 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
Hi everyone,
I've a question concerning Autopano Giga/Pro (don't matter) in particular the GPU acceleration.
Is the GPU acceleration only used for the realtime editing of the panorama (the wiki seems to indicate this) ?
Or can the GPU acceleration be used to accelerate the rendering process (assembling and blending) ?
Thanks a lot
Offline
Lumbermatt wrote:
Is the GPU acceleration only used for the realtime editing of the panorama (the wiki seems to indicate this) ?
Yes. Afaik it doesn´t work with rendering.
best, Klaus
Offline
I confirm this : it is not used yet for rendering.
Offline
There are several threads in which Alexandre wrote that some time in the future GPU *could* (and I'd wager it *would* if it could) be used for rendering. So I think it's just a matter of time, especially knowing how much processing power modern GPUs have...
Offline
Yes, we have a huge development task planed for v3 about that. GPU everywhere.
Offline
cool, then its time to get 4x GPUs or more
:-)
Henrik
Offline
tived wrote:
cool, then its time to get 4x GPUs or more
:-)
Agree! 
best, Klaus
Offline
My question indeed was concerning the new computer I want to setup mainly for panorama generation.
I plan to use the new Sandy Bridge plateform and if there were an opportunity to use the GPU I would have added a dedicated GPU.
Until the GPU calculation is supported, I'll use the IGP of the SB CPU. By the way, I have to choose between the 2500K (core i5) and 2600K (core i7). The main difference are the 8 logical cores (Hyperthreading) of the i7 in comparaison to the i5 (and 80 bucks more of course).
Does somebody have a feedback concerning the gain that hyperthreading provides on autopano's calculation times ?
Offline
For current autopano generation, I will always advise to get a better CPU if you can and not spent too much money on the GPU. The CPU is far better used than the GPU.
We use hyperthreading from core i7. So it does stitch faster with HT but not as much as it would have been if it was physical core.
Offline
AlexandreJ wrote:
Yes, we have a huge development task planed for v3 about that. GPU everywhere.
Hi Alexandre - sorry to bump this old thread, but it seemed as good a place as any to ask the quetion.
Are you talking about Nvidia Tesla support with that comment?
Offline
Not only. We will not rely on CUDA as technology to be able to do that but on OpenCL, so any CPU / GPU could used ( even heterogenous architecture with both ).
Offline
AlexandreJ wrote:
Not only. We will not rely on CUDA as technology to be able to do that but on OpenCL, so any CPU / GPU could used ( even heterogenous architecture with both ).
Woah.
Just reading up about Tesla's (448 cores!) as my brother has a machine with one right now, and is probably going to be putting another 3 on the board.
What kind of timeline are we looking at here? I think I may have to start planning my next PC!
Offline
Not soon. It won't be on v3.0 but on a minor update (3.*).
Offline
gddxb wrote:
AlexandreJ wrote:
Yes, we have a huge development task planed for v3 about that. GPU everywhere.
Hi Alexandre - sorry to bump this old thread, but it seemed as good a place as any to ask the quetion.
Are you talking about Nvidia Tesla support with that comment?
Put in a Tesla board and it needs 500 Watt........
Will that board have/need/come with its own external powersupply?
Don t get me wrong, I think GPU processing is great when your CPU is on 100% all the time and you don t want to replace it. But if (re)placing your GPU means you also need to replace you powersupply and learn to live with a much higher energybill, I am not sure if I want to go there.
In terms of calculation power, how would this card performed compared to say a i7 920 CPU??
Would be great if the post at http://www.kolor.com/forum/p87033-2011- … -58#p87033 can be expanded with a GPU section
Offline
Really, I cannot give any performance measurement yet. There's too much work to do first because reaching this area.
So for the moment, stick on CPU first. This evolution using GPU to speed up the process is not coming soon.
Offline
HansKeesom wrote:
gddxb wrote:
AlexandreJ wrote:
Yes, we have a huge development task planed for v3 about that. GPU everywhere.
Hi Alexandre - sorry to bump this old thread, but it seemed as good a place as any to ask the quetion.
Are you talking about Nvidia Tesla support with that comment?Put in a Tesla board and it needs 500 Watt........
Will that board have/need/come with its own external powersupply?
Don t get me wrong, I think GPU processing is great when your CPU is on 100% all the time and you don t want to replace it. But if (re)placing your GPU means you also need to replace you powersupply and learn to live with a much higher energybill, I am not sure if I want to go there.
In terms of calculation power, how would this card performed compared to say a i7 920 CPU??
Would be great if the post at http://www.kolor.com/forum/p87033-2011- … -58#p87033 can be expanded with a GPU section
There's some performance indications in the paper linked to here (PDF link):
http://developer.download.nvidia.com/pr … essing.pdf
That was from a couple of years ago though, and I have no idea regarding relative performance to CPU's.
Offline
AlexandreJ wrote:
Really, I cannot give any performance measurement yet. There's too much work to do first because reaching this area.
So for the moment, stick on CPU first. This evolution using GPU to speed up the process is not coming soon.
Hi Alexandre,
I was not so much asking for a precise answer :-) I must assume you have made and estimate of how much faster things will be before you started coding for it.
From the link of gddxb I get the impression that we can count on a factor 5 x to 10 x compared to a Intel Quad-core Xeon E5540 (Nehalem) 2.53 GHz.
Not sure how a XEOn compares to my i& but 5 to 10 times sounds good to me, that s for sure :-)
Last edited by HansKeesom (2011-08-23 16:14:53)
Offline
AlexandreJ wrote:
Not soon. It won't be on v3.0 but on a minor update (3.*).
Now that we have v3.0 I am eagerly waiting for the minor upgrade that will make the big leap in gpu processing. I can only dream of near instant preview........ Putting money aside already for this :-)
Offline
what's that Hans, 2x nVidia GTX-690 4GB :-)
Henrik
PS: You turn them on, and the neighbourhoods lights dims! 
Last edited by tived (2012-12-17 07:22:40)
Offline
tived wrote:
what's that Hans, 2x nVidia GTX-690 4GB :-)
Henrik
PS: You turn them on, and the neighbourhoods lights dims!
His electricity supplier sends him a personal greeting card and present basket (oops, hamper, right word?) every year for christmas.
greetings from germany
Chris
Offline
I don't know whats happening elsewhere but here in Western Australia Electricity has gone up by 35-50%
in the last year or so - and we are not doing much for alternative energy sources? Makes you wonder... we have almost 10 months of sunshine but very little is done with solar, it blows every day at middag, but we have very very few windmills, ohh yes, thats right, the cows gets confused or something like that?
I guess I better stick to my GTX-580 ;-)
Henrik
Offline
I plan to upgrade from my GTX260 (182Watt) to GTX650Ti (110Watt) so you also can save money if you upgrade ;-))
Georg
Offline
Actually some the newer GPUs use less power then the old ones
Henrik
Offline
