Active cooled Tesla GPU from Maxwell or Pascal architecture existing ?

Is there any active cooled Tesla GPU with PCIe connection (so a GPU I can put into a ‘stanadrd’ workstation (like a HP Z800) available, which has Maxwell or Pascal architecture ?

The new Tesla P100 (PCiE version) seems to be also passive cooled. So far it seems for me that the ‘newest’ Tesla which is active cooled is the Telsa K40. This is also indicated from the list at http://www.thinkmate.com/hardware/co-processors

Not sure what you’re after, exactly, but Quadro M6000 is a Maxwell architecture GPU with fan. It would have roughly similar capabilities to a Tesla M40.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814133623

We need reliable GPUs which are desigend for 24/7 computing throughout several years. Teslas are marketed that way. We recommend currently Tesla K20 / K40 to customers. I was just wondering whether there are newer Tesla GPUs which we could add to the list. Passive cooled cards do not help us much cause in our aplication field (media / broadcast) workstations are the standard and not rack server.

Are high-end Quadros (X5000, X6000) also designed for 24/7 computing ?

I’m not aware of Tesla GPUs newer than K40c which are actively cooled, at this time.

If you need a product recommendation for a particular use-case, my suggestion would be to contact your NVIDIA rep, rather than basing your product decisions on comments from a public forum.

I believe Quadro GPUs are reliable and capable of continuous operation for any supported professional use-case. Compute APIs such CUDA are definitely a supported use-case for Quadro GPUs. Media/broadcast workstations are a focus vertical for Quadro.

A bit of anecdotal evidence: I have personally used various Quadro models in workstations (HP, Dell) with CUDA in a 24/7/365 fashion (often under heavy load) for years without any issues. This included both high-end and low-end Quadro GPUs.

The only potential issue I have run into is that the PSUs in some workstations seem to be a bit under-dimensioned for operating high-end GPUs. But so far I have never killed a PSU this way, while I have killed (at least) a handful of PSUs in consumer rigs.