Pascal is a new GPU microarchitecture developed by NVIDIA. I will be informing you of what Pascal is, what its benefits are, and the two new Pascal architecture-based GPUs that were announced at NVIDIA’s private event last week, and everything else announced during the event in another article later this week. I will also include what I believe this means for different types of consumers.
What is a microarchitecture?
A microarchitecture is the electrical circuitry of a CPU, GPU, or some digital signal processors. Microarchitectures, or just architectures, are the heart of all modern electronics. For example: shrinking the size of their transistors or increasing the instructions per cycle is what is making new technologies a lot more energy-efficient, powerful, and smaller.
What is Pascal?
Pascal is NVIDIA’s new microarchitecture that has been in development for over 2 years. NVIDIA’s CEO, Jen-Hsun Huang, stated that several thousand engineers worked on it, and the company poured several billion dollars into research and development.
Jen-Hsun made a joke after revealing the research and development budget by saying, “I’m pretty sure you can go to Mars.”, referring to how the cost of developing this architecture is equal to that of getting humanity to Mars. That’s insane.
Pascal brings many new advancements to the GPU market. It brings new, exclusive technologies that can enhance your everyday life even though you may not realize it. Pascal is being used for not only making computers that will go inside cars for self driving technologies, but also for applications like deep learning and servers.
NVIDIA recently released information on their gaming lineup of their Pascal GPUs (since they had already revealed the server cards). These were the GeForce GTX 1080 and 1070.
GeForce GTX 1080 and GeForce GTX 1070
NVIDIA announced both the GeForce GTX 1080 and GeForce GTX 1070. The Founders Edition cards will cost $699.99 and $449.99 respectively and will be released May 27th and June 10th. According to Jen-Hsun, the GeForce GTX 1080 supposedly is three times more energy-efficient than a GeForce GTX Titan X.
* My speculation.
? Unknown information.
GeForce GTX 1080 in Use
NVIDIA showed the GeForce GTX 1080 in use at the event a few times. The first time the GPU was displayed in use was when the character study (when the combination of physical traits and qualities are analyzed for the characters) of Paragon was presented to the audience. The whole thing was happening in real time at 60 frames per second.
What is surprising is how the GPU was able to perform a demanding task like this in real time.It is so demanding because the GPU had to render each character as a 3D object and render each ray of light and their microscopic interactions with these objects all at the same time and in realtime. This is usually done in 5 frames per second by a whole server.
Movie and gaming studios have been struggling with their content creation process for years because of its extremely demanding processing power and time taken to complete larger renders. With this new Pascal architecture it allows studios to spend less of their time and money on tasks such as these.
Small studios and individual content creators also get great benefits. The relatively low price of the card is also great for these types of consumers who don’t have large budgets. The low heat output is also useful for small form factor cases because the card can get very high clock speeds without heat being a problem in these small form factor cases.
What was amazing was the specs that the card was running on. The card was overclocked to 2114 MHz and had an effective memory clock of 5508 MHz at only 67 degrees Celsius on an air cooler. The only way any other GPU would reach that kind of overclocking is by cooling it with liquid nitrogen. This really shows the advancements and efficiency that the Pascal architecture can bring to the market.
The second time NVIDIA showed the card in use was when some gameplay of DOOM was presented running at 1920x1080p on Ultra settings. DOOM was also using Vulkan, a popular 3D computer graphics API. The card was running at a smooth 60 frames per second which doesn’t sound very impressive.
What really was impressive was when the developers of DOOM decided to remove the frame cap so we were able to see what kind of frame rates the card can really achieve. When you take a look at the top right you can see that the card was jumping from 110 to 150 frames per second.
With this type of jump in performance and efficiency, this is great for all types of gamers as well. Gamers who can spend money on more GPU’s can now spend less money on them just to be able to play a game at ultra settings and at 4K resolution just to get decent framerates.
Gamers who can’t spend a lot of money get much more performance for a lower price increasing satisfaction of this section in the market. The energy efficiency saves gamers from having to buy high wattage power supplies to just be able to run multiple GPU’s but also save some money from their electric bills.
Game developers will know that they can now make their games even more demanding because all of their intended customers will have the right hardware to be able to run a game at even higher settings. Gamers can now get even better games because of these Pascal based GPUs.
The pascal cards introduce and upgrade to SLI or Scalable Link Interface. SLI is the brand name for the connection between NVIDIA GPUs. SLI HB doubles the bandwidth of the Maxwell cards’ SLI which improves multi-gpu performance overall by allowing more information to move at one time.
In conclusion, Pascal is the future of content creation and gaming. NVIDIA has done an amazing job with this architecture and has put so much work into this. Stay tuned later this week or early next week on an article about the other things announced at the event NVIDIA. Feel free to criticize my writing as this is my first article so that the next one can be even better. I have two questions you can answer in the comments:
- Do you think NVIDIA will make GP200 cards?
- Do you think NVIDIA will make a pascal based Titan card? And if so, do you think it will be a GP100 card, GP200 card, or one of each?
If you need me to explain what I mean by GP200 and GP100, then feel free to ask me to in the comments.