Hi end gpu or HI end cpu?
While C1 uses gpu power openCL im wondering
how to incerse speed of c1 at lowest cost
1) update a graphic card from 9600GT to GTX 560 96 to 384 cores mote than 4 times faster
or update cpu
2) amd 4x3,7 Ghz to intel 2500k about 50% more performance bit more expensive
Can any one make a compassion of performance at OpenCL on/off or downclock Gpu cores to 10% 50% 70% gpu power
Im really wonder the results. May by super-performance openCL graphic card is the best price to performance ratio even in slower cpu-s. If so it means u can update just a graphic without need to reinstalling windows.
can anyone make such test ?
how to incerse speed of c1 at lowest cost
1) update a graphic card from 9600GT to GTX 560 96 to 384 cores mote than 4 times faster
or update cpu
2) amd 4x3,7 Ghz to intel 2500k about 50% more performance bit more expensive
Can any one make a compassion of performance at OpenCL on/off or downclock Gpu cores to 10% 50% 70% gpu power
Im really wonder the results. May by super-performance openCL graphic card is the best price to performance ratio even in slower cpu-s. If so it means u can update just a graphic without need to reinstalling windows.
can anyone make such test ?
0
-
You had probably already solved the problem in some way, but for record:
With C1 6.3.2 64bit on Windows 7 system I did an update through following steps:
1) AMD X2, dual-core @ 2.4 GHz + non-OpenCL graphics
2) Changed GPU to nVidia GTS 250
3) Changed MB to Intel i5 2500K (3.4 GHz quad core), kept the GPU.
All time worked with 17 Mpx Canon RAW files.
Original configuration was OK with C1 4.xx, but with 6.xx it was sluggish and unusable.
Updating the GPU did not change the experience significantly. It sometimes made the things slightly faster, but (especially earlier C1 6.xxs releases) often crashed. Things like local adjustments were totally unusable (click and wait a second or two).
After updating the CPU the experience was totally different. Local adjustments works at usable speed (almost no lags), import and rendering is way faster. Whereas before I had to think through the adjustments, change them, wait, now I can work in real time. The speedup in RAW-JPEG conversion (with per-image crops, adjustments) is roughly 5-10 times after the CPU upgrade for me.
So either one has to upgrade to a Fermi GPU at least, or the GPU power is not really utilized as long as the CPUs can't prepare the work for it in time. Therefore I suggest first CPU, then GPU upgrade.0 -
I've done some further tests and looking around:
(1) Only thing the OpenCL is ever used for is to faster apply adjustments for onscreen display (e.g. transformations of the proxy file). C1 does not use it for proxy generation or final rendering (confirmed with C1 support) as of 6.3.2..
(2) With a decent CPU (e.g. i5 is enough), OpenCL is probably NEVER used, because even with GTX 560 I did not see any notice in the logs (except for GPU initizialization) nor change in interactive/subjective speed while working with the images.
I contacted support about this as I find the 1720 knowledgebase article misleading in this case.
There is definitely no use in going into combination of faster CPU+GPU and there's not much sense in going into fast GPU alone as you can get the fast CPU for a comparable price.
So from my point of view it is just a trendy label on the software in this case (to avoid the "marketing bullshit" formulation).0 -
[quote="Vaclav1" wrote:
I've done some further tests and looking around:
(1) Only thing the OpenCL is ever used for is to faster apply adjustments for onscreen display (e.g. transformations of the proxy file). C1 does not use it for proxy generation or final rendering (confirmed with C1 support) as of 6.3.2..
(2) With a decent CPU (e.g. i5 is enough), OpenCL is probably NEVER used, because even with GTX 560 I did not see any notice in the logs (except for GPU initizialization) nor change in interactive/subjective speed while working with the images.
I contacted support about this as I find the 1720 knowledgebase article misleading in this case.
There is definitely no use in going into combination of faster CPU+GPU and there's not much sense in going into fast GPU alone as you can get the fast CPU for a comparable price.
So from my point of view it is just a trendy label on the software in this case (to avoid the "marketing bullshit" formulation).
Thanks for the tests, no need to buy an extra GPU now.0 -
Vaclav,
Thanks for the tests.
I am considering an upgrade. My existing photo editing dedicated system was built as an interim solution whilst watching for how software developments might move the recommended specifications.
Running 6.3.2 on 32bit Windows 7 (and setting the use 3Gb Ram) the system seems reasonably stable even for large batch edits . Speed, on an old 2 core 3Ghz Pentium processor with 4Gb RAM installed, is mostly OK (ie acceptable) although sometimes the processes seem to go walkabout for a while. Faster would be nice, sometimes it would be especially usable, although mostly I doubt I could improve my effectiveness much .... except when using local adjustments when things really become slow. (In some cases, I'm not yet entirely sure what they are, 3 or more layers and a certain combination of adjustments seems to lead to instability which I think is something that really related to processing speed and memory management.)
Now I appreciate that there is a lot going on in there - anything to do with multiple layers is likely to be hard work - but I would hope that a better specced machine would deal with that. Your experiments seem to suggest it would, which is great, but did you run any assessment for the Local Adjustments functionality performance?
I'm thinking that your observation that an i5 CPU would be enough looks rather attractive from a cost point of view but if Local Adjustments (which I assume will be extended over time) would benefit even more from an i7, for example, then it might make more sense to go i7 (or equivalent) now rather than at some point in the future. So if i5 would still be enough the Motherboard/CPU cost difference might be enough to cover the cost of adding a SSD to the spec. rather than relying on spinning disks for everything.
I'm not surrently dealing with very large files so your comments on GPU usage suggest I have little to gain from investment there. In any case a Graphics card is an easy upgrade for later if necessary.
Thanks in advance for any further advice you can offer.
Grant0
Post ist für Kommentare geschlossen.
Kommentare
4 Kommentare