Matrix Multiply slower on RT than PC?

[Jeremy_Marquis]

Doh.  Again with going too fast.  The picture below explains why all results are the same and parallelization did nothing.

Apparently, I can't subtract.

[altenbach]

---

Jeremy_Marquis wrote:

Apparently, I can't subtract.

---

Yes, I mentioned it. 😄

I recommend to use "high resolution relative seconds", because you get a negative value if you subtract wrong. Easier to notice. 😮

(You also get finer resolution and the results are in seconds instead of ms.)

[wjdwyer]

To do some math-intensive work, am trying to figure what would be an optimal configuration to purcahse.  The speed-up between Parallel FOR loop (4 cores) and Parallel FOR loop (32 cores) is only 2x.

For matrix multiply, is there a LV bottleneck for regarding the physical vs. virtual cores?  Given that your machine has 8-cores/socket, might the speed-up correspond to one-socket and physical cores?

[Reyquaza]

Hi wjdwyer,

It looks like this thread is a little over a year old, and your question is a bit different than the original question. I would recommend creating a new thread for your question, as it'll be more likely to be seen and answered that way.