• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA Tesla T4

Soldato
Joined
10 Apr 2015
Posts
4,300
Location
Hungerford, UK, Earth
Ive been disposing of random unwanted IT at work, and come across an NVIDIA Tesla T4. I assume its not much use in the home environment. I installed it, and the nvidia control panel detected it and bonded the Cuda cores with the existing GTX1080... Im not sure if this gains a lot in gaming?.. Im sure the fleabay prices are inflated, and a person wanting this kind of card wouldnt be looking there.

Is there anything else it could be used for? will rendering software recognize it?
 
I think they are used for graphics acceleration on server based compute workloads and AI

So things like Citrix virtual desktops or AI/analytics code from Rstudio, it only really serves a purpose to a consumer as someone learning or coding in these area's imho others may want to add more
 
Seems to have pretty decent specs, better than my 1660 super. Having no graphics output is a major snag of course.

I would use it for distributed computing tasks (as I do for my main graphics card, throttled down). 'Boinc' tasks - the things that started with seti@home, and has branched out: I'm currently running a couple of climate calculations for Climateprediction.net, world community grid (medical research, generally third world), and a few under the 'Science United' umbrella (science and medical) on my CPU cores. The GPU element of this is under GPUGrid - generally science.

It could almost certainly be used for bitcoin (or whatever the current coin is) mining too. With 2560 cuda cores but only 70W it should be very cost efficient. I dont mine, so cant say for certain, but it it's identified as a graphics card by the os, it should work I think. And no need to give it a dummy VDU to display to (is this even neccessary?).
 
It's good for any kind of ML experimenting. I use Google Colab to run stuff in the cloud and one of the cards they provide is the Tesla T4. It's old and slow compared to newer cards but better than my laptop. Or my desktop GPU :p
 
will rendering software recognize it?
Quick google says Blender's Cycles renderer will support it.

However they appear to not be any faster than GeForce cards based on the same architecture, since the Tesla cards are focused on double-precision floating point acceleration which isn't of any benefit. Dunno whether the same applies to the like of Vray or Octance.
 
Is there anything else it could be used for? will rendering software recognize it?

They can be used for 3d rendering, but a quick check on completed listing prices - actual completed sales (including USA where quadros and teslas always sell for less due to volume) suggest that people must be buying them for something else. As you can get a 2nd hand 3090 for less.
 
Quick google says Blender's Cycles renderer will support it.

However they appear to not be any faster than GeForce cards based on the same architecture, since the Tesla cards are focused on double-precision floating point acceleration which isn't of any benefit. Dunno whether the same applies to the like of Vray or Octance.

looks like octane will support it, i was thinking of trying this package to go with sketchup pro... i can install it and try it see if it actually sees it.. althought i dont know how AMD and Nvidia drivers will cope in the same system.
 
I tried thea render, which let me pick the CPU and the Tesla for rendering. although i think the software is so slow for just a 720p image, and if i end the task the CPU stops processing, but the Tesla keeps going at 100% until i reboot the PC. its also RED HOT, the backplate is to hot to touch, which makes me think it should be in a rack mount cases with the fan arrays blowing directly into it
 
Back
Top Bottom