160p or 4K. As the resolution increases, the load on the graphics card also increases, and this leads to lower frame rates during gaming and more system resource utilization. Therefore, if you want to play games at higher resolutions, you need to pick a powerful GPU that can handle this load. VRAM. VRAM is the memory on the graphics card itself. Do not confuse it with RAM, which is the system memory. This memory
is where the GPU temporarily stores data. The higher the VRAM, the more data the graphics card can process. However, if the VRAM is lower than required, you may experience FPS drops, stuttering, and significant performance issues. The likelihood of encountering such issues is quite high due to the high VRAM usage of modern games such as The Last of Us. Of course, the optimization of the game that you're planning to play is very important as well. Games with
optimization issues can lead to issues regarding VRAMm usage. We mostly encounter such issues in games that were ported from console over to PC. By the way, as resolution increases, the VRAM usage increases as well. Due to this, I recommend that you go with a graphics card that has at least 8 GB of VRAM if you're planning to game at 1080p and at least 12 GB of VRAM if you're planning to game at 1440p. Raw power. Raw power is the real performance the GPU offers without assistive technology such as DLSS. So,
the performance you get from the game without any additional pieces of software reflects the raw power of your graphics card. This power is especially important in professional use cases such as rendering, modeling, and tasks requiring calculations. VRAM and raw power are two factors that complement each other. Bottlenecks occur if one of
them is strong and the other is weak. For example, the RTX 3070 Ti has more raw power than the 4060 Ti, but we can't see its full performance as it only has 8 GB of VRAM. On the other hand, the 4060 Ti has 16 GB of VRAM, which is pretty good, but its power can sometimes be insufficient. One of the cards that
can set this balance very well is the RX7700 XT. It offers a great balance with a high amount of raw power and 12 GB of VRAM. So, I highly recommend this graphics card. Power consumption. The amount of power the graphics card draws from the power supply is very important. As the power consumption of a card increases, the temperatures of the GPU will rise. So, you will have to turn
towards graphics cards with better cooling. To find out this value, we usually look at these. TGP, total graphics power, refers to the maximum power the GPU can pull. TBP, total board power, refers to the amount of power the graphics card pulls when all components of the card, fan, lighting, VRAM, etc. are included. What about TDP, thermal
design power? The TDP value mostly concerns the cooling side of things. This is more commonly used compared to the other two, but the TGP and TBP give the clearest result. You can also enter the Tech PowerUp website to see the PSU value that is recommended for your graphics card. But don't forget the value this website gives can sometimes be just on the limit and sometimes be too much. Instead, I think that the best choice is asking for help from others on forums. PCI Express version. PCI Express is the interface where the motherboard exchanges data with the graphics card.
This means that despite how high the graphics card's processing power might be, you can still experience slowdowns if the graphics card is unable to transfer the data to the motherboard fast enough. But is this really something you need to worry about all the time? No. It becomes important only in some special circumstances. Expressions such as PCI Express X4, X8, and X16 refer to the physical datab bus width. On the other hand, expressions such as PCI Express 3.0, 4.0, and 5.0
show how fast this data bus can transfer data. But some graphics cards come with low data bus widths. For example, the RX6500 XT only uses the PCI Express X4 lane. If you use this graphics card with an old motherboard that, for example, only supports PCI Express 3.0, 0 the
bandwidth will noticeably go down and you will experience visible FPS drops for cards like the RX6500 XT. PCI Express 4.0 support is a must. What about high-end cards? Cards such as the RTX 4090 and the RX7900 XTX usually use the X16 lane. In these cards, the performance difference between PCI Express 3.0 or 4.0 is barely reflected as the bandwidth is already sufficient.
To summarize, if you are buying a GPU with a low data bus width, such as the RX6400 or 6500 XT, it is very important that your motherboard supports PCI Express 4.0. In the mid tier and the high-end, the PCI Express version doesn't create a huge difference, but a more recent motherboard is always better. Plus, as most modern graphics cards are PCI Express 4.0 x16 compatible, you just need to ensure that your PC components are compatible with each other. Now you know what to pay
attention to when picking a graphics card, but how will you find out how the card performs in real life? The best way to find this out is to watch performance tests on YouTube. You can check out graphics card comparison videos. This way, it will be easier for you to decide on which graphics card to buy if you are stuck between two graphics cards. Alternatively, you can check out in-depth reviews of individual graphics cards. You can find these out from graphics card reviews. In these videos,
it isn't just the gaming performance that is being tested. There are also tests for production workloads such as video editing and 3D rendering. This way, you can learn about the general purpose performance and not just the gaming performance. Alternatively, you can also get the opinions of experienced users by creating threads on hardware forums. You'll usually get fast and helpful replies. All right, let's wrap up the video. If you like the video,
don't forget to hit the like button. If not, feel free to dislike. Stay mysterious until the next video. Take care and bye. [Music]
2025-05-05 01:07
