v20 with eGPU?
any ideas? just read the digital transitions "mac shopping" recommendations and GPU is listed as most important...
i heard in the past that eGPU with v12 wasnt really a big improvement,i understand that there are a wide range of external cards available and some might work better then others...
thinking of the (expensive but elegant) blackmagic pro solution, does anyone have experience?
i heard in the past that eGPU with v12 wasnt really a big improvement,i understand that there are a wide range of external cards available and some might work better then others...
thinking of the (expensive but elegant) blackmagic pro solution, does anyone have experience?
0
-
Since there are a number of threads noting that C1 20 doesn't seem to be making much use of discrete GPUs in late model MBPs running Catalina and Mojave, I'd be surprised if a super-powerful eGPU is much help either. 0 -
Nature Isme wrote:
Since there are a number of threads noting that C1 20 doesn't seem to be making much use of discrete GPUs in late model MBPs running Catalina and Mojave, I'd be surprised if a super-powerful eGPU is much help either.
I beg to differ - I use v20 in Mojave on Mac Pro 2013 with 2xD700 & 64GB and when I use eGPU (Vega Frontier) C1 does get a boost from the external card when applying edits and rendering previews. Not much, but it's noticeable when I use C1 without eGPU on same catalogues. I have not forced C1 to prefer eGPU, monitor is plugged into external card and this way C1 uses all 3 cards simultaneously.
Sadly, v20 still uses (an outdated version) of OpenCL and not Metal to speed things up, maybe that transition would yield better performance gains and extend more computational tasks to graphic card(s).
Last thing might be cost, my eGPU setup hit ±$800 mark (card + eGPU box), but I got a really good deal on that Mac Pro so it didn't hurt too much and I use it all the time to minimize thermal load on the machine.0 -
Why do you use eGPU on a MacPro ?
Go for the Sapphire Pulse R580 (approved by Apple) in one of the PCIe slots (+/- 200 euros), and you will even get the Boot Screen on Mojave.
Besides, the card is Metal-compatible.0 -
i guess there is a lot of trial and error out there and i will have to just get one and check it out myself....
i am assuming adobe uses metal now? it really is a joke how much faster LR CC is compared to C1...0 -
I honestly do not know about what adobe does now.
I am not a pro, so i can’t say that every minute matters.
As long as the system is reasonably reactive (not as Nikon infamous Capture NX-D...), I can live with it if it works well.
OpenCL or Metal ?
I opted for a third-party card endorsed by Apple simply for my peace of mind.
We will see what the next updates will bring (there’s quite a bunch of work on the table...)0 -
PSS wrote:
i guess there is a lot of trial and error out there and i will have to just get one and check it out myself....
i am assuming adobe uses metal now? it really is a joke how much faster LR CC is compared to C1...
I dont know what Adobe is using for LR but CO20 still uses a dated OpenCL, so it is a joke but can't laugh0 -
fatihayoglu wrote:
PSS wrote:
i guess there is a lot of trial and error out there and i will have to just get one and check it out myself....
i am assuming adobe uses metal now? it really is a joke how much faster LR CC is compared to C1...
I dont know what Adobe is using for LR but CO20 still uses a dated OpenCL, so it is a joke but can't laugh
What difference would a later version of OpenCL make?0 -
I don’t know, maybe updated optimized code for recent hardware rather than 5 yrs old one. Otherwise what’s the need for an updated version, right? 0 -
i'm using CO 20 on a mac Mini with Blackmagic eGPU Pro. It's definitely using it. It's important to have the "prefer external GPU" option on. When adjusting image the usage is around 30-50%. When exporting it's near 100%.
On my Macbook Pro it's only using the AMD GPU if i switch off the system-setting "automatic change of graphic-mode“ (wich enforces the AMD GPU to be used at all times) .
I was in contact with the support about it. I demonstrated in a video the behavior. Since then i haven't heard back from then.
It would be good if they could use Metal2 instead of openCL .. i'm sure the performance would increase a lot.0 -
verstaerker wrote:
i'm using CO 20 on a mac Mini with Blackmagic eGPU Pro. It's definitely using it. It's important to have the "prefer external GPU" option on. When adjusting image the usage is around 30-50%. When exporting it's near 100%.
On my Macbook Pro it's only using the AMD GPU if i switch off the system-setting "automatic change of graphic-mode“ (wich enforces the AMD GPU to be used at all times) .
I was in contact with the support about it. I demonstrated in a video the behavior. Since then i haven't heard back from then.
It would be good if they could use Metal2 instead of openCL .. i'm sure the performance would increase a lot.
I think we were all hoping for switch to metal with v20....oh well, maybe next year...
The mini probably needs the egpu anyway, I am not too concerned with export times, my concern is faster redraw and snappier adjustments....I would expect things to be instant with the egpu vs ok(with a slight lag) with my mbp...
I also do some basic video work which would definitely e easier with the egpu...
having to manually go into system settings sound like a pain...
are you running the egpu into the mini and the monitor from the egpu?0 -
fatihayoglu wrote:
I don’t know, maybe updated optimized code for recent hardware rather than 5 yrs old one. Otherwise what’s the need for an updated version, right?
Well one reason for for a new version of anything is to make some new features available so that more types of process can use the 'standard'.
IF the hardware and all levels of software manufacturers adopt the new standard for new products then people buying new products may find some benefit. How much benefit may be unknown. If one has a lot of coding to create in addition to what you already have, only useful to your client base if they upgrade hardware and all sorts of software, the commercial rationale AND the benefit to customers may be something of a mirage.
IF the performance benefits are always going to be huge - say 200% and saving people hours of their time - then you may be able to convince people it is worth the cost of upgrading.
If it's closer to 10% benefit and mostly of interest to hardware geeks for testing results ... maybe not so exciting.
I doubt the lowly GPU on my Notebook would be able to gain much - probably nothing - from a later version of OpenCL and the code that the developers would need to add to make use of it in some way.
As windows user I think I can be fairly sure that it will never make use of Metal - whatever that is.
For most things on my machine whether or not I am using OpenCL does not make much difference at all - yet it is still perfectly usable.
These new standards can often seem to be little more than a marketing tool to shift more product by persuading people to keep "at the cutting edge."
A similar example might be a proposal to change out SSD drives for NVMe drives for the speed advantage.
The number are huge - NVMe drives I/O is, typically , a lot fast than traditional SSD SATA3 drives.
In practical terms where will that speed advantage make a difference?
Should we all rush out and change to NVMe drives because they are faster?
Can I justify converting my NAS (2x8Terrabyte drives) to NVMe because the existing drives are slow IF I use them as a primary storage source working with C1?
Grant0 -
my startup (and main drive) has been NVMe for a few years and yes it is hard to go back to even SSD....i still use some actual spinning HDs but for all that stuff my data is in the cloud now because it is faster then HD....
i dont know anything about coding but i do know that a leaner os and updated apps make things faster....LR vs LR CC...night and day....pretty much anything on an iPad pro....the app and machine just gets out of the way....
i had a tether demo with the hasselblad app and their X system....the full raw files pop up so fast, its hard to notice any delay once you take your eyes off the camera finder....
in general iOS is not there yet, the hardware is, but i am not sure i would want to work with a touch screen exclusively...
i have used C1 for 15? years now, i definitely appreciate it but at some point the entire thing needs to be re-written or maybe just ported over to an app in iOS? with full integration to desktop?
this is not about bleeding edge, its about staying relevant....
i was able to ditch my mac pro for a mbp a few years ago...looking at the egpu to make that move even more permanent....who says an iPad pro wont replace the mbp in a few years? external monitor, external gpu, storage in cloud, i am all for it....0 -
PSS wrote:
my startup (and main drive) has been NVMe for a few years and yes it is hard to go back to even SSD....i still use some actual spinning HDs but for all that stuff my data is in the cloud now because it is faster then HD....
i dont know anything about coding but i do know that a leaner os and updated apps make things faster....LR vs LR CC...night and day....pretty much anything on an iPad pro....the app and machine just gets out of the way....
i had a tether demo with the hasselblad app and their X system....the full raw files pop up so fast, its hard to notice any delay once you take your eyes off the camera finder....
in general iOS is not there yet, the hardware is, but i am not sure i would want to work with a touch screen exclusively...
i have used C1 for 15? years now, i definitely appreciate it but at some point the entire thing needs to be re-written or maybe just ported over to an app in iOS? with full integration to desktop?
this is not about bleeding edge, its about staying relevant....
i was able to ditch my mac pro for a mbp a few years ago...looking at the egpu to make that move even more permanent....who says an iPad pro wont replace the mbp in a few years? external monitor, external gpu, storage in cloud, i am all for it....
Staying relevant in photographic terms may mean changing everything to top end mobile phone standards and doing stuff with software to simulate "traditional" photographic images until things get to the point where people no longer care whether what they are looking at is a "real image" or cgi.
Once printed media is a distant memory and all marketing is undertaken via social media why would anything else be necessary - or relevant?0 -
Staying relevant in photographic terms may mean changing everything to top end mobile phone standards and doing stuff with software to simulate "traditional" photographic images until things get to the point where people no longer care whether what they are looking at is a "real image" or cgi.
Once printed media is a distant memory and all marketing is undertaken via social media why would anything else be necessary - or relevant?
That seems like a philosophical question...
I could not care less how the image is taken or created, it’s all tools...and the more they get out of the way, the better...
personally I would love to use the smallest, lightest capture device possible and the most portable, inconspicuous machine to organize and manipulate the images...all with the highest possible quality of course...
If that happens to be a pair of glasses and some foldable slab with voice commands (that actually work and seem natural to me)....ok with me....
Until then I will try to make my life as easy as possible and deliver the best IQ in the least amount of time0 -
PSS wrote:
Staying relevant in photographic terms may mean changing everything to top end mobile phone standards and doing stuff with software to simulate "traditional" photographic images until things get to the point where people no longer care whether what they are looking at is a "real image" or cgi.
Once printed media is a distant memory and all marketing is undertaken via social media why would anything else be necessary - or relevant?
That seems like a philosophical question...
I could not care less how the image is taken or created, it’s all tools...and the more they get out of the way, the better...
personally I would love to use the smallest, lightest capture device possible and the most portable, inconspicuous machine to organize and manipulate the images...all with the highest possible quality of course...
If that happens to be a pair of glasses and some foldable slab with voice commands (that actually work and seem natural to me)....ok with me....
Until then I will try to make my life as easy as possible and deliver the best IQ in the least amount of time
That's all fine.
But for what end purpose?0 -
SFA wrote:
fatihayoglu wrote:
I don’t know, maybe updated optimized code for recent hardware rather than 5 yrs old one. Otherwise what’s the need for an updated version, right?
Well one reason for for a new version of anything is to make some new features available so that more types of process can use the 'standard'.
IF the hardware and all levels of software manufacturers adopt the new standard for new products then people buying new products may find some benefit. How much benefit may be unknown. If one has a lot of coding to create in addition to what you already have, only useful to your client base if they upgrade hardware and all sorts of software, the commercial rationale AND the benefit to customers may be something of a mirage.
IF the performance benefits are always going to be huge - say 200% and saving people hours of their time - then you may be able to convince people it is worth the cost of upgrading.
If it's closer to 10% benefit and mostly of interest to hardware geeks for testing results ... maybe not so exciting.
I doubt the lowly GPU on my Notebook would be able to gain much - probably nothing - from a later version of OpenCL and the code that the developers would need to add to make use of it in some way.
As windows user I think I can be fairly sure that it will never make use of Metal - whatever that is.
For most things on my machine whether or not I am using OpenCL does not make much difference at all - yet it is still perfectly usable.
These new standards can often seem to be little more than a marketing tool to shift more product by persuading people to keep "at the cutting edge."
A similar example might be a proposal to change out SSD drives for NVMe drives for the speed advantage.
The number are huge - NVMe drives I/O is, typically , a lot fast than traditional SSD SATA3 drives.
In practical terms where will that speed advantage make a difference?
Should we all rush out and change to NVMe drives because they are faster?
Can I justify converting my NAS (2x8Terrabyte drives) to NVMe because the existing drives are slow IF I use them as a primary storage source working with C1?
Grant
That’s lots of IF’s and blind defending CO for not utilizing current technologies. Since I bought my late 2015 top spec iMac 5K, CO is extremely slow on basic adjustments, that’s about 4 years now. Many people have and still are complaining about the performance of their top spec Macs. Every single time, CO blames Apple because of not putting strong GPU in these machines but we all know that when we bought these machines, they were capable of top class video editing. So, sorry but, if a machine can edit video, for basic photo editing it should fly. Period. No excuse or whatsoever. I’m not talking about DAM or anything. And also, I have realized CO does not utilities GPU for adjustments. Then it is poor or lazy coding, sorry, no excuse there as well. Also Apple has stopped or going to stop with the next release to support OpenCL, that means their code won’t be optimized for OpenCL, one can expect poor performance from applications that demands GPU use. Metal is a framework, written by Apple for Apple HW. We, MAC users, expect all the softwares should support Metal so we work on a optimized system. So it’s not about a company brings something to sell more, it’s a framework. When Java went from version 5 to 6, it wasn’t to sell more products. The code was optimized and cleaned. If you are using a donkey years old code, it contains full of bugs etc. then you can expect bad performance.0 -
PSS wrote:
I think we were all hoping for switch to metal with v20....oh well, maybe next year...
The mini probably needs the egpu anyway, I am not too concerned with export times, my concern is faster redraw and snappier adjustments....I would expect things to be instant with the egpu vs ok(with a slight lag) with my mbp...
I also do some basic video work which would definitely e easier with the egpu...
having to manually go into system settings sound like a pain...
are you running the egpu into the mini and the monitor from the egpu?
The redraw speed when doing adjustments are totally fine when the correct GPU is actually used.
On my MBP this is only the case if i did the mentioned switch in the system settings or using gfxCardStatus tool.
It's almost instant for typical adjustments. (lets ignore pairing masks .. that's always slow).
on the mini i run the Monitor on the eGPU. Thats better in almost any usecase. Especially when doing video editing i noticed quite lower fps when having the monitor on the mini.0 -
fatihayoglu wrote:
That’s lots of IF’s and blind defending CO for not utilizing current technologies. Since I bought my late 2015 top spec iMac 5K, CO is extremely slow on basic adjustments, that’s about 4 years now. Many people have and still are complaining about the performance of their top spec Macs. Every single time, CO blames Apple because of not putting strong GPU in these machines but we all know that when we bought these machines, they were capable of top class video editing. So, sorry but, if a machine can edit video, for basic photo editing it should fly. Period. No excuse or whatsoever. I’m not talking about DAM or anything. And also, I have realized CO does not utilities GPU for adjustments. Then it is poor or lazy coding, sorry, no excuse there as well. Also Apple has stopped or going to stop with the next release to support OpenCL, that means their code won’t be optimized for OpenCL, one can expect poor performance from applications that demands GPU use. Metal is a framework, written by Apple for Apple HW. We, MAC users, expect all the softwares should support Metal so we work on a optimized system. So it’s not about a company brings something to sell more, it’s a framework. When Java went from version 5 to 6, it wasn’t to sell more products. The code was optimized and cleaned. If you are using a donkey years old code, it contains full of bugs etc. then you can expect bad performance.
You make a lot of assumptions. Notably that new code (really additional code options when talking OpenCL as far as I can tell) has less bugs than old code. I seriously doubt that but if you have evidence from a broad base of developments I will happily change my mind.
Java was actually a great example of where developer would freeze their code to a "known good" (for them) version in order to avoid having to constantly battle with conflicting updates. They would ensure that their installation always included a version of Java they were OK with.
Certainly at some point they would need to move on - but there is no point in a smaller developer battling the latest (and allegedly greatest) of everything unless some specific benefit accrues to their product and they hope that makes it worth battling with the early adopter issues. Adobe may have the resources and influence to take things on, not least because they can spread any benefits across a very wide product range, some of which is in the development tools sphere of influence anyway. So they have to make the effort to stay in that market and once done it's logical to roll out the results through the rest of their products.
For many other developers that is not so obviously the case.
Nothing I have read about OpenCL suggest that any version, widely adopted or not, is a panacea for for high speed processing in any situation.
Indeed Apple, which initially created OpenCL, seems to be intent on moving on to other things that suit its purpose and that potentially means that dual platform vendors may see less point in chasing OpenCL for its portability - one of the main marketing points for the technology.
That and the apparent effect that increasing the technology base and complexity of any product tens to mean that optimum results only come with very exact tuning of parameters to available features and functions. Without that disappointment may follow.
That's a big overhead and carrying it will be of potential (but not certain) benefit to a small number of existing users and a some others who end up upgrading their hardware to new equipment that itself is keeping up with the technology in ways that go beyond just including the ability to support a recent iteration of the code if anyone has used it.
And a basic internal communication system configuration that can deal with the data volumes at the speeds that allows the technology to appear to be effective to a human user.
I keep looking for something that suggests that the great wow factor and excellent "bang for buck" result from OpenCL2.x.
If anyone has a link I'd love to read what it says.
Grant0 -
verstaerker wrote:
i'm using CO 20 on a mac Mini with Blackmagic eGPU Pro. It's definitely using it. It's important to have the "prefer external GPU" option on. When adjusting image the usage is around 30-50%. When exporting it's near 100%.
On my Macbook Pro it's only using the AMD GPU if i switch off the system-setting "automatic change of graphic-mode“ (wich enforces the AMD GPU to be used at all times) .
I was in contact with the support about it. I demonstrated in a video the behavior. Since then i haven't heard back from then.
It would be good if they could use Metal2 instead of openCL .. i'm sure the performance would increase a lot.
What is this option prefer external GPU ? How do we have this option on ?0 -
KrishnaKotti wrote:
What is this option prefer external GPU ? How do we have this option on ?
in finder on the app-icon, rightclick Information ... there is a checkbox "prefer external GPU"0 -
btw i'm using CO now on a MacPro 2019 with a VegaII .. this runs like a dream. Absolutely great.
Except the masking tool with AutoAdjustment on0 -
Since my Mac Mini my eGPU isn't fully used during exporting pictures. I opened a thread: https://forum.phaseone.com/En/viewtopic.php?f=71&t=37458&p=177162#p177162
Does anyone experience something similar? Funny fact is, that the Mac Mini is a little bit faster than the Macbook Pro. But I would like to know why the eGPU isn't used. Only difference between the systems is OS and the size of internal the SSD.0 -
Whitesnake wrote:
Since my Mac Mini my eGPU isn't fully used during exporting pictures. I opened a thread: https://forum.phaseone.com/En/viewtopic.php?f=71&t=37458&p=177162#p177162
Does anyone experience something similar? Funny fact is, that the Mac Mini is a little bit faster than the Macbook Pro. But I would like to know why the eGPU isn't used. Only difference between the systems is OS and the size of internal the SSD.
Isn't used or isn't FULLY used?
Usage of a GPU does not imply bypassing all other computer processes not that the the available data transfers rates in the computer and to and from the GPU can force 100% utilisation.
Furthermore there may be power demands that have to be managed and heat controls to apply.
Given the nature of the operation you may also need to consider the reporting resolution of whatever you are using to monitor performance.
Grant0 -
Isn't fully used (it's used approx. 60 %). The Macbook Pro used 100% of the eGPU. 0 -
Whitesnake wrote:
The Macbook Pro used 100% of the eGPU.
For what does the Macbook use the eGPU 100%?
By that I mean "What is the Macbook doing (by itself) that requires 100% use of the eGPU?"0
Post is closed for comments.
Comments
25 comments