5 Replies Latest reply on Jun 8, 2017 9:02 AM by JJMack

    Brush performance, not using CPU efficiently.

    KimSx

      Hi,

       

      Got a new system and Photoshop CC 2017.1.1 is generally pretty fast, however I do a lot of painting and noticed that CPU heavy brushes aren't performing so well. When monitoring it I got this. It seems Photoshop barely makes use of more than a single thread? This was captured while using a brush that showed heavy lag (spinning circles while it tried to catch up), meanwhile the CPU is practically sleeping. Is there a way to change this, or does Photoshop not support multiple cores properly for brushes?

      brushperformance.jpg

        • 1. Re: Brush performance, not using CPU efficiently.
          JJMack Most Valuable Participant

          Multiple threads can only be used when processing can be broken into to threads the can be performer in parallel some processes are sequential in nature you can not blend in paint if you do not first pickup what the paint should be blended with and what it blends with changes with brush spacing.  How can a process like that be broken into parallel threads that can execute at the same time.

          • 2. Re: Brush performance, not using CPU efficiently.
            KimSx Level 1

            Thanks, makes sense. It does outsource some brush operations to the other threads, and it would be nice if that could be even more, since the stroke placement is rarely the most taxing part for the CPU, rather stroke modifiers such as texture, dual brush ,etc. If at all possible, that is. Processors with more threads and lower clock frequency get punished even more.

            • 3. Re: Brush performance, not using CPU efficiently.
              JJMack Most Valuable Participant

              Punished hardly. My machine has Dual 6 core 2GHZ Xeons so when thing can be threaded I have 24 Processors running full tilt performance is not bad.   My machine set be back $2000 not bad.  It is true I could have Purchase dual 8 Core 3.6 GHZ for $8000 and have better performance. However, I do not consider keeping $6000 in my pocket punishment.   I do nothing that needs the much processing power routinely. Photoshop surely does not require that much compute power routinely..

              • 4. Re: Brush performance, not using CPU efficiently.
                KimSx Level 1

                So you don't need it, good for you. That doesn't mean others don't. There's a lot of painters out there using PS. Matte painting in 8k, 16 bit with 100+ layers, and you'd wish for this too. I'm not a programmer but I'm thinking there is room for improvement - this is what I get in other graphical software when I'm pushing it:

                lwperformance.jpg

                • 5. Re: Brush performance, not using CPU efficiently.
                  JJMack Most Valuable Participant

                  You not a programmer. You have no ideas of how Photoshop is programmed you know other programs use multi threading  so Photoshop has room for improvement.  Here is what I see using Photoshop. You need to realize Photoshop has an interactive User Interface. Post of the time Photoshop is waiting on you. Most likely more than you waiting on Photoshop.   You wait on Photoshop while your long running Photoshop Automation processes tun and when things need to be done a as sequential process that involve processing a huge number of pixels because of size or layers of numbers of smart filters that need to be reapplied.  Even when you have a Fast processors and lot of free machine resources  Processing a huge amount of pixels can be quite time consuming.  This Screen capture shows the CPU utilization of Photoshop the only active Application I was using on my machine. You can see Photoshop was using little of my processors till i used surface blur on a huge layer.  Photoshop had no problem  multi threading Surface blur.  All 24 of my logical Possessors went form no utilization to 100% utilization for around 50 seconds.  It took 50 seconds to surface blur my huge single layer document.

                  Capture.jpg