New Posts  All Forums:

Posts by xlink

That's a really good deal. Where can I buy this 65 foot TV?
At the end that's what it comes down to. You could throw more cores at a handful of tasks, do calculations based on expected stochastic outcomes and whatnot but at the end of the day that fails to efficiently improve scaling.So yes, you can do more things, more seamlessly. You can't do one thing more quickly. How one conceptualizes "thing" will get increasingly nuanced with time though. There are a lot of tradeoffs and no silver bullets in sight. In the near future,...
Excellent clarification.When you say solve probabilistically, would you be referring to essentially running out multiple possibilities and selecting the one which happens to be true in lieu of branch prediction?
They're interrelated. Gustafson's law is about solving bigger, more complex problems in similar time. Amdahl's law is about how quickly you can solve a set problem.Gustafson's law is more of a best case scenario. Running a machine learning algorithm would fall more into this camp.Running a physics simulation would fall somewhere between the two.Imagine a tower made of bricks that is knocked over. Each brick might have its calculations done by a single core. There will come...
There are interdependencies.It's not a matter of the software being made proper. There are very real limits for which there is no easy work around. Imagine software being split into 10000 threads. This would be near some theoretically perfect level. If one of those threads takes 10x as long as the others... you'll be limited by the performance of that thread and your system will be lagging while waiting for it to...
The latency, as measured in TIME as opposed to cycles, is a function of the ratio between cycles/Hz.Double the Hz and you can double the number of cycles required for a given latency in ns.The latency is higher, but there's more bandwidth overall and the CPUs are designed to favor bandwidth over latency AND to be insensitive to memory performance.
What's next, you referencing AT&T's 67Bn (debt included) purchase of Directv?
Your borderline ad hominem does nothing to refute what I'd said. My claim ultimately reduces down to the differences being smaller than previously.The differences in the memory subsystems between the xb1 and ps4 appear similar to the prior generation.The xbox still has fast cache and slower main memory and the ps, on average, has higher bandwidth/latency main memory.Heck you could argue that the ps4 now having a homogenous memory bank reduces the complexity - no more rambus.
They have a lot more in common now than did the last few generations. This narrows the scope of tradeoffs required during development for multiplatform games while generally attenuating the drawbacks.The previous gen of consoles...did not have a full unified shader architecture despite being release when PC cards just started to have those (8800GTX and 2900XTX) http://en.wikipedia.org/wiki/Unified_shader_modelused PowerPC processorshad very little RAMEach console itself...
In this case, the consoles have A LOT in common with PCs. Developing for one means a relatively easy port to the other. The cost/time/quality tradeoffs aren't as severe as they'd previously been, especially given the diminishing returns that are being experienced on the hardware side.512MB shared RAM is a lot harder to develop for than 8GB RAM which even on the PC side is often adequate(even with additional overhead) these days.
New Posts  All Forums: