The new performance testing tool, available through Steam, provides a coarse evaluation of your overall system. We put four configurations through their paces.
The new performance testing tool, available through Steam, provides a coarse evaluation of your overall system. We put four configurations through their paces.
It’s been only a few days since HTC and Valve revealed the final retail model of the HTX Vive VR Headset, including a somewhat eye-watering price-tag. Naturally no-one with any sense should be contemplating a $799 purchase without some indication of whether your system will even be capable of meeting the hardware’s stringent requirements, including greater than High Definition displays running at a locked 90fps. Consequently Valve have released the SteamVR Performance Test, a tool available through Steam, which makes a coarse assessment of your gaming rig.
The SteamVR Performance Test is available to download at http://store.steampowered.com/app/323910/. Once installed you can find it in the tool section of your Steam library: simply navigate to the Library, then click ‘Games’ and select ‘Tools’ from the drop-down menu. The application will take roughly two minutes to run through a sequence within Valve’s Aperture Robot Repair VR demo – a little longer on mechanical HDDs than SSDs – and quickly spit out an easily digestible summary.
We ran the test on four system configurations to get some idea of where the boundaries of ‘Capable’ are on hardware we had to hand. One is a clear failure case, with a GPU well below the recommended specifications as well as an out-of-date CPU, but others fall within the envelope of Capable (i.e. will work, but graphical fidelity settings may need to be tuned down) to VR Ready.
First Result – The Failure Case: NVIDIA GeForce GTX 660 2GB
First up we have our workaday system, comprising an Intel Core i5-3570K, NVIDIA GTX 660 2GB and 8GB DDR3 RAM; it represents a system around three years old, bought for gaming but in dire need of an upgrade. Naturally, it doesn’t perform well.
SteamVR’s Performance Test marks this one down as a clear fail, but even here there’s interesting data. Clearly the GPU fails – the GeForce GTX 660 was only ever lower-mid-tier, and pushing >HD frames at 90fps is highly unrealistic. What is noteworthy however is that the 3570K is Green; the recommended CPU is Intel’s 4590K, a full generation more advanced but in the same performance bracket. This could mean that many Ivybridge and even Sandybridge-era systems will be VR-capable with a GPU upgrade.
Today’s Capable Mid-range: AMD Radeon R9 285 2GB
The Radeon R9 285 is some way below the recommended R9 290, but was the closest GPU to the R9 290 in AMD’s product stack this time last year. Here it’s matched with Intel’s Core i7-5930K which is one of our testing system CPUs, a model which is clearly comfortably above the i5-4590K recommendation, and 16GB DDR4 RAM.
Once again, some useful info. Far from being a straight fail, the Radeon R9 285-based system is classed as ‘Capable’. It scores a 3.5(Medium) on the Average Fidelity measure, indicating that medium settings would be recommended for VR use, and 2.6% of all frames were rendered below the 90fps minimum. Gamers who already have an R9 285, R9 380 (a near-identical model) or R9 380X may be able to hold off on an immediate upgrade if they’re willing to pay around with in-game optimisations for higher performance.
The Recommended: ASUS R9 290 DirectCU II OC 4GB
Pairing with the Intel i7 CPU this time is an OC model of AMD’s R9 290, a small step up in product numbering but a major one in architecture and overall performance. This card has a Hawaii Pro GPU at its heart, is factory overclocked, and should be fairly close in performance to AMD’s current ‘bang for your buck’ king the Radeon R9 390 8GB.
Finally breached the ‘Ready’ barrier. Here we get far more consistent high-fidelity frames, and no longer drop any on the way to a smooth VR experience. For those counting at home, SteamVR rates this configuration with a 6.4(High); high-fidelity VR gaming should be order of the day with this GPU so long as you meet other requirements.
Following the summer refresh the nearest equivalent GPU is the slightly faster Radeon R9 390, a GPU which also comes with double the frame buffer size (8GB) as standard. That’s likely to have small but noticeable impact in high-resolution gaming, so your speculative R9 390 would be a great purchase for staying within some semblance of a budget.
The Recommended II: GIGABYTE GeForce GTX 970 WINDFORCE 3X 4GB
The other side of the recommended coin is NVIDIA’s mid-range king: the GeForce GTX 970. This GPU is still current, making use of NVIDIA’s latest Maxwell architecture even if it’s not as powerful as the GTX 980, 980Ti or Titan X.
Once again, the recommended spec is very much Ready for VR. No frames drop below the 90fps threshold, and frame fidelity is consistently higher than the R9 290 – our GTX 970 gets a score of 6.9(High). Even in the accompanying graph, coarse representation as it is, performance for the GTX 970 is noticeably smoother and higher than the older R9 290. Anyone lucky enough to have a GTX 970 should be well on their way to a VR-capable system without the need for an expensive upgrade (apart from the headset itself, of course).
Additional results
In the past 18 hours we’ve been provided with a few more results.
1) A system running Intel’s Core i7-4770K and NVIDIA’s GTX 780 was only able to muster a result of 5.8(Medium). That’s somewhat surprising given the strong performance of AMD’s R9 290, which is of similar vintage, and is worthy of further investigation if you’re currently running that GPU. Although on paper the GTX 780 should have comparable performance to the GTX 970 it would be wise not to assume that’s the case.
2) AMD generated in-house testing results for a range of both their competitors’ and their own GPUs. Here are the results:
AMD 2 x Radeon R9 Nano – 11
NVIDIA GeForce GTX 980 TI – 11
AMD Radeon R9 Fury X – 9.6
AMD Radeon R9 Fury – 9.2
NVIDIA GeForce GTX 980 – 8.1
AMD Radeon R9 Nano – 8
AMD Radeon R9 390x – 7.8
AMD Radeon R9 390 – 7
AMD GeForce GTX 970 – 6.5
*Test system: Intel i7-6700K, 2x4GB DDR4-2666 RAM, Z170 motherboard, Radeon Software 16.1.1, NVIDIA driver 361.91, Win10 64-bit.
NVIDIA GeForce GTX 980 TI – 11
AMD Radeon R9 Fury X – 9.6
AMD Radeon R9 Fury – 9.2
NVIDIA GeForce GTX 980 – 8.1
AMD Radeon R9 Nano – 8
AMD Radeon R9 390x – 7.8
AMD Radeon R9 390 – 7
AMD GeForce GTX 970 – 6.5
*Test system: Intel i7-6700K, 2x4GB DDR4-2666 RAM, Z170 motherboard, Radeon Software 16.1.1, NVIDIA driver 361.91, Win10 64-bit.
The SteamVR Aperture demo can make use of AMD’s Affinity multi-GPU technology, which is part of the LiquidVR toolkit; using this tech one GPU renders each eye, resulting in more scalable performance. AMD’s testing indicates that two R9 Nano’s working in tandem was able to hit an 11 (likely near-constant Very High image fidelity), and as with all Nano systems would be strong VR performers in a small-footprint system. Implementation of Affinity is still a work in progress, but results are promising.
Officially, AMD claims that all Radeon R9 390, Nano and Fury-class GPUs attain a VR-Ready categorisation through the test; we have no reason to doubt these claims. Naturally take their own results with a pinch of salt – we cannot independently verify them, and indeed our GTX 970-system hit a score of 6.9 with the same drivers.
A Word of Caution
Within the SteamVR results output Valve make it clear that testing does not account for the CPU cost due to positional tracking, or indeed other processes running concurrently with VR operation (for example: streaming). Furthermore it doesn’t test USB functionality, a looming weakness in many otherwise competent systems that relatively few interested parties are currently aware of. Hopefully the tool will continue to be developed to better simulate realistic workloads and probe for functionality of this type in the future.
Download the SteamVR Performance Tool for yourself at http://store.steampowered.com/app/323910/ and give it a bash, you may be surprised by the results.
