Running Linux from an M.2 disk

[AlveKatt]

I am about to order the computer, tomorrow, that people here have been helping me plan.

 

I just realized there is at least one very important question left: M.2 being a rather new technology, can I use it as a System disk when the system is Running Linux?

 

The plan is to use the latest version of Linux Mint, 17.03, which uses the Ubuntu 14.04 core.

 

If the answer is no I will settle for a normal SSD. While I do not doubt that windows 10 is a OS with good features I have big issues with microsofts business model of forced updates, dubious phone home DRM and general lack of respect of its users privacy.

[Stefan1024]

If it's a SATA SSD yes (Samsung 850 Evo) , an NVMe one (Samsung 950) migth be tricky.

[AlveKatt]

Ok, that was exactly what I didn't want to hear. The samsung 950 was my planned one.

 

Could you define "tricky" a bit more?

[Technous285]

I have to ask one thing: What chipset/CPU family is the new system going to be running on?

Skylake/100-series boards (eg: Z170's) are JUST getting some Linux support with Kernel 4.4 and at the moment distros with that are still few and far between. Linux Mint 17.3 is still running Kernel 3.19 whilst (as an example) KaOS 2016.03 (March 2016 release) is on Kernel 4.4.5.

[AlveKatt]

13 minutes ago, Technous285 said:

 

I have to ask one thing: What chipset/CPU family is the new system going to be running on?

Skylake/100-series boards (eg: Z170's) are JUST getting some Linux support with Kernel 4.4 and at the moment distros with that are still few and far between. Linux Mint 17.3 is still running Kernel 3.19 whilst (as an example) KaOS 2016.03 (March 2016 release) is on Kernel 4.4.5.

 

Intel Core I7 5820K with 28 PCIe lanes.

Asus X99-A Motherboard.

 

Here is a complete list: https://www.komplett.se/k/shoplist.aspx?mode=receive&si=1675149&su=B66D99DB-DCF8-410F-94FE-5F1E1968EE12

 

But, sounds like I should forgo the M.2 Drive then?

[Stefan1024]

1 minute ago, AlveKatt said:

 

Intel Core I7 5820K with 28 PCIe lanes.

Asus X99-A Motherboard.

 

Here is a complete list: https://www.komplett.se/k/shoplist.aspx?mode=receive&si=1675149&su=B66D99DB-DCF8-410F-94FE-5F1E1968EE12

 

But, sounds like I should forgo the M.2 Drive then?

Do you have a need for a super fast NVMe drive? Will you notice the difference in the applications you are using?

[AlveKatt]

I don't know. My current problem is editing high poly objects in Blender is lagging. I am in the middle of making assets to a personal animation project that I can't finish on my old computer because, in no particular order, I can't do texture painting on high poly models in the 3d view, moving bones in the character rigs lags something fierce and trying to use the sculpt tools on half complex objects results in Blender bogging down to a halt. Painting and sculpting is impossible with any lag whatsoever, the results of every stroke gets extremely unpredictable, basically, you get a mess.

 

Maybe I'm over reacting, my current system is nine years old:

I just really want to animate, sculpt and paint. I love sculpting with Polymer clays in the real world, and I really want the drawing tablet sculpting tools in Blender to be viable.

 

But fluid positioning of bones in high poly characters is the most important part. People in the Blender forums claim that that particular part isn't really helped by the GPU, so I thought I would get the M.2 just in case.

[Stefan1024]

14 minutes ago, AlveKatt said:

- snip -

Please quote people so they get a notification.

 

Monitor the utilisation of your old PC during modeling. I assume your CPU and RAM is maxed out. The 5820X is an extrem performance boost compared to your old PC and the SSD is usually not the bottleneck in CAD modeling. As long as you have an SSD and not a HDD it should run just fine.

 

Also a Titan X is jut a waste of money compared to the GTX980Ti, even 3D modeling can't use more than 6 GBytes VRAM as far as I know.

[AlveKatt]

1 minute ago, Stefan1024 said:

Please quote people so they get a notification.

 

Monitor the utilisation of your old PC during modeling. I assume your CPU and RAM is maxed out. The 5820X is an extrem performance boost compared to your old PC and the SSD is usually not the bottleneck in CAD modeling. As long as you have an SSD and not a HDD it should run just fine.

 

Also a Titan X is jut a waste of money compared to the GTX980Ti, even 3D modeling can't use more than 6 GBytes VRAM as far as I know.

Does the 980TI have double precision? I saw a Linus video where he said the Titan X was the only card that had that.

Also, making 3d-animated shorts aren't like making a game, you don't need, or want, to optimize everything to run in Real time.

 

But thanks, I think I should do a little research about 980TI vs Titan X in my specific use scenario.

[Stefan1024]

1 minute ago, AlveKatt said:

Does the 980TI have double precision? I saw a Linus video where he said the Titan X was the only card that had that.

Also, making 3d-animated shorts aren't like making a game, you don't need, or want, to optimize everything to run in Real time.

 

But thanks, I think I should do a little research about 980TI vs Titan X in my specific use scenario.

Maxwell in gneral is terrible at DP compute. The Titan Black was better at DP than the 7xx series, but the Titan X is not compared to 9xx.

Also all GPUs can do DP, but some at 1/2 the SP speed and others like the Titan X at 1/64.

 

Quote

To refresh your memory, the Titan X comes with 8 billion transistors, 3072 CUDA cores, 7 TFLOPS in single-precision and only 0.2 TFLOPS in double-precision performance. The Titan Black has single-precision performance of 5.1 TFLOPS and double-precision performance of 1.3 TFLOPS.

 

[AlveKatt]

26 minutes ago, Stefan1024 said:

Maxwell in gneral is terrible at DP compute. The Titan Black was better at DP than the 7xx series, but the Titan X is not compared to 9xx.

Also all GPUs can do DP, but some at 1/2 the SP speed and others like the Titan X at 1/64.

 

 

Ok, I am sitting here thinking there has to be a reason for the Titan Xs existence. I am trying to figure out how to find benchmark tests for my specific user case. The extra Cuda cores will speed up rendering on the card. My Current project uses Blender INternal rendering Rather than the Cycles rendering engine, so I can't render this project on the card, But I want the option to render cycles projects too.

 

I need to figure out how important Double precision is too. I thought it would help in scenes with grander scales, where I animate complex characters in environments with objects whose relative size is a lot bigger, with distances that are large. This is important because I am a space opera fanatic, and I need to be able to work in a scene that has that grand scale. But to be honest, I realize don't really know what double precision does.

[Stefan1024]

3 minutes ago, AlveKatt said:

Ok, I am sitting here thinking there has to be a reason for the Titan Xs existence. I am trying to figure out how to find benchmark tests for my specific user case. The extra Cuda cores will speed up rendering on the card. My Current project uses Blender INternal rendering Rather than the Cycles rendering engine, so I can't render this project on the card, But I want the option to render cycles projects too.

 

I need to figure out how important Double precision is too. I thought it would help in scenes with grander scales, where I animate complex characters in environments with objects whose relative size is a lot bigger, with distances that are large. This is important because I am a space opera fanatic, and I need to be able to work in a scene that has that grand scale. But to be honest, I realize don't really know what double precision does.

It is made for market segentation to maximize margin. But also there are GPGPU applications that just need a lot of VRAM. But they are rare.

 

When representing frictional numbers binary you have to approximate them. You can do it "normaly" ( float, normal precision, 7 digit, SP) or with a better accuracy ( double, double precision, 15 digit, DP). For graphical stuff, SP is enougth as your eye can see the difference. For most compute stuff that's also true. But sometimes you like to ran long high precise calculation and you don't get awa with SP.

 

Do your research, it's important if your applications use SP or DP.

[AlveKatt]

9 minutes ago, Stefan1024 said:

It is made for market segentation to maximize margin. But also there are GPGPU applications that just need a lot of VRAM. But they are rare.

 

When representing frictional numbers binary you have to approximate them. You can do it "normaly" ( float, normal precision, 7 digit, SP) or with a better accuracy ( double, double precision, 15 digit, DP). For graphical stuff, SP is enougth as your eye can see the difference. For most compute stuff that's also true. But sometimes you like to ran long high precise calculation and you don't get awa with SP.

 

Do your research, it's important if your applications use SP or DP.

It seems Blender uses Single precision. Sorry for wasting your time, but it's a lot to take in right now. Some threads I read say that complex scenes can eat up those 12 gigabytes of memory pretty fast. The question I need to consider is if I as a single creator could do such complex scenes. I'm frankly not sure I am that good. Getting the titan X might be a way of future proofing if I ever find myself working with a team of hobbyists like myself. Getting the 980ti might mean I could put what I save against getting one of those awesome bent screens with G-sync and a lot wider workspace, that would be a boon with the way Blenders interface works, well, still would need to save a couple of thousand SEK more, but still.

 

Argh, decision angst! Would love your further input, but the next reply from me might be a while. I should really distract myself for a while, let the subconsious mull it over. I need a break, or I will break. Thank you!

[AlveKatt]

23 hours ago, Stefan1024 said:

It is made for market segentation to maximize margin. But also there are GPGPU applications that just need a lot of VRAM. But they are rare.

 

When representing frictional numbers binary you have to approximate them. You can do it "normaly" ( float, normal precision, 7 digit, SP) or with a better accuracy ( double, double precision, 15 digit, DP). For graphical stuff, SP is enougth as your eye can see the difference. For most compute stuff that's also true. But sometimes you like to ran long high precise calculation and you don't get awa with SP.

 

Do your research, it's important if your applications use SP or DP.

 

I'm pretty sure going to go for the Titan x. Because a lot of polygons take up a lot of video ram according to the people at the blender artists forum.

And I realized that when the time comes to animate I want all the models I have made will be in the same scene. Including interiors. And because I love curves, and smooth curves makes for an extreme rize in polygons. Even the suppposed to be human made battleship I am working on has curves, not like the human ships in Halo that are all angular faces and looking like machine guns.

 

http://alvekatt.deviantart.com/art/Battleship-WIP2-598227489

http://alvekatt.deviantart.com/art/Biomechanoid-ship-596422132

http://alvekatt.deviantart.com/art/Feliniod-alien-full-body-595214101

http://alvekatt.deviantart.com/art/Xenomorph-opera-563937863

 

The battleship interior will need to have a lot of curves too, to fit the outside esthetic, I want the viewer to know you are inside that particular ship when you see it, even if the sequentiality of the scenes doesn't make it apparent. The biomechanoid ship is, well, Organic.

 

Thank you so much for all the help.

[Stefan1024]

4 hours ago, AlveKatt said:

 

I'm pretty sure going to go for the Titan x. Because a lot of polygons take up a lot of video ram according to the people at the blender artists forum.

And I realized that when the time comes to animate I want all the models I have made will be in the same scene. Including interiors. And because I love curves, and smooth curves makes for an extreme rize in polygons. Even the suppposed to be human made battleship I am working on has curves, not like the human ships in Halo that are all angular faces and looking like machine guns.

 

http://alvekatt.deviantart.com/art/Battleship-WIP2-598227489

http://alvekatt.deviantart.com/art/Biomechanoid-ship-596422132

http://alvekatt.deviantart.com/art/Feliniod-alien-full-body-595214101

http://alvekatt.deviantart.com/art/Xenomorph-opera-563937863

 

The battleship interior will need to have a lot of curves too, to fit the outside esthetic, I want the viewer to know you are inside that particular ship when you see it, even if the sequentiality of the scenes doesn't make it apparent. The biomechanoid ship is, well, Organic.

 

Thank you so much for all the help.

You are wellcome.

 

The models looks a nice, but a bit wired to me. But if you have a lot of them you may need more than 6 GByte (there are also 390x with 8 GByte for 1/3 of the price).