Sony made a new depth sensor for smart phone (https://www.sony.net/SonyInfo/News/Press/201706/17... which they are going to implement in their next Flagship. I think once it gets depth sensor the quality of scans could be awesome. Absolutely love sony and how they innovate and think differently.
Serious question: Why is "AnandTech" still using a years old version of Agisoft Photoscan (v1.0) instead of the latest version which is faster thanks to improved multi-threading, has more processing functions running through the OpenCL path (which obviously can run on CPUs & GPUs) and now also has a CUDA path for NVidia GPUs ? This renders this specific benchmark totally useless. Don't you use the latest HW drivers available at the time of review? Yeah? Then why don't you use the latest version of a software/benchmark/whatever when there has been several new versions released in the past years?
Nah..after a certain amount of time it's totally ridiculous to use an obsolete version of a software. Everybody would be up in arms if they did the same with the games used in their benchmarking suite. Will anybody accept if they only ever used the launch version of a game even 3 years later when said game has been patched dozens of times since its release? Nope
I got a new version of Photoscan a couple of weeks ago, and I've been on the road since. I plan to run both old and new side by side for a while until we get a good amount of data. But to answer the first question, we initially (from 2014) used a special version made just for us that allowed us to automate the testing. Since then we dropped out of contact as there wasn't any real need to update the benchmark (as some benchmarks like CineBench R15 are super old when Cinema 4D R19 is out), but it has got on my radar again, and thankfully our contact at Agisoft was still working there. Part of my to-do list after IFA is to script that up: the mechanism is slightly different to before, so I'll have to optimize around some upcoming projects.
'Because Ian is a shill' - this week I'm shilling for Intel, eh? The other week it was AMD. If you're going to call out that we have bad practices, perhaps try something original.
Especially when I can state there are over 80 different CPU tests in our database. More than half have some level of threading/multithreading. Agisoft specifically has different stages of thread/frequency/cache/interconnect bottlenecks, as there are four separate stages, each with their own requirements.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
11 Comments
Back to Article
Threska - Saturday, September 2, 2017 - link
Since one can't patent math this should be showing up in more places.wiineeth - Saturday, September 2, 2017 - link
Sony made a new depth sensor for smart phone (https://www.sony.net/SonyInfo/News/Press/201706/17... which they are going to implement in their next Flagship. I think once it gets depth sensor the quality of scans could be awesome.Absolutely love sony and how they innovate and think differently.
Dr. Swag - Saturday, September 2, 2017 - link
Think differentJudgeDredd - Saturday, September 2, 2017 - link
Serious question: Why is "AnandTech" still using a years old version of Agisoft Photoscan (v1.0) instead of the latest version which is faster thanks to improved multi-threading, has more processing functions running through the OpenCL path (which obviously can run on CPUs & GPUs) and now also has a CUDA path for NVidia GPUs ? This renders this specific benchmark totally useless. Don't you use the latest HW drivers available at the time of review? Yeah? Then why don't you use the latest version of a software/benchmark/whatever when there has been several new versions released in the past years?wpwoodjr - Saturday, September 2, 2017 - link
To maintain consistency with past reviewsJudgeDredd - Saturday, September 2, 2017 - link
Nah..after a certain amount of time it's totally ridiculous to use an obsolete version of a software. Everybody would be up in arms if they did the same with the games used in their benchmarking suite. Will anybody accept if they only ever used the launch version of a game even 3 years later when said game has been patched dozens of times since its release? NopeIan Cutress - Saturday, September 2, 2017 - link
I got a new version of Photoscan a couple of weeks ago, and I've been on the road since. I plan to run both old and new side by side for a while until we get a good amount of data. But to answer the first question, we initially (from 2014) used a special version made just for us that allowed us to automate the testing. Since then we dropped out of contact as there wasn't any real need to update the benchmark (as some benchmarks like CineBench R15 are super old when Cinema 4D R19 is out), but it has got on my radar again, and thankfully our contact at Agisoft was still working there. Part of my to-do list after IFA is to script that up: the mechanism is slightly different to before, so I'll have to optimize around some upcoming projects.JudgeDredd - Saturday, September 2, 2017 - link
Thanks a lot for the update Ian. Can't wait for you guys to run some benches on the latest version.Lolimaster - Saturday, September 2, 2017 - link
To make intel look good (the excuse being their benchmark compare tool) focusing mostly on single thread/turbo.Ian Cutress - Saturday, September 2, 2017 - link
'Because Ian is a shill' - this week I'm shilling for Intel, eh? The other week it was AMD. If you're going to call out that we have bad practices, perhaps try something original.Especially when I can state there are over 80 different CPU tests in our database. More than half have some level of threading/multithreading. Agisoft specifically has different stages of thread/frequency/cache/interconnect bottlenecks, as there are four separate stages, each with their own requirements.
Tigran - Saturday, September 2, 2017 - link
Ian, did you know you look like Stellan Skarsgård? ))