W9100 FirePRO graphics workstation card

Graphics Cards

One of the more powerful graphics cards available today (06/2017) is the FirePRO W9100 workstation card from AMD.  It delivers a total of 2816 stream processors and a choice of 32Gb or 16Gb of memory.  The W9100 provides ports for multi-tasking across 6x 4K monitors.

W9100 GPU card Specifications

  • Engine Clock: 930 MHz
  • Memory: 16GB GDDR5
  • 512-bit interface
  • 6 Mini Display Port 1.2 outputs
  • 320 GB/s bandwidth
  • 2,816 stream processors
  • 5.24 TFLOPS single precision performance
  • 2.62 TFLOPS double precision performance
  • 275 W TDP
  • 18.9 Gflops/Watt

This card debuted in April 2014 and it’s pretty difficult to get a hold of one of these cards in 2017.  In 2016, the card was updated with 32 Gb of memory, but otherwise unchanged.

New advances in technology

New advances in smaller memory is paving the way for even higher densities on modern cards, but demand seems to be out-pacing supply for quality cards.  Higher densities is also producing cards that are more efficient, with higher bandwidth.  In the past, total graphics memory was the big number, but lately it seems like processing cores and bandwidth are taking priority.

The new Vega cards coming out promise a much higher level of efficiency, but the rarity of HBM2 memory may be a limiting factor, reducing availability for the average user.  Thankfully, once the logistics are resolved, we can expect some very powerful cards at a fraction of the cost of previous models.

Pricing / Availability

Currently, W9100 cards are rare, and when they are available, they commonly are priced from $3k up to $5k, regardless of the memory profile.  All around the world, if you look on Amazon or other websites, the W9100 card is listed as out of stock.  They’re practically impossible to get at a reasonable price, so the demand is extremely high right now.

Newer AMD cards are being offered at a significantly lower price, but they’re not expected to hit shelves until August 2017 at the earliest.  Newer cards are geared much more specifically, providing extremely high double-precision processing numbers for mathematicians and other scientists in the form of the Vega Frontier edition.  Perhaps AMD is hoping to reduce the high-demand of math processing GPU’s effect on gamer’s video card supplies?


As with any new technology, the numbers may sound impressive, but what really matters is actual benchmark numbers.  How do these server cards perform in real-world use?  How well are the various features handled in actual compiler usage?  How do the improvements effect actual computation?  We’ll have to wait to see since these new cards have yet to be made available.