NVIDIA specs VS the competition.

Why have NVIDIA less specs more performance vs the competition?

Hi my name is Pedro Figueroa student of Atlantic University College (atlanticu.edu)
The Master Degree I’m working is on the area of Computer Graphic Designer with a concentration on Design

 The thesis is about GPU, video cards or graphics cards and I would like to know why

you “NVIDIA” have less specs on the GPU and the Video Card like: TDP or Power requirements,
V-RAM or Memory, CUDA cores or processors and Memory Interface Width, Memory Bandwidth(GB/sec).

Having less specs than the competition you “NVIDIA” achieve better performance,
that is because of a software implementation rather than Hardware?

I would gladly appreciate your HELP and truth!!!

I am not aware of any comprehensive study in this regard for GPUs, but for any modern processor architecture (CPU or GPU) application-level performance is usually a function of numerous complex interactions inside hardware, and between hardware and software. This creates the potential for many bottlenecks that prevent one from achieving the theoretical throughput of particular functional units.

For example, on processors that have separate FADD and FMUL units, theoretical FLOP/s ratings can only be approximated if the instruction stream delivers 50% additions and 50% multiplications, globally as well as locally. I worked on one x86 processor in the past where getting close to the theoretical FLOP/s rating required paying close attention to the length of instructions, to avoid becoming limited by instruction decode. In various architectures, achieving full computational throughput requires careful register allocation in the instruction sequence to avoid bank conflicts in the register file.

Software engineering excellence period. (No, I don’t work for NVIDIA, though I’m an unyielding fan)

I think one big thing NVIDIA does is reconfigure their drivers for each game/app. When you look at the release logs, you often see things like 50% more FPS on the Witcher on GTX 9xx cards. It’s hard to imagine that won’t be to the detriment of other games if applied universally. 1 very common optimization is to use different algorithms/data structures for sparse and dense data (e.g. use a hash table for sparse data and a direct mapping table for dense data). I’ve done this with great speedup before. So much is possible from specialization.

Again, I don’t work for NVIDIA, so I’m extrapolating from their driver reliability record.

Personally, I don’t remember ever getting a blue screen due to an NVIDIA driver. Also, when I submitted CUDA bug reports, they worked on even very obscure ones.

I don’t want to disparage AMD engineers in any way. I’d say the difference is mostly due to company culture. I do wonder to what extent is the excellence a mutual, collective effort or is it the will of one person… - Admiral Rickover, did you join NVIDIA? Are you the person behind 0 blue screens, just like how there’s been 0 nuclear accidents in the US navy?

Nvidia lists their GPU specs, so not sure to what you are referring.

As far as performance this list (which is an average of submitted 3D Marks scores) shows Nvidia vs the “competition”;

http://www.videocardbenchmark.net/high_end_gpus.html

Nvidia still holds the tops spots by a considerable margin.

And the “competition” is doing really well these days, as you can tell from their stock price:

http://bigcharts.marketwatch.com/quickchart/quickchart.asp?symb=amd&insttype=&freq=1&show=&time=8

The “truth” hurts!

Hi and thanks to njuffa and Uncle Joe for your quick and well explained answers, however the more answers the better!

CudaaduC I do not know what is your side, if it is 2D & 3D Artist or a programmer but we are not just talking about gaming performance. I have use both cards and NVIDIA is much better on most of the softwares used by 2D & 3D Artist in the industry.

Why have NVIDIA
less specs more performance vs the competition?

Hi my name is Pedro Figueroa
student of Atlantic University College
(atlanticu.edu)
The Master Degree I’m working is on the area of Computer Graphic Designer with a concentration on Design.

The thesis is about GPU, video cards or graphics cards and I would like to know why
you “NVIDIA” have less specs on the GPU and the Video Card like: TDP or Power requirements,
V-RAM or Memory, CUDA cores or processors and Memory Interface Width, Memory Bandwidth(GB/sec).

Having less specs than the competition you “NVIDIA” achieve better performance,
that is because of a software implementation rather than Hardware?

I would gladly appreciate your HELP and truth!!!

Whoops, I’ve been mistaken. The lack of blue screens has probably more to do with Microsoft requiring the driver being implemented in user space than in the kernel.

WDDM on Wikipedia: