RTX 2070 Block Diagram
The block diagram for the TU106-400 GPU is above. It’s the same structure as what we discussed with TU102, except hacked in half for a total of three GPCs. The rest of the containerization remains the same, and we’d recommend watching our Turing architecture discussion for more of that. In short, there’s still one overarching command processor at top that deals with shoving things into GPU memory from the PCIe bus and tasking the workload out, then there are three GPCs, each of which contains six TPCs, each of which contains two SMs, which then contains 64 FPUs. The end total is 2304 FP32 FPUs. Read the Turing architecture deep-dive for more information on this.
EVGA’s RTX 2070 Black & Nvidia’s Revised Sampling Guidelines
EVGA’s RTX 2070 is a $500 model which carries no additional branding, it’s just “RTX 2070.” Technically, we’re told that the card is an EVGA RTX 2070 Black edition, but we’d assume the limited margin on a $500 unit prohibited EVGA from investing in the additional ink.
For this round of review units, Nvidia is asking partners to handle product sampling. This isn’t abnormal for non-halo launches; however, in addition to this, Nvidia also required that board partners ship a $500 (MSRP) SKU to reviewers and noted that more expensive models must be accompanied with MSRP SKUs. From what nVidia tells us, this is to help combat issues with partners sending only their most expensive models, resulting in skewed price-to-performance reporting that never sees cards at MSRP. We don’t find this wholly unreasonable, given that we end up buying most of the ultra low-end models typically since partners aren’t too keen to share their cheapest cooling solutions with the press. All of this results in a bit of a battle between partners and Nvidia: The partners rely on NV for performance of the GPU, and so must differentiate from competitors in cooler quality; the supplier relies on partners to not paint a picture of being priced out of affordability, and so wants the MSRP units to actually be dispatched. Both are reasonable needs and wants, it just so happens that the supplier has the ultimate choice in the direction followed.
For this launch, you’ll likely see more $500 SKUs than we would have otherwise, and the more expensive coolers won’t get as much coverage until later.
Points of Interest for 2070: Memory Bandwidth
As we progress through the benchmark charts for this review, we’d recommend paying close attention to scaling performance across resolutions. A good point of comparison would be the GTX 1080 and RTX 2070, where performance disparities emerge at 1080p and often converge at 4K. In these scenarios, memory bandwidth is potentially a limiter of the RTX 2070’s full stride. Memory bandwidth requirements increase faster than the FLOPS requirement increases, in some cases. Memory overclocking demonstrates this well.
The testing methodology has completely changed from our last GPU reviews, which were probably for the GTX 1070 Ti series cards. Most notably, we have overhauled the host test bench and had updated with new games. Our games selection is a careful one: Time is finite, and having analyzed our previous testing methodologies, we identified shortcomings where we were ultimately wasting time by testing too many games that didn’t provide meaningfully different data from our other tested titles. In order to better optimize our time available and test “smarter” (rather than “more,” which was one of our previous goals), we have selected games based upon the following criteria:
- Game Engine: Most games run on the same group of popular engines. By choosing one game from each major engine (e.g. Unreal Engine), we can ensure that we are representing a wide sweep of games that just use the built-in engine-level optimizations
- API: We have chosen a select group of DirectX 11 and DirectX 12 API integrations, as these are the most prevalent at this time. We will include more Vulkan API testing as more games ship with Vulkan
- Popularity: Is it something people actually play?
- Longevity: Regardless of popularity, how long can we reasonably expect that a game will go without updates? Updating games can hurt comparative data from past tests, which impacts our ability to cross-compare new data and old, as old data may no longer be comparable post-patch
Game graphics settings are defined in their respective charts.
We are also testing most games at all three popular resolutions – at least, we are for the high-end. This includes 4K, 1440p, and 1080p, which allows us to determine GPU scalability across multiple monitor types. More importantly, this allows us to start pinpointing the reason for performance uplift, rather than just saying there is performance uplift. If we know that performance boosts harder at 4K than 1080p, we might be able to call this indicative of a ROPs advantage, for instance. Understanding why performance behaves the way it does is critical for future expansion of our own knowledge, and thus prepares our content for smarter analysis in the future.
For the test bench proper, we are now using the following components: