Tag Archives: nows

Technology

We test GeForce Now’s new “3080” upgrade, discover unmatched cloud-gaming power

Leave a comment

Enlarge / GeForce Now works on all of these devices. But you’ll want to double check whether your ideal combination of hardware, screen, and Ethernet connection will get you up to either 1440p resolution and 120 fps, or 2160p resolution and 60 fps. If so, GeForce Now’s new 3080 subscription tier might be perfect for you.

Nvidia

The prospect of buying a reasonable new GPU in 2021 remains a crapshoot, and that says nothing about your hopes of buying a higher-end option anywhere near MSRP values. In a chip-shortage universe, there’s not a ton we can do to change this unfortunate reality, outside of asking greedy cryptominers to please donate their high-end GPUs to people who want to play games with the things.

For some people, cloud gaming might be a good alternative. This concept lets gamers connect their much weaker hardware (netbooks, set-top boxes) to supercomputer farms. So long as they can maintain a decent broadband connection and endure hits to button-tap latency (and bandwidth overages), they can, on paper, expect higher-end gaming. But so far, we haven’t seen impressive computing power in that marketplace. Stadia in particular launched as a woefully underpowered service, while the biggest PC-centric cloud option, Nvidia GeForce Now, has a mix of power limitations and usability frustrations.

This week, Nvidia moves forward with its most intriguing cloud-gaming service upgrade yet: GeForce Now 3080, named after its powerful RTX 3080 GPUs. Preorders for that service are now officially live, and depending on your willingness to compromise, you might want to look into it.

We’ve tested its pre-release version for the past week, and the results have, quite frankly, been dreamy. This $198/year service tier works on two fronts: it opens up connections to more powerful Nvidia servers, and it unlocks more options on the local end for anyone using the service. The result is a white-hot stunner that rivals the computing power you can muster with a locally owned RTX 3080 Ti.

How GeForce Now fits into the stream-iverse

The catch, of course, is that GeForce Now is still the most unwieldy cloud-gaming option on the market. To its credit, the service is also the most flexible and storefront-agnostic.

Thus, before I get to the best parts of Nvidia’s new “GeForce Now 3080” option—its faster performance, its higher maximum resolution, and its higher maximum frame rate—I should set the stage for how the service works and compares to its contemporaries, so bear with me.

  • GeForce Now revolves around accessing games you purchased at other storefronts.

  • To add a game to your GeForce Now library, manually search for it, then “add to library.” You’ll have to buy it elsewhere. Nvidia GeForce merely checks whether you paid for it—though it does conveniently pull up your save file from those storefronts’ cloud save features.

  • Use GeForce Now’s Steam library-scanning feature to automatically gather games you bought via Steam. This doesn’t work for EGS, Ubisoft Connect, or EA Origin. (Artifact? That works on GeForce Now? Huh.)

  • Search the service’s full slate of compatible games. It’s a bit unwieldy.

  • Type in something like “free-to-play,” and GeForce will get you up and gaming without requiring a single purchase.

Most cloud-gaming services demand that you rely on their store ecosystems in one way or another. You can only play games on Google Stadia if you buy those games’ Stadia-exclusive versions (or access freebies via the paid Stadia Pro subscription service). If you want to stream games within Xbox Game Streaming, you have to pay for Xbox Game Pass Ultimate, and you can only stream that service’s selection of approximately 200 games—as opposed to additional Xbox games you individually purchase. And Amazon Luna offers a variety of “channels,” each with individual costs and unique content, that you can pick and stack the same way you might do with video-streaming subscription services.

The cost of GeForce Now, conversely, has nothing to do with games you might buy or borrow and everything to do with the Nvidia hardware you’re leasing in the cloud. In some ways, GeForce Now is just a cloud computer that you can use as you see fit. When you use GeForce Now, you log into other storefronts on its server farm, load games you’ve already purchased, and play them using their profiles and save files. Nvidia’s cloud-gaming service doesn’t care where or how you buy games. It just wants to power them.

One big catch, however, is that some game publishers do not allow Nvidia to stream their games. (Remember: when you buy a game via an online storefront, you’re only paying for access to a license. This, among other things, means publishers can yank your access around in exactly this way.) Upon the service’s 2019 launch, Nvidia was forced to remove games that it originally supported after certain publishers cried foul—particularly games from Activision Blizzard’s Battle.net service. In good news, over time, many more games have been added to the service from the following storefronts, now totaling a little over 1,100 games:

  • Steam
  • Epic Games Store
  • Ubisoft Connect
  • EA Origin

Up until this week, GeForce Now only had two tiers: $98/year or free. The latter includes performance downgrades and required waits in server queues, so if too many people are using the service, you have to wait behind paying customers. That free option is a decent way to basically confirm that your ideal streaming device—a smartphone, a set-top box, or a weak netbook—can connect to the service and translate your gamepad taps or keyboard-and-mouse frenzies to cloud-streamed video games. But it’s not great for image quality or computing power.

RTX 3080 tier wins, even at a higher resolution

The paid version, meanwhile, includes rudimentary “Nvidia RTX” support. Its server instances include Nvidia’s proprietary GPU cores that are dedicated to ray tracing and Deep Learning Super-Sampling (DLSS), but only a few per instance, as powered by an RTX-upgraded variant of Nvidia’s Tesla T10 server-grade GPU. The results are generally powerful enough to get average, modern PC games up to a steady 1080p, 60 fps refresh, usually with a number of graphical bells and whistles enabled.

As I’ve previously attested, if you’re within the right geographic range of Nvidia’s servers and have a low-ping wired Ethernet connection, you can expect all-but-unflinching performance while playing with mouse-and-keyboard on a variety of shooters on the service. But 1080p resolution at 60 fps and medium settings is basically what the rest of the streaming fray offers. How much more juice can the same Nvidia app ecosystem muster, especially if Nvidia itself, manufacturer of so many high-end GPUs, applies its own hardware upgrade?

The best way to answer that is to let a few of its compatible games do the talking. These are the exact same PC versions of games that you might install on your own computer, after all, and some come with built-in benchmark sequences. Thus, I ran a few tests on the existing $98/year service, dubbed the “founders” tier, before Nvidia invited me to a pre-release test of the $198/year “3080” tier so I could compare the sheer power of both server options.

  • Assassin’s Creed Valhalla‘s benchmark results on the GeForce Now 3080 service tier. “Ultra” graphics settings, 1440p resolution.

  • Assassin’s Creed Valhalla‘s benchmark results on the GeForce Now founders service tier. “Ultra” graphics settings, 1080p resolution.

  • Watch Dogs Legion‘s ray tracing-intensive benchmark results on the GeForce Now 3080 service tier. “Ultra” graphics settings, RT at near-max, 1440p resolution, with DLSS enabled in “quality” mode.

  • Watch Dogs Legion,‘s ray tracing-intensive benchmark results on the GeForce Now founders service tier. “Ultra” graphics settings, RT at near-max, 1080p resolution, with DLSS enabled in “quality” mode.

The above benchmarks for the computationally brutal Assassin’s Creed Valhalla (no ray tracing) and Watch Dogs Legion (substantial ray tracing) are explained in their captions. To summarize: all tests from the newer 3080 service tier are run at a higher 1440p resolution, yet they still soundly outpace the same tests run at a lower 1080p resolution on the service’s founders tier. Sadly, we couldn’t run these tests with a frame time chart attached, so we’re left with Ubisoft’s vague, squiggly line charts. Still, all of those benchmarks do come with crucial “lowest 1 percent” counts, and when those are higher (which they are, by a large margin, in the 3080 tier), you can expect fewer frame time stutters and refresh rate dips.

Read original article here

Science

If you ever wanted to help find new planets, now’s your chance

Leave a comment

Enlarge / The telescope array of the Next Generation Transit Survey.

If you’ve ever wanted to search for distant worlds, your time has come. The team behind a planet-hunting telescope array called the Next-Generation Transit Survey (NGTS) is looking for help with the large volume of data the instrument has produced. The NGTS scans large areas of the sky with a collection of small, robotic telescopes to detect dips in stars’ light that are caused by a planet passing between the stars and Earth.

The team now has a lot of data, which it has sifted through using computers. But computers have difficulty distinguishing a likely planet from various sources of noise, so the researchers are asking the public to double-check the computers and provide a final call on what a signal is.

Public transits

One of the most successful means of searching for exoplanets has been the transit method, in which a telescope repeatedly observes the amount of light originating from a star. If a planet wanders in front of that star, the amount of light will dip slightly. These dips have a very stereotypical shape if you plot them over time in what’s called a light curve, with a fairly steep drop as the planet swings in front of the star, followed by a long, flat reduction.

Enlarge / Exoplants can produce dips in the light from a star that tend to have steep sides and a flattened bottom.

But there are other phenomena that can cause similar-looking dips in a light curve, plus a bunch of factors that create noise that computers have a hard time distinguishing from a signal. For example, the equivalent of sunspots (sunexospots? exosunspots?) on other stars will rotate across the field of view, and many stars experience short-term variations in their activity. These things can combine to make for a noisy signal with numerous dips. In addition, the telescopes have regular periods of downtime in which they briefly stop observations, which also confuses the software analysis.

Finally, NGTS is on Earth and suffers from noise due to the atmosphere—something that the Kepler telescope did not have to deal with.

As a result of these difficulties, the NGTS team has many potential transits that have been flagged as interesting-looking by its software. But they need to be verified before they’re accepted as actual transits, and the NGST team simply isn’t large enough to do the work. So it has teamed up with the Zooniverse citizen science platform to get the public’s help.

What kind of dip is that?

Those interested in participating can navigate to the project’s page on the Zooniverse platform and click the button in the “Get started” section. If you’re a new user, this will start a tutorial that explains the different types of light curves flagged by computer algorithms. These include things like gaps in the data, a mischaracterized curve, and the chaotic mess caused by stellar variability.

Enlarge / Before starting, there’s a tutorial showing you examples of all the different options you’ll have to choose from.

Is this a U- or V-shaped curve? With relatively sparse data, it can be hard to tell.

You’ll also have to make calls on whether a light curve has a flat bottom or is V-shaped. This is harder than it may sound because the NGTS doesn’t have the same frequency of imaging as nice, neat Kepler curves. Instead, as seen above, the curve has fewer individual data points in it, and the atmosphere produces more scatter between them.

But don’t worry too much about getting it wrong. Past Zooniverse projects have typically had multiple volunteers scan each image to ensure they didn’t put too much weight on any one individual’s judgment. And anything that looks interesting will ultimately be examined by someone who does astronomy for a living.

Read original article here