suggestion Productivity Benchmarks

zamalek · July 15, 2016

New here, hi :).

Let's face it: one of the major reasons that we all use PC is because it's useful for more than gaming. We can do our taxes on that shit. What I happen to do on that shit is software development and, so far as the Microsoft stack goes, quite a lot goes into planning a sexy development PC. I want a sexy development PC and a sexy gaming PC, and so things like "8 cores is too much" doesn't really apply to me. IDE responsiveness etc. affects my job and information like that simply doesn't exist: it's all hypothetical (high Mbps is good for compilation, GHz is good for IDE). My IDE uses every single piece of your hardware - including the GPU. Rendering a 2D UI incurs many draw calls, and so it's not purely about fillrate in this case. Developers can be absurdly specific about the hardware in their PC:

"Don't ask me why I need six cores and 24GB. To paraphrase a Zen master, if you have to ask, you do not know. "

Why not include the compilation time of, say, Chromium in your benchmarks or maybe the framerate of Visual Studio with a complex WPF designer open? While I'm generally uninterested in Bitcoin, the hash rate of GPUs is a really good indicator of their performance for non-gaming workloads. But, I also want to know the performance of games because I do that too. Maybe it's blindingly fast in Visual Studio but unusable for VR.

Which brings me on to my next point: VR-readiness opinions would be great. My 6-year interval between builds is right about now and I want trustworthy opinions on how well something copes with VR. The RX 480 was being touted as a VR-ready device and so that makes talking about VR potential in the review of the device quite important. Normally I'd jump for green team, but VR-readiness at such a low price is just plain practical compared to that sweet, sweet 1080. It's a hard decision to make and I would have really appreciated some of that honest opinion that LTT is renowned for.

BrightCandle · July 15, 2016

It handles VR as well as a 970 will, it's the minimum specification. As a developer you don't typically want a minimum spec to do something because typically we get stuff working first, then we focus on making it perform well and in that order. So having a bit of performance left over with a faster card is a good way to save a bit of productivity and delay having to do something about it until we get to the stage where we have enough of the software in place to understand what we should optimise. On the other hand if you want to make sure you never drop a frame then using minimum spec might help ensure you deal with the issues as they are introduced as any frame drops will be making you quite unwell.

I don't like developing with minimum spec, I like to have overhead to run a debugger or a profiler and still get decent performance and not have to close down other tools as well as running servers locally as well.

LogicalDrm · July 16, 2016

Testing this is another thing to add. And most people won't care that much about GPU performance in productivity software. I think Linus has taken productivity as part of CPU testing already, by stating how fast it encodes videos.