Revit 2019 and IOMeter
This page contains two benchmarks that we have implemented for the first time into this CPU review. We are always looking for benchmarks that provide more of a real-world look at performance instead of just synthetic workloads and are excited about these test. That being said, we are always looking for feedback and suggestions as to our benchmark suite!
Revit 2019 – RFO Benchmark
RFO Benchmark is a community developed series of scripts that can be used to evaluate system-level performance in Revit, the popular building information modeling software from Autodesk mainly used by those in the Architecture fields.
This benchmark tests several different aspects of Revit functionality, which stress the system in various ways including the following subtests:
- Update (converting a file to the Revit 2019 format from a previous version)
- Model Creation
- Export (converting the document to a format for raster printing)
- 3D Render
- Graphics (refreshing and rotating within the viewport of Revit)
In Revit, we see a split between the HEDT processors and the mainstream processors depending on what action is being done. Tasks like Model Exporting and Rendering take advantage of the extra available cores, while the rest of the functions are single threaded, and hence faster on the faster clock speed CPUs
In general, it seems like the faster single-threaded mainstream CPUs like the i9-9900K are actually the most suitable for Revit, despite it being a "workstation-level" application.
One of the more overlooked aspects of single-threaded performance is disk throughput. As disks get faster and faster, a single application thread can struggle to achieve full bandwidth in some scenarios.
In order to evaluate this, we are running a RAM disk of 1024MB on the host machine using the Softperfect RAM disk application, and then testing throughput in IOMeter with a 100% random 4K workload. Since a RAM disk is the fastest possible storage device, this will give us an idea of how IOPS scale with single-threaded processor performance.
Our IOMeter results show that Intel still has the advantage for maximum IOPS at a queue depth of 1 (also known as a single thread). The Core i9-7900X is almost 80% faster than the Threadripper 2920X in this test.
Enabling Dynamic Local mode results in a 22% drop in Queue Depth 1 Maximum IOPS for the Threadripper 2970WX.