NVIDIA LDAT and PCAT, new tools to have full control over the GPU

NVIDIA LDAT and PCAT

With the evolution of GPUs, the number of frames per second has risen and the system latencies, as a consequence of the greater speed with which the images follow one another on the screen, have been reduced, so much so as to highlight how some components of the configuration are still obstacle to those who want the maximum possible reactivity, no matter if for pure personal pleasure or to give the best in competitive titles. There is thus a world of optimizations that include the possible congestion of the render queue, a drawback that in titles that exploit the GPU intensively can lead to considerable leaps in latency. It is therefore essential to keep an eye on a configuration designed for maximum reactivity, even if technologies such as Reflex Boost take care of it automatically.

But the actual gain depends on several factors and this is where the LDAT sensor comes into play, a latency analysis tool for GeForce cards that is paired with the PCAT kit, designed instead to measure l actual consumption of NVIDIA video cards. A few days ago we received them both and we tested them to detect the latencies and consumption of a configuration equipped with a GeForce RTX 3070, an ASUS Chakram mouse and an ASUS ROG Swift from 360 Hz, already equipped with a latency detection module but important for testing given the ability to go up to 360 frames per second.

Precision tools, yet easy to use

NVIDIA tools ready to be installed Designed for competitive and technical gamers, the two NVIDIA tools were not designed for sales but for internal testing and insiders. In case of strong requests, however, the small LDAT could arrive on the market as opposed to the PCAT device, probably destined to remain a device dedicated to editors, overclockers and technicians. And it is something that does not surprise us. On the other hand, the measurements of the PCAT are exclusively related to consumption and do not have a direct impact on gaming, unlike the LDAT, useful for detecting the latency of the system, excluding or including the mouse, so that you can perhaps improve your configuration. or detect problems, for example, with a PC station provided for a tournament. It is no coincidence that it has been designed to be easily installed on any monitor and in any condition.

The sensor, already updated to a more advanced second version, consists of a box which, applied to the monitor with a band elastic, allows you to measure the actual time that passes from the click of the mouse to the reaction of our character on the screen. By exploiting it, therefore, we can evaluate factors such as the impact on the latency of NVIDIA Reflex technology, the response time of the monitor and the delay of the mouse, in order to choose the best in a landscape that offers increasingly responsive screens and mice that promise a minimum response time. Parameters of this type had a relative weight when there were no high-speed connections or synchronization technologies capable of guaranteeing perfect distribution of the frames, but now they become important in the context of creating systems and configurations capable of guaranteeing lightning-fast response. at our commands.

LDAT sensor, still disconnected LDAT installation, as we said, is quite simple and designed to be easy to fix and move along the screen of any gaming monitor. The side of the LDAT box that looks at the screen, where there is the brightness sensor that also returns the value of the latter, is covered with a thickness of soft rubber, while on the opposite one, added with the second version, there is it is a key that simulates the main one of the mouse and is useful if you do not have a compatible mouse or if you want to exclude the peripheral from the latency calculation. On the front, however, there is a connection for the mouse, which can also be detected using the audio, provided obviously you connect a fairly sensitive microphone to the 3.5mm input placed at the top of the device, which allows you to use any mouse. At the bottom, finally, there is the USB output necessary to connect the instrument to the PC, where it is necessary to install a software, integrated with FrameView starting from version 1.1, to be able to view the latencies obtained with a single click or even by throwing batteries. of clicks after setting from the software the number of shots we want to sample and the delay between one click and the next.

LDAT, installation and measurements

Connected and ready to test with Overwatch In all this, the most complex operation concerns the positioning along the screen of the box with the sensor that must match up with a part of the image that clearly reacts to the click. It is therefore usually positioned in correspondence with the muzzle of a weapon capable of firing instantly and therefore suitable for tests on the latency of a game. But games like Overwatch, updated to include Reflex technology these days, also have a special mode to make LDAT work easier. By activating it, a square appears on the left of the screen that lights up white when the chosen hero shoots. It is therefore not necessary to try to find the perfect triangulation with the tip of the weapon, even if the operation is still slightly more complicated than having a monitor with NVIDIA Reflex Latency module, just like the ASUS ROG Swift 360Hz used for testing. . But it must be said that one of the advantages of LDAT is that it can be used with any monitor, with any brand of GPU and with a large number of games.

For the tests we have chosen Overwatch, as we said updated with full reflex support and subject to rendering queue saturation issues at high framerates. Perfect therefore to test the Boost mode which has advantages over setting a limit to the framerate to improve latency. In fact, it works to display only the number of images per second that can be managed by the GPU in use, in this case a KFA2 GeForce RTX 3070, still trying to reach maximum fluidity. And that is why it needs some implementation work by developers which on the one hand reduces its spread, but on the other hand does not sacrifice the performance of the video card.

Moving on to the benchmarks, we loaded Overwatch in low detail and in 1080p, with the render rate at 100%. Enough to get to 360 frames per second, actually taking advantage of the ASUS ROG Swift 360 and detecting, with a battery of 40 detections, an average latency of 15.61 milliseconds, mouse included, without activating the Reflex technology. An already excellent value that has been reduced slightly by ticking the Reflex item which has brought us to 15.00 split milliseconds, with minimum values ​​of about 7 milliseconds. And it is here that we resorted to the Boost which together with Reflex took us to 13.8 milliseconds, with a minimum peak of 5.8 milliseconds, highlighting a gain, albeit minimal. But that didn't surprise us.

! Function () {"use strict"; window.addEventListener ("message", (function (a) {if (void 0! == a.data [" datawrapper-height "]) for (var e in a.data [" datawrapper-height "]) {var t = document.getElementById (" datawrapper-chart - "+ e) || document.querySelector (" iframe [src * = '"+ e +"'] "); t && (t.style.height = a.data [" datawrapper-height "] [e] +" px ")}}))} ();

Where the framerate is very high the gains are lower, but Reflex offers a bonus even in the range of 360 frames per second, thanks to the Boost technology that prevents the crowding of frames and that, in the moments in which the framerate hovers around 250fps, comes to lower the latency from 24 to 18 milliseconds. Many gamers, however, play at lower refreshes, despite monitor refreshes on the rise, with 360 Hz coming to next-generation laptops as well. Furthermore, the resolution is increasing and in 4K, where the framerate is necessarily lower, Reflex guarantees important bonuses, although the lower number of frames reduces the risk of queues by minimizing the usefulness of the Boost technology.

Let's talk , again with Overwatch but this time at maximum detail and in Ultra HD, by a leap, at 60fps locked, from about 87 milliseconds without Reflex to 37 milliseconds by activating Reflex technology. An abyss. And with the framerate locked at 60fps the gain was evident even in 1080p. In fact, we went from 37.4 milliseconds to 25.3, obtaining a significant increase in reactivity. But as we said, blocking the framerate at low levels reduces the Boost technology which in these cases has rewarded us, in the best of cases, with a gain of a fraction of a millisecond. The advantages, however, are evident as are the conditions to make the most of them and obtain an effective benefit in terms of reactivity, not very relevant in a graphic adventure, but rather important in an excited first-person shooter.

PCAT, installation and measurements

In the foreground the Riser and behind the PCAT module with the power cables As anticipated, PCAT has less to do with gaming and looks directly at reviewers and technicians, measuring thanks to two cards the actual consumption of a video card. It is therefore not designed for trade even if it does not require particular efforts to be installed and used, as it does not need welding or special precautions. It can also be fixed to a test bench, but it can also be simply placed on the motherboard, connected to the small Riser that is inserted into the PCIe 4.0 slot and connected to the power supply destined for the GPU.

It actually comes in the way between card and power supply and can also support very powerful cards thanks to 6 8-pin power connectors, three to be used at the input and three obviously at the output. Our KFA2 GeForce RTX 3070 only has one, which made it very easy for us, but the tool is obviously designed to support models designed for overclocking high-end cards as well. There are no 12-pin converters for the inputs of the new NVIDIA Founders, but on the other hand it is a solution designed to be used easily with any video card.

The active kit with OLED screen that shows consumption Once installed it is obviously possible to control the parameters with a special software, which allows you to individually view the consumption of the 12V auxiliary connectors and the consumption through the PCIe slot, which enjoys integration with FrameView, also in this case from version 1.1 onwards. In addition, NVIDIA has provided us with a complex Excel template that allows you to load test logs and organizes all data in a legible way, starting from the application used for the test to arrive at parameters such as framerate, temperatures, frequencies, efficiency, use of the CPU and obviously consumption, complete with graphics.

Our surveys with graphics in plain sight But the most delicious thing is the small OLED screen installed on the PCAT module, complete with NVIDIA LOGO when switched on, which shows all parameters in real time. In a few moments it is therefore possible to verify any overclocking effects or other as well as obviously checking that the card works as expected. In our case we had no issues with 15W on average in Windows, around 55W in Overwatch, and 171W peaks in The Witcher 3 in stock mode and 197W in overclock mode. With the built-in stress test we instead recorded a peak of 204W, but these are not always reliable maximum values. The measurement method makes it advisable to carry out prolonged benchmarks, however very simple and intuitive to carry out with a tool that, although not designed for the common user, can be used without any problem by anyone.

Have you noticed any errors?