New Posts  All Forums:

Posts by xlink

This would be different.In the case of Core 2, Intel basically took another set of parts, Core Duo/Pentium M and then relaunched with with both higher performance AND higher clock speed. Using P4 as a predecessor paints a muddy story as it was explicitly designed around high clock speeds and less around performance.
Mind sharing a link? I'm legitimately curious. I saw the first gen of bulldozer as requiring more transistors and more power per unit of performance relative to the 1060s
Part of me just died a little.
It depends upon the task you're trying to do.If the task is anything that ISN'T embarrassingly parallel(e.g. most numerical simulations), you'll run into a similar issue.https://en.wikipedia.org/wiki/Embarrassingly_parallelYes, you can send the same CPU to go on and process other tasks, but that won't speed up the original task - it'll still be bound by the slowest thread. If you can nearly-infinitely divide a problem, that's great but many interesting problems do not have...
I work at a major telco in their TV group.I agree with this entirely.The future is going to be highly improved streaming of "low fidelity" content via unicast and "high fidelity" multicast.Low fidelity is a RELATIVE term. In 5 years it'll be 1080p. In 10 years probably 4k + 60Hz + 3D or virtual a la oculus or similar.My opinions are not my employer's.Older customers are happy with the traditional linear TV service they grew up with. They're hooked.People of my generation...
The figurine vaguely reminds me of my ex.
It depends.Assume you have 100 cores and 100 tasks. Assume completion time for each task is iid exponential with mean 1... you'd expect the max of a single run to be around 4.7 units of time.The implication there is that you end up being limited by single threaded performance for sequential tasks that have a limit to how parallelizeable the workloads can get.look up amdahl's lawhttps://en.wikipedia.org/wiki/Amdahl's_law
AMD has been playing the core count race for a while now... 4x4 platform which was essentially the brute force approach to competing with Conroe. 2x 6 core CPUs via MCM. It didn't start with Bulldozer.http://arstechnica.com/gadgets/2006/11/8327/ http://arstechnica.com/gadgets/2009/09/amd-makes-tradeoffs-in-upcoming-12-core-server-cpu/
For what it's worth, I sincerely hope that Zen is a home-run, I'm holding out on my next upgrade until benchmarks are out. I was a huge AMD fan back in the day, absolutely loved my Athlon XP, A64, Opteron 165 and Phenom II BE systems and I have been sad that AMD has fallen behind in the last half decade.
I like how this new design will allow for huge improvements in the area of... MOAR CORES.
New Posts  All Forums: