Agisoft Metashape
Agisoft Metashape => General => Topic started by: Wishgranter on January 06, 2014, 01:08:03 AM
-
Hi all, wish all good and fruitful year 2014 ;)
and we have managed to bring interesting stuff for the community here...
The http://www.anandtech.com/ decided to use Photoscan in 2014 benchmark suite, so they will test a lot of CPus and GPUs under controlled environment so all of us will see what the best price/performance devices are suited for your projects. the results will be released later this year, we expect that dual CPUs will be tested in 1-2Q too. ( 4P systems later ?? ) So stay tuned when it is released.........
This way want thanx to Ian Cutress from Anandtech that he and his team is inspired with this interesting software from Agisoft team and willing to use it in benchmark apps for the 2014 year.
http://www.anandtech.com/ - its the site where i get a lot of the indepht information about how the hardware stuff work and so can help many times with hardware problems and issues....
-
Hi mate,
I wish you to have a good year as well.. ;)
thx for the news. please let us know when the result is released.
Thanks to Ian Cutress ;D
-
Exciting!! Looking forward to seeing the results!
-
First set of results with the GIGABYTE BRIX Pro:
http://www.anandtech.com/show/7648/gigabyte-brix-pro/3
It seems:
Stage 1 prefers cores and MHz
Stage 2 needs fast memory / GPUs
Stage 3 likes high IPC and memory
Stage 4 likes cores and MHz
This was only a first set of quick results put together for a preview. Will hopefully have more by the end of the month on more platforms / DRAM combinations.
-
As we are looking for some new machines I appriciate your work and all updates regarding new results.
BTW - did someone try photoscan on the new Mac Pro?
-
I might get a chance to try the Mac Pro next week. The boss just reviewed one, but I have to get him to give me remote access to test. The 1866 C13 memory is going to kill performance though. ECC JEDEC standards are pretty lax, which makes me wonder if you can put normal non-ECC memory in it for a bit of a boost (and the system will enable XMP too).
-
That's interesting Ian, Thx a lot..
By the way, what about the full system specs.?
-
Agreed, very cool to have Agisoft in a standard set of benchmarks.
I do hope you use a project with a decent photo size. The 'official' benchmark file we use on this forums does not have large enough photos in my opinion. Especially for the Dense Cloud stage, the results in Million Samples/sec deviates way too much with results on real life projects. Because of the small photo size on the sample project, I think too much time is spend on setting up and overhead, and not enough on the actual number crunching.
For example, the AMD R9 290 GPU scores around 750 million samples/sec with this sample project, but on big projects (a set of 300 images with 21MP size) the average score is around 850. And I've processed projects where the same card scores over 1000.
-
Our benchmark uses the 50 image building tutorial set from here: http://www.agisoft.ru/tutorials/photoscan/02/
We use the following settings:
## Align Photos: High, disabled, 40000
## Build Dense Cloud: Medium, Aggressive
## Build Model: Arbitrary, Dense, Interpolation=Enabled, Medium
## Build Texture: Generic, Mosaic, 4096, 1
By recommendation from Agisoft, the GPU numbers have two CPU cores disabled each. Meaning that Stage 1/3/4, which do not use the GPU, might look slower than when the GPU is not used.
The test takes between 15 minutes (6-core Ivy Bridge-E) and an hour (Haswell mobile CPU) usually.
Basically we have to balance benchmark time - we can't have something that runs for a day. Wouldn't get paid otherwise! Especially as my normal benchmark routine takes 20-30 hours to get through (we are a computer component review website).
Even if you feel you cannot take the results quantitatively, take them qualitatively - it at least shows X > Y.
-
Yup with the GPU benches its a bit complicated, its because depht map reconstruction depend on few more things as just Mpix value. it depend on scene composition, so same Mpix value will give different Mpix/sec performance as the scene change. That mean if you do GIS scene and say some staue-building reconstruction it will give a bit different results. From my observations is best if is used HIGH or ULTRA settings for the reconstruction then is the GPU used to 100 %. so when set to medium the GPU is used say to just 60-70%, but it depend even on CPU speed.
But for other stages is OK with the settings......
On my Dual Xeon setup i get best results with all cores disabled, because of the interconnect saturation as data move over PCI-E bus adn CPU bus. but thats a Dual CPU, single CPU setup could get a bit better results..... Its just my observation, it could be a bit different on new E5v2 CPUs as then improved some CPU stuff....
For standard tests its enought with this. to have more precise Dual CPu results a bigger dataset could/should be used. If i get a bit more FREE time, we will digg intro this matter with IAN, so stay tuned and enjoy the single CPU + GPU setup results....
As in BUG REPORT forum thread is mentioned one problem with 80-160 Core quad CPU system, so if everything go and problem is solved we could do test even on so ULTRA system for you.....
-
And as IAN say, they are a bit balanced test. they cannot run benchmark for too much time. its a review and not specialized company that do indepht benches for specific app for free.....
as we get results in we could do specialized benchmark for the most powerfull CPU + GPU setup. say fastest CPU + Fastest GPU, Dual CPU, Quad CPU and multi GPU with the fastest GPUs with one heavy scene. so we get results for the say 3 systems, and see the eficiency of CPU + GPU scaling. becasue ifg you add second CPu you get not 100% speedup, just say 80-85 %, usinng 4 CPUs you get 70% speedup per CPU added to system......
-
.....
By recommendation from Agisoft, the GPU numbers have two CPU cores disabled each. Meaning that Stage 1/3/4, which do not use the GPU, might look slower than when the GPU is not used.
.....
I have two computers, each with a quad core I7 and a GPU. PhotoScan on each is set with "2 CPUs disabled". During much of the Align Photos and Build Mesh all CPUs are running at or very close to 100% according to Windows Performance Monitor and Open Hardware Monitor. There is no difference in performance during these stages when the number of CPUs "disabled" in PhotoScan is changed.
-
Indeed, the number of available cores only applies to the Dense Cloud phase. I can tell for sure because I got it set to use zero cores, and it still Aligns fine.
My comment about the project size was not to say your benchmark is not useful (on the contrary, I love it!), I just wanted to warn you that the results you get from a small project can be slightly skewed because Photoscan is spending a relatively short time on actually crunching numbers.
There are also phases in the Mesh Generation where the process runs entirely on a single core.
-
I might get a chance to try the Mac Pro next week. The boss just reviewed one, but I have to get him to give me remote access to test. The 1866 C13 memory is going to kill performance though. ECC JEDEC standards are pretty lax, which makes me wonder if you can put normal non-ECC memory in it for a bit of a boost (and the system will enable XMP too).
I love this thread! Glad so see you here, Ian :D
Wishgranter - splendid initiative!
Looking forward to upcoming tests. Maybe to see what I did wrong ::) since I will have to purchase my own workstation very soon - don't have time to wait for the test results :-\
Right now I am pondering the memory issue.
Aiming for the modern workstation motherboard ASUS Z9PE-D8 WS, 2xSocket-2011 which has 8 memory slots. (Only registered/buffered memory when greater than 64GB.)
http://www.asus.com/Motherboards/Z9PED8_WS/#specifications
Either I could do
Kingston DDR3 1333MHz ECC/REG 256GB (8x32GB) LRDIMM 240-pin, 1.35v, CL9
or
Kingston DDR3 1600MHz ECC/REG 128GB (8x16GB) RDIMM 240-pin, CL11
Speculating that maybe not so many of my upcoming projects might need 256GB, I am tempted to go for the faster 128GB. But how much performance difference should I expect between these two options, in theory? Assuming the projects generaly would stay within the 128GB limit, and if bigger, splitting in chunks (does that sound reasonable?).
Processors likely to be 2x Intel Xeon E5-2687W v2 (8 cores each, 3,4GHz) and two watercooled R9 290X GPU.
Posting here since I figure it?s actually on topic and of general interest :)
-
Based on my memory testing of Ivy Bridge and Haswell consumer platforms, there are simple rules when it comes to memory (if money is no object).
1. Calculate the Performance Index (PI). PI = MHz / CAS Latency
2. Choose the one with the higher PI.
3. If your options have similar PI, choose the one with the higher MHz.
Sources:
Haswell Memory Scaling: http://www.anandtech.com/show/7364/memory-scaling-on-haswell
Ivy Bridge Memory Scaling: http://www.anandtech.com/show/6372/
Here's a general graph on a memory intensive benchmark (Dirt3), FPS vs PI:
(http://images.anandtech.com/doci/7364/D3%20Min_575px.png)
When I get around to doing a memory scaling article on a few upcoming platforms and scenarios, I will have data to back it up. I should have access to a dual socket motherboard within the next few months, and some high end CPUs (either 8, 10 or 12 cores) to do some other testing for you, but please bear with me :)
-
Wow Ian, thanks for prompt response!
So the two options I mentioned have almost equal PI, indicating the 1600MHz EC11 would probably not outperform 1333 EC9, not by much anyway.
So, all in all, considering that 256GB allows for bigger projects than 128GB, I am now leaning towards the 256GB/1333.
Memory [GB] PI PI * GB / $ USD
Kingston DDR3 1600MHz ECC/REG 128GB EC11 128 145,5 8,6 2158
Kingston DDR3 1333MHz ECC/REG 256GB EC9 256 148,1 8,6 4432
I for sure will stay tuned for your upcoming tests.
Tom :)
-
Ian,
I am really glad you included Photoscan in benchmarks on Anandtech! This is one of few pieces of software that can stress hardware in a number of interesting ways. And PS is getting more and more popular by the minute! :)
I am particularly interested in how memory speeds improve PS performance - it seems to be quite the bottleneck in some stages.
I am also curious to know if/whether full utilization of CPU and two GPUs at the same time causes enough thermal issues to result in throttling. I saw that Anand in his review of new Mac Pro, in order to fully saturate the hardware, had to run two programs (Furmark and Prime, and it DID cause slowdown), well now Photoscan alone could do the trick (Build Dense Point Cloud step). I can't wait to see whether Mac Pro's cooling solution is sufficient for Photoscan. And lastly, it would be great to see how does dual D700 compare in OpenCL performance to dual 7970s or 290s on regular PC builds.
Thanks Ian!
-
Latest update to these results.
Click the link to see breakdown:
http://anandtech.com/show/7752/msi-x79agd45-plus-review-building-up/7
(http://images.anandtech.com/graphs/graph7752/61251.png)
I'm currently sitting here with a 12 core Xeon with 24 threads, 32GB DRAM 1600 C11, and it doesn't seem to be any faster than a 4960X. This is because the single threaded speed of the Xeon is poor, and despite having 2x threads, is down 25% in MHz for the multithreaded stuff. I will add these results on my Xeon article, coming soon (hopefully).
-
Some more results, including with 12/8 core Xeons:
For a breakdown of each of the stages:
http://www.anandtech.com/show/7914/msi-a88xme35-review/8
(http://images.anandtech.com/graphs/graph7914/62367.png)
-
Very interesting results.
Where can I download the same image set to test myself?
The samples on the agisoft page seem to be 31 & 32 images?
-
It's the Building Reconstruction set on http://www.agisoft.ru/tutorials/photoscan/
-
Results for the building dataset and the new AMD W9100 card....
With high setting I have:[/size]Using device: Hawaii, 44 compute units, 16384 MB global memory max work group size 256 max work item sizes [256, 256, 256] max mem alloc size 16127 MBfinished depth reconstruction in 724.2 secondsDevice 1 performance: 872.468 million samples/sec (Hawaii)[/color]
-
Thank you for sharing, Ian, Wishgranter and all of you.
Now I would like to share the benchmark of my recent workstation build.
ASUS Z9 P8 WS, 2* Xeon 2687 v2, 3* R9 290X, 1* Quadro K5000.
1333 MHz RDIMM CL9. Win 8.1 64
(https://dl.dropboxusercontent.com/u/36968875/PS_benchmark.PNG)
Interesting observation:
The Build Dense Cloud computing (82 sec) in the above table was done with all CPU cores deselected (0/32 CPU cores in the OpenCL preference).
Selecting the recommended 24/32 resulted in 94 seconds processing time.
Selecting 32/32 active cores resulted in 101 seconds processing time.
/ Tom
-
And as for Building Dense Cloud in High, all other same as above:
Device 1 performance: 1062.14 million samples/sec (Hawaii)
Device 2 performance: 1041.21 million samples/sec (Hawaii)
Device 3 performance: 1043.3 million samples/sec (Hawaii)
Device 4 performance: 426.897 million samples/sec (Quadro K5000)
Total performance: 3573.54 million samples/sec
Finished processing in 299.175 sec
Ultra high
Device 1 performance: 1115.57 million samples/sec (Hawaii)
Device 2 performance: 1099.39 million samples/sec (Hawaii)
Device 3 performance: 1111.38 million samples/sec (Hawaii)
Device 4 performance: 438.738 million samples/sec (Quadro K5000)
Total performance: 3765.07 million samples/sec
Finished processing in 2080.23 sec
/ Tom
-
Thanx Mr Map, so its not only my Dual setup that have better results with all CPUs disabled.. on single CPUs seems that just few cores disabling get better results......
-
Hi.
Just sharing this if useful:
All these tests were run on
Dell Precision T7600
Intel® Xeon(R) CPU E5-2620 0 @ 2.00GHz ? 12 core
Memory: 125.8 GB
GPU was turned on, 1 GPU to replace 1 CPU core as recommended.
Where more GPU than CPU available no CPU used.
Running the bench mark scene BUild Dense Cloud at Ultra High, Moderate Filtering
1x GeForce GTX 680 (8 Cores @ 1124Mhz, 2047MB) Finished processing in 1289.69 sec (21min 29sec)
Device 2 performance: 632.531 million samples/sec (GeForce GTX 680)
1x GeForce GTX 780 (12 Cores @ 1019Mhz, 3071MB) Finished processing in 984.936 sec (16m 25sec)
Device 1 performance: 823.705 million samples/sec ( GeForce GTX 780 )
1x GeForce GTX TITAN (14 Cores @ 928Mhz, 6143MB) Finished processing in 959.494 sec (16m)
Device 2 performance: 840.561 million samples/sec (GeForce GTX TITAN)
2x Quadro 5000 (11 Cores @ 1026Mhz, 2559MB Finished processing in 946.803 sec (15min 46sec)
Device 1 performance: 392.461 million samples/sec (Quadro 5000)
Device 2 performance: 396.724 million samples/sec (Quadro 5000)
1xGeForce GTX 780 (12 Cores @ 1032Mhz, 3071MB)
1xQuadro 5000 (11 Cores @ 1026Mhz, 2559MB Finished processing in 654.824 sec (10min 54sec)
Device 1 performance: 913.684 million samples/sec (GeForce GTX 780)
Device 2 performance: 397.382 million samples/sec (Quadro 5000)
Hope this is useful
Cheers
-P
-
Hi.
Just sharing this if useful:
All these tests were run on
Dell Precision T7600
...
Hi @pjennes,
It would be interesting if you would also publish the associated GPU performances, ie samples per second, please.
Cheers / Tom
-
I should have a new update to my CPU tests in the next couple of weeks.
If anyone is interested, here are my stage settings for the benchmark:
## parameters used:
## Align Photos: High, disabled, 40000
## Build Dense Cloud: Medium, Aggressive
## Build Model: Arbitrary, Dense, Interpolation=Enabled, Medium
## Build Texture: Generic, Mosaic, 4096, 1
-
Hi @pjenness,
It would be interesting if you would also publish the associated GPU performances, ie samples per second, please.
Cheers / Tom
Hiya
I have update my post above to inlcude the samples.
Let me know if more infor needed
Cheers
-P
-
http://www.anandtech.com/show/7963/the-intel-haswell-refresh-review-core-i7-4790-i5-4690-and-i3-4360-tested/3
(http://images.anandtech.com/graphs/graph7963/63169.png)
I have two sets of E5 processors which I will test after Computex in June.
-
Has anyone had the chance to test out any of the the new MacPro flavors with dual GPU's?
-
I have added Photoscan to AnandTech's results database.
I just added the results from about a dozen CPUs that were not in the database, and I have a fair number lined up over the coming months.
Link for all results: http://www.anandtech.com/bench/CPU/1057
Graph of Total Time:
(http://images.anandtech.com/graphs/graph7611/65098.png)
Links to:
Stage 1: http://images.anandtech.com/graphs/graph7611/65094.png
Stage 2: http://images.anandtech.com/graphs/graph7611/65095.png
Stage 3: http://images.anandtech.com/graphs/graph7611/65096.png
Stage 4: http://images.anandtech.com/graphs/graph7611/65097.png
CPU Mapping Speed: http://images.anandtech.com/graphs/graph7611/65099.png
-
8)
Thanx IAN.........
-
Hi all,
thanks for publishing the benchmark-tests.
Having the most recent up-to-date and powerful hardware will make you really happy when working with agisoft, but it may also strain your budget nastily.
For those who are looking for computing power but having only restricted budget, I would like to report my experiences with second hand / used older generation hardware:
D.... T7500 Precision Workstation, Dual hexacore Xeaon X5680 with 3.3 GHz, 48 GB 1333 MHz Cl 9 RAM, Dual nVidia Quadro 4000, 2 GB RAM., manufactured in 2010
Pricing EU / Germany: 1200 - 1600 Euro
First test run with agisoft showed, that processor cooling was absolutely insufficient for running under full load for a longer time. Core temp higher than 80 - 90 deg Celsius was common, but which I think is not healthy over a longer period of time.
I modified the cooling system in he following way:
- removed all the plastic stuff from inside
- removed both processor heatsinks (mainboard and riser-card, which is for the second CPU)
- installed two Noctua NH-U9DX 1366 tower cooler with fans, which sustainly solved the overheating problem
Core temp under full load over longer period of time does now seldom reach more than 60 deg C.
Here are the benchmark results for the 4 stages, settings and images as described in the other posts above (4 of 24 virtual cores disabled), Hypershreading and all Cores activated in Bios
1. 205.3 sec
2. 161.8 sec
3. 231.5 sec
4. 109.9 sec
Total of 708 sec or 11.8 min, Performances for step 2: CPU 181.631 million samples/sec, GPU1 249.646 million samples/sec, GPU2 254.201 samples/sec, total 684.86 .
The machine is running Ubuntu 12.04.4 LTS with recent updates installed.
Cheers
-
Hey,
i'm testing my new machine and here are some interesting benchmarks i've made today:
I7 4790K @ 4,4GHz
32GB DDR3 TridentX @ 2400MHz
2* PowerColor R9 290 PCS+
Some SSDs too...
Cooler Master HAF XB Case
3* very noisy but nice Delta Fans (30W each... :o)
H100i watercooling for CPU
All the stuff is very well cooled CPU: 40°C GPU: 43°C
Settings:
building.psz
Building dense cloud / medium / aggressive
8/8 CPU cores | 0/2 R9 290
Device 1 performance: 177.953 million samples/sec (CPU)
Total performance: 177.953 million samples/sec
Finished processing in 416.586 sec (exit code 1)
6/8 CPU cores | 1/2 R9 290
Device 1 performance: 151.705 million samples/sec (CPU)
Device 2 performance: 863.408 million samples/sec (Hawaii)
Total performance: 1015.11 million samples/sec
Finished processing in 174.813 sec (exit code 1)
4/8 CPU cores | 2/2 R9 290
Device 1 performance: 127.262 million samples/sec (CPU)
Device 2 performance: 715.734 million samples/sec (Hawaii)
Device 3 performance: 632.521 million samples/sec (Hawaii)
Total performance: 1475.52 million samples/sec
Finished processing in 118.62 sec (exit code 1)
0/8 CPU cores | 2/2 R9 290
Device 1 performance: 868.328 million samples/sec (Hawaii)
Device 2 performance: 741.236 million samples/sec (Hawaii)
Total performance: 1609.56 million samples/sec
Finished processing in 112.827 sec (exit code 1)
Same project,building dense cloud, but 'high':
4/8 CPU cores | 2/2 R9 290
Device 1 performance: 146.476 million samples/sec (CPU)
Device 2 performance: 882.266 million samples/sec (Hawaii)
Device 3 performance: 917.818 million samples/sec (Hawaii)
Total performance: 1946.56 million samples/sec
Finished processing in 503.681 sec (exit code 1)
0/8 CPU cores | 2/2 R9 290
Device 1 performance: 907.5 million samples/sec (Hawaii)
Device 2 performance: 903.321 million samples/sec (Hawaii)
Total performance: 1810.82 million samples/sec
Finished processing in 485.123 sec (exit code 1)
Some interesting facts...
My company notebook is a XMG with 880M 8GB - while testing with the monument benchmark i got
these results against my normal render machine.
Medium setting, building dense cloud:
Notebook: 266sec. (GTX880M 8GB)
Desktop: 240sec. (1* R9 290 PCS+ 4GB)
Desktop: 123sec. (2* R9 290 PCS+ 4GB)
Ultra high setting, building dense cloud
Notebook: 8622sec. (GTX880M 8GB)
Desktop: 3723sec. (2*R 290 PCS+ 4GB)
The GTX880M (f**k... this is mobile!) has a lot of power...
Hope this helps :)
Cheers
Daniel
-
I've just been handed a VM by our IT department primarily for testing Photoscan. Interesting weeks ahead ;D
PC: Dell Precision T5600 , 635W, Xeon E5-2630 (Six Core, 2.30GHz Turbo, 15MB, 7.2 GT/s), 32GB (4x8GB) 1600MHz DDR3 ECC RDIMM, 3 GB NVIDIA Quadro K4000
VM (via PCoIP): 16 vCPU cores and 64 GB RAM + equivalent to Quadro K5000 GPU (dedicated)
-
Hello!
I just got the Asus Geforce GTX 970 Strix and have been running some quick tests.
I've used the building.psz with Build Dense Cloud at Medium and Aggressive. 6 out of 8 cores enabled.
First run:
Device 2 performance: 732.496 million samples/sec (GeForce GTX 970)
Second run:
Device 2 performance: 735.672 million samples/sec (GeForce GTX 970)
Third run:
Device 2 performance: 742.626 million samples/sec (GeForce GTX 970)
I then tried running it at High and I got:
Device 2 performance: 854.403 million samples/sec (GeForce GTX 970)
I also did one test with 0 out of 8 cores enabled (at medium) and I got:
Device 1 performance: 843.676 million samples/sec (GeForce GTX 970)
This is all with Nvidia driver GEFORCE 344.11.
I also ran my old GTX 570 and it scored between 525-545 million samples/sec.
Best, Magnus.
-
Hey,
a few month ago i put together a workstation for Photoscanon paper for our Organisation.
It featured a i7-4960X, 64GB DDR3 as main components.
Now we finally got the OK from the purchasing office and got the money.
In the meantime intel released its new Haswell-E processors.
The 5960X and the 5930k... i now wonder if i should stick with my old configuration or if i should put
together a new one? The 5960X offers more cores but is more expensive. The 5930K is a lot cheaper than the 4960X, but how would the performance compare in Photoscan. I also read that both new cpu wont beat the 4960X on single thread performance.
Another Problem is the budget, both new cpu require a new, more expensive MB and RAM. My budget is quite fixed so if i got a i7-5960X i would definatly have to get rid of 32GB RAM for now.
With the cheaper 5930K i could stick with 64GB, maybe even get a better gpu as before.
what would you suggest?
The System will be used for 3D documentation of smaller artifacts (but many) and for orthophoto & DEM generation of smaller areas created with a balloon or drone.
-
...
In the meantime intel released its new Haswell-E processors.
...
With the cheaper 5930K i could stick with 64GB, maybe even get a better gpu as before.
what would you suggest?
I find my main limit to be available RAM. Many of the x99x-boards have 8 memory slots and will support up to 128GB. I would go with the 5930K and 16GB RAM chips - if they are available soon, so that i could upgrade if needed. I guess Photoscan will benefit from the faster memory-speeds also.
-
Have just built a new Rig for home
Intel i7-5930K CPU @ 3.50
Asus Ramapage V motherboard
32 GB Ram (for now)
Gigabyte 980GTX G1 4GD
Gigabyte 970GTX G1 (Setup as eGPU inside a viDock -> expressCard PCie interface)
Wasnt sure what scene to benchmark so did the statue one. Ultra High and Aggressive.
Using device: GeForce GTX 970, 13 compute units, 4096 MB global memory
Using device: GeForce GTX 980, 16 compute units, 4096 MB global memory
finished depth reconstruction in 2288.13 seconds
Device 1 performance: 616.547 million samples/sec (GeForce GTX 970)
Device 2 performance: 1012.03 million samples/sec (GeForce GTX 980)
Using 1.1.0 pre-release (as current release didnt work with the 9xx series or nvidia drivers)
Hope that is useful info
Cheers
-Paul
-
Is there any new word on the GTX980's? I just see that one user had success, and some were going to be doing some benchmarks. Anandtech doesnt seem to have GPU on the charts for Photoscan, just CPU.
Anyway, I am spec'ing out a workstation for the company and am looking at 2-3 GTX980's, or 2 K6000, or 2 TITAN. (with 64-128 GB RAM, and 1-2 Xeon's)
Just need to know if there are any known issues. Any suggestions as well?
Thanks!
Edit: I forgot to include links, my comments are in reference to multiple threads (one on the new GTX cards, and one on the 1.1.0 pre-release testing, and thought I would post it in the benchmarks thread)
http://www.agisoft.com/forum/index.php?topic=2878.0
http://www.agisoft.com/forum/index.php?topic=2883.120
-
0/0 cores / 2/2 r290x
Device 1 performance: 932.164 million samples/sec (Hawaii)
Device 2 performance: 926.353 million samples/sec (Hawaii)
Total performance: 1858.52 million samples/sec
0/0 cores / 2/2 gtx980
Device 1 performance: 1007.78 million samples/sec (GeForce GTX 980)
Device 2 performance: 1003.94 million samples/sec (GeForce GTX 980)
Total performance: 2011.73 million samples/sec
monument test file, resolution high...
nice cards!
-
thanks!
-
I'm curious, and I thought this thread might be the one to post in. I had a gtx 680 sitting around and popped it in thinking I would get some extra processing power, and I did, but the results were strange. It seems that while dense point cloud reconstruction was indeed slightly quicker (.12x quicker), alignment was slower. I may have mixed something up in the testing, so I thought I'd ask if disabling a CPU core in order to utilize another CUDA card results in such poor performance improvements or if my results seem wonky... Thanks.
-
Hello driftertravel,
Disabling CPU cores in the OpenCL tab of PhotoScan Preferences window has effects only on depth maps generation step, if any OpenCL device is checked on.
-
CUDA cards don't show any improvement over basic openCL cards like the GTX's. I am using a quadro k4000 and two 580 GTX's and the 580's are exactly double the performance of the quadro. Since, as Alexey mentioned to me in an email recently:
As for the graphic cards, professional class cards from Quadro series are optimized mostly for CAD applications and double-precision calculations that are not actually used by PhotoScan. So high-end gamer cards show much better performance due to higher number of GPU cores and frequency. Tesla cards will show similar performance to the recommended gamer-class GPUs, but will cost much more.
-
Any test using dual Xeons yet? Considering a build with 2x Intel Xeon E5-2660 v3. Anyone know if Agisoft will benefit greatly from the 40 hyper threaded cores? Or should i just save a ton of money and go for a i7-5960X?
-
Hello Igor,
For dual-Xeon systems I can suggest to use 6-8 core Xeons with highest possible frequency. Using 40 cores with 2.2 GHz could be even slower than desktop six-core i7.
We haven't yet made direct comparison between 5960X and 5930K, but assume that second option may be faster due to higher CPU frequency, even the number of cores is six versus eight.
-
So in other words its a complete waste of money going with dual Xeons for Agisoft? A single CPU with maximum available clock speed is the best option? In that case the clear winner should be the i7 5930 with 6 cores and 3.5GHZ. Intel Xeon E5-1650 v3 is 6 core at 3.5GHZ and 15mb Cash should also be a contender? Throwing in 2 E5-1650 v3 will not increase performance significantly?
-
Hello Igor,
Dual-Xeon with high CPU frequency will be faster than single CPU, but as I've already said, it's better to use 6-8 core CPUs. Also Xeon-based configuration will allow you to install more than 64 GB RAM, if it is required by your tasks.
For example, dual Intel Xeon E5-2667v2 configuration was about 20-40% faster than single E5-2667v2 on Align Photos stage on our test aerial datasets.
-
Ok that clarifies it. Thank´s for the explanation.
-
I have just tested my new MacPro with the Building-Benchmark Scene:
New MacPro, 6-Core, 64GB Ram, Dual D700
Dense cloud Medium-setting 0/12 agressive:
Device 1 performance: 667.114 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Device 2 performance: 693.546 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Total performance: 1360.66 million samples/sec
Finished processing in 82.2601 sec (exit code 1)
Dense Cloud High-setting:
Device 1 performance: 819.165 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Device 2 performance: 834.194 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Total performance: 1653.36 million samples/sec
Finished processing in 434.088 sec (exit code 1)
Dense cloud Ultra-setting 0/12 mild:
Device 1 performance: 804.472 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Device 2 performance: 832.658 million samples/sec (ATI Radeon HD - FirePro D700 Compute Engine)
Total performance: 1637.13 million samples/sec
Finished processing in 3347.54 sec (exit code 1)
-
Alexey, you said dual CPU will improve Alignment stage with 20-40%. How About build dense cloud and build mesh stages. How much roughly if any improvement should i expect?
-
Hello Igor,
During our tests (note that we haven't tested a lot of dual-Xeon configurations on various projects) for dense cloud generation (except depth maps estimation) we have about 30% speed up for 2x Xeons and mesh generation stage speed up was only about 10-15%.
But I assume that values may vary depending on the CPU, scene type and even OS.
You can check the following topic, though it is quite old: http://www.agisoft.com/forum/index.php?topic=1330.0
-
Anandtech is using Photoscan to benchmark CPUs so you can pretty much find all your answers there.
Check out the recent 14core Haswell Xeons for instance: http://anandtech.com/show/8730/intel-haswellep-xeon-14-core-review-e52695-v3-and-e52697-v3/2
For the record, all of their Agisoft benchmarks were most probably recorded using pre 1.1 version, but I don't think there should be any game-changing differences between 1.0.x and 1.1.x benchamrks.
-Andrew
-
Thanks for the link. .Not very impressive results for the Xeons given the high price Xeon ES2697 is slightly slower when running it as dual CPU incredibly enough according to the test.
Looks like the CPU i already have the i7 4930 is actually a very smart choice. The top 12 core Xeon ES2690 is around 14% faster but is 10x more expensive. Glad i asked on the forum before bying as it seems investing in expensive Xeons could have been a huge loss of money for nothing. I use a lot of other software though so have to consider this to.
Disappointing though that there seems to be no way of getting a descent speed increase in Agisoft no matter what CPU you use. Aligning 1500 36 mpix images on my I7 4930 high setting right now and the processing time is 55 hours.... Investing 3000 USD in the top Xeon cpu and the align might be finished in 47 hours instead of 55. That is not nearly enough to justify the investment.
I will have to go Xeon for the RAM though. So trying to figure out which one. ES1650 6core runs at 3.5GHZ so should be fast, no tested though? Probably should just buy a single CPU and not double?
-
Hello Igor,
55 hours for 1500 is quite a lot. Are you using any preselection? Also using lower values for key-point and tie-point limits will results in faster processing.
-
Thanks for asking! Im using these settings
High setting - Diabled (not generic)
1000 tie point limit.
40 000 point limit
No image pair pre selection.
55 hours is a lot but if that is what it takes i have to wait :-) That is the projected time, usually it seems to get finished faster though but still 45 hours plus i would think. If you have suggestions on how to speed this up with out losing accuracy I'm very interested though.
-
Glad i asked on the forum before bying as it seems investing in expensive Xeons could have been a huge loss of money for nothing.
Not all Xeon CPUs are that expensive, take the E5-2620v3 that I have. Performance wise, it is about 10% less powerful as your 4930K, and it only cost $339 new. Now while the i7 is faster on single core processes, the Xeon can a max of 768GB and almost every motherboard out there that supports Xeon CPUs will do 4 x16 GPUs.
Also one other thing that a lot of people might not know. Reducing memory hard faults will speed up your processing times. What is a hard fault you might ask? Well its the reason Agisoft recommends upwards of 700GB of RAM for some scans. Take a look at the link below for an explination.
http://www.brighthub.com/computing/windows-platform/articles/52249.aspx
-
Interesting with the hard faults. I am running a huge image set now of 1500 images as i mentioned above. Its been running for 12 hours now and 0 hard faults per second is reported by Windows. CPU clock speed is 118% of normal and all cores at 99-100% so seems my system is running fine and the cooling is enough. I have 64gB DDR3 and around 24 GB is used right now.
I have 55 hours remaining to align so what i said before is wrong. Total time for me to align is 67 hours not 55 for the 1500 images. This is just one of 4 chunks of similar size. The last chunk though:--) All the others are already aligned, merged and dense cloud generated. It has been a pretty monumental processing task though . That is why I'm interested in more speed.
-
Generally you wont see that many hard faults per second but the fact that you have some here and there will slow things down. My first batch of RAM should be here tomorrow and the rest later this moth. In theory there shouldbe some speed gains by eliminating the dumping of RAM to a temp spot on the HDD. I will test this out by running the same 100 photos, 500 photos, and 1000 photo sets on 16GB, 32GB, 64GB and 128GB. Once I am done I will post my results to see if I am correct.
-
Just finished the build tonight, so optimizations are still a ways off, but:
System:
ASUS Z9PE-D8 WS
2x Xeon 2687W ES V2 (3.2ghz base, 3.8 turbo)
RAM 64GB Corsair Vengeance 1600MHZ
4x GTX 780 TI 3GB Video
2x Samsung 850 EVO SSD in RAID0
2x Seagate 3TB 7200RPM in RAID0
2x Seagate 3TB 7200RPM in RAID1
2x WD Black 1TB 7200RPM in RAID1
Building Test #1 (0 CPU Cores enabled):
Stage 1 Alignment:
Matching - 2 min 22 sec (142 sec)
Alignment - 2 sec
Stage 2 Dense Point Cloud:
Time - 25 sec
Stage 3 Model:
Processing Time - 3 min 33 sec (213 sec)
Stage 4 Texture:
UV Mapping - 23 sec
Blending - 19 sec
Total Time: 7.06 Min (424 sec)
Building Test #2 (24/32 CPU Cores enabled):
Stage 1 Alignment:
Matching - 2 min 25 sec (145 sec)
Alignment - 2 sec
Stage 2 Dense Point Cloud:
Time - 29 sec
Stage 3 Model:
Processing Time - 3 min 26 sec (206 sec)
Stage 4 Texture:
UV Mapping - 28 sec
Blending - 21 sec
Total Time: 7.18 Min (431 sec)
So, it looks like the 1600mhz ram is killing me in stage 3, it's only running triple channel, but I'm going to try some quad channel 2133 CAS 11 next week if I can get it to OC properly on this motherboard. I'll post an update if these scores improve. The chips are unlocked, so I may overclock them a mite bit to 3.4/4.0(turbo) or so and see how much it helps. They seem to run about 3.6 when Photoscan is cranking acording to CPU-Z. Not quite double the performance of 1 2687W, but from 12.74min to 7.06 is nothing to scoff at. Also doubles the performance of an i7 4960X, and this rig did not cost twice what that one would (I canabalized some parts from an old build, some things would have changed were I buying this stuff new), but price/performance I'm pretty happy. Plus cinebench R15 scores are 2176 CPU and 60.38fps in OpenGL, so that's nice.
Edit: Tweaked some things, including the memory timing which was off. Pulled 4.18min on the test.
-
This is quite amazing compare to Annadtechs tests! You are 3 times faster than there fastest rig tested.
-
This is quite amazing compare to Annadtechs tests! You are 3 times faster than there fastest rig tested.
Do the anandtech test cover also GPU performance in agisoft? I found the cpu performance graphs, but i didnt find any graphs for gpu performance, it seems they are testing GPUs only in games, is that true or did i missed the graphs/tables?
-
Yes, I'm very interested in GPU benchmarks too
-
Finally had some time to bench a few of my rigs. Here are the results:
E5-2620V3
0/12 High/Medium (alignment/dense cloud)
Device 1 performance: 763.04 million samples/sec (GeForce GTX 780)
Device 2 performance: 772.025 million samples/sec (GeForce GTX 780)
Total performance: 1535.07 million samples/sec
Finished processing in 769.993 sec (exit code 1)
E5-2620V3
0/12 High/High (alignment/dense cloud)
Device 1 performance: 849.039 million samples/sec (GeForce GTX 780)
Device 2 performance: 864.494 million samples/sec (GeForce GTX 780)
Total performance: 1713.53 million samples/sec
Finished processing in 1781.51 sec (exit code 1)
E5-2620V3
8/12 High/Medium (alignment/dense cloud)
Device 1 performance: 75.8052 million samples/sec (CPU)
Device 2 performance: 604.505 million samples/sec (GeForce GTX 780)
Device 3 performance: 620.462 million samples/sec (GeForce GTX 780
Total performance: 1300.77 million samples/sec
Finished processing in 790.722 sec (exit code 1)
E5-2620V3
8/12 High/High (alignment/dense cloud)
Device 1 performance: 75.7513 million samples/sec (CPU)
Device 2 performance: 729.364 million samples/sec (GeForce GTX 780)
Device 3 performance: 744.578 million samples/sec (GeForce GTX 780)
Total performance: 1549.69 million samples/sec
Finished processing in 1917.92 sec (exit code 1)
E5-1650V3 (Stock Speed)
0/12 High/Medium (alignment/dense cloud)
Device 1 performance: 643.007 million samples/sec (GeForce GTX 580)
Device 2 performance: 814.515 million samples/sec (GeForce GTX 780)
Total performance: 1457.52 million samples/sec
Finished processing in 594.65 sec (exit code 1)
E5-1650V3 (Stock Speed)
6/12 High/Medium (alignment/dense cloud)
Device 1 performance: 88.3696 million samples/sec (CPU)
Device 2 performance: 557.637 million samples/sec (GeForce GTX 580)
Device 3 performance: 686.651 million samples/sec (GeForce GTX 780)
Total performance: 1332.66 million samples/sec
Finished processing in 596.527 sec (exit code 1)
E5-1650V3 (Stock Speed)
6/12 High/Medium (alignment/dense cloud)
Device 1 performance: 68.545 million samples/sec (CPU)
Device 2 performance: 546.319 million samples/sec (GeForce GTX 580)
Device 3 performance: 680.935 million samples/sec (GeForce GTX 780)
Device 4 performance: 634.991 million samples/sec (GeForce GTX 780)
Total performance: 1930.79 million samples/sec
Finished processing in 594.9 sec (exit code 1)
E5-1650V3 (Stock Speed)
6/12 High/High (alignment/dense cloud)
Device 1 performance: 85.7756 million samples/sec (CPU)
Device 2 performance: 612.198 million samples/sec (GeForce GTX 580)
Device 3 performance: 749.642 million samples/sec (GeForce GTX 780)
Device 4 performance: 796.494 million samples/sec (GeForce GTX 780)
Total performance: 2244.11 million samples/sec
Finished processing in 1424.95 sec (exit code 1)
E5-1650V3 (Overclocked to 4.0GHz)
6/12 High/High (alignment/dense cloud)
Device 1 performance: 87.2289 million samples/sec (CPU)
Device 2 performance: 650.225 million samples/sec (GeForce GTX 580)
Device 3 performance: 772.937 million samples/sec (GeForce GTX 780)
Device 4 performance: 799.533 million samples/sec (GeForce GTX 780)
Total performance: 2309.92 million samples/sec
Finished processing in 1340.55 sec (exit code 1)
Ill be running more tests on the overclocked machine later this coming week. Im trying to get it to around 4.4 or 4.5GHs so we will see how things look once its completely finished.
-
Can we not, as community, all run a test of a single set and publish our times and hardware specs :-)
-
Can we not, as community, all run a test of a single set and publish our times and hardware specs :-)
The same set of data has been ran through all the benchmarks on this entire post, the 50 image building sample set that Agisoft has on the downloads page. I just ran it at two different settings high/medium and high/high to get a more accurate results.
-
So two of GTX 580 should outperform one GTX 780 for the same price?
-
So two of GTX 580 should outperform one GTX 780 for the same price?
yep
-
Nvidia Titan X Benchmark
The building sample data was used http://www.agisoft.com/downloads/sample-data/
I uses the following settings: (Same as Anand Tech but tie points set to 0 and CPU 0/12)
## Align Photos: High, disabled, 40000. 0 tie points
## Build Dense Cloud: Medium, Aggressive
## Build Model: Arbitrary, Dense, Interpolation=Enabled, 0 face count
## Build Texture: Generic, Mosaic, 4096, 1
System info
Microsoft Windows 7 Professional
Processor Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz
64GB RAM
2x Titan X 12gb cards. SLI enabled
Align photos
Finished processing in 294.814 sec (exit code 1)
Dense cloud
All CPU cores disabled in agisoft prefrences. 0/12
finished depth reconstruction in 80.37 seconds
Device 1 performance: 1004.91 million samples/sec (GeForce GTX TITAN X)
Device 2 performance: 994.448 million samples/sec (GeForce GTX TITAN X)
Total performance: 1999.36 million samples/sec
Generating dense point cloud...
selected 50 cameras in 0.22 sec
working volume: 1755x2763x1220
tiles: 1x1x1
selected 50 cameras
preloading data... done in 0.325 sec
filtering depth maps... done in 43.934 sec
preloading data... done in 7.956 sec
accumulating data... done in 1.201 sec
building point cloud... done in 0.608 sec
5706624 points extracted
Finished processing in 137.235 sec (exit code 1)
Mesh
Finished processing in 144.503 sec (exit code 1)
Build Texture
9056814 faces extracted in 32.709s
Calculating vertex colors...
processing nodes... done in 0.016 sec
calculating colors... done in 20.289 sec
Finished processing in 144.503 sec (exit code 1)
Parameterizing texture atlas...
Packing 2117 charts...
Blending textures...
blending textures... ************************************************** done in 49.94 sec
postprocessing atlas... done in 0.117 sec
Finished processing in 154.71 sec (exit code 1)
-
Ran the test on Ultra and this time the load went up to around 80% on the cards. Much better performance now. Under boost the clock speed stayed at between 1300-1316mhz on the 2 EVGA Titan X Superclocked cards.
finished depth reconstruction in 2004.43 seconds
Device 1 performance: 1326.94 million samples/sec (GeForce GTX TITAN X)
Device 2 performance: 1279.33 million samples/sec (GeForce GTX TITAN X)
Total performance: 2606.27 million samples/sec
-
Nvidia Titan X Benchmark
The building sample data was used http://www.agisoft.com/downloads/sample-data/
I uses the following settings: (Same as Anand Tech but tie points set to 0 and CPU 0/12)
## Align Photos: High, disabled, 40000. 0 tie points
## Build Dense Cloud: Medium, Aggressive
## Build Model: Arbitrary, Dense, Interpolation=Enabled, 0 face count
## Build Texture: Generic, Mosaic, 4096, 1
System info
Microsoft Windows 7 Professional
Processor Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz
64GB RAM
2x Titan X 12gb cards. SLI enabled
Align photos
Finished processing in 294.814 sec (exit code 1)
Dense cloud
All CPU cores disabled in agisoft prefrences. 0/12
finished depth reconstruction in 80.37 seconds
Device 1 performance: 1004.91 million samples/sec (GeForce GTX TITAN X)
Device 2 performance: 994.448 million samples/sec (GeForce GTX TITAN X)
Total performance: 1999.36 million samples/sec
Generating dense point cloud...
selected 50 cameras in 0.22 sec
working volume: 1755x2763x1220
tiles: 1x1x1
selected 50 cameras
preloading data... done in 0.325 sec
filtering depth maps... done in 43.934 sec
preloading data... done in 7.956 sec
accumulating data... done in 1.201 sec
building point cloud... done in 0.608 sec
5706624 points extracted
Finished processing in 137.235 sec (exit code 1)
EDIT
Ran the test on Ultra and this time the load went up to around 80% on the cards. Much better performance now. Under boost the clock speed stayed at between 1300-1316mhz on the 2 EVGA Titan X Superclocked cards.
finished depth reconstruction in 2004.43 seconds
Device 1 performance: 1326.94 million samples/sec (GeForce GTX TITAN X)
Device 2 performance: 1279.33 million samples/sec (GeForce GTX TITAN X)
Total performance: 2606.27 million samples/sec
Mesh
Finished processing in 144.503 sec (exit code 1)
Build Texture
9056814 faces extracted in 32.709s
Calculating vertex colors...
processing nodes... done in 0.016 sec
calculating colors... done in 20.289 sec
Finished processing in 144.503 sec (exit code 1)
Parameterizing texture atlas...
Packing 2117 charts...
Blending textures...
blending textures... ************************************************** done in 49.94 sec
postprocessing atlas... done in 0.117 sec
Finished processing in 154.71 sec (exit code 1)
-
Ran another test this time with SLI disabled and only using 1 card in Agisoft. All CPU cores disabled and exactley the same settings as before. High setting on dense cloud instead om medium that i used for the first test.. Pretty good performance i would say! The GPU load went up to 95% now for some reason. Still hitting 1316mhz as max clock speed so no change there.
finished depth reconstruction in 541.615 seconds
Device 1 performance: 1534.46 million samples/sec (GeForce GTX TITAN X)
Total performance: 1534.46 million samples/sec
-
Another test on High dense cloud setting. This time SLI diabled but both GPUs enabled in Agisoft. For some reason the performance drops a bit per card compared to using a single card.. Wonder why that is? I don´t think its heat related as the cards still run on 1316 mhz, same as running a single card. Titan X seems to be around 40%-60% faster then 980 in Agisoft and only use 250W. So if you are building a serious hard core system Titan X might be worth it after all even for Agisoft?
2x Titan X EVGA Superclocked. Dense cloud High.
finished depth reconstruction in 322.303 seconds
Device 1 performance: 1353.73 million samples/sec (GeForce GTX TITAN X)
Device 2 performance: 1350.39 million samples/sec (GeForce GTX TITAN X)
Total performance: 2704.12 million samples/sec
1x Titan X EVGA Superclocked.Dense cloud High.
finished depth reconstruction in 541.615 seconds
Device 1 performance: 1534.46 million samples/sec (GeForce GTX TITAN X)
Total performance: 1534.46 million samples/sec
-
Interesting stuff! I have a few questions:
What sort of performance could I expect from an Nvidia 970m with 6GB DDR5 Ram?
How can I test that photoscan is using the GPU?
How many CPU cores should I disable for this card? Just 1?
Should I use the Intel 4600 gpu, built into my i7? As far as I know, it uses system RAM.
I have http://www.xoticpc.com/sager-np8278s-clevo-p170sma-p-6985.html
Thank-you!
-
You can check the performance of the card in the Agisoft console. Its reported there once the dense cloud has finished processing. The recommnedation is to diable 1 core per GPU. So in your case disable 1 core. By default one core should already be disabled. For fast cards such as 980 and Titans the recomendation is to disable all cores as the GPU is faster than the CPU, at least that is how i have understod it.
No idea what performance your card has. 980 is around 900 milion samples per sec i have herd so a bit less than that. Test it and report back. You can also disable all cores and test the same data set again and see if you get an iprovement.
-
Will do!
My laptop also has a build in Intel HD 4600 gpu. Should I enable this as well?
-
You can check the performance of the card in the Agisoft console. Its reported there once the dense cloud has finished processing. The recommnedation is to diable 1 core per GPU. So in your case disable 1 core. By default one core should already be disabled. For fast cards such as 980 and Titans the recomendation is to disable all cores as the GPU is faster than the CPU, at least that is how i have understod it.
No idea what performance your card has. 980 is around 900 milion samples per sec i have herd so a bit less than that. Test it and report back. You can also disable all cores and test the same data set again and see if you get an iprovement.
Also, which numbers should I be looking at? Testing now. I already noticed that Photoscan was using the Intel4600, instead of the 970m. Changed that in the NVIDIA control panel, and it's running off the 970m now. I've disabled 2 cores, and am testing with 52 cameras, that are geo-referenced. Testing a high accuracy dense cloud. I'll disable only 1 core next, and see what the difference is.
[UPDATE] With 2 cores disabled it completes the task in 1991.04 seconds. With 1 core disabled it completes the task in 1837.46 seconds.
-
Try to disable all the cores and see what happens. How many million samples per second do you get? The console will report that is you scroll up a bit.
-
With all cores disabled, and 53 cameras it builds a high quality dense cloud in 1722 seconds, and at 595.855 samples per second.
With 1 core disabled, and 53 cameras it builds in 2075 seconds, at 539.855 samples per second.
Does this mean that my GPU is faster than my CPU? Should I leave all cores disabled?
-
Yes i think you have a fast enough GPU to leave all cores disabled. You seem to get better performance that way. Very good result for a laptop.
-
Wow! The new 900m series is within comparable distance to the desktop cards. I've been very impressed with its gaming ability so far.
Can you explain to me how the processing works? I see that even with all cores disabled the CPU still gets used.
-
Sorry , only know how to use the software but sill learning. The developers have to answer that. Alexey maybe can explain a bit more?
-
the CPU need to prepare data for GPU... adn it need to move it to GPU.
The M version of Nvidia is now almost at 80-90% to the desktop counterpart... seach on http://www.anandtech.com/tag/gpus
-
the CPU need to prepare data for GPU... adn it need to move it to GPU.
The M version of Nvidia is now almost at 80-90% to the desktop counterpart... seach on http://www.anandtech.com/tag/gpus
Great, thanks!
Ja, I've seen how close the performance gap is getting. Great news for us mobile workers!
-
Dear Community ;)
I have started to crate a table for recording the results of this thread. Maybe you could create there a good overview for all Agisoft Users. (maybe you could also write down your experience in other wiki topics/pages)
http://wiki.agisoft.com/wiki/Hardware
-
hi MWittenebel
here its almost al together so just look from beginning for numbers...
-
Dear Community ;)
I have started to crate a table for recording the results of this thread. Maybe you could create there a good overview for all Agisoft Users. (maybe you could also write down your experience in other wiki topics/pages)
http://wiki.agisoft.com/wiki/Hardware
It would be even better if you said what is the meaning of the numbers, because it's not obvious that they are million samples/sec. To be honest it took me sometime to figure it out.
I will try to launch a test with our new dual k5200 setup to add a new input in the table.
Best,
-
Hi, I am new to the forum, this will be my very first post and i choose to revive this "necro" thread in the hope that it would be updated every time. Since i believe it is one of the utmost important topic for processing in agisoft. Knowledge is power, and processing power IS power.
Anyway, due to inherent randomness in workload and workflow i suggest from now on to use PugetBench data and flow to standardize the result and makes it apple to apple comparison. Here is my result on my laptop with AMD Ryzen 5980HS 32GB RAM and RTX3080 Laptop (16GB VRAM) running on Windows 10, internal GPU disabled in Metashape..
PugetBench Standard
Agisoft Metashape Professional Version: 1.8.5
CPU: AMD64 Family 25 Model 80 Stepping 0, AuthenticAMD
Number of GPUs Found: 2
GPU Model(s): AMD Radeon(TM) Graphics (gfx90c), NVIDIA GeForce RTX 3080 Laptop GPU
Project: Rock Model
Align Photos: 32.0
Build Depth Maps: 17.9
Build Dense Cloud: 16.0
Build Mesh from Dense Cloud: 27.9
Decimate Mesh: 4.6
Build Texture: 24.7
Total Processing Time: 123.1
Project: School Map
Align Photos: 27.7
Build Depth Maps: 54.9
Build Dense Cloud: 41.4
Build Tiled Model: 241.6
Build DEM: 2.7
Build Orthomosaic: 48.1
Total Processing Time: 416.4
PugetBench Extended
Agisoft Metashape Professional Version: 1.8.5
CPU: AMD64 Family 25 Model 80 Stepping 0, AuthenticAMD
Number of GPUs Found: 2
GPU Model(s): AMD Radeon(TM) Graphics (gfx90c), NVIDIA GeForce RTX 3080 Laptop GPU
Project: Park Map
Align Photos: 668.5
Build Depth Maps: 891.4
Build Dense Cloud: 806.5
Build Tiled Model: 3240.0
Build DEM: 36.8
Build Orthomosaic: 676.1
Total Processing Time: 6319.3
Project: School Model
Align Photos: 484.1
Build Depth Maps: 515.6
Build Dense Cloud: 989.1
Build Mesh from Dense Cloud: 192.2
Decimate Mesh: 44.8
Build Texture: 335.2
Total Processing Time: 2561.0
Please continue this thread. If i found positive feedbacks I will also post some other hardware combination results mainly from my other PC.