CPU core performance scaling
CPU Perf scaling
On this page we show some performance relative to CPU cores, e.g. differences based on the number of CPU cores. Our X99 based motherboard can independently control the number of CPU cores of our Core i7 5960X in the sense they can be disabled / enabled.
So basically with our motherboard we start to disable CPU cores as shown in the above BIOS screenshot. We leave hyper-threading enabled. Our CPU per core is clocked at 4.4 GHz and we drop from 8 towards 4 towards 2 and even 1 cpu core enabled. I'll be using FCAT for measurements as that way we can display the tiniest of differences in performance.
In the first example below we take the most stringent graphics setting we can select, this is Very High Quality with FXAA on a GeForce GTX 980. This is Quad HD (2560x1440) and we use the first 30 seconds of the intro start where you walk towards the mountain in a blizzard.
The above chart is an FPS plot relative to time-frames. At 50% we see that 50% of the frames are rendered at roughly 45 FPS okay ? Even with one fricking CPU core the game performs as fast as one with eight cores. Let's zoom in with frame latency though:
So aside from one frame-drop (means nothing) we see very consistent performance right ? So this is the difference in-between 1-2-4 and 8 cores. I call that NIL. Things for AMD however are completely different, and is the root cause of the performance issues:
Here again we see an FPS plot relative to percentiles. This is the Radeon R9 Fury. Watch and observe it closely, the red line is the one core configuration ... it is the fastest with 43 FPs at 50% of the rendered frames = Average FPS. It is then followed by the 8 core install at 39 FPS, then the 4 core is lower at 36 FPS and finally the 2 core setup does 33 FPS on average. In total there's a 10 FPS differential !
I think we found a root core of the performance issues, they seem to be related to the number of CPU cores. It is mind boggling to see that with the Radeon the 1 core setup remains to be the fastest. Here you will also notice that with a 2 core processor setup we have micro-stuttering going on whereas the 1, 4 and 8 core setup are fine. After seeing this it is without doubt that AMD will need to do a thing or two with driver optimizations.
Now lets combine the two 'similar priced' brand products:
So yeah, the root cause of the weird Radeon performance has to be CPU optimization / utilization / API overhead related. Above again the FPS relative to percentiles (higher is better).
And here 30 seconds frame-time latency, where lower is better. The outcome of this test was very interesting to say at the last.
In the above video you can see that intro sequence FCAT recording (each rendered frame is labeled with a color you can see to the right) with merely cpu CPU active on the GeForce GTX 980. QHD FCAT recordings take 450 MB/sec to record, sound is disbaled.