How Well AMD R9 Nano Fares Against GTX Titan X? Let’s Do Some Math!
AMD unveiled three brand new high-end graphics cards at E3 2015, and out of those possibly the star of the show was the Radeon R9 Nano. This 6-inch long, 175W card is based on AMD’s latest GPU Fiji with 4GB HBM1. The company claims the R9 Nano offers up to two times the performance-per-watt over the previous flagship, the R9 290X, which makes it the most powerful mini-itx graphics card ever. But, one might ask how fast is it exactly?
Well, AMD did not directly share a precise performace figure for the card in question. But using simple Math skills coupled with our previous knowledge of the subject, we can figure out how well Nano should perform when it launches later in the summer. So let’s start doing some geek work without any delay.
How fast a graphics card could perform depends on its “compute power” – which refers to how many FP32 TFLOPS a card could achieve. It means if we know the peak FP32 TFLOPS of R9 Nano, we’ll be able to compare it to those of other cards, thus getting an exact idea of its compute power.
We know that performance-per-watt is defined as the peak FP32 TFLOPS divided by the Typical Board Power, i.e., Perf/W = FP32/TDP. Since we already know the peak FP32 performance of the R9 290X and its TDP, we can go ahead and calculate its power efficiency as well (Via WCCFtech).
Peak FP32 (R9 290X) = 5.6 TFLOPs; and its TDP is 250W
So, Perf/W = 5.6 TFLOPs/250W, or 5600 GFLOPs/250W
Or, Perf/W = 22.4 GFLOPs/W
We also know that the R9 Nano has 2X the perf/watt of the R9 290X.
Perf/W (R9 Nano) = 2X (5.6TFLOP/250W) = 2X 22.4 GFLOPs/W
Or, Perf/W = 44.8 GFLOPs/W.
We also happen to know the TDP for the Nano, i.e., 175W as mentioned at the start of the article. Now, using the same Perf/watt equation, we can extrapolate the peak FP32 TFLOPS of R9 Nano.
R9 Nano’s Perf/W = FP32 in GFLOPs (unknown) / TDP (175)
44.8 = FP32 (unknown) / 175
44.8 x 175 = FP32 (unknown)
Or, FP32 = 44.8 x 175 = 7840 GFLOPs or
FP32 (R9 Nano) = 7.84 TFLOPs.
We’re not sure whether AMD can actually hit this 7.84 TFLOPs figure with the R9 Nano. But in the light of the additional data points the Sunnyvale, Calif-based tech firm revealed earlier, it should probably reach around the 7.3TFLOPs mark.
Compared to the GeForce GTX Titan X, the R9 Nano would edge out the Nvidia’s flagship (having 6.14-6.6 TFLOPs) in FP32 Compute and deliver nearly twice the GFLOPs/W.
It’s worth mentioning here that the compute power doesn’t directly correlate with gaming performance. Especially when we talk about different architectures, i.e., AMD’s GCN and Nvidia’s Maxwell. Except when comparing chips based on the same architecture. For example the R9 Nano, R9 Fury X and the console chips. In which case, it does translate to gaming performance in every scenario.
Have something interesting to add to the story? Sound off in the comments below!