Video

the 20 gb of ram are better for AI and home LLM's /stable diffusion than the other card. 8 gb is right on the border of being useful for ai applications in my experience.

yea man theat si veri nasti but vreii good to nay ..nevre so i agel nevar tha si lai a scam ..3200$ froe licine tha shood be iligal man ..yea fafkig a 40gbs nope

6:55 Did the Path change? I only have install.d, postinst.d and postrm.d in the /etc/kernel directory.

I picked up similar tools on AliExpress, for about half the price. Delivery in 7 days, which is getting pretty good

i own a 3080m. one of the best gpu ive owned soo far. desktop performance at 100w power draw. cant wait to get a 4080m

That beer sounded really good

I was able to set one machine up without issue - basically following the naming structure GPUP-02 I tried to replicate this and change it to GPUP-03 and receive an error and the VM is not creating the partition.

For anyone taking engineering in college, I think the elecom huge wireless is a must. Of course it should be mentioned all the benefits of a trackball mouse for ergonomics (especially with marble scroll downloaded), but it is the superior mouse for portable use. I can sit on a stool with my laptop on my lap and the huge laying sideways on my stomach and use it perfectly. You absolutely may be able to get by in CS with a touchpad, but for most engineers, we will encounter some sort of 3d modeling software whether it be fusion or solidworks. It is integral to be able to use the right click, left click, and middle click while moving the mouse for most of these programs. This type of mouse allows for that while also not requiring a flat surface. Additionally, the trackball allows for an incredibly intuitive rotate view function as you can physically rotate your ball as an analogue of your model, I'd rate it above a space mouse even.

Could I use this for VR?

""either way, there's a whole lot of coupling going on"" I LOST it 🤣🤣🤣🤣

does it run GTA5

Almost 10 minutes of ranting before you even get to the tutorial!

it does worth it especially when you're out of town and don't want to bring your massive heavy 1400x 71717au, 17,000gpu pc with you

ramdisk maybe in addition to, or in lieu of, some of that local storage since you have 2x 8 slots? Likely deals with the random I/O from games better than ssds over sata. Don't dispose of the 8x ssds though; keep them for persistent non-networked storage. Use ramdisks temporarily during a gaming session. Dunno. Just an idea. Albeit ddr4 is not cheap, I know. Like $600+ bucks for 512gb kit (8x 64gb). At least $1200+ for dual socket is hefty.

Do you have a wiring diagram for pinout or instructions?

Could you give each VM access to a shared storage on HDD where you could bulk store games that are not being played? Then when you need a game you just move it to the VM drive that's on SSD?

I hope you get into more detail how you passthrough the video cards. I have seen so many ways of how to passthrough gpus with proxmox and proxmox has added way more features and I don't know what the "correct" way is anymore...thank you!

someone does a shunt mod service for the a4000 ada. apparently it runs faster than a 3070ti and pulls just over 100w under load.

I don't have any problems using a P106-090 with 6GB as an eGPU using the latest Nvidia drivers.

Did you go with RAIDZ-2 for this server?

I like my 2013.. Just wish the GPU could get swapped out though with something more modern. I bought an eGPU which was very expensive using an NVIDIA card… Right up until Apple blocked it. Did the workaround but it required regular mucking around to get it working as it would frequently stop working and require setting back up. TB2 and the 2.7Ghz 12 core is just fine for me. A lot of what i do requires CUDA cores and being able to keep an NVIDIA GPU up to date when necessary would make this last for many more years. Id switch this to running Windows but then id lose my Adobe CS6 suite.

Thank you for going through all the troubles testing this out. I would love to have a similar setup for my home lab. I choose to do it in 4u instead of 2U as that makes cooling at reasonable noise levels more doable but I love using old server gear for it as it just has so many more features than consumer gear. I am not saying that everyone needs 128 PCIE lanes you get with EPYC but only having 24-28 lanes on consumer sucks because after you add a single GPU and storage you are basically done as far as expansion.

great video where is the written instruction you promised ,i thought the video will be how you made the cables

You got so many things wrong with such confidence

"why would Asus do this?" - I guess you aren't up to speed on current events when it comes to Asus...

you know somethings about nvidia licensing appliance u need right for the newer types of cards? else hit me up if you need some help

I'm in love with the idea of getting one of these, but which one is the best buy right now? I'm trying to build a gaming rig on a budget, but I'm a bit confused about all the options. Also I saw the review of the RX580. Is that still a good option, or whats the new cheap best GPU? Please link to any recommendations on Aliexpress.

Who else is torn between running TrueNAS as a VM on Proxmox or running TrueNAS on bare metal as a file server and a couple VMs?

waiting for that all gpu benchmarks!!

great video but nah not gonna support the scammy ASUS their RMA is attrocious... So sad to see a company go down the drain... when I grew up they were considered the best on the market

nice

This is exciting! Would you mind testing some consumer cards in this context? I'm starting to amass a collection of last gen budget GPUs (3060s, 1660 supers, Vega 56); if you have the cards, how do these perform in a virtualized context compared to an enterprise GPU?

The only channel about computers that recommends that you should start drinking like a pro. 🙂 I do not know if that use of words will make you a alcoholic very soon.

imo it's almost always easier to build your own cables than depin those damn things

This is crazy cool. :-)

You could probably sell thoose cables. Also since when do you have V100

I love the idea of a test suite for enterprise cards. Just please, include some AI stuff. Text gen web UI and Stable Diffusion. I'm not an AI fan, but it needs to be run locally or the big companies will steamroll everyone and take control.

Loving this series on data center GPUs. So interesting to see what these older cards can do! I would really appreciate some local AI benchmarks as well. Tasks like benchmarking Whisper or local LLM tokens per second would be great. A lot of us out here are wondering how well we can run AI at home on these cards.

You don't have RAID setup on your camera?

a dollar to troll? what a deal!

How about we just forego the gender vocabulary and call one end the “pin” side and the other the “receiver” side?

When you test the tesla v100s, you should test them with and without SLI. Edited because of a stupid typo

From my experience the performance of the GPUs scale better the more they are split apart. i.e. 6 VMs do overall better than 4 VMs. I always found myself being held back by how much VRAM the card has. On that note Kepler (K40 / K80) is basically e-waste for the power to performance these days. I'd like to say the same for Maxwell (M40) too. Pascal bare minimum for power to performance, and anything newer feels too expensive for the age of the hardware. Guess I can blame AI craze for that.

Male/Female is always looked at from the metal pins, not the housing. Think, what needs to be stiff....

Kizakura Kuromame is my favorite beer. It's a black bean beer, that with certain foods or a pallet cleanser gives an amazing flavor. It's pretty hard to come by here in the US, but there are a small number of select grocery stores that have it, or you can find it in Kyoto's Kizakura Kappa Country if you're in Japan :). They also have a selection of other beers, an IPA, golden ale, and I forget the other (maybe a blonde?). Also, Sapporo Black is decent, but the Sapporo Gold is better imo. I'm less of a hop-heavy IPA (I only like conifers in my gin, thank you much), so these Japanese beers are much more clean and smooth drinkers. If you want the true experience, just fly to LA and go to Ootoro, spend at least $300, and weep for the lack of divine cuisine in the US of A.

Camera: om nom nom nom

I gotta ask, for your GPU VMS have you found a workaround for anti cheats or are they limited to just non anti cheat enabled games Asking cause I have GPU PT VM on proxmox

The male-female designation goes like this...male goes into female, duh. But you can have male terminals in a female plug and vice versa. One example of a female terminal in a male plug is the ATX power cable, and the hard drive molex connector on IDE drives. Gonna tell ya, those terminals SUCK ASS. UL requires a pull of 35 lbs on 16 ga wire but those terminals have a tendency to fail unless it's crimped for a 45+lb pull test. One thing I recommend is to put a dab of solder on the crimp, it helps especially if you're hand crimping. Sauce: I worked in a wire mill for 10 years.

I’m so stoked for the GPU benchmarks, can’t wait

Now make a high availability cloud gaming cluster 🥵