Jump to content

the 100,000 dollar pc??

JCBiggs

Where is that thing!!     I watched and I'm interested in knowing how they are gonna get 8k display out to the editors desks.  Assuming they put the PC in the server room. 

 

To bad you couldnt have multiple thunderbolt 3.0 cards and use the optical thunderbolt cables and docks for USB and video.  Then you just need 6 cables running from the server room to the workstation den.  wouldnt be 8k but i think it would do 5?

 

 

]

Link to comment
Share on other sites

Link to post
Share on other sites

The $100K machine currently resides in Linus's office.

 

I don't know where they plan on putting it, maybe in that conference room or in the editing den.

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Crunchy Dragon said:

The $100K machine currently resides in Linus's office.

 

I don't know where they plan on putting it, maybe in that conference room or in the editing den.

Im in a company that will be growing its number of engineers dramatically in the next 2 years. Im very interested in that build.   I dont think we would ever go so far as 100k,  but Id take a look at a scaled down version for sure. 

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, JCBiggs said:

Im in a company that will be growing its number of engineers dramatically in the next 2 years. Im very interested in that build.   I dont think we would ever go so far as 100k,  but Id take a look at a scaled down version for sure. 

A scaled down version definitely would be cool.

 

I've been toying with the idea of investing in a bunch of LGA 1366 server hardware to make virtualized gaming rigs the way Linus has done on a few occasions, but it's pretty pricey even when using X58 hardware.

 

Maybe if I ever need that much editing power(or ever have that many editors), I would seriously consider a scaled down version of what they have going on.

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

6 minutes ago, Crunchy Dragon said:

A scaled down version definitely would be cool.

 

I've been toying with the idea of investing in a bunch of LGA 1366 server hardware to make virtualized gaming rigs the way Linus has done on a few occasions, but it's pretty pricey even when using X58 hardware.

 

Maybe if I ever need that much editing power(or ever have that many editors), I would seriously consider a scaled down version of what they have going on.

is there any reason the Thunderbolt 3.0 dock and fiber optic cables wouldnt work on a smaller system that didnt need 8k displays?   it seems like it might actually be simpler since you wouldn't need the daughter board or have to working about the 'imou'?  groups 

 

I like virtualization, but i hate working in a virtual environment.  working in an environment with a bare metal hypervisor that doesnt feel like a remote desktop would be great.   Plus 1 box to maintain/upgrade instead of 4-8.   The bigger issue for me is the shared CPU capacity.  

Link to comment
Share on other sites

Link to post
Share on other sites

honestly, just making 7 workstations would be better.

15k each can get you a 16c EPYC /10c Xeon+ nvidia GP100/p6000.

more like a 24c epyc actually.

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, JCBiggs said:

is there any reason the Thunderbolt 3.0 dock and fiber optic cables wouldnt work on a smaller system that didnt need 8k displays?   it seems like it might actually be simpler since you wouldn't need the daughter board or have to working about the 'imou'?  groups 

 

I like virtualization, but i hate working in a virtual environment.  working in an environment with a bare metal hypervisor that doesnt feel like a remote desktop would be great.   Plus 1 box to maintain/upgrade instead of 4-8.   The bigger issue for me is the shared CPU capacity.  

Thunderbolt 3 docks and cables should work on any system that supports the technology.

Linus is using a server board though, and server hardware is not meant to be messed around with to the extent that consumer hardware is. Even getting Thunderbolt 3 AIC support on a server board would be a challenge in and of itself.

 

It may be simpler in theory, but in practice, Linus's whole plan with the daughterboard and running the GPUs on that makes the most sense to me.

 

Personally, when you've got that much processing power, shared CPU capacity doesn't become much of a problem.

Just have enough cores dedicated to the hypervisor to run that, and dedicate the others to the VMs in the capacity you need.

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Firewrath9 said:

honestly, just making 7 workstations would be better.

15k each can get you a 16c EPYC /10c Xeon+ nvidia GP100/p6000.

more like a 24c epyc actually.

True, but this is Linus Tech Tips. xD

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, Crunchy Dragon said:

Thunderbolt 3 docks and cables should work on any system that supports the technology.

Linus is using a server board though, and server hardware is not meant to be messed around with to the extent that consumer hardware is. Even getting Thunderbolt 3 AIC support on a server board would be a challenge in and of itself.

 

It may be simpler in theory, but in practice, Linus's whole plan with the daughterboard and running the GPUs on that makes the most sense to me.

 

Personally, when you've got that much processing power, shared CPU capacity doesn't become much of a problem.

Just have enough cores dedicated to the hypervisor to run that, and dedicate the others to the VMs in the capacity you need.

I actaully meant i prefer the shared CPU capacity.  you  know why have 7 or 8 i7 8700s sitting around idling for 20 hours  a day, when i can have  a couple super high end cpus that will actually give you more performance, but spread it out over the day to the ones that are actually using it (since in my line of work its unlikely 2 people will be cranking their cpu at the same time for several hours)   this should lead to better peformance with more efficiency,

 

I see what you mean  when it comes to the issues with thunderbolt.  Id almost be willing to bet you could get away with a consumer board though.  again, smaller scale.   hell id be interested in trying this for the new pc im building for the house. 

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, JCBiggs said:

I see what you mean  when it comes to the issues with thunderbolt.  Id almost be willing to bet you could get away with a consumer board though.  again, smaller scale.   hell id be interested in trying this for the new pc im building for the house. 

Oh, definitely. If you scale it down, Thunderbolt isn't a big problem.

 

Just as long as you can get the hardware necessary to run whatever you need to comfortably.

 

I also like the idea of having thin clients all over a house, rather than multiple towers. Might be something worth looking into.

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, Crunchy Dragon said:

Oh, definitely. If you scale it down, Thunderbolt isn't a big problem.

 

Just as long as you can get the hardware necessary to run whatever you need to comfortably.

 

I also like the idea of having thin clients all over a house, rather than multiple towers. Might be something worth looking into.

I tried thin clients once. i hated them.  I have been thinking about using them in the walls for smart house controls, but that would be about the extent.  I like the idea of the actual PC experience  the way you guys are doing.  its really just doing linus's personal rig but with the twist of multiple stations.      

 

Im not that familiar with unraid, or hypervisors to be honest, but cant you let the hypervisor control how much of the cpu you get automatically?  lets say you have 56 cores, and your the only one working today, could you get all 56?  

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, JCBiggs said:

Im not that familiar with unraid, or hypervisors to be honest, but cant you let the hypervisor control how much of the cpu you get automatically?  lets say you have 56 cores, and your the only one working today, could you get all 56?  

If you're in charge of the hypervisor, theoretically.

 

I forget if Linus mentioned that in the last update video, but I feel like he did say that.

 

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, Crunchy Dragon said:

If you're in charge of the hypervisor, theoretically.

 

I forget if Linus mentioned that in the last update video, but I feel like he did say that.

 

10-4.  id just like to see it scale.  give each user a couple of cores for basic operation and then when someone is crunching toolpaths of doing FEA they ramp up into more/all the cores.  then maybe split them with a second person if they start doing work.   anyways.  think Im gonna look into this more.   be a fun home project. 

Link to comment
Share on other sites

Link to post
Share on other sites

4 minutes ago, JCBiggs said:

10-4.  id just like to see it scale.  give each user a couple of cores for basic operation and then when someone is crunching toolpaths of doing FEA they ramp up into more/all the cores.  then maybe split them with a second person if they start doing work.   anyways.  think Im gonna look into this more.   be a fun home project. 

That would definitely be more efficient, which is why I don't see Linus overlooking something like that.

 

He's all about efficiency and getting crap done, even though he doesn't necessarily portray that on camera.

Quote or tag me( @Crunchy Dragon) if you want me to see your reply

If a post solved your problem/answered your question, please consider marking it as "solved"

Community Standards // Join Floatplane!

Link to comment
Share on other sites

Link to post
Share on other sites

good to see they have an x16 card not just the x4 linus is using!  That could turn one PCIe x16 into 4 thunderbolt 3.0 AICs (which are x4 each)  

https://www.onestopsystems.com/product/pcie-x16-gen-3-host-target-kit-6810

 

crap... you know i completely forgot about the thunderbolt header you have to connect. 

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 weeks later...
On 9/13/2018 at 7:06 PM, JCBiggs said:

Im not that familiar with unraid, or hypervisors to be honest, but cant you let the hypervisor control how much of the cpu you get automatically?  lets say you have 56 cores, and your the only one working today, could you get all 56?  

You can but you have to shut down the VM then start it up after changing the processor config. It's not automatic and hypervisors won't let you change the processor config when the VM is already running.

 

On 9/13/2018 at 6:50 PM, JCBiggs said:

is there any reason the Thunderbolt 3.0 dock and fiber optic cables wouldnt work on a smaller system that didnt need 8k displays?   it seems like it might actually be simpler since you wouldn't need the daughter board or have to working about the 'imou'?  groups 

 

I like virtualization, but i hate working in a virtual environment.  working in an environment with a bare metal hypervisor that doesnt feel like a remote desktop would be great.   Plus 1 box to maintain/upgrade instead of 4-8.   The bigger issue for me is the shared CPU capacity.  

Last I heard there aren't TB3 optical cables currently. TB2 there are, but that doesn't seem to be the case with 3. Anyways, TB3 makes things more complicated for virtualization (too complicated really). The server board needs to have TB3 support, you might be able to pass through the TB3 controller, but it's possible that it only works for PCIe devices, ie, the monitor ports on the TB3 dock might not work. Maybe you could pass through a GPU on TB3 and use that for the monitor output. But then the dock plus PCIe box and GPU inside would almost be a PC by itself.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×