Introducing New Passively Cooled Quadro RTX 8000 and RTX 6000 GPUs Optimized for the Data Center

Auer

[H]ard|Gawd
Joined
Nov 2, 2018
Messages
1,972
http://blog.pny.com/blogpnycom/intr...-rtx-6000-gpus-optimized-for-the-data-center?

passive-blog-header.jpg?width=1106&name=passive-blog-header.jpg


The introduction of passively cooled Quadro RTX 8000 and RTX 6000 GPUs enable the integration of RTX-powered solutions into existing (or future) server enclosures suitable for NVIDIA Tesla full-height and dual-slot width passively cooled boards – a change that significantly widens systems compatibility – and a first for any ultra-high-end Quadro product.

NVIDIA Quadro RTX features such as RT Cores for cinematic quality production ray-tracing, multi-precision Tensor Cores that accelerate AI/DL/ML/MV and big data analytics, or AI enhanced design and visualization tools, support for virtual GPU (vGPU) that brings Quadro performance with data center security to devices ranging from tablets to non-Quadro equipped mobile or desktop PCs, and ideal attributes for at the edge eXtended Reality (XR) AR and VR deployments in 5G and Wi-Fi 6 environments, are compelling and beneficial reasons for institutions to purchase NVIDIA RTX Servers.

NVIDIA authorizes and supports the utilization of NVIDIA Quadro RTX 8000 and RTX 6000 (actively or passively cooled) professional graphics boards in server enclosures for data center deployment to realize these use cases. All offer similar GPU performance, but the RTX 8000 option offers an unprecedented 48 GB of GPU memory, while the NVIDIA Quadro RTX 6000 provides 24 GB. All utilize ultra-fast GDDR6 with optional ECC, and NVLink offers GPU memory pooling for two cards, providing 96 GB or 48 GB respectively, along with performance scaling since GPU core counts are effectively doubled.

Passively cooled versions of the RTX 8000 and RTX 6000 are available across a much wider range of server chassis than their actively cooled counterparts, making Quadro RTX in the data center more accessible than ever before to offer previously unrealizable levels of performance and paradigm shifting capabilities – all with Quadro IT manageability.
 
why do they do this? because expected ro run 24/7 and the fan is one less thing to break
Server U racks typically have several high rpm fans like the Deltas that push a ton of air through out the chassis. The air is funneled through all components like CPUs which also don't have fans generally. And yes it is for reliability since the fans can easily be swapped out without tools.
 
I wonder why they call this a Quadro. Doesn't nvidia call their dedicated compute cards "Teslas?"

Also, calling it passively cooled seems pretty disinegnuous to me. The card may not have fans, but it clearly relies on the fans in the chassis, which would be monstrously loud.
 
Server U racks typically have several high rpm fans like the Deltas that push a ton of air through out the chassis. The air is funneled through all components like CPUs which also don't have fans generally. And yes it is for reliability since the fans can easily be swapped out without tools.


ahh ty. i forgot to put a '?" so it sounded like i was stating a fact. i have no clue haha.
 
I wonder why they call this a Quadro. Doesn't nvidia call their dedicated compute cards "Teslas?"
Looks like this is a unique product from PNY because you're right, Tesla are for data centers and HPC while Quadro are for workstations. PNY probably came out with these since they're cheaper than Tesla.
 
I wonder why they call this a Quadro. Doesn't nvidia call their dedicated compute cards "Teslas?"

Also, calling it passively cooled seems pretty disinegnuous to me. The card may not have fans, but it clearly relies on the fans in the chassis, which would be monstrously loud.
Yes nVidia frequently reminds us all that anything running a server OS should be running a Tesla card and that they do not support that Quadro's in Server OS's. That said, the drivers and the functions work perfectly fine and many programs that you would want to run in a data center or on a centralised server do not like the Tesla cards as much and the pricetag on the Quadro's is much more palatable. For instance I have a drafting/animation server that students use a terminal serive to access from chromebooks. I can spend $12,000 on Quadro's or I can spend $40,000 on Tesla's. The Quadro's in this case do a better job, at a much lower price point.
 
Back
Top