Many cores Vs High clock level

Discussion in 'hardware' started by pb1, Sep 21, 2022.

  1. imdb

    imdb Registered Member

    Joined:
    Nov 2, 2011
    Posts:
    4,208
    totally agree with @xxJackxx and @Bill_Bright .
    just buy the fastest ssd you can afford and a compatible mobo.
     
  2. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    I guess the question that should have been asked first... are you ONLY doing restores with this machine? If so throw the money at the fastest drive and worry about nothing else. If you're going to do ANYTHING else go i5.
     
  3. Brian K

    Brian K Imaging Specialist

    Joined:
    Jan 28, 2005
    Posts:
    12,113
    Location:
    NSW, Australia
    pb1,

    The fastest restores are with NVMe to NVMe SSDs. HD to HD restores are much slower. CPUs seem less important.
    My main computer has an i7 CPU and an OS restore takes less than 20 seconds.
    My test computer has an i3 CPU and an OS restore takes less than 30 seconds. There is less data in the test OS. About two thirds of the main computer.

    The test computer is about 10 years old and has EVO instead of Pro Samsung NVMe SSDs. It has less RAM than my main computer.

    Edit... I replaced the MB, RAM and CPU a few years ago so it's not really a 10 year old computer.
     
    Last edited: Sep 21, 2022
  4. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,943
    Location:
    The Pond - USA
    Some observations... large cache is most beneficial to multi-threaded applications but will help all processing attempts if code is written correctly (as in-line as possible). Restore operations for both IFW and REFLECT never use more that 1-core when restoring to internal or USB3-based disks and 2-cores when imaging to an internal disk (1-core to USB3-based disks). Since both of the above imaging apps have a DEFAULT compression algorithm, CPU speed will definitely help that more than multi-core operation. Of course the compression part of existing codes can be greatly improved using multi-threaded processes, but so far, neither Macrium nor Terabyte have thought to do so (maybe futures to speed up the products).

    A lot of the above observations made by many of the posters are very true, especially the no-compression operations, but since you seem to really want a snapshot replacement operation (non-Rollback RX), the most important decision will be the imaging app followed by your disk speeds. To my knowledge, there are only (2) imaging apps on the market that can do what are called "difference restores"... basically restoring only the differences between what's already on the disk and your last known "snapshot" image. Those apps are "Drive Image Backup & Restore" (old IFW) and "Macrium Reflect." All others seem to restore entire images from the "snapshot" reference.

    The only way I've seen to even come close to "difference restores" using SATA SSDs is an NVMe to NVMe environment using PCIe v3/v4 connection environments... this comes very close to Rollback RX snapshot speeds. Unless you're worried about future applications requirements, a high-speed (maybe even overclocked), large cached quad core CPU with 16gB of RAM should be more than sufficient for your new "snapshot" machine.

    My own personal desktop System uses all SATA SSDs for DATA purposes and a PCI-e v4-based NVMe running about 3.5gB/s. I do a lot of snapshot restores and the speed experience is just slightly slower than Rollback RX (which I used for years until difference restoration appeared in Macrium Reflect in 2015)... more than adequate for my needs.
     
    Last edited: Sep 21, 2022
  5. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Thank you kindly Sir, the above facts cleared up a lot in my mind, things that was unclear before and was the reason for my posting. I am using both MR and IFW but are gonig to make IFW the main program for the new Pc. The intention is to get insanely speedy restores with the combo of IFW and a Pc that is fast in the "right way" - for just that.
     
  6. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    I suppose, besides using 1 core, it only uses base frequency, NOTHING else. No Turbo.
     
  7. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    I hate to run this in circles but turbo should be utilized for any amount of cores if the workload is sufficient and the temperature is acceptable. It doesn't need to max out all cores for turbo to work. If you have access to a machine with turbo and watch Task Manager on the performance tab you will see the speed of the CPU fluctuate greatly with any kind of use.
     
  8. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,943
    Location:
    The Pond - USA
  9. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Well, that both answers the question, and not, since it is not detailed to the/a, specific situation, just a sweeping generalization.

    Then it becomes a question of - How much is sufficient workload? A simple restore, that is done in an even pace with no extra loads suddenly poping up, probably will not use any Turbo since the strain is not that high. In difference to do work with the Pc and on and off starting and ending a lot of actions. That is my assumption of how it probably works. If that is correct, then theTurbo will not be used in restore mode instead the Pc will work from Base frequency.

    Right or wrong?!
    Evaluate please to a decisive answer.
     
  10. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    Any. You may want to refer back the the Intel article I posted previously. The CPU will throttle back and forth between the minimum and maximum clock speed as needed, and will probably run somewhere in between most of the time. It certainly does not favor the minimum speed all of the time.
     
  11. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Ok. Thanks.

    As mentioned earlier in the thread, cache matters, the L3 cache is 2MB larger in the i5, the others are also larger, does that matter much, or little?

    My thinking is; does the higher Turbo and larger cache, compensate for the lower base frequency, 2.4ghz in the i5, than in the i3, which is clocked to 3.0.

    How would you evaluate that?
     
  12. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    How much the cache matters depends on what you are doing. For backups/restorations probably not much. The i5 will rarely be running at minimum speed. Ultimately when needed it will throttle itself to faster clock speed than the i3. The i5 does not lose anywhere but price.
     
  13. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Sure? 2.4 vs 3.0 is... a bit. I am thinking that the i5 therefore will be a tad slower when restoring, which will oppose my purpose.
     
  14. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    I disagree but it is difficult to explain without you having an example to look at. The best I can do is this... I have an i9-10980hk in my laptop. The base clock is 2.4 GHz. I just opened Task Manager and went to the performance tab. It shows the real-time clock-speed. When I first opened it the CPU was running at 4.26 GHz. It can max at 5.3 GHz. After I let it sit for a minute it slowed to 1.69 GHz. In between it fluctuated to multiple speeds in between. You probably have to see this in person to "get" how it works. But again, the i3 is probably never going to outperform the i5. In any case both would get the job done. It is all a case of how much difference it will make and is it worth the difference in price. I don't think anyone can answer that without testing it.
     
  15. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Well, several has said the same as you now did. But fact is, in test of the 2 processors on the net, this i3 outperforms the i5 in single core, so... . And the difference between the 2 is only 0.2ghz in Turbo mode. I think that for restore purposes i will not win anything with the new Pc, otherwise, much. It is better, but not totally. But testing is of course the only decisive way, but that is hard to come by. Hence the thread.

    Anyway, thank you for your time and attention, appreciated. I have become wiser due to the info in this thread.
     
  16. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,943
    Location:
    The Pond - USA
    In the case of Macrium REFLECT... seems the future may be here now. v8.1 (currently in BETA) looks like it uses all available cores for its basic CPU required tasks (configurable). An example shown by the Developers demonstrates an increase in FULL imaging speed of about 4X (no idea of disk configuration for this example).

    Best to stay tuned for additional interesting changes in this app! The above new pipelined parallel processing implementation will be available shortly for restorations as well.
     
  17. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,943
    Location:
    The Pond - USA
    A colleague of mine who has access to Macrium's v8.1 REFLECT Beta ran a quick test for me. The friend imaged a 45gB "used" DATA image from a 3.2gB/s NVMe SSD to a 500+mB/s SATA SSD and the difference from Macrium v8.0 to v8.1 was a little over 3X in speed... FYI!
     
  18. imdb

    imdb Registered Member

    Joined:
    Nov 2, 2011
    Posts:
    4,208
    wow, that's blazing speed. very impressive. thanks for sharing, trf. looking forward to its stable release. :thumb:
     
  19. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Ha, wow i say! That is thrilling news.
    Thanks for that info.
     
  20. imdb

    imdb Registered Member

    Joined:
    Nov 2, 2011
    Posts:
    4,208
    so what did you do? buy a new rig?
     
  21. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    Nice. I just received my new 2TB external SSD for backups today. I'm looking forward to much faster backups.
     
  22. imdb

    imdb Registered Member

    Joined:
    Nov 2, 2011
    Posts:
    4,208
    enjoy it. :thumb:
     
  23. pb1

    pb1 Registered Member

    Joined:
    Apr 4, 2014
    Posts:
    1,268
    Location:
    sweden
    Not yet, i was thinking in December around Black friday or Christmas but now, with this info, the need for a costly one is gone. So this was interesting news in many respects. So even if it will be Macrium v.9.0, this event of mine will be cheapier than it was planed.
     
  24. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,623
    Location:
    USA
    Thanks.
    Not wanting to hijack this discussion but for anyone that was curious I just did a fresh image from a disk with 232 GB of data. 12 minutes. The old external HDD would have taken more than an hour and a half. This thing is now 100% SSD internal and external. Well worth the upgrade.
     
  25. imdb

    imdb Registered Member

    Joined:
    Nov 2, 2011
    Posts:
    4,208
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.