GIGABYTE Releases its Latest GPU Computing Server

[btarunr]

GIGABYTE Technology, a leading creator of high performance server hardware, is happy to announce today the release of its G210-H4G. Developed in partnership with Carri Systems, a French specialist of GPU computing & HPC solutions, this product is a 2U rackmount housing 4 blades capable of receiving one NVIDIA GRID, Tesla or AMD FirePro card each.

In the context of an unfolding GPU virtualization market and the vast amount of new possibilities that it opens, the G210-H4G is an elegant solution designed with simplicity and flexibility in mind to let organizations, big and small, enjoy the benefits of this fast-growing computing trend.



Compatible with all the models from NVIDIA's GRID and Tesla product families, and the latest AMD FirePro S9150, this product is a standard 2U rackmount built with 4 blades. Each blade is built with a GA-6LISL motherboard that can receive an Intel Xeon E3-1200 V3 processor, with 4 hot-swappable bays for 2.5" HDDs or SSDs, and with a special PCIe x16 (Gen3 x16 bus) slot large enough to receive a computing card up to 267mm length and 300W TDP. Each blade has also been specifically designed to deliver an optimal airflow to and around the computing card, in order to give it the proper thermal environment to operate at its best capabilities.

With such specifications, the G210-H4G is an ideal product to unlock the enormous potential offered by NVIDIA and AMD's professional cards in many types of scenarios, notably for virtual desktops running graphics intensive applications and cloud gaming as well.

Supported Computing/GPU Cards

• NVIDIA GRID K340
• NVIDIA Tesla K40
• NVIDIA Tesla K20X
• NVIDIA Tesla K20
• NVIDIA Tesla K10
• AMD FirePro S9150

For more information, visit the product page.

View at TechPowerUp Main Site

 

[librin.so.1]

MFW I see HDMI, DP and 2x DVI ports within the blade's enclosure.

 

[Aquinus]

MFW I see HDMI, DP and 2x DVI ports within the blade's enclosure.

That's because there is an expectation that the Tesla card is going to be used to do GPGPU. You would use the VGA port on the board itself, this is a server not a workstation.

 

[librin.so.1]

That's because there is an expectation that the Tesla card is going to be used to do GPGPU. You would use the VGA port on the board itself, this is a server not a workstation.

I know, lol.
But...
1. Due to the nature of GPGPU cards, I always assumed they have no graphics output connectors whatsoever in the first place.
2. It's still an odd see and for a conventional mind to digest.

 

[Lysd3xic]

Judging by the video output and SLI finger (both useless for GPGPU applications) I would imagine that the card in the picture is a Tesla using a PCB made for a GTX or Quadro card, probably to save money when making them.

 

[librin.so.1]

Judging by the video output and SLI finger (both useless for GPGPU applications) I would imagine that the card in the picture is a Tesla using a PCB made for a GTX or Quadro card, probably to save money when making them.

>still soldering those graphics output connectors on instead of, well... not soldering any of that stuff on
>saving money

doesn't really add up, no.

 

[Aquinus]

It could be a prototype and not the finished product.

 

[librin.so.1]

touché, @Aquinus, touché...

 

[Death Star]

The older Teslas have generic VGA output that's abstracted from the compute GPU. We have a C2075 in my lab with a DVI out. None of the GPUs characterized as a "compute module" have video output though.

That's definitely just a gaming card with the outer sheath and heat spreaders removed. I don't see a reason why a gaming card couldn't in principle be used in something like this, but the software side for handling multiple users is different for consumer cards than compute modules. This was a very hard lesson in queuing when I set up two remote machines with Titans and 290Xs -_- Also, unless you write your own software most commercial software for configurations like this doesn't support consumer cards.