Author

Topic: GPU Benchmark - vs CPU benchmark (Read 156 times)

newbie
Activity: 11
Merit: 4
September 22, 2018, 11:20:20 AM
#1
While not alt coin 100% specific, or mining 100% specific, - i'm working on a white paper related to both so i hope the question in this spot is appropriate.

What benchmark tool (windows or linux) can anyone suggest that would provide a numeric value of output that shows both CPU and GPU that could relate to each other in the same score type.  

what I am looking for is very similar to how NiceHash algorithm checks for best output on a specific miner's computer (showing the Hash per second on the CPU vs each gpu).  The difference for me is I'm simply looking to get that type of calculation comparison in a way that may not be so cryptocurrency focused (i.e. using Nicehash as my bench mark tool).
 

This is for a white paper i'm working on that is not crypto specific but still relates.  I'm trying to find something that shows the relationship between GPU models (RX580, Vega 64, and GTX 1080 TI in my case) and any cpu to show how more efficient on power consumption the GPU is in doing a specific benchmark.  (I'm monitoring power to compare versus the 'score' of whatever benchmark I can use).

The only caveat is the benchmark tool I use needs to be publicly accepted - not home grown.  I've gone through LinPack, PassMark, BaseMark, UserBenchMark, Unigine_Heaven, and a few others.   The common thing I'm running into is they state the 'gpu' success in 1 metric (say Frames per second) versus the computer/cpu score in another metric (say flops per second).  Linpack has some gpu related items - but only for CUDA (wont work on the AMD also).  Writing my own benchmark (say a simple hash per second test say with cryptonight or something - unfortunately isn't going to meet the needs of a publicly accepted benchmark suite to use as my baseline).

My goal is to get the exact same benchmark number (say fps) compared against both to show the relationship.  I also realize that different benchmarks that are cpu specific vs gpu specific throw the results in favor of whatever they are written at.  I'm trying to visually show that a GPU is about 7 x more power efficient at the 'same' benchmarked mathematical task as say the CPU.

Sorry for the techo-babble-  but I seen some very smart posts on here and this I figured was my best shot to find one of you to point me in the write direction for my white paper research element I'm trying to produce.

thank you.
Jump to: