NicholasBoccio

Well, at least Linus still have $500 budget to create a public apology video if he wants to.

Here is what Linus said on the billet monoblock:
"Like it's a cool product. But it's a bad product."
That's explicit from the WAN Show, timestamping it from the GN video (29m48s, if it decides to embed instead):
https://youtu.be/FGW3TPytTjc?t=1788
 
Billet said they had no guarantees on how it would perform on a 4090 and that it was designed to work on a 3090 Ti. 
LMG testing on a 4090, positioning that the product was bad, and then refusing to re-test after Billet was able to get a hold of a 4090, discuss how it was going to perform like trash on a 4090 because of the design changes (again, this monoblock was never designed to work on a 4090) and it had a massive 1 millimeter gap that would kill cooling performance, and LMG standing by the results is unconscionable on LMG's part.
 
 LMG tested a product incorrectly and stood by it. Period. There's no excusing them testing a product wrong when the manufacturer sent a GPU specific monoblock AND the required GPU to test it with. LMG apparently used or misplaced the 3090 Ti that Billet sent (apparently misplaced, because as of this time LMG STILL hasn't located it), and then tested the monoblock with the wrong GPU.

"We can't spend a few hundred bucks to make sure we're not irreparably damage a companies reputation"
Seriously? To make 100% sure before you fuck over a startup you can't spend an hour or two to double check your results?
And then you SOLD the block at an auction causing them to be stalled as they no longer have their best prototype! What the absolute fuck????

GN did actually offer returns, even after months, to anyone affected.
 
Here's GN's response to that issue - I started the video where they explain the situation.
1. They immediately marked the mat as "Out of Stock" so that no on else would buy them.
2. They attempted to find a means to fix the issue by correcting the mats and failed to find one.
3. They created pinout cards to send to affected customers, which include not only the correction, but additional pinout info. They also made metal pins featuring the corrected pinout. Anyone affected was sent an email where they could get a kit featuring these. Anyone who didn't want a kit was allowed to return the mod mat for a refund instead.
4. They did then decide to sell the remaining mod mats, however, with the full correction kit, and with complete transparency on their website about what you were getting.

*edit* I actually got to watch the video when I got home, and its pretty rough. I've seen a few inaccuracies in LTT videos the last few months that were small, some GN didn't even show, obvious QA issues, but the kind of errors GN shows go beyond basic QA.
 
Everyone knows GN's the goat when it comes to actual data, LTT's forte into data is really secondary to the entertainment priority the channel has always had.
 
GN's accusations are valid here.
 
*another edit*
The response was well done in my opinion. corporate, but that's probably for the better. We'll see what the results are in their next videos, which'll be a solid week away for good reason.

Can you explain more? I like the simplicity of this method - but I have not read about this sort of method. Thanks for your help.

Now I am running Win 10 Pro. 

I currently have a DL580 4x E7-4870 & 128GB RAM sitting doing nothing.
 
Currently, I am using a Lenovo P330 Tiny i7-8700T 32gb RAM w/ Quadro P620
I have a BlueIris server running AI (from this guide https://www.youtube.com/watch?v=fwoonl5JKgo) running 13 cameras (See my network and camera setup here https://community.ui.com/stories/Home-Ubiquiti-Setup-Security-networking-and-more/4f0a698f-4d02-481c-9d48-17d9c72a2327) which works well, however the AI program uses the CPU as it is installed on Windows, and cannot use the GPU on Windows. In order to use the GPU for AI, it must be installed on a Linux system. This has reduced the processing time 10x for other users. Currently the CPU takes about 500ms to process 1 image, which is quick, but it does cause the CPU to max at 100% when processing images and recording. My initial tests show that the AI program is utilizing most of the CPU when it is maxed.
 
Here is my question: Should I install Proxmox on the P330 Tiny with a windows and linux VM and passthru the GPU to the Linux VM for the AI program and see if that helps reduce the CPU load when it's processing images? I wouldn't use this box for anything else.
 
The other option I wanted to get opinions on, was to buy an HP branded GPU and install that into the DL580 and pass that thru to a Linux VM and have the AI run from there and just remove the processing entirely from the BlueIris box. 
 
The reason I would not run BlueIris on the DL580 is that I don't like the idea of the cameras not recording for 5-10 minutes while the server reboots, when the P330 takes 15 seconds. 
 
All comments and criticisms are welcomed.
 

I currently have a DL580 4x E7-4870 & 128GB RAM sitting doing nothing.
 
Currently, I am using a Lenovo P330 Tiny i7-8700T 32gb RAM w/ Quadro P620
I have a BlueIris server running AI (from this guide https://www.youtube.com/watch?v=fwoonl5JKgo) running 13 cameras (See my network and camera setup here https://community.ui.com/stories/Home-Ubiquiti-Setup-Security-networking-and-more/4f0a698f-4d02-481c-9d48-17d9c72a2327) which works well, however the AI program uses the CPU as it is installed on Windows, and cannot use the GPU on Windows. In order to use the GPU for AI, it must be installed on a Linux system. This has reduced the processing time 10x for other users. Currently the CPU takes about 500ms to process 1 image, which is quick, but it does cause the CPU to max at 100% when processing images and recording. My initial tests show that the AI program is utilizing most of the CPU when it is maxed.
 
Here is my question: Should I install Proxmox on the P330 Tiny with a windows and linux VM and passthru the GPU to the Linux VM for the AI program and see if that helps reduce the CPU load when it's processing images? I wouldn't use this box for anything else.
 
The other option I wanted to get opinions on, was to buy an HP branded GPU and install that into the DL580 and pass that thru to a Linux VM and have the AI run from there and just remove the processing entirely from the BlueIris box. 
 
The reason I would not run BlueIris on the DL580 is that I don't like the idea of the cameras not recording for 5-10 minutes while the server reboots, when the P330 takes 15 seconds. 
 
All comments and criticisms are welcomed.
 

Lol anyone from east coast USA going. I know it's kind of far but I'll pay for gas.

I added my rental. @Canada EH, good point. Everyone should make sure that you have your tickets printed, return flight details printed, and we need to add each other in our phones just to be safe. However, IME, going into another country for a trade show is really easy.
 

If you are going to LTX 2018 there is plenty of other ways of getting to the convention other than driving or flying. How about by train!
For LTX 2017 I took the train using Amtrak from Fargo, ND to Vancouver, BC with a layover connection in Seattle. The train ride was around 36 hours or so for me. One-way. But I had the vacation time, and the experience was more pleasant than flying. Mainly because I'm a bigger guy.
 
Reasons to Travel by Train
An average coach class Amtrak seat is bigger than an average first-class airline seat, and it reclines. You get to see some of the greatest views by train other than by flying. Refueling and restocking stops (smoke stops) so you can stretch your legs. ALWAYS stay next to the train or you WILL be left behind and you will NOT be given a refund. I've witnessed people being left behind myself. They cry like babies. Don't be a stupid crybaby.  In some cases (including mine) traveling by train was much cheaper than flying. For me, it was 1/2 the cost of an airline seat ticket. Most Amtrak trains travel at or more than 79MPH. (You will see slower speeds going through rough terrain like mountains.)   
Cons
Longer travel time. My travel time included an 8-hour layover in Seattle. Which was during the day. So I went and saw some of Seattle while waiting for my connection. Got to see the Space Needle and Boeing's Museum of Flight.  MUCH longer travel time if using VIA rail. (Tracks are not as up to par as America's and therefore have a maximum of 55MPH on long distance Canadian routes ) To be clear this is for people who have the time to travel long distances by train. The journey time is much longer than flying. But if traveling more than 600 miles it is much faster than driving as the train doesn't stop at a hotel due to driver fatigue. It switches drivers at certain stops and continues the journey. This sounds complicated and a waste of time but again it is a great way to travel. 
 
Once you get to Pacific Central Station in Vancouver, which is served by both Canada's VIA Rail and America's Amtrak, it's less than a half a block away from Translink's Skytrain, Expo Line, Main-Street Science World Station. Pacific Central Station to Science World Station.pdf From there you travel north to Waterfront Station. Skytrain System.pdf Transfer to Skytrain's Canada Line and make your way south to Lansdowne Station which puts within 5 blocks of the Richmond Olympic Oval. Now you can either walk or take the C94 connecting bus to directly in front of the Richmond Olympic Oval. Science World to Richmond Olympic Oval.pdf
 
Your only concern now is a hotel. Don't worry there's a map for that. Click the "View on Map" drop-down arrow when loaded. https://www.tourbytransit.com/vancouver/accommodations#mapAnchor
I stayed at the Met Hotel in New Westminster for LTX 2017. It is literally right next to a Skytrain station and is reasonably priced. http://www.themethotel.com/ This hotel is still viable even after changing LTX venues this year. Just a tad bit longer travel by Skytrain. 
 
For American citizens only. YOU MUST HAVE A PASSPORT OR YOU WILL BE DENIED ENTRY INTO CANADA. Using any mode of transportation to travel internationally requires your passport. Canadians need no worry if you travel by VIA.   
 
How is this possible.
Amtrak is an American rail service. But provides international train service to select cities in Canada including Vancouver. All you need is your passport and a reason. ALWAYS BE HONEST WITH BORDER SECURITY AND DECLARE ANYTHING THAT THEY MAY BE INTERESTED IN IF YOU FEEL IT VIOLATES THE DECLARATION FORM GIVEN TO YOU. I told them that I was going to a tech convention and that I had a bag of beef jerky. Meat is banned from being entered into Canada. But since it was dead cooked meat I was ok. Nothing organically fresh is allowed to cross the border. (Unless it was something you caught while hunting or fishing.) Either from America to Canada, or Canada to America.
 
Unfortunately due to credit card debt that I used on my last trip to LTX. Spent damn near a $1000 bucks. NOT including the cost of traveling by train because I got them points yo. I won't be able to go to LTX 2018. But I feel that taking the train is a much under-appreciated mode of travel in North America and deserves to be taken into consideration when traveling. 
 
Have fun at LTX 2018 and safe travels, Evan G.
 
Websites to visit:
https://www.translink.ca/
https://www.amtrak.com/home
http://www.viarail.ca/en
https://www.tourbytransit.com/vancouver/accommodations
http://www.themethotel.com/
GOOGLE MAPS IS YOUR FRIEND
 
 

Thanks @kirashi
@LinusTech @Slick @CPotter