Show Posts
This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.
All of my CPUs were always undervolted and I never had problems with calculation errors even during several hours of computing.
It is several times higher probabilty of errors in the system RAM modules affected with cosmic and electromagnetic radiation after several hours if modules don't have ECC correction. In that case money saved on CPU cooling can be used to buy RAM modules with ECC correction, like server computers have. Ryzen CPUs support this feature.
In case of exaggerated concerns and be 1000% sure, you can always leave a slightly higher voltage value than it is necessary for the proper operation of the processor. But it is still huge difference and big save in energy and heat compared to what we get from the factory with default settings. Manufacturers are keeping huge reserve because there is no time to test all chips to exact voltage needs.
Chips used in smartphones have much much less reserves in voltages compared to desktop CPUs, and are working several days without errors.
Undervolting is safe procedure, because it is about to find the lowest voltage which is not causing freezing and not the first one which causes issues. It is also adviced keep 0.02V reserve after finding the right voltage value.
Using ECO mode only helps to decrease power consumption and temperatures, but it also masivelly decrease performance because it is not based on undervolting principles.
https://youtu.be/RlMq1VEWNIM?t=84 that cinenebch is good representation how multicore performance dropped after eco mode enable.
What is even better is lock multicore frequencies to certain fixed value, e.g. 4.4/4.6/4.8GHz and do not use automatic boost single core frequencies...this makes undervolting much easier and more effective.
Finding the lowest voltage for CPU is much easier in ryzen master app then adjusting values in BIOS and restarting computer after each change. So it is 100% worth to do, otherwise we are significantly overpaying the performance. With undervolting we can also save money because much expensive water cooling solutions is not needed.
Hi,
AMD 7xxx is better choice, because 13900k is mix of 8 performance cores and 16 efficient cores(without multithreading) and to match AMD performance it needs to be clocked almost to max boost frequencies, which produces a lot of heat.
Even better choice is AMD 7950X(non 3D), because that extra 64MB cache will highly probably not improve performance in metashape at all.
That cache is good only for certain specific games, which are heavily vectorizing data in CPU caches.
To get maximum potential performance you will need to undervolt the CPU in bios or in AMD software, because on default you can easily stuck on ~base multicore frequencies, high temperatures and high power consumption ...and that means lower performance in the end. This applies to all modern CPUs in the last ~10 years.
This may seem like a minimal thing to look for, but the change of "Build Dense Cloud" to "Build Point Cloud" is fine, but now the order in the Workflow dropdown is out of order it seems and can be confusing. it is nice to have the workflow items in order of common workflow process order. Now Build Point Cloud is below Build Tiled Model.
Dear smiller,
Thank you for your feedback.
Usually, when building a model and a tiled model, users can skip building a point cloud and use depth maps as source data. Users who need a point cloud as a result usually process aerial data, after that they build DTM/DSM, Orthomosaic. Therefore, the workflow is divided into such sections and the order of commands has been changed.
