AMD launches FirePro W9100 with 16GB GDDR5

[piesquared]

Yep, the only drawback of Hawaii is its transistor density. Had it been lower it could have completely trashed GK110 in any metric.

IMO the higher density was the correct design decision though as the larger die would cost more to produce. Slight less leakage there yes but a decent cooler can manage that.
They also managed to cram a 512 bit bus in that thing.

 

[Vesku]

Would be good to have more options... Maybe a blower type for the multi-card and rack-style configurations, and the open-air cooler for single-card configurations in a tower style workstation.

For dense racks they often don't bother with blowers instead they'll use that saved space for a bigger heatsink with air flow provided by high flow fans in the front of the case.

Rackmount ready S9000

http://www.amd.com/PublishingImages...G/firepro_technology_s9000_angle_5x5_375w.png

Rackmount ready S10000

Doubt clocks will be pushed too hard, as the server market is efficiency conscious.

 

[piesquared]

It seems FirePro is gaining a lot of attention especially with this new W9100 beast.

There's some big names in the industry at NAB involved with AMD and the new Ultra Workstation, or Visual Supercomputer is brilliant. The move to OpenCL appears to be well past the tipping point and accelerating. FirePro is almost certainly going to be taking big market share and making a lot of hay.

http://www.fireuser.com/

 

[norseamd]

yall welcome to buy one of these and send it to me with shipping paid of course

would this do anything for amd in the hpc environment?

 

[VulgarDisplay]

yall welcome to buy one of these and send it to me with shipping paid of course

would this do anything for amd in the hpc environment?

Until Open CL gets better documentation and easier to work with overall CUDA is pretty well entrenched.

 

[3DVagabond]

It seems FirePro is gaining a lot of attention especially with this new W9100 beast.

There's some big names in the industry at NAB involved with AMD and the new Ultra Workstation, or Visual Supercomputer is brilliant. The move to OpenCL appears to be well past the tipping point and accelerating. FirePro is almost certainly going to be taking big market share and making a lot of hay.

http://www.fireuser.com/

Easy to get companies to let you slap their logo on your product. Money, some free SWAG, and help with OpenCL implementation for their software and your in. AMD is making inroads and they've released some very competitive hardware.

It's not easy to pry the competition's hardware out of people's work stations, though. And when you do you had better make sure their experience is first rate. People are using their computers to earn a living. Down time is unacceptable. You certainly can't take weeks and months to find solutions for bugs, like you can in the consumer space.

Autodesk and Adobe are huge and have a lot of customers (almost everyone uses their software to some extent). We'll have to see how well the support goes. nVidia has been doing this for a long time and most of the programmers for these companies are going to have Quadro cards in their computers.

 

[piesquared]

AMD could have NV by the sac here. If NV was banking on 20nm and are now having trouble with it, Hawaii could rule the professional, HPC and server space for some time and enough time to gain a lot of traction. By the time NV has a response, AMD will be onto the next series. It's not surprising given NV's brutal track record of transitioning to a new node. And that was with AMD and TSMC blazing the way first. The last time NV got to a node first it didn't turn out so well and they've followed ever since. FirePro looks to eat quite a bit of NV's lunch.

OpenCL here we come.

Building a 150 TFLOPS cluster with Accelerators in 2014

Green 500

You might have noticed the big differences between the GFLOPS/Watt. Where this is important, is the Green 500 – the list of energy efficient supercomputers. The goal of today’s supercomputers is that they are mentioned in the top 10 of both lists. If you build an efficient cluster (say 2 CPUs + 4 GPUs), you can get to 70-80% of max DGEMM performance. Below is a list for 75%:
•AMD FirePro – 7.10 GFLOPS/Watt DGEMM -> 5.33 GFLOPS/Watt @ 75%
•NVIDIA Tesla – 5.65 GFLOPS/Watt DGEMM -> 4.24 GFLOPS/Watt @ 75%
•Intel XeonPhi – 3.56 GFLOPS/Watt DGEMM ->2.67 GFLOPS/Watt @ 75%

Currently this list is lead by a cluster with K20X GPUs, steaming out 4.50 GFLOPS/Watt, which even has 86% of max DGEMM.

In other words: if the FirePro gets out in time, then the green 500 could be full of FirePro GPUs.

The winner

As there are only three offers, they are all winners. What matters is the order.
1.AMD FirePro – 16GB fast memory, clear winner in DGEMM performance. Bad: not available yet.
2.NVIDIA Tesla – Second to everything from FirePro (bandwidth, memory size, GFLOPS, price). Bad: outdated OpenCL-support.
3.Intel XeonPhi – Same as FirePro if it comes to memory, but 60% slower in DGEMM and 50% less efficient. Bad: 300 Watt for a server.

I am happy that after years of NVIDIA leading the pack, AMD is the clear winner now. As AMD is the foremost supporters of OpenCL, this could seriously democratise HPC for the times to come.

Rory Read is executing like a mofo.

AMD- What a Difference a Year Makes