Moore's Law in photos

[mr.Curmudgeon]

http://www.technologyreview.com/article/21886/

From 1 to 758 million transistors in 50 years.   

[SavannahLion]

Shame they don't run any faster.

It's the software, not the hardware. All that eye candy, background management, and hardware abstraction chew up a lot of resources.

[Samstag]

Shame they don't run any faster.

It's the software, not the hardware. All that eye candy, background management, and hardware abstraction chew up a lot of resources.

In my line of work we have systems run by 25-year old computers next to systems running on bleeding edge multi-core multi-processor hardware.  In my experience the older the computer, the faster it will boot.  Unless it runs linux.

[ahofle]

There should be a corollary to Moore's law showing that any advances in hardware will be completely overrun by software 'advances' for a net performance gain of 0. 

[mr.Curmudgeon]

Doh!

[SavannahLion]

Shame they don't run any faster.

It's the software, not the hardware. All that eye candy, background management, and hardware abstraction chew up a lot of resources.

In my line of work we have systems run by 25-year old computers next to systems running on bleeding edge multi-core multi-processor hardware.  In my experience the older the computer, the faster it will boot.  Unless it runs linux.

Yeah, even as a (suffering) software developer that was a point of contention with me. I remember when Unreal hit the market and all I had at the time was a 233Mhz (or was it a 133MHz?) AMD. I really wanted to run the game but couldn't afford the hefty upgrade. My only option was to slim down the software. I trimmed so much crap out of the system that by the time I was done, my benchmarks were consistently in the upper 20% and my FPS were as smooth as silk. I was running games with better frame rates than my classmates running the latest CPU's. After HL came out and I installed twin Voodoo2's that I got on sale, it left the faster systems in the dust. I kept that PC for at least two years beyond its life expectency. I think I tossed it when a HDD upgrade left the system suffering.

Nowadays, it's a massive amount of effort just to eek out a meager 10% performance bump. It would be less time consuming to pour in a $1000 worth of hardware each year just to keep those teeth sharp.
 

There should be a corollary to Moore's law showing that any advances in hardware will be completely overrun by software 'advances' for a net performance gain of 0. 

Wirth's Law.

[patrickl]

It all depends on how you define "performance". Software designers tend to be fixed on the amount of time that they (can) use for the computer to give a rsponse (ie "every frame it needs the screen needs to be redrawn" or "in a maximum of 2 seconds the database retrieval needs to be done"). So yeah this timed performance is a constant, but the complexity of what is returned within that fixed time has advanced tremendously.

It's like a store replacing a tiny van with a truck and then complaining the old van moved a tray of beer quicker than the truck moves a weeks supply of beer .

[danny_galaga]

great pics 

i like the 'mondrian' period myself...

[Ed_McCarron]

great pics 

i like the 'mondrian' period myself...

I was just impressed that MrC posted pics that didn't make me wanna hurl...

[SavannahLion]

It all depends on how you define "performance". Software designers tend to be fixed on the amount of time that they (can) use for the computer to give a rsponse (ie "every frame it needs the screen needs to be redrawn" or "in a maximum of 2 seconds the database retrieval needs to be done"). So yeah this timed performance is a constant, but the complexity of what is returned within that fixed time has advanced tremendously.

It's like a store replacing a tiny van with a truck and then complaining the old van moved a tray of beer quicker than the truck moves a weeks supply of beer .

I think it's more like a big ass truck that has an overnight sleeper, satelite radio, and a mobile bathroom to move that weeks supply of beer. It's not the data that's the problem, it's what's used to manipulate that data.

Look at it this way. Let's take program X that's slimmer and program Y that's a little chubby and program Z that's the slimmest of all. The reason I might like X is the extensibility inherent with the design. If I wanted to, I can add some plugins to achieve additional features. Y on the other hand tries to include every conceivable function a user might want even if 90% of the end users never actually use the functions. Program Z does one thing and nothing else, but does that one thing very very well. They all process the same data.

The problem with most modern programmers is that many want to create program Y and very few create programs like X or Z. I can live with a program sucking down the necessary resources digesting 9 GB of data. It's data I want to have processed. I can't deal with a program that sucks down 200MB or more of RAM loading a giant monolithic block of code that's capable of turning on my sink, purchasing gas and balancing my budget when all I want it to do is grab and display a 200k block of data from a file.

With a plug-in architecture, the software might end up eating more resources if I add all the plug-ins to achieve the same functionality as the monolithic version. But the catch is, I don't need all the functionality of the monolithic program therefor I still end up being ahead of the resource game. Both with a smaller program and a program that fits exactly what I need it to do.

And if I really had to worry about overhead and needed to eek out every scrap of resources, then switching over to Z would be the thing to do.

Most programmers don't necessarily need to focus on eeking out 10% performance increases on their refactors. I'm just saying there needs to be a dramatic change in the current school of thought that "hardware is cheap" and writing monolithic software to fill limited hardware resources.