Naming ConventionsDid you know that AMD has both an RX 480 and an RX 560, and the 480 is considerably faster? Or that Nvidia has a GTX 960 that is faster than its 1030? Read on and we will explain…
First one or two digits: Generation
The first one or two digits designates the generation of the card. This will be a 7, 9, 10, 16 (more on that later), or 20 (the only RTX line).
GeForce RTX cards follow the same conventions to the GTX cards, except they start with “20”. So the RTX 2060 is the successor to the GTX 1060.
Nvidia recently released “Super” versions of their RTX cards. This seems to be roughly equivalent to a “Ti” level upgrade- more cores, higher clockspeeds, etc.
Last two digits: Class/Speed
The series tells you where the card falls in Nvidia’s hierarchy. 80 is always the top. So a GTX 980 is an “80 class”, and a GTX 1060 is a “60 class”.
For some cards, Nvidia will release a “Ti” designation. Think of this like you might a “+” on a report card. The “Ti” will be better than the series it is attached too, but not as good as the next step up. So a GTX 1070 Ti is faster than a GTX 1070, but not quite as fast as a GTX 1080.
To compare across generations
As a rule of thumb, a jump in the generation of a card is worth about the same as a jump in class. So, very roughly, a GTX 1050 matches up with a GTX 960. There are exceptions to this, which we will get into below, but this rule is a good enough starting point.
What about Ray Tracing?
Ray tracing has been around for some time now in the movie world. If you’ve ever seen a shiny superhero or villian walk around with realistic lighting effects, there’s a decent chance that his reflections and shadows have been rendered through a ray tracing algorithm.
At a really high level, this means that instead of “rasterizing”, or using the old pythagorean theorum to render a 3D object on a 2D monitor, a computer is actually modeling thousands of individual rays of light. They will bounce off of surfaces based on the surfaces’ characteristics, and reflect other colors, light sources, etc.
With the RTX 20xx cards, Nvidia is supporting ray tracing taking place on their GPU.
In theory, this is amazing. Unfortunately, in practice, game developers need to support ray tracing, and there is a bit of a “chicken or egg” problem. Developers aren’t going to spend a lot of time supporting something that hasn’t been widely adopted, and people aren’t going to pay a premium for a feature that isn’t supported in games.
So, for the time being, we are not assigning a lot of value to ray tracing in our card rankings. As more games start to adopt ray tracing, this value will increase.
What does “7nm process” mean?
CPUs and GPUs consist of bunch of little transistors- think binary electical gates that are either open or closed. Their size is expressed in nanometers (“nm”). AMD’s Radeon VII’s transistors are 7nm, while Nvidia’s Turing’s transistors come in a little bigger (12nm).
All else being equal, smaller transistors should be better. If the transistors are smaller, you can pack more into the same space, and theoretically they should be more efficient.
However, there are a lot of factors outside of transistor size that determine the speed and efficiency of a GPU. For example, Nvidia’s Turing chips are more efficient and outperform the Radeon VII even with their larger transistors.
AMD has been releasing graphics cards for a long time, and has changed their naming structure a few times. Like Nvidia with the GTX line, if you’re looking for an AMD graphics card in 2019, you’re going to be looking at a Radeon series card.
RX 4xx and 5xx
These are pretty simple, the first digit in the number represents the generation. So an RX 580 is a 5th generation chip.
The final two digits represent the series. Like with Nvidia, bigger numbers are faster. So an RX 580 is faster than an RX 570.
Unlike Nvidia, there are only minor differences between 4th and 5th generation. So an RX 480 is just about as fast as an RX 580, and is faster than an RX 570.
RX Vega Series
There are currently only two RX Vega cards- the RX Vega 56 and the RX Vega 64. They “56” and “64” refer to the number of “compute units” in each card.
The Radeon VII is AMD’s flagship GPU. The number “7” is significant; it is the first card to be produced with a 7nm process, which is a huge jump. The Vega 56 and Vega 64 are both 14nm cards.
AMD recently (7/7/19) released their RX 5700 series GPUs. Prior to release, these were referred to as the “Navi” generation. The 5700 and 5700 XT models are both midrange chips- competitive with the RTX 2060 and 2060 Super.
ON TO THE RANKINGS!
We have benchmarked each card using a variety of resolutions accross a number of different games. The resulting “Frames Per Second” for each game has been translated into a score that we’ve called the “CompUSA Benchmark Unit” or CBU.
Since we update this list with new entries over time, it doesn’t make a lot of sense to have a “Tier 1” that turns into “Tier 2” every 6 months. So instead of using numbers, we’re naming each tier after a member of the animal kingdom, with a more powerful predator appearing to represent the next level of performance. Obviously within each tier, there will be cards that are faster or slower, but the intra-tier differences are marginal compared to the jump in performance between tiers.
As of the time of this writing, the highest CBU score comes from the Titan RTX at 2,828. That represents enough horsepower to play just about any game at max settings and 4k resolution at 60+ FPS! The lowest current-gen score that we’ve observed comes from the Intel UHD 620 chip that comes integrated onto Whisky Lake processors. This scores 172, which is enough power to play more modest games at low settings, 720p, at about 20-30 FPS.
SHARKS WITH LASERS TIER
This tier of cards is powerful enough to play just about any current game at 4k resolution with max settings and still pump out about 60 FPS. The only setting that *could* cause hiccups for these guys is ray tracing. Even monster systems can struggle to stay above 60 FPS when ray tracing is enabled.
- Titan RTX – 2,828 CBU 100% 100%
- RTX 2080 Ti – 2786 CBU 98.5% 98.5%
- RTX 2080 – 2573 CBU 91% 91%
- RTX 2070 Super -2523 CBU 90.5% 90.5%
This tier of cards has enough horsepower to play anything at 1440p with max settings and still get 60+ FPS. Certain titles should also run at 4k with these cards.
- RTX 2070 – 2361 CBU 83.5% 83.5%
- Radeon VII – 2349 CBU 82.5% 82.5%
- GTX 1080 Ti – 2329 CBU 82.5% 82.5%
- RTX 2060 Super – 2317 CBU 82.5% 82.5%
- Radeon RX 5700 XT – 2308 CBU 82% 82%
- Radeon RX 5700 – 2206 CBU 79% 79%
- RTX 2060 – 2151 CBU 76% 76%
- GTX 1080 – 2032 CBU 72% 72%
- GTX 1070 Ti – 2018 CBU 71% 71%
The Squid tier of cards is powerful enough to deliver a 60+ FPS experience at 1080p with virtually any game at max settings. If you have a 1080p monitor, these guys represent the best “bang for you buck”.
- Radeon Vega 64 – 1937 CBU 69% 69%
- GTX 1660 Ti – 1923 CBU 68% 68%
- Radeon Vega 56 – 1920 CBU 67.5% 67.5%
- GTX 980 Ti – 1860 CBU 66% 66%
- Titan XP – 1748 CBU 65.5% 65.5%
- GTX 1660 – 1730 CBU 61% 61%
- Radeon RX 590 – 1522 CBU 54% 54%
SMALL FISH TIER
This is the last tier at which 1080p gaming is really feasible for most titles. If you’re currently stuck with integrated graphics, these will be significant step up, but they will not be “future proof”. Believe it or not, this tier is roughly on par with the PS4 and Xbox One.
In some cases, there will be multiple memory configurations for these cards. The RX 580 and 570 both come with either 4GB or 8GB. The additional RAM will help with higher resolutions.
- GTX 1060 – 1280 CBU 53% 53%
- GTX 970 – 1409 CBU 50% 50%
- Radeon RX 580 – 1380 CBU 49% 49%
- Radeon RX 480 – 1328 CBU 47% 47%
- GTX 1650 – 1268 CBU 45% 45%
- Radeon RX 570 – 1241 CBU 44% 44%
- Radeon RX 470 – 1125 CBU 40% 40%
This tier of cards is enough to wring some enjoyment out of most titles, but you will have to make some concessions regarding resolution and detail. You’re probably going to be gaming at 720p with medium settings to get an acceptable frame rate.
- GTX 1050 Ti – 982 CBU 35% 35%
- GTX 960 – 951 CBU 33.5% 33.5%
- GTX 1050 – 865 CBU 31% 31%
- GTX 950 – 854 CBU 27% 27%
- Radeon RX 560 – 717 CBU 25% 25%
- Radeon RX 460 – 708 CBU 25% 25%
- Radeon RX 550 – 564 CBU 20% 20%
These GPUs (mostly integrated) are enough to get the signal from your PC to your monitor. Any gaming is going to have to be done at low resolution/low settings. One thing that this tier is GREAT for is low power consuption. For example, if all you want to do is run an HTPC, the GT 1030 is actually a spectacular card.
- Vega 11 (integrated) – 372 CBU 13% 13%
- GT 1030 – 362 CBU 12.5% 12.5%
- MX150 (integrated) – 358 CBU 12.5% 12.5%
- Vega 8 (integrated) – 284 CBU 10% 10%
- Vega 10 (integrated) – 266 CBU 9.5% 9.5%
- Vega 3 (integrated) – 218 CBU 8.5% 8.5%
- Intel UHD 630 (integrated) – 197 CBU 7% 7%
- Intel UHD 620 (integrated) – 172 CBU 6% 6%
Frequently Asked Questions
What kind of GPU do I need if I just want to surf the web, watch Netflix, etc?
If you are doing anything besides gaming or photo/video editing, just about any GPU in a modern computer will be just fine. An RTX 2080 isn’t going to watch Netflix any faster than any other card.