Virtual Reality Flight Simulation Computer; Selection of Components for Optimum Performance

Virtual Reality Flight Simulation Computer; Selection of Components for Optimum Performance

By Bruckler, J., Garcia, J., Garcia, J.M., & Hutchison, R.

Summary of Findings:

Upgrading the base system CPU from an Intel i7-9900K CPU to an Intel i9-9900K CPU made a significant improvement to system performance. Adding RAM from 16GB to 32 GB made a noticeable improvement to read/write speeds and overall performance. Upgrading the RAM to a faster clock from 3600 to 4400 mHtz did not improve performance. We think that the limiting factor was not the faster clock RAM but the limitation was in the MB and recommend not purchasing faster clock RAM until confirming your MB can utilize the faster clock. Changing the SSD from Samsung 860 EVO to the Samsung 970 EVO NVMe drive significantly improved the system’s performance. The system tested here, as configured, consistently benchmarked at over 200 fps and over 10,000 gflps.

This project looks at what hardware is best to optimize the performance of a computer designed to provide processing power to a virtual reality flight simulator. We tested various configurations of motherboards, processors, video cards, RAM and hard drives, both SSD and M.2 NVMe. As noted, some additional hardware is recommended, and some is not.

Base System Configiration:

Mother Board: Gigabyte Z3980 Designare
Power Supply: Thermaltake 80 Plus Bronze 850 Watt
CPU Cooler: Corsair H115i Platinum Hydro Series
Video Card: GeForce RTX 2080 6GB and 11GB
Samsung 860 EVO SSD and Samsung 870 M.2 NVME

CPU Hardware Tests performed:

The base system tested was the Intel i7-9900K CPU. The i7 was benchmark tested with the free version of Novabench. The initial test reported a CPU score of 948 of which all downstream tests was compared. The test was repeated with the Intel i9-9900K CPU. The I9 was benchmarked with a CPU score of 1778. Initial base configuration included the same Samsung 860 EVO SSD and 16GB of Corsair Vengeance 3600 RAM.

RAM Hardware Tests performed:

The base system was then tested with Patriot Viper Series DDR4 4400MHz with two 8 GB sticks for a total of 16 GB system RAM. The Patriot Viper RAM was benchmark tested with benchmark software. Benchmark reported a 23534 MB/s for read/write speeds with a RAM rating score of 260. Tests were conducted with the same amount of RAM but with a slower clock rate of 3600. The slower RAM tested was Corsair DDR4 3600MHz with two 8 GB sticks for a total of 16 GB system RAM. The Corsair RAM was benchmark tested with Novabench. Benchmark reported a 24,403 MB/s for read/write speeds with a RAM rating score of 263. The overall difference was not significant to warrant the cost of the more expensive faster RAM. However, after the testing was complete, we noticed that the MB had a slower clock rating then the RAM. While not conclusive, we think that the limitation of the RAM was not the RAM itself but the limitation was the bandwidth of the MB.

We then re- tested after adding 16GB more of the Corsair DDR4 3600MHz with now four 8 GB sticks for a total of 32 GB system RAM. The Corsair RAM was benchmark at 26,349 MB/s for read/write speeds with a RAM rating score of 311. The overall difference was significant and warrants the additional cost for 16 GB more RAM to 32 GB. We will update this report shortly with results with 64GB RAM.

Solid State Drive Hardware Tests performed:

The base system was tested with Samsung 860 EVO 1 TB SATA SSD. The Samsung 860 EVO SSD was benchmarked at 512 MB/s for read speeds and 525 MB/s write speeds with a hard drive rating score of 90. The base system was then tested with Samsung 970 EVO 500 GB NVMe drive. The Samsung 970 EVO NVMe was benchmarked at 2,169 MB/s for read speeds and 2,420 MB/s write speeds with a hard drive rating score of 263. This definitely supports the cost justification to upgrade from SSD to M.2 NVMe.

Final System Bench tests and Final Configuration: Intel i9 9900K, Gigabyte Z390 Designare, 32GB Corsair 3600 DDR4 RAM; Samsung 512GB M.2 NVMe SSD, Thermaltake 80 Plus 850 Watt PS, Corsair H115i Platinum Hydro water cooler, Nvidia GeForce RTX 2080 6B, Win 10 Pro, installed on a open test bench.

Novabench Results:

Overall: 3618
CPU: 1778
RAM: 311
RAM Speed: 26349 MB/s
GPU: 1263
Disk: 263
Disk Read Speed: 2169 MB/s
Disk Write Speed: 2420 MB/s write
FPS: 224
GFLOPS: 11746

Note: Previously our company has tested our systems with the Nvidia GeForce 2080Ti 11GB GPU and 64GB of Corsair RAM. There is an incredibly significant increase in the gflps rating which increases to over 16,000 gpflps with the 11GB 2080 GPU. There is minimal improvement when going from 32GB RAM to 64GB RAM. The other benchmark tests are not fundamentally impacted by the more expensive video card, or the additional RAM. Our recommendation for production level VR computers for the DoD is to upgrade the GPU to the more expensive 11GB card and add the extra 32GB of RAM. For a personal computer and a home VR system, we do not think the extra nearly $1000 worth the expense.

About Dr. Roger Hutchison.