Announcement

Collapse
No announcement yet.

Ryzen 5 vs Core i5

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #61
    Gary,
    my thoughts. I would still go with the cheaper one.
    You don't get twice the performance when half of the CPU cores are sat idle almost all of the time.
    There may be a slight improvement in performance but you'd probably not notice it.

    "A couple of years ago I bought an i3 machine and it seemed a bit slow or unresponsive to the user. It ran my apps just fine but the PC itself seemed a bit sluggish. How does your comment bear on that observation?"

    Maybe you had something like an antivirus program running in the background which slowed everything down?
    Maybe it has a slow disk drive?
    Maybe it was short of RAM?
    It's difficult to come to any conclusions about a computer from years ago that I've never seen.



    If you were doing something that really needed a lot of CPU power then you'd write multi-threaded programs to make sure all CPU cores were utilised, but with single threaded programs it makes no sense to buy the CPU with more fast cores when they will rarely be used.

    The more likely cause of sluggishness with those 2 CPUs is that they're intended for laptops.
    Unless they have proper cooling then, when running CPU intensive programs, they'll quickly reach maximum temperature and throttle back to prevent overheating, but that will apply to any laptop CPU.

    Comment


    • #62
      Howdy, Stuart!

      I assume you mean everything going on besides my app?

      If, as the picture suggests, the (background) CPU usages are very small, would that translate to a worse user experience? What is the practical/observable effect of what you describe?

      Thanks for the comment!

      Comment


      • #63
        I noticed Dell completely skipped PCIe 4.0 withThunderbolt on desktop computers.
        I am guessing new motherboards are coming out around Christmas with PCIe 5.0.
        This should really bump up the performance with newer NVMe SSD's and virtual machines.
        I would also think Windows 11 should be taken into consideration.
        !1th generation (and coming higher) CPU needed from Intel for PCIe 4.0 for those waiting to upgrade machines.

        Comment


        • #64
          If you look at the number of threads the OS is using, you will see the virtue of additional cores. You may only write single core apps but the OS is runnng many things at the same time. Just as an example, I still use a now antique 6 core i7 that I have clocked at 4 gig but a 12 core Xeon I also have in an identical board running at its default 3 - 3.5 gig feels faster and out performs the old i7 on any thread work. Give an OS more threads and it leave more room for you own apps.
          hutch at movsd dot com
          The MASM Forum

          www.masm32.com

          Comment


          • #65
            It's like deciding if your family needs 4 Ferraris or 8 Ferraris.
            I'm sure there are rare occasions where the 8th Ferrari might be useful but is it really worth the expense of the 4 extra Ferraris when they'll just sit idle most of the time alongside 2 of the four Ferraris you already have?

            Comment


            • #66
              Howdy Paul!

              I'm of a mind to go as you suggest - using the less expensive i3 over an i5. Other than my apps, my low vision users don't do much more than access the Internet. I think the feeling of "sluggish" is more because I'm used to an i7 and my expectations are much higher.

              Also I'd expect that an i3 of today would perform better than the i3 of 3 years ago. That favors giving the latest generation a try.

              I think I'll go ahead and buy a PC with a recent i3 and give it a test drive. I don't want my customers to pay more $$ for a PC if they would find the lower cost system totally acceptable. I don't want to impose my own personal preference over to them when they might not be able to tell the difference.


              Howdy, Stuart!

              feels faster
              is at the heart of this discussion. I've used a Lenovo stick to run my software, so I know that almost any processor will run my apps. But a Lenovo-stick PC just won't feel as fast as an i3/i5/i7 machine, at least, not to you and I who are used to faster machines.

              Comment


              • #67
                Originally posted by Gary
                Other than my apps, my low vision users don't do much more than access the Internet.
                Web browsers use many threads. I have my internet cache on a RAM disk.

                Comment


                • #68
                  Originally posted by David Roberts View Post
                  Web browsers use many threads. I have my internet cache on a RAM disk.
                  Many threads isn't an issue in this case. Windows has always used many threads.
                  According to Task Manager, the laptop I'm typing this on is currently running 2200 threads but CPU utilisation is only 1%.
                  Most of the threads are idle and don't need much of the CPU.

                  The background threads are almost irrelevant.

                  If this was a single core CPU then I'd have 1% of a single core used for background tasks and 99% of a single core available for my single thread application.

                  If this was a 2 core CPU then I'd have 1% of 2 cores (equal to 2% of a single core) used for background tasks and 100% of a single core available for my single thread application.
                  My single thread application would be 1% faster.

                  This is actually a 4 core CPU.
                  I have 1% of 4 cores (equivalent to 4% of 1 core) doing background tasks and 100% available for my single threaded program and 2 cores sitting idle.

                  I still only run 1% faster but I have 2 idle cores with nothing to do and a third core idle for 96% of the time.

                  That's a bit simplified as Windows will actually distribute the 4% across more than 1 core, but it highlights the issue.

                  Comment


                  • #69
                    Originally posted by Paul
                    Many threads isn't an issue in this case. Windows has always used many threads.
                    I know and very few processes use four or more threads. I am using Firefox right now with only one tab open, and it is using four threads. Four threads running asynchronously must be better than four threads running synchronously.
                    Last edited by David Roberts; 26 Jul 2021, 05:15 PM. Reason: 'and' instead of 'but'

                    Comment


                    • #70
                      4 threads using 0.1% of a CPU doesn't really make any difference to Gary's 1 thread that might briefly need 100% CPU.
                      The 4 threads might restrict Gary's code to 99.9% instead of 100% of a CPU core but his users are not going to notice that.
                      Should he pay $150 more to reduce the 0.1% problem?

                      Comment


                      • #71
                        I am not concerned about Gary's single threaded applications - I am talking about multi-threaded applications. When multi-core PCs started to become popular, web browsers started to become multi-threaded. A single core machine would not do Google Chrome any favours as we start ramping up the number of open tabs.

                        Comment


                        • #72
                          Paul made a good point, cook a processor and it gets slow as the CPU throttles down to keep the maximum temperature down. On big desktops the solution is easy, use a BIG liquid cooler or a massive air cooler but with a laptop, decent cooling solutions are hard to come by. If you can find a way to stick a fan into the guts of it, it will work better for longer and I have heard of accessory coolers for higher end laptops.
                          hutch at movsd dot com
                          The MASM Forum

                          www.masm32.com

                          Comment


                          • #73
                            I thought of this thread when I came across this link today :
                            https://www.techpowerup.com/288144/windows-11-performance-issues-on-ryzen-fixed-by-updates-from-microsoft-and-amd

                            "
                            Microsoft and AMD on Thursday released software updates that fix the two performance issues affecting AMD Ryzen processors with Windows 11. The two issues were abnormally high L3 cache latency, and a broken "Preferred Cores" system. The companies had assessed that the issues impact performance of Ryzen processors on Windows 11 by as much as 15%."

                            Comment


                            • #74
                              Windows has always used many threads.
                              Thirty-Two bit Windows.
                              Michael Mattias
                              Tal Systems (retired)
                              Port Washington WI USA
                              [email protected]
                              http://www.talsystems.com

                              Comment


                              • #75
                                Gary if your client's use office apps or browsers, you so much better going for an AMD and getting that extra performance. Avoid mechanical drives at all costs and you will get a good 8 year life from the machine. I could not recommend an i3 to any family or friends.

                                Comment


                                • #76
                                  where large ram works great is as Steve Hutchesson mention the use of ram disk
                                  if your programs are installed in ram disk'
                                  every time one shut down puter the ram disk if saved an reloaded auto at startup'
                                  my 2

                                  Comment


                                  • #77
                                    I use Radeon Ramdisk as it seems to be the best I could find and it does not interfere with how the OS works. I used another for a while but it broke the OS update so it had to go. I built my old box 6 years ago with 8 x 8gig 2400 modules but recently one of them failed and crashed the box so I bought 4 x 16 gig 2666 modules to replace the old ones. It just happened that I had another two identical modules and rather than leave them in a draw, I plonked them in as well to get 96 gig of memory.

                                    Now while it may not be normal to install very large amounts of memory in a work station, I work with video regularly and some of the software needs temporary disk space and to reduce disk load time, putting the temporary directories on ramdisk improves the load time while for multiple gigabytes of video, can be substantial. I have used 16 gig sized ramdisks for a long time and it gives you a high speed playpen for testing software.

                                    Now as James mentioned, when you shut the computer down, the content of the ram disk is written to a temporary file and when you next start the computer, the content is written back to the ramdisk which is really handly if you need to have an application that is always fast to load. Now it gives you an unusual effect, a gradient of storage speed for different tasks.

                                    Ramdisk
                                    NVMe drives
                                    SATA drives
                                    HDDs

                                    Fastest is smallest, slowest has largest capacity in gradients of size.

                                    New stuff is coming out that is a lot faster again. My CPUs dictate the memory speed where later ones can use much faster memory again. Some NVMe drives are gen 4 and are faster than the Samsung drives I use which top out at about 3.5 gig/sec. SATA drives top out at about 550 megabytes/sec and most mechanical hard disks aren't much faster than about 200 meg/sec.

                                    Its the server folks who play with really big memory, 768 gig is possible with one CPU and with some of the later server hardware has massive core counts. Server technology is another planet from workstations but occasionally you can use server CPUs in a work station if you can find the board. The X99 socket 2011-3 allows you to use E5 26xxV3 Xeon processors which are now really cheap from China.
                                    hutch at movsd dot com
                                    The MASM Forum

                                    www.masm32.com

                                    Comment


                                    • #78
                                      Steve, is the Radeon Ramdisk licence permanent or renewed yearly?
                                      as the salmon fish is compelled to go back to it's birthplace to spawn, so comes a day when man is compelled to go back to it's source.. GOD

                                      Comment


                                      • #79
                                        Originally posted by Steve Hutchesson View Post
                                        I use Radeon Ramdisk [cut] I work with video regularly
                                        Bottom line, any of the vendors will give you better performance than SATA SSD / spinning rust.

                                        But, when picking a ramdisk vendor, be conscious of how you will use that drive? Each vendor has strengths and weaknesses. If you're going for sequential processing (video), one vendor may be a better choice for that, but not the best choice for random (IOPS). For example, one might be 30% higher in IOPS, but lower in sequential (see comparison).
                                        • If you're just looking to save some time, follow Steve's 'gradient of storage' and pick a vendor with the features you like.
                                        • If you're trying to minimize a particular processing job (my case), look up the synthetic benchmarks, and make sure you benchmark it yourself with your app on your machine. Your CPU / memory type (ddr3/4) / speed all play as significant a factor as the ramdisk software itself.

                                        Comment


                                        • #80
                                          Hi Ray,

                                          I only have a couple of apps that test ramdisk speed, these are the results I get on this current box. Memory is DDR4 2666 clocked in the BIOS at 2999. The ramdisk is set to 8 gig but on my main box which is undergoing maintainance, I normally use 16 gig. I get the impression that directly allocate memory is a lot faster as it does not transfer through disk IO mechanisms. Where I can test it, the results tend to be all over the place.

                                          Code:
                                          -----------------------------------------------------------------------
                                          CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
                                                                    Crystal Dew World : https://crystalmark.info/
                                          -----------------------------------------------------------------------
                                          * MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
                                          * KB = 1000 bytes, KiB = 1024 bytes
                                          
                                             Sequential Read (Q= 32,T= 1) :  3674.906 MB/s
                                            Sequential Write (Q= 32,T= 1) :  4039.311 MB/s
                                            Random Read 4KiB (Q=  8,T= 8) :  1261.757 MB/s [ 308046.1 IOPS]
                                           Random Write 4KiB (Q=  8,T= 8) :  1060.365 MB/s [ 258878.2 IOPS]
                                            Random Read 4KiB (Q= 32,T= 1) :   381.025 MB/s [  93023.7 IOPS]
                                           Random Write 4KiB (Q= 32,T= 1) :   302.405 MB/s [  73829.3 IOPS]
                                            Random Read 4KiB (Q=  1,T= 1) :   390.193 MB/s [  95262.0 IOPS]
                                           Random Write 4KiB (Q=  1,T= 1) :   342.761 MB/s [  83681.9 IOPS]
                                          
                                            Test : 1024 MiB [A: 0.4% (30.3/7997.0 MiB)] (x5)  [Interval=5 sec]
                                            Date : 2021/10/27 9:27:11
                                              OS : Windows 10 Professional [10.0 Build 19042] (x64)
                                          
                                          -----------------------------------------------------------------------
                                          
                                          HDSPEED
                                          [Test Stop 27/10/2021 9:32:31 AM] errors: 0 average: 5.7 GB
                                          hutch at movsd dot com
                                          The MASM Forum

                                          www.masm32.com

                                          Comment

                                          Working...
                                          X