In addition to the global price drop on all graphics cards, MSI is now also reducing the MSRP of the RTX 3050 Ventus 2X (OC) from its lowest range. The phrase “affordable graphics card” thus takes on a new spin, and if you’ve read somewhere that the performance characteristics of Ventus graphics cards aren’t worth much, forget it. We certainly don’t see it that way, and few RTX 3050s for this money can boast a more functionally attractive design.
Gaming tests
The largest sample of tests is from games. This is quite natural given that GeForce and Radeons, i.e. cards primarily intended for gaming use, will mostly be tested.
We chose the test games primarily to ensure the balance between the titles better optimized for the GPU of one manufacturer (AMD) or the other one (Nvidia). But we also took into account the popularity of the titles so that you could find your own results in the charts. Emphasis was also placed on genre diversity. Games such as RTS, FPS, TPS, car racing as well as a flight simulator, traditional RPG and sports games are represented by the most played football game. You can find a list of test games in the library of chapters (9–32), with each game having its own chapter, sometimes even two (chapters) for the best possible clarity, but this has its good reason, which we will share with you in the following text.
Before we start the gaming tests, each graphics card will pass the tests in 3D Mark to warm up to operating temperature. That’s good synthetics to start with.
We’re testing performance in games across three resolutions with an aspect ratio of 16:9 – FHD (1920 × 1080 px), QHD (2560 × 1440 px) and UHD (3840 × 2160 px) and always with the highest graphic settings, which can be set the same on all current GeForce and Radeon graphics cards. We turned off proprietary settings for the objectivity of the conclusions, and the settings with ray-tracing graphics are tested separately, as lower class GPUs do not support them. You will find their results in the complementary chapters. In addition to native ray-tracing, also after deploying Nvidia DLSS (2.0) and AMD FidelityFX CAS.
If the game has a built-in benchmark, we use that one (the only exception is Forza Horizon 4, where due to its instability – it used to crash here and there – we drive on our track), in other cases the measurements take place on the games’ own scenes. From those we capture the times of consecutive frames in tables (CSV) via OCAT, which FLAT interprets into intelligible fps speech. Both of these applications are from the workshop of colleagues from the gpureport.cz magazine. In addition to the average frame rate, we also write the minimum in the graphs. That contributes significantly to the overall gaming experience. For the highest possible accuracy, all measurements are repeated three times and the final results form their average value.
Computational tests
Testing the graphics card comprehensively, even in terms of computing power, is more difficult than drawing conclusions from the gaming environment. Just because such tests are usually associated with expensive software that you don’t just buy for the editorial office. On the other hand, we’ve found ways to bring the available computing performance to you. On the one hand, thanks to well-built benchmarks, there are also some freely available and at the same time relevant applications, and thirdly, we have invested something in the paid ones.
The tests begin with ComputeBench, which computes various simulations (including game graphics). Then we move on to the popular SPECviewperf benchmark (2020), which integrates partial operations from popular 2D and 3D applications, including 3Ds max and SolidWorks. Details on this test package can be found at spec.org. From the same team also comes SPECworkstation 3, where GPU acceleration is in the Caffe and Folding@Home tests. You can also find the results of the LuxMark 3.1 3D render in the graphs, and the remarkable GPGPU theoretical test also includes AIDA64 with FLOPS, IOPS and memory speed measurements.
For obvious reasons, 3D rendering makes the largest portion of the tests. This is also the case, for example, in the Blender practical tests (2.91). In addition to Cycles, we will also test the cards in Eevee and radeon ProRender renderers (let AMD have a related test, as most are optimized for Nvidia cards with proprietary CUDA and OptiX frameworks). Of course, an add-on for V-ray would also be interesting, but at the moment the editorial office can’t afford it, we may manage to get a “press” license in time, though, we’ll see. We want to expand application tests in the future. Definitely with some advanced AI testing (we haven’t come up with a reasonable way yet), including noise reduction (there would be some ideas already, but we haven’t incorporated those due to time constraints).
Graphics cards can also be tested well in photo editing. To get an idea of the performance in the popular Photoshop, we’re using a script in PugetBench, which simulates real work with various filters. Among them are those that use GPU acceleration. A comprehensive benchmark suggesting the performance of raster and vector graphics is then also used in alternative Affinity Photo. In Lightroom, there are remarkable color corrections (Enhance Details) of raw uncompressed photos. We apply these in batches to a 1 GB archive. All of these tasks can be accelerated by both GeForce and Radeon.
From another perspective, there are decryption tests in Hashcat with a selection of AES, MD5, NTLMv2, SHA1, SHA2-256/512 and WPA-EAPOL-PBKDF2 ciphers. Finally, in the OBS and XSplit broadcast applications, we measure how much the game performance will be reduced while recording. It is no longer provided by shaders, but by coders (AMD VCE and Nvidia Nvenc). These tests show how much spare performance each card has for typical online streaming.
There are, of course, more hardware acceleration options, typically for video editing and conversion. However, this is purely in the hands of encoders, which are always the same within one generation of cards from one manufacturer, so there is no point in testing them on every graphics card. It is different across generations and tests of this type will sooner or later appear. Just fine-tuning the metric is left, where the output will always have the same bitrate and pixel match. This is important for objective comparisons, because the encoder of one company/card may be faster in a particular profile with the same settings, but at the expense of the lower quality that another encoder has (but may not have, it’s just an example).
Would it be possible to perform thermal tests of the upcoming GPUs using the same noise levels – just like in CPU cooler and fan tests? With the current methodology, there is a multitude of results, but it’s often impossible to decide which result is better. Some GPUs operate at less than 60 degrees (an overkill) but to achieve it, they run twice as loud. The many game noise and thermal tests could be replaced with just coil whine tests, as it changes a lot in different environments (specific game, being in menu/playing).
The original proposal of the test methodology looked exactly as you describe (we obviously have the same taste :)), but we eventually dropped it. There were several reasons for that. But the main reason was time. If I had taken the cooler apart in such detail and tested it in several modes (instead of one) it would have been unmanageable in terms of time. It’s already too much now. With coolers and fans it’s easier, as they are not as complicated devices as graphics cards, where a lot of time is taken up by performance tests.
The time consumption of a single graphics card test in its current form, by our standards for accuracy, is approaching 300 hours of net time. With a more detailed analysis of the cooler, it would be quite a bit more, and in addition to the modes normalized by the various noise levels, we would still need the default one for evaluation in the end, which is based on the video card’s BIOS settings and is key for the vast majority of users.
And then there’s one more thing that causes some risk when testing at fixed noise levels. The GPU temperature is usually significantly lower than the VRAM and memory temperatures, which are quite difficult to keep track of. Sure, hardware monitoring and, in case of critical values, stop the test (typically at low speed) and evaluate it as a failure, but… capacity-wise, we unfortunately can’t afford this marginal type of tests, which have only a very small audience, in the interest of self-preservation. Although personally, I’d love to pursue it. So perhaps sometime in the future when we can afford to split the work between more people.