i7 or Ryzen?
With the recent AMD Ryzen developments I am wondering what would make a better COP workstation: top of the bill i7700k or a Ryzen 1700X? These are similarly priced. The Intel has a better single core performance, but the AMD obviously has a better multicore performance (having the double amount of cores). So what would be a better CPU?
0
-
Well, if you're not going to use a GPU to process your images, I'd be interested in knowing the answer too, as my 2500K is getting long in the tooth and I'm looking at upgrading near the end of the year..
As C1 is multi threaded, I'd say that Ryzen should fair well against the 7700K, might be worth contacting someone like jayztwocents and asking if he'd do some tests?0 -
I just put together a system for still and video based around the i6800k.
It has two very fast SSDs and 64Gb RAM and a 1070 card.
Without using a stopwatch, I reckon it's 8-10 times faster than my old i5 system.
D.0 -
I'm interested in the difference, but mostly on the effect when doing layers (brushing) and with a decent GPU. 0 -
[quote="Dinarius" wrote:
I just put together a system for still and video based around the i6800k.
It has two very fast SSDs and 64Gb RAM and a 1070 card.
Without using a stopwatch, I reckon it's 8-10 times faster than my old i5 system.
D.
What was the setup of you're I5? 8-10 times is quite a lot, especially because a lot is done with the GPU.0 -
For me the biggest speed saving is in outputting files.
Canon files (5D - 175Mb/16bit) save in about 3-4 seconds.
It was 20-30 on the old system.
In terms of rendering within, say, Photoshop, saving of changes to files (e.g. a lot of retouching) is virtually instantaneous.
I have not yet compared it with multi-layered files.
Suffice to say, I'll have a lot more time for other things now! 😊
D.0 -
[quote="Dinarius" wrote:
1070 card.
My R9 390 is actually faster than a 1070.. 😉 I wasn't actually talking about file output but masking etc.. 😊
Having watched a few content creation YouTube vid showdown's, think I will be heading AMD by the end of the year! 👿0 -
When you look at the performance as e.g. given on http://www.videocardbenchmark.net/ you see that the R9 390 is slower than the 1070 (by a margin of 20-25%). Any clues on why you experience another behavior? Has this been discussed before? 0 -
Purchased Ryzen 7 1700 16gb PC 3200 RAM Radeon 380x M.2 SSD two days ago.
C1 worked very fast and uses all cores.
Processing of 36mp files are lighting fast (via GPU).
So far i cannot imagine an operation to slow-down C1 on this setup (but redraw after 100% zoom takes a fraction of second i.e. still visible - maybe it can be eliminated by bigger Preview Size).
For AMD/ATi vs nVidia GPU my observation says AMD supports OpenCL better - nVidia rely mostly on their proprietary CUDA.
C1 uses OpenCL.0 -
[quote="NN174596UL" wrote:
Purchased Ryzen 7 1700 16gb PC 3200 RAM Radeon 380x M.2 SSD two days ago.
C1 worked very fast and uses all cores.
Processing of 36mp files are lighting fast (via GPU).
So far i cannot imagine an operation to slow-down C1 on this setup (but redraw after 100% zoom takes a fraction of second i.e. still visible - maybe it can be eliminated by bigger Preview Size).
For AMD/ATi vs nVidia GPU my observation says AMD supports OpenCL better - nVidia rely mostly on their proprietary CUDA.
C1 uses OpenCL.
What's you're experience when drawing masks with already multiple layers with local adjustments?
Are all threads used in this scenario and is it very responsive?
(I have the impression that in this scenario the GPU is only used a little bit and the burden falls on the CPU.)0 -
С1 uses all cores and threads when adjusting according to Windows Task Manager.
For processing it uses GPU extensively.0 -
The performance improvement (175Mb 5D MkIV files) in C1 on my new machine has been astonishing; the improvement in Hasselblad Phocus (files not a whole lot bigger at 220Mb) much less so.
Not sure if it's something that Phocus doesn't use, which C1 does. Or, whether it's something I haven't changed/selected in Preferences or elsewhere.
D.0 -
[quote="NN174596UL" wrote:
С1 uses all cores and threads when adjusting according to Windows Task Manager.
For processing it uses GPU extensively.
Thanks for the test.0 -
[quote="Dinarius" wrote:
The performance improvement (175Mb 5D MkIV files) in C1 on my new machine has been astonishing; the improvement in Hasselblad Phocus (files not a whole lot bigger at 220Mb) much less so.
Not sure if it's something that Phocus doesn't use, which C1 does. Or, whether it's something I haven't changed/selected in Preferences or elsewhere.
D.
This could be the use of a far better GPU by C1. Exporting files (processing them to tif (without compression) or jpeg) uses the GPU a lot. You can test this in C1 by turning of the GPU in preferences and time the difference. Don't use a big batch for it 😉0 -
Suddenly, C1 loses OpenCL for GPU (it still Auto in Preferences) and ICOC.xml in ProgramData folder says it's ok
<device>
<Index>0</Index>
<Platform>AMD Accelerated Parallel Processing</Platform>
<Version>OpenCL 2.0 AMD-APP (2117.14)</Version>
<Device>Tonga</Device>
<Driver>2117.14 (VM)</Driver>
<GlobalMemoryMB>4096</GlobalMemoryMB>
<NumberOfKernels>1111</NumberOfKernels>
<NumberOfCompiledKernels>1111</NumberOfCompiledKernels>
<Benchmark>0.130908</Benchmark>
<Status>Benchmarked:CL_SUCCESS</Status>
</device>
BUT it's not working - redrawing it's laggy, processing tooks dozen of seconds and GPU usage monitor of the MSI Afterburner software says it's not uses much.
Nothing changed in the system.0 -
[quote="Dinarius"] wrote:
For me the biggest speed saving is in outputting files.
Canon files (5D - 175Mb/16bit) save in about 3-4 seconds.
It was 20-30 on the old system.
In terms of rendering within, say, Photoshop, saving of changes to files (e.g. a lot of retouching) is virtually instantaneous.
I have not yet compared it with multi-layered files.
Suffice to say, I'll have a lot more time for other things now! 😊
D.[/quote]
I have i7 920 2,66ghz (4 cire/8 threads 8 years old) and two r9 280x.
Outputting nef d800 to tif 16bit uncumpressed(206 Mb) takes less than two seconds each file
I have a little lag when zooming 100% because I think it's made by the processor0 -
[quote="Alain" wrote:
[quote="NN174596UL" wrote:
С1 uses all cores and threads when adjusting according to Windows Task Manager.
For processing it uses GPU extensively.
Thanks for the test.
In my setup Cpu uses all threads almost to 100% and little gpu (17-20% on average).
I have little-known almost no lag in the pencil strokes with 3 layers. Adjusting exposure of layers with no lag0 -
[quote="NN174596UL" wrote:
Suddenly, C1 loses OpenCL for GPU (it still Auto in Preferences) and ICOC.xml in ProgramData folder says it's ok
<device>
<Index>0</Index>
<Platform>AMD Accelerated Parallel Processing</Platform>
<Version>OpenCL 2.0 AMD-APP (2117.14)</Version>
<Device>Tonga</Device>
<Driver>2117.14 (VM)</Driver>
<GlobalMemoryMB>4096</GlobalMemoryMB>
<NumberOfKernels>1111</NumberOfKernels>
<NumberOfCompiledKernels>1111</NumberOfCompiledKernels>
<Benchmark>0.130908</Benchmark>
<Status>Benchmarked:CL_SUCCESS</Status>
</device>
BUT it's not working - redrawing it's laggy, processing tooks dozen of seconds and GPU usage monitor of the MSI Afterburner software says it's not uses much.
Nothing changed in the system.
I just bumped into this thread searching for Ryzen/Capture One Pro compatibility. I'm moving on from the Mac - getting sick of waiting for Apple to release new desktop machines, and still there's no guarantee they're going to continue supporting pro users even after this next round.
anyway, I've been hemming and hawing over a Ryzen 1700 overclocked to 3.9GHz (if I can get there), or an Intel 5820K safely overclocked to 4.5GHz, pretty confident I can get there based on a lot of reading.
Have you figured out if you've run into a GPU/C1P problem, or a CPU/system/C1P problem?
Thanks,
c0 -
[quote="NN174596UL" wrote:
Suddenly, C1 loses OpenCL for GPU (it still Auto in Preferences) and ICOC.xml in ProgramData folder says it's ok
<device>
<Index>0</Index>
<Platform>AMD Accelerated Parallel Processing</Platform>
<Version>OpenCL 2.0 AMD-APP (2117.14)</Version>
<Device>Tonga</Device>
<Driver>2117.14 (VM)</Driver>
<GlobalMemoryMB>4096</GlobalMemoryMB>
<NumberOfKernels>1111</NumberOfKernels>
<NumberOfCompiledKernels>1111</NumberOfCompiledKernels>
<Benchmark>0.130908</Benchmark>
<Status>Benchmarked:CL_SUCCESS</Status>
</device>
BUT it's not working - redrawing it's laggy, processing tooks dozen of seconds and GPU usage monitor of the MSI Afterburner software says it's not uses much.
Nothing changed in the system.
For the record, while in 'Preferences/General/Hardware Acceleration (Use OpenCL for)', it does say "Auto" for both Display and Processing; it clearly states underneath this (albeit in greyed-out type) that, "Capture One is using hardware acceleration." At least it says so in my setup.
So, while C1 may not allow the manual activation of Open CL, I can only assume that it will avail of it if your system will work with it.
Apart from lens profile issues, which I have written about elsewhere, C1 has never worked better for me.
D.0 -
Auto IS manual in so far as it can be set.
OpenCL is not a general "use for everything" facility - it's the ability to use another processor, dedicated to graphical type calculations, for some specific processing tasks under certain circumstances.
"Auto", in that context, means that C1 will assess the potential benefits of using your OpenCL capable processor for the task you are setting and use it in some way if it the parameter suggest there likely to be benefits.
"Never" means that this decision has been made by the user and no attempt at hardware acceleration will be made.
Before either of those come into effect C1 checks the system for the availability of OpenCL capable GPUs and assesses the processing capability of any it finds. IF they do not have enough potential to make them worth using they are not considered further.
If it (or they in some cases) do offer some potential benefit they will be used when that benefit can be delivered.
Once co-opted for potential use the possible vagaries of the GPU drivers may come into play ... but that is a different subject EXCEPT in the case when setting "Use Hardware Acceleration" is set to "Never" in order to eliminate possible problems caused by a particular GPU card or its currently installed driver software.
For what it is worth my very low power GPU is assessed as usable (just) and does register some number changes in a GPU Monitor when I start C1.
With V10 in play I see no value changes when editing an image. If processing there may be a small increas ein a couple of values and twitch in the Monitor's histogram early in the process as files are added to the process queue but that is about all. All cores of the CPU are shown as running at 100%. (These observations made with a small test catalogue, smallish raw files. 4 variants selected for output.)
If I run V9.3 the activity levels reported are significantly higher.
I have no way of properly understanding any significance for those results, partly because output processing has changed between V9 and V10 to some extent.
HTH.
Grant0 -
I discontinued building with AMD processors some years ago when I discovered the Adobe Creative Suite would not install on it. After six months or so Adobe released a patch that resolved the issue. I'm sure the new AMD processors are excellent, but I'll pay the extra for Intel. 0 -
I recently tested an i7 6900k (8 cores and 16 threads)3,2ghz (stock) with capture one pro 10 and an ssd samsung 850 evo sata.
Only using the proccessor i exported to tiff 16 bit (aprox 206mb) 30 nefs from d800 in 69 seconds.
Using my i7 920 with 2 r9 280x takes 55s. only cpu without hyperthreading in my computer takes 363 seconds.
i also noticed almost the same lag when zoomig to 100% in a non sequential image selected0 -
Intel have announced a new family of processors that start with if and are topped out by i9s.
One more thing to add to the list of considerations.0 -
I am working with C1 10 Pro + Ryzen 1700 + Radeon R9 Nano every day. I never have had such a silent, fast, effizient system. During my work on Images with C1, I have a lot of programs running (like IMatch 2017, X1-jsearch, Boinc), which all use CPU-Ressources. I never have to wait for C1 because it is busy. Even if I put styles on plenty of images.
My recommendation would be the cheap Ryzen 1700 = 8C/16T with 65 W TDP. More Considerations will come: Threadripper with up to 16 C/32 T.0 -
So far OpenCL problems is driver problem too, after new GPU driver update is better to delete this OpenCL settings in Program Data folder and let C1 compile it from scratch.
Now issue is gone and my system worked as intended (Ryzen 1700 16gb DDR4 2933mhz, Radeon 380x, Win10 x64) - fast and efficient0 -
[quote="SFA" wrote:
Intel have announced a new family of processors that start with if and are topped out by i9s.
One more thing to add to the list of considerations.
Intel panicked with the release of Threadripper, KaybelakeX is such a mess... and CoffeeLakeX coming later this year as well .. lets not even go into the $2K CPU pricing!0 -
Hi,
I will go with Threadripper.
As shown today 7900X (10 Core) is a no go.
My CPU 3930K (4.2 ghz, Consumes up to 220 Watts OCed)
I use two W8100 Workstation GPUs <--- Professional Drivers, Consumer Drivers can be Buggy)
My System has the same Speed as a 6700K, 16 GB, RX480:
1102 Images Wedding Export almost 26 min on both. (Up to 4 Layers, 1DX II & 1DX)
RX 480 on 6700K No OC is as fast as Two W8100 on 3930K Oced to 4.2ghz.
But RX480 on 3930K 4.2ghz is as fast as One W8100 on Same CPU, so Conclusion 3930K is the Limiter.0 -
[quote="Bobtographer" wrote:
[quote="SFA" wrote:
Intel have announced a new family of processors that start with if and are topped out by i9s.
One more thing to add to the list of considerations.
Intel panicked with the release of Threadripper, KaybelakeX is such a mess... and CoffeeLakeX coming later this year as well .. lets not even go into the $2K CPU pricing!
Regardless if Intel panicked or not, they announced one helluva an mid-tier line of CPUS with the 7800x, 7820x and 7900x. If this is what it takes to kick Intel in the arse to release some raging CPU's, yay! It's a win for everyone.
I have zero interest in the 18 core beast, so I couldn't care less if it costs $2,000. But $600 for an 28 lane 8-core machine that can easily/safely overclock to 4.8GHz, yikes!
I also have zero interest in the 16 core Threadripper, but I am interested to see how the 10 core Threadripper shakes out. I already priced out numerous builds on pcpartpicker.com, including a 10-core i9 for $2,850.
https://pcpartpicker.com/user/craigjohn/saved/XnVkLk
That's a few hundred less than the price I'd pay for a 27" 5k 4.2GHz iMac with a 512GB SSD and 32GB of ram (upgraded from OWC)...
If the 10 core Threadripper can keep pace (or close to it) with the Intel 7900x, and it costs all of $550, BOOM! Score. Just hoping it's ready to go rather than being in Beta testing mode when it gets released to the public.
Here are some cinebenchR15 numbers I've compiled...
My current computer (2009 Mac Pro Quad Core):
Intel Xeon W3520 @ 2.66GHz - multi core = 432 - single core = 88
Intel 7900X 3.3GHz - multi core = 2165 - $1,000 USD - single core = 190
Intel 7900X 4.0GHz - multi core = 2170 - $1,000 USD
Intel 7900X 4.6GHz - multi core = 2485 - $1,000 USD
Intel 7820X at 3.6GHz - multi core = 1750 - $600 USD - single core = 180
Intel 7820X at 4.8GHz - multi core = 2020 - $600 USD - single core = 210
Intel 7800X at 3.5GHz - multi core = 1300 - $390 USD - single core = 175
Ryzen 1700 at 3.0GHz - multi core = 1400 - $300 USD - single core = 140
Ryzen 1700 at 3.9GHz - multi core = 1770 - $300 USD - single core = 160
Ryzen 1700X at 3.4GHz - multi core = 1510 - $350 USD - single core = 140
Ryzen 1800X at 3.6GHz - multi core = 1625 - $440 USD - single core = 155
Ryzen 1800X at 3.9GHz - multi core = 1705 - $440 USD - single core = 160
**If you note, there is zero reason to buy the 1800X or even the 1700X CPU from Ryzen, as the $300 1700 safely overclocks to the same point as it's more expensive siblings.
Considering I use Photoshop quite a bit after Capture One Pro, having a higher single core score is equally beneficial as multicore, and Ryzen lags a bit behind in single core scores compared to Intel. Ryzen also lags in overclocking, and the new Intel X chips are geared for overclocking.
As of right now, that 7820x is looking fantastic...
This 7820X build is only $2,440:
https://pcpartpicker.com/list/zpNJQV
This 7800X build is only $2,230:
https://pcpartpicker.com/user/craigjohn ... iew=Ntxpgs
For couple of hundred extra bucks, that 8 core is looking sweet.
But still, I'm waiting for Threadripper later this summer before I pull the trigger.0
Post is closed for comments.
Comments
27 comments