Welcome Stranger to OCC!Login | Register

Wolfenstein II: The New Colossus Performance Analysis

   -   
Category: Gaming
Price: $59.99
» Discuss this article (0)

Introduction:

It has been a bit, but I am finally back in Wolfenstein II: The New Colossus, this time to look at how it performs across multiple GPUs. While it is my RX Vega 64 I use as my daily driver, I still have a GTX 770, GTX 980, and GTX 1080 (and GTX 1070, but I skipped it, which I will explain later). The GTX 770 is getting a bit old now, but it is also the NVIDIA GPU given on the game's minimum requirements, so I wanted to give it a shot. (Actually I just checked and it is the 4 GB version listed as the minimum spec, and what I have is the 2 GB version. Spoiler, it ran, but you will have to look at the page to find out how well.)

For data collection I used OCAT, as I have before, and had it set to a run length of 300 seconds (five minutes). I like having runs this long as it can let me just play without having to worry about staying to a strict path, but just close to it. With enough samples, data becomes more representative, so deviations from the path of one test should become less and less significant when compared to another. I also am still using R for processing the resulting CSVs.

I did two runs on each GPU; one in the Roswell, Downtown district and one in the Roswell, Underground district. These can both be repeatedly accessed after completing the game, and repetition is important. The Downtown district is a relatively small map, but it is also set outside while the Underground district is, obviously, indoors. The Underground district is also a much larger map, so I was not able to complete a run of it in five minutes, without being fast or lucky. Roswell, Downtown I was often completing before the timer ended, so I just ran around the map randomly, keeping things busy.

Something AMD has been doing for some years now is putting out significant driver package updates. The update for 2017 was Adrenalin and among its new features was the ability to monitor and log various performance metrics. Unfortunately frame time is not among these, and the polling rate is at best one second, but various measurements, including power and temperature, can be recorded this way. I think these are interesting metrics to consider, so for the Vega 64 runs I also recorded this information and figured out how to add it to the graphs. Another feature of current AMD drivers is Radeon Chill, which will dynamically reduce the frame rate of a game to reduce power consumption, and I did some runs with this enabled as well, for the fun of it.

As this monitoring and logging capability is not a part of the NVIDIA drivers, I do not have comparable data for those GPUs. I am aware there is software out there that could get it for me, but I want to keep the process lean and right now I do not have such software installed. I also personally like the idea of keeping to what is presented to me, meaning what the drivers offer and not going past that, but I did make an exception with the NVIDIA GPUs. The exception is MSI Afterburner, which just has a more aggressive fan profile than the drivers use, but no other setting is tweaked. AMD drivers, however, have the Wattman feature allowing its GPUs to be tweaked without a third party tool. After getting the RX Vega 64 I spent the time to find how far I can undervolt it, eventually settling on 965 mV. I have also raised the fan speed limit to 3000 RPM (default 2400 RPM) and pushed the power limit up to +50%. I also did runs with stock settings, except for the custom fan limit. I will discuss these settings and the performance logging more in the Vega 64 section, because there are some points worth knowing.

With that gone through, here are my specs and the drivers:

It is very important to note two things about The New Colossus. One is that it uses the low-level Vulkan graphics API. I am not an expert on graphics APIs, but here is a quick summary of my understanding: low level means less abstraction work is needed to send instructions to the GPU, or any hardware. The analogy I like to use (and it could be incorrect) is an API is like a library for a computer, and whenever a function or operation is requested, it will check this library to figure out how it is done and then abstract it to run on any hardware. A low level API reduces this abstraction, so the instruction gets to the hardware faster and there is less efficiency lost by the abstraction process. Less overhead and the instructions themselves possibly being better optimized makes for a powerful combination. Vulkan also offers the added bonus of enabling better multi-threading support, compared to older APIs. This means the many threads modern CPUs offer can be better put to use.

The other point to make about the game, and this is one many other people also make, is that it appears to be a very well optimized implementation of Vulkan. I am sure we can all think of at least one poorly optimized title where the performance is lagging behind what the technology is capable of. As the game original released some time ago, it has certainly become more optimized through patches but also driver updates.

Now we can get to the actual analysis.




Related Products
Random Pic
© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.2108829021   (xlweb1)