 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
All yoking aside, it is becoming an issue how much electricity one uses. I used to have my workstation running 24/7 but now try to turn it of when it is not in use. I can turn it onn and controll it remotely even through internet so no need to keep it running all the time. I would love to see an option "turn computer off when done" for the batch rendering process.
i have read about GPU's that use only 2 watts in iddle state, not every card does that. So certainly something to consider.
Offline
tived wrote:
what's that Hans, 2x nVidia GTX-690 4GB :-)
Henrik
PS: You turn them on, and the neighbourhoods lights dims!
Doing fine on my 85 euro hd 6570 which also has 4 GB. No problem spending 1000 euro but then I still have a 4 GB card, although faster.
Looking forward to http://www.brightsideofnews.com/news/20 … rate.aspx. :-)
Offline
Hans,
You said you can turn it on and off, via the internet? How? In Particular how would you turn it on?
I use teamviewer to remotely access it which will let me turn it off, but I am struggling to work out how to turn them on.
Thanks
Henrik
PS: There is always something else to learn
Offline
tived wrote:
Hans,
You said you can turn it on and off, via the internet? How? In Particular how would you turn it on?
I use teamviewer to remotely access it which will let me turn it off, but I am struggling to work out how to turn them on.
Thanks
Henrik
PS: There is always something else to learn
something called an ip power 9258 http://pimzos.nl/shop/aviosys/ipsenseco … er9258.php
It has one power inlet and four power outlets.
Through it's www-userinterface you can turn of and on these outlets. Also you can have the box ping a certain ip-adres and restart an outlet when not receiving a response for a certain while.
alternative is to look for the wake-on lan protocal that some network cards will support.
Offline
Thanks Hans but would those not require a static IP address ?
Merry Xmas
Henrik
Offline
gkaefer wrote:
I plan to upgrade from my GTX260 (182Watt) to GTX650Ti (110Watt) so you also can save money if you upgrade ;-))
Georg
I've got a GTX670, which use 170W at max. That's not my main concern 
Leif
Offline
What is your main concern the? :-)
Ps: I have jumped on m4/3 kit oly like your with a few primes,
Offline
general question, maybe someone can answer:
now I use one graphic card with 1 GPU and 700mb... and plan to upgrade to GTX 650ti with 1 GPU using 2GB GDDR5 able to combine up to three such cards via SLI.
so If I combine 2 cards, does than autopano does see & use:
1 GPU with 2 GB memory,
2 x 1 GPU with 2 x 2 GB memory,
1 GPU with 4 GB memory or
2 GPU with 4 GB memory
and is the GPU used at all - or only the GDDR5 memory?
is the number of CUDAs or any relevance?
its not for rendering used - I know.
its used for preview only - I know.
is it possible to implement the jobs needed for decoding raw images by autopano so the GPU can do the job.
(currently looking at the cores I assume the raw decoding in autopano is also not threaded to multiple CPU cores)
thanks a lot,
Georg
Offline
Finding the balance between speed of GPUS and the amount and the speed of the memory on the graphics card
Hen
Offline
tived wrote:
Finding the balance between speed of GPUS and the amount and the speed of the memory on the graphics card
Hen
That didn't make sense :-)
It's about eliminating any and all bottlenecks, you might be able to buy a cheap card with slower processor on it but with more ram, but that's only half the story.
Ideally it should be lots/max ram and fast processor. BUT we all have to compromise and this is where cards like Hans bought fits in, and it has seemed to have made a difference for not a whole lot of money.
Measure your current setup, compare it to other systems which is similar to what you aspire to buy and then do the cost/performance evaluation.
Henrik
Offline
tived wrote:
Thanks Hans but would those not require a static IP address ?
Merry Xmas
Henrik
Nope, as the box has DDNS or Dynamic DNS. It can connect to a DDNS provider and tell it on which IP it is. Never used however as I have a fixed IP?
To be honest I mostly use it to remotely restart when it hangs itself. I dont need to save power as the maxhine uses 150 watt only when not being used, which include all external backupdrivers.
Last edited by HansKeesom (2012-12-22 21:46:38)
Offline
That makes a big diff, I used to have a static ip but no longer , makes it onerous step in the process
Offline
tived wrote:
tived wrote:
Finding the balance between speed of GPUS and the amount and the speed of the memory on the graphics card
HenThat didn't make sense :-)
It's about eliminating any and all bottlenecks, you might be able to buy a cheap card with slower processor on it but with more ram, but that's only half the story.
Ideally it should be lots/max ram and fast processor. BUT we all have to compromise and this is where cards like Hans bought fits in, and it has seemed to have made a difference for not a whole lot of money.
Measure your current setup, compare it to other systems which is similar to what you aspire to buy and then do the cost/performance evaluation.
Henrik
I did some testing with my card (http://www.sapphiretech.com/presentatio … &lid=1)
It is important to keep in mind the datastream that goes from disk -> RAM -> GPU RAM
I found that when running of a normal SATA-connected SSD, there is not really a difference in preview speed between using the GPU or not. The only difference is the loading on screen seems faster when using the GPU.
When running from a PCIE-SSD there is a larger difference I could notice. Instead of a 100 seconds without GPU I get 80 seconds.
20% might not seems much but it should be added that with the GPU I can work on 3 to 4 panoramas in parallel, as long as I don't zoom in all of them. Working on 4 panoramas at the same time really means I am busy all the time and the computer keeping up with me. This I can't do when GPU is off.
CPU is mostly on 65%, GPU on 55%
This means some room for improvement although I realise that it will be hard to get the GPU up to 100% as I can imagine APG cannot be programma to do so.
Based on the above and the datastream disk -> RAM -> GPU RAM in mind I think there are three ways for me to speed up things
1 faster data between RAM -> GPU
2 More memory on GPU
3 More memory in RAM
1. the card I have is only 128 bit so there should be something to gain when moving data from RAM to GPU RAM
2. the more RAM on the GPU the more data can stay there and needs not to be regained from RAM.
3. The more RAM the less data needs to be regained from Disc.
Wondering if a second GPU equal to what I already have will address 1 & 2. I guess only one way to find out. I hope to get an answer on that one soon see http://www.kolor.com/forum/p109068-toda … 42#p109068
[Update 25-12-2012 : o well, for 85 euro I am not gonna wait, ordered the second GPU and hope to share some results in a couple of days.
Last edited by HansKeesom (2012-12-25 21:43:20)
Offline
