Cheat as you cheat can?
It's easy to explain why we've changed our benchmark selection slightly compared to the recently released CPU tests. Since we have to compare several graphics cards from different manufacturers, the special driver optimization for such benchmarks is also important.
We have therefore deliberately refrained from using the SPECviewperf freeware benchmark, because it has a real image of the actual workstation or CAD performance, including the workloads it contains, are about as much as cheap artificial honey with real beekeeping honey. In addition, some of the programs it contains do not actually start at all or do not start at all. limited with consumer graphics cards or drivers.
Driver offers for specially dead-optimized benchmark suites are unfortunately the order of the day; both manufacturers, mind you. But on this very subject we already had a small paragraph on the previous page. Therefore, these statements are all too often completely worthless and the distances between the consumer and pro cards are sometimes so severe that one wants to hit the wall smoothly with one's head (e.g. Maya). Good for PR and marketing, but unfortunately only mirror fencing.
One of the parade horses from the collection of now-driver-optimized benchmarks is Cinebench OpenGL. We have reconsidered with current drivers and find the result (and the way of the drivers up to it) quite worth questioning. We will certainly not use this as long as we have to put different drivers against each other.
Solidworks 2015
Then rather an honest comparison with the complete installation of one of the usual standard software packages like Solidworks. If we compare the Radeon Vega FE with the Quadro P6000, you get about 90% of the performance for approx. 30% of the price. Even if the P6000 is ahead in the end, the Radeon Vega FE does not make a bad picture. It operates at roughly the same level as a Quadro P5000 (equivalent to the GeForce GTX 1080), but also costs just over half.
We can also see from the individual benchmarks that there are no real outliers in the individual sub-composite scores. For this purpose, we have compared everything once again in percentage terms for the better over-the-counter, with the Quadro P6000 always being the 100% mark.
Creo 3.0
Creo 3.0 also runs on the Radeon Vega FE, even really good by the way! In none of the sub-areas is the Radeon Vega FE slower, on the contrary. You can see what real application optimization can make when you look at software in all its complexity. But here it is a real performance in everyday life and not a multi-second benchmark sequence, about whose image quality one could then also argue about. However, the quality of the graphics output was at a similarly high level for both cards.
Also with Creo 3.0, the normalized comparison to the Quadro P6000 shows the differences most clearly. A whopping 20% extra power in total is a real house brand – for now. But can be sure that Nvidia will again optimize the driver to follow suit. This in turn suits every user, because even with the professional drivers it is like with the respective game profiles of gaming graphics cards and their drivers.
2015 3ds Max
The appearance could not be more balanced, despite the many and even larger workloads. In the GPU Composite Score, both cards are almost equal at the end.
From a purely percentage point of view, the partial gap in the affected areas is always less than 10% and even turns into a 6% lead in an important sub-sector. This in turn saves the Radeon Vega FE in the end on the same step, because 0.6% difference is almost already in the range of measuring tolerances.
AutoCAD 2017 – 3D Performance
Let's move on to AutoCAD and the fact that instead of OpenGL and large, own graphics libraries, we simply use DirectX for rendering – in a rather unexcited form, which usually relies directly on Microsoft's libraries using a simple wrapper. In terms of performance, the Radeon Vega FE ranks fairly solidly between the GeForce GTX 1080 and the GTX 1070, with a tendency towards the GTX 1080.
Should this very real, hardly optimizable (and therefore also difficult to manipulate by the manufacturers) evaluation of the Radeon Vega FE now exemplify all DirectX-based applications and even games that do not have their own special game profile in the driver Bring? That would be rather a disappointment, to say the least, but it is not yet all evening. Let's see.
- 1 - Einführung und Übersicht
- 2 - Details zu Architektur und HBM2-Speicher
- 3 - Demontage, Kühler und Interposer-Details
- 4 - Platinendesign und Detailinformationen
- 5 - Benchmark-Intro, 2D-Troughput und -Performance
- 6 - 3D Workstation- und Design-Benchmarks
- 7 - Gaming-Performance: DirectX 11
- 8 - Gaming-Performance: DirectX12
- 9 - Gaming-Performance: Vulkan/OpenGL 4.5
- 10 - Leistungsaufnahme im Detail
- 11 - Takt, Temperaturen und Geräuschemission
- 12 - Zusammenfassung und Fazit
Kommentieren