Skip to main content

Workstation recommendation please

Comments

112 comments

  • Christian Gruner
    [quote="BeO" wrote:
    NVIDIA QUADRO FX 880M, 1 GB, C1 OpenCL benchmark around 3.0


    I think you will see that GPU like the AMD R9 290x (benchmark around 0.08), have a lot of punch. It basically opens a whole new world, even on a very slow CPU.

    [quote="BeO" wrote:
    - If gfx card is supported (VRAM and benchmark) and setting is set to AUTO, then the GPUs are used by some OPENCL code, right?


    OpenCL code is only executed on the GPU('s). If the card is supported, and the setting is on auto, GPU will be used for all GPU-supported operations in CO (i.e. Fuji Xtrans images are not supported, and thus will be worked on by the CPU).

    [quote="BeO" wrote:
    - if harware acceleration is set to NEVER, then (instead of the GPU) the CPU is used, right?
    Which code, OPENCL or another language?


    Correct. Language used is C++, so this can be used on multiple OS'.

    [quote="BeO" wrote:
    - Is OPENCL code executed on the CPU at all?


    No, only on the GPU. OpenCL is simply not mature enough, so it would be too risky to put "all money on 1 horse".

    [quote="BeO" wrote:
    If external benchmarks represent C1 performance sufficiently, what's the point in having an internal benchmark.


    Internally, CO uses the benchmark and RAM-amount to enable or disable features, codepaths and so on. The benchmark has never been developed for the end-user specifically, but as an internal tool for CO.

    [quote="BeO" wrote:
    What should I take as a decision criterion e.g. Mac vs. PC, or Xeon vs. Intel Core, if performance matters, and if you are saying that Xeons are not great C1 performers?


    Look at the benchmark page I posted a few posts back. As mentioned previously, the numbers there are sufficient info for you to choose a CPU or GPU from.

    [quote="BeO" wrote:
    Imho, it really is time for Phase One to work on and publish decent hardware / OS recommendations especially regarding performance.


    As posted in other threads and also in this thread, is fairly easy to summerize:
    CPU:
    Use the benchmark-page posted earlier. Use the "Video composition" benchmarks. Those resemble CO file processing the most.
    Generally speaking, buy the newest generation processors available, and with those,the more cores, the better.

    GPU:
    Get a card with as many Cuda cores (nvidia) / stream processors (amd) as possible.
    Go for as high mem bandwidth as possible.
    Be sure to have 2 gb of vram or more.
    Generally AMD cards perform OpenCL operations better than Nvidia per money spent.
    Also use benchmark page posted, same benchmark option as with CPU.

    RAM
    As much as possible, and as fast as possible.

    Disk
    Preferably SSD disks with high rating on sustained read and write.
    0
  • NNN635397352773288520
    The quadro is because of the 10 bit option and its has a low working temperatur and that the drivers are tested for our kind of work, but yes it cost alot compared to what it can. ๐Ÿ˜„
    0
  • Christian Gruner
    [quote="NNN635397352773288520" wrote:
    The quadro is because of the 10 bit option and its has a low working temperatur and that the drivers are tested for our kind of work, but yes it cost alot compared to what it can. ๐Ÿ˜„


    For Capture One use, 10 bit support is a waste of money.
    Drivers are only verified/certified for CAD work, not for graphical work. This is basically what you pay for in a card like this.

    The low working temperature is a result of very limited computational power. And if you do go up in the cards that seem to be performing "ok" like Quadro M5000 or FirePro W8000, they basically have the same formfactor and noisy fans as larger gaming cards.
    0
  • BeO
    Top Commenter
    Thanks Christian,

    very structured answers and straight to the point.

    Is this the filter you would suggest for the CPU?
    https://compubench.com/result.jsp?bench ... ase=device

    Cheers
    BeO
    0
  • Christian Gruner
    [quote="BeO" wrote:
    Thanks Christian,

    very structured answers and straight to the point.

    Is this the filter you would suggest for the CPU?
    https://compubench.com/result.jsp?bench ... ase=device

    Cheers
    BeO


    Yep!
    0
  • BeO
    Top Commenter
    Ok, thank you ๐Ÿ˜Š
    0
  • Alain Decamps
    [quote="Christian Gruner" wrote:
    [quote="BeO" wrote:
    Thanks Christian,

    very structured answers and straight to the point.

    Is this the filter you would suggest for the CPU?
    https://compubench.com/result.jsp?bench ... ase=device

    Cheers
    BeO


    Yep!


    The benchmark gives the i7 5930K 40% faster that the i7 5820k, which is very unlikely. The differences are, as far as I know, a 6% clock speed increase and 40 PCI-e 3 lanes versus 20. Anandtech cinebench 15 multithreaded also gives a difference of about 6%.
    0
  • Alain Decamps
    [quote="SFA" wrote:
    ...

    The simpler approach is to make an assumption that product development for Photo editors (as with games) will seek to make use of as much computing resource as is made available and so the more resource you have the better the performance is likely to be. Of course there may always be some weak link in an otherwise powerful system but the general principle is "buy the best you can afford".

    In summary, the question of benchmarking has so many variables, including the content that you are intending to process during the years you expect to use a new machine, that any question about recommendations is similar to asking "How long is a piece of string." That is a question I would not wish to answer with a definitive statement.

    Grant

    Grant

    I agree with you're general opinion about benchmarks, but. Most of the time the budget is not unlimited and choices have to be made. Some generic effect could be given, without being to specific.

    I'l given some generic questions :

    For example I've read (from you about co7) that hyper threading makes a really noticeable difference, but Christian has written that it makes almost no difference. I don't know it as an i4 2500k user (btw. I'm very interested what the effect is while doing large soft brushes with local adjustment layers), but it does make a difference what CPU to choose. A i7 6700K or i5 (both 4 cores) or go for a i7 5820k/ i7 5930k with 6 cores.

    Does the RAM interface matter 2-channels (i7 6700k) or 4 channels (i7 5820k), RAM speed?

    Does the interface to the GPU matter, for example: PCIe X8 3.0 or PCIe X16 3.0 (could be one "fast" card or two "slower" cards)?

    Placing the active set of RAW-files on a SSD is a no-brainer, but does a very fast PCIe SSD matter?

    BTW. Except for drawing soft local adjustment brushes I don't have a speed problem with CO9 on my old and not top the line i5-2500K. I can use extra memory,
    0
  • Speedy_xx
    Checking ImgCore log today: does "OpenCL benchMark : 2.204896" means anything in reality? Is it comparible between different workstations?
    0
  • BeO
    Top Commenter
    - this benchmark figure is said to be an P1 internal figure for an important step of the image processing pipeline (doo't know which one)
    - not necessarly representative for the whole image processing process
    - C1 uses opencl only with your graphics card, so it is a measure for a graphics card only
    - if you have multiple grapics cards, each one has its own figure in the log
    - the lower the better.
    - 2.2 is slow.
    - the figure is somewhat linear, e.g. 1 is somehow twice as good as 2.
    - good figures are way below 1 or even below 0.25, 0.1 is very good (scanning forum's posts)

    cheers
    BeO
    0
  • Robert Whetton
    [quote="Christian Gruner" wrote:
    [quote="NNN635397352773288520" wrote:

    Buy an AMD gamer-card instead. Basically go for as many Stream Processors and the highest mem bandwidth available within your budget. It is that easy to choose ๐Ÿ˜‰

    R9 390 running on a 5 year old i5 2500K
    2016-01-13 15:27:50.980> OpenCL benchMark : 0.066074
    0
  • BeO
    Top Commenter
    Christian,
    I am about to select a graphics card, currently in favour of ASUS STRIX Radeon R7 370. However the R7 370 is not listed as supporteed in the article 1720 for C1 v8.

    Is there a specific reason other than P1 did not test it? Are there dependencies what C1 supports like e.g. the graphics chip (Bonaire, Hawaii etc.) or do you think I have a good chance that this card will actually run...?

    Thanks,
    BeO
    0
  • Christian Gruner
    [quote="BeO" wrote:
    Christian,
    I am about to select a graphics card, currently in favour of ASUS STRIX Radeon R7 370. However the R7 370 is not listed as supporteed in the article 1720 for C1 v8.

    Is there a specific reason other than P1 did not test it? Are there dependencies what C1 supports like e.g. the graphics chip (Bonaire, Hawaii etc.) or do you think I have a good chance that this card will actually run...?

    Thanks,
    BeO


    I can't see a reason why it would not run ๐Ÿ˜Š
    0
  • BeO
    Top Commenter
    Thank you!
    0
  • Alain Decamps
    [quote="Christian Gruner" wrote:
    ...
    RAM
    As much as possible, and as fast as possible.

    ...


    Hi instead of replacing my co station, a quite a lot of work and money, for about 50-100% speed increase, I think about upgrading the RAM. I currezntly only have 8GB and it's very clear to me that moving to 16GB will allow longer working.

    I currently have a i5-2500k with ddr3 1333 cl9 and either just could increase the size or replace it with ddr3 1833 cl10, which would give about 40% more -theoretical- bandwidth. The price difference is minimal, it's only more testing ๐Ÿ˜‰

    Would the speed increase of the RAM inside the PC be noticeable? I read that the sandy bridge memory controller is quite good, on the other hand seems co one of the few applications that uses lot's of memory bandwidth.

    The GPU has a score of about 0.145.

    For me I'm only looking of the speed while brushing local adjustments. It's not optimal even when there's still quite some RAM free in the OS.

    Alain
    0
  • BeO
    Top Commenter
    The internal gfx card ntelร‚ยฎ HD Graphics 3000 has 0.145? Not a bad figure I would say, so I assume you have an additional discrete card too? If that's the case, can you disable the slower GPU?
    I've read here in the forum that for two GpuS with different benchmark figures the slower GPU would slow down the faster one...

    RAM: I doubt you will get 40% of brushing performance. However, if the price difference is minimal I would test the faster RAM, you can slow down the MHZ in case of troubles, can you?

    EDIT: Anyway, 16GB is better than 8GB ๐Ÿ˜„
    0
  • Alain Decamps
    [quote="BeO" wrote:
    The internal gfx card ntelร‚ยฎ HD Graphics 3000 has 0.145? Not a bad figure I would say, so I assume you have an additional discrete card too? If that's the case, can you disable the slower GPU?
    I've read here in the forum that for two GpuS with different benchmark figures the slower GPU would slow down the faster one...

    RAM: I doubt you will get 40% of brushing performance. However, if the price difference is minimal I would test the faster RAM, you can slow down the MHZ in case of troubles, can you?

    EDIT: Anyway, 16GB is better than 8GB ๐Ÿ˜„


    The 0.145 is from a discrete AMD 7870 card.
    0
  • BeO
    Top Commenter
    [quote="Alain" wrote:


    The 0.145 is from a discrete AMD 7870 card.


    That's what I thought. Have you disabled the internal GPU?
    Edit: It's maybe not used by C1 anyway.
    0
  • Alain Decamps
    [quote="BeO" wrote:
    ...
    RAM: I doubt you will get 40% of brushing performance. However, if the price difference is minimal I would test the faster RAM, you can slow down the MHZ in case of troubles, can you?

    EDIT: Anyway, 16GB is better than 8GB ๐Ÿ˜„


    I also doubt that it would give 40% extra performance myself ๐Ÿ˜‰

    I'm only considering the extra memory because I'm hesitant for upgrading my computer, microsoft has clearly stated that win7 support on intel skylake will only be for a short while. Given there very aggressive upgrade to win10 strategy, I'm sure they will "destroy" win7 on skylake after that very short period. Unfortunaly CO doesn't run on linux.

    When I place extra memory I have to test it, whether I'm adding the same type to get to 16GB or placing new fast 16GB memory. That "cost" is more important to me than the price difference.
    0
  • BeO
    Top Commenter
    So, what's the alternative...
    RAM, gfx card, new CPU, new board + CPU (haswell?), MAC...?

    I just ordered a new complete computer i7-5xxxx with win 7.
    I don't like win 10 either ("big data" mentality of MS) but I am afraid sooner or later we won't have a choice if we stick to Windows...

    I assume a speedy brush is best achieved by a faster gfx card. I have ordered a similar specced (as yours) but Geforce (so probably slower as yours), which should give me better performance as I have right now; I'll wait for the next generation gfx cards to upgrade, if necessary.

    cheers
    BeO
    0
  • Christian Gruner
    [quote="Alain" wrote:
    Would the speed increase of the RAM inside the PC be noticeable?


    I don't think you will feel this at all, but you might be able to feel 8 vs 16 gb, if you run multiple mem-dependant app, like CO, PS, Premiere and so on.
    0
  • BeO
    Top Commenter
    I have 8 GB currently and use layers. C1 slows down significantly when total RAM usage goes up to 7.9 GB which is often the case with several layers. Besides C1 I have only FireFox open on a regular basis, PS sometimes.
    cheers
    0
  • Alain Decamps
    [quote="BeO" wrote:
    I have 8 GB currently and use layers. C1 slows down significantly when total RAM usage goes up to 7.9 GB which is often the case with several layers. Besides C1 I have only FireFox open on a regular basis, PS sometimes.
    cheers


    I don't feel a slowdown when editing 1 image, but when editing a nr. of images I see that CO uses more and more RAM.
    0
  • Alain Decamps
    [quote="Christian Gruner" wrote:
    [quote="Alain" wrote:
    Would the speed increase of the RAM inside the PC be noticeable?


    I don't think you will feel this at all, but you might be able to feel 8 vs 16 gb, if you run multiple mem-dependant app, like CO, PS, Premiere and so on.


    Thanks.
    0
  • Alain Decamps
    [quote="BeO" wrote:
    So, what's the alternative...
    RAM, gfx card, new CPU, new board + CPU (haswell?), MAC...?

    I just ordered a new complete computer i7-5xxxx with win 7.
    I don't like win 10 either ("big data" mentality of MS) but I am afraid sooner or later we won't have a choice if we stick to Windows...

    I assume a speedy brush is best achieved by a faster gfx card. I have ordered a similar specced (as yours) but Geforce (so probably slower as yours), which should give me better performance as I have right now; I'll wait for the next generation gfx cards to upgrade, if necessary.

    cheers
    BeO


    Maybe I still go to a i7-5820k with a recent x99 board. This januari there are some nice MB's announced (for example the Gigabyte x99 SLI P. -the P is important, new version), those support a fast m.2 PCIE 4x SSD.
    0
  • BeO
    Top Commenter
    i7-5820K
    MSI X99A SLI PLUS, incl.USB 3.1
    M.2 (PCIe) SSD (samsung 950 pro, 512GB)
    Sata3 SSD (samsung 850 evo, 1 TB)
    32 GB RAM (2666)
    Geforce 960
    Not yet arrived...

    cheers
    BeO
    0
  • SFA
    [quote="BeO" wrote:
    i7-5820K
    MSI X99A SLI PLUS, incl.USB 3.1
    M.2 (PCIe) SSD (samsung 950 pro, 512GB)
    Sata3 SSD (samsung 850 evo, 1 TB)
    32 GB RAM (2666)
    Geforce 960
    Not yet arrived...

    cheers
    BeO


    Hmm.

    I will be very interested in your comments when you have it up and running.


    Grant
    0
  • Wesley
    [quote="BeO" wrote:

    I don't like win 10 either ("big data" mentality of MS) but I am afraid sooner or later we won't have a choice if we stick to Windows...


    I built a custom PC with Win 10 for my brother.
    You can turn off all that send data to Microsoft so not a big deal.
    This is coming from a person that haven't used Windows since XP on my old custom rig (Mac user now).
    So I did a fair amount of research.
    I'm thinking the reviews I saw before buying it were people that just clicked on the quick default installation instead of the custom button.

    A legit annoyance would be the forced OS updates (can be delayed though).
    0
  • BeO
    Top Commenter
    Yes Grant, I intend to report.

    [quote="Wesley" wrote:

    You can turn off all that send data to Microsoft so not a big deal.

    Hm, I wouldn't be too sure about that.




    Also, many of the MS updates are no longer documented according to this article:


    and
    Windows 10 November Update Resets Settings to Re-Enable Data Collection


    However, privacy in WIndows 7 and 8.1 might also be "compromised" by certain updates:


    Where are we heading to, if I can't even trust those who I pay for...
    0
  • Alain Decamps
    [quote="BeO" wrote:
    i7-5820K
    MSI X99A SLI PLUS, incl.USB 3.1
    M.2 (PCIe) SSD (samsung 950 pro, 512GB)
    Sata3 SSD (samsung 850 evo, 1 TB)
    32 GB RAM (2666)
    Geforce 960
    Not yet arrived...

    cheers
    BeO


    That msi MB has the right specs for a rather low price. Very interesting.
    0

Please sign in to leave a comment.