Ryzen 1700x and Vega 64 Build - Benchmarks and One Oddity

[Zoolook]

Hi - I posted this over at AMD-OSX as well, so apologies if you've seen it there. I am trying to resolve an issue with one game, Rise of the Tomb Raider, but attached is my system information and full suite oof benchmarks for context.

 

So here are my Ryzentosh Specs (it was previously a gaming PC but with an nVidia card, which I swapped for the Vega 64)

AMD Ryzen 1700x (3.4ghz) - Water Cooled
Aorus Pro B450 WiFi Motherboard
16GB Corsair RAM @ 3200mhz (I'm planning to upgrade to 32GB)
Colormaster Vega 64 (8GB)
Fenvi T919 WiFi and BlueTooth Adapter
Samsung EVO 970 NVMe 512GB SSD (Boot)
WD 7200 2TB and WD 5400 4TB (Time Machine)
Lian Li Case

So first off - pretty much everything works. App Store, FaceTime, Handoff, Airplay, SideCar with an iPad (this one was a nice surprise). I have not tested Logic Pro which is important to me, but I'll get to that in a later post. The system runs cool, although the AMD Power Management isn't quite as precise as on Windows.

 

 

So here are the Benchmark Results: Geekbench, Cinebench, Novabench, Unigine Valley, Unigine Heaven and all 3 Tomb Raider Games.

Geekbench - Doesn't really tell us anything we don't know, although I will say it's quite forgiving of overheating Intel chips as it doesn't stress them too much. Intel still wins in Single Core (8th Gen Intel Vs 1st gen Ryzen) but loses in multicore (Cinebench shows a much wider gap because thermal throttling comes into play). On the GPU side, the fact a Vega 64 crushes a modest under clocked Polaris-based chip should not be a surprise - it's comfortably 2 - 3 times faster.

 

 

Cinebench R20 - This is probably one of the best raw CPU benchmarks out there given it will max it out for a sustained period of time. My 8850 always throttles on this test, unless you disable TB, but this is also pretty much what happens in real-life usage so it's a good test. Again in Single Core mode, Intel has the edge and the chip can maintain its turbo boost for almost the entire test. But in Multicore, Ryzen is 70% faster thanks to two extra cores and no thermal issues. The multiplier ratio is also FAR better with AMD, again thanks to the fact it maintains exactly the same clock speed. The Intel chip really might as well only have 4 cores.

 

 

Novabench - So I cam across this benchmark recently, and it's clearly a little old, and it also has an oddity in that the Vega 64 does really badly. I included it because it's clear than some software out there is tuned to actual Mac configurations, and might not perform well on an unofficial configuration.

On CPU it's showing 20% faster, which for a short light test might be about right - smaller than Cinebench but larger than Geekbench. Disk score I can believe, Apple makes really nice disk IO and optimizes the software, but the GPU score is inexplicable to me.

 

 

Unigine Valley and Heaven - So two old favorites that still run on Mac. This is mostly testing the GPU, and it's between 90% and 126% faster with the newer test showing the better result. It's a shame Superposition isn't available on Mac as I think that would show a much larger difference given the VRAM it uses - if I had a 1440p or 4k monitor, there would be a larger difference. The min fps issue on the Ryzentosh would probably manifest on any Mac no matter how powerful, because for wherever reason, macOS really struggles to pre-cache textures, so the tests always start with a stutter but are pretty consistent afterwards.

 

 

Tomb Raider (2013) and Shadow of the Tomb Raider - OK so I left Rise of the Tomb Raider out until afterwards, given it's the big mystery. But these two more or less follow the script, but there are some clues as to what might be going on with RoTR.

2013's TR shows more than 3x the performance on the Ryzentosh with a min frame rate just under the magic 60fps. I have to say playing the game with v-sync on at 60hz saw zero stutters at all, so I think that min fps is just caching again. Even so, the difference is huge.

On SoTTR it's not quite as clear cut, although the frames rendered is 2x as many on the Ryzentosh and the average FPS bares this out. However what's telling (and not shown below) is that on my MBP the benchmark says the result is 99% GPU limited, but on the Ryzentosh it says it's 0% GPU limited (so obviously 100% CPU limited).

The details support this, with the GPU being consistently 3x faster, but the CPU is up to 33% slower. I am not sure if this is because it's only using a few cores or if it's an optimization issue. I suspect the latter because in monitoring the CPU during a benchmark, all 8 cores seem active, although they're not steady. Even so, the game is extremely smooth in playing with only the occasional stutter on the very highest settings (even nVidia's Purehair!)

 

 

Shadow of The Tomb Raider - ok so I saved the best (worst?) until last. I have played around with this for hours and cannot fathom why it performs so badly. Even on the LOWEST settings, I get the same results more or less. All 3 scenes stutter badly near the start, almost freezing, and nothing I do changes this. Unfortunately the same occurs during gameplay, random freezing or inexplicable slowdowns. Ironically, it was THIS game that made me want to build a Ryzentosh as I started casually playing it on my MBP but it would grind to a halt after 10 mins of play due to thermal issues.

So this is disappointing. If anyone has any idea why this is occurring, or has a similar system but doesn't have this issue, please get in touch.