R.I.P: FireStream (2006 - 2012)

It goes without saying that with GCN AMD has significantly improved their GPU compute capabilities across the board. Consumer compute has already benefitted through the Radeon HD 7000 series, while professional graphics will begin benefiting with the FirePro W series, and GCN will be laying the foundation for the Heterogeneous System Architecture (HSA) in 2014. All of AMD’s product lines are benefiting from GCN… all of them but one: FireStream.

For those of you not familiar with FireStream, FireStream is (or rather was) AMD’s line of dedicated GPU compute products. Initially launched in 2006 and based on AMD’s R580 GPU, the FireStream was AMD’s first product geared exclusively towards GPU computing. Since 2006 AMD has regularly updated the product line, releasing new cards based on the RV670, RV770, and RV870 GPUs.

The most recent refresh of the product was the release of the FireStream 9300 series in 2010, which saw the FireStream family move to AMD’s first meaningfully capable OpenCL GPU, the RV870. Since then AMD ended up choosing to skip a 2011 refresh of the product based on their Cayman (VLIW4) GPU, which was a somewhat odd move at the time. While VLIW4 is not the kind of superior compute architecture that GCN is, it was still fundamentally designed to improve AMD’s compute performance, which it did thanks to the use of narrower SIMDs that allowed for a partial shift away from ILP to TLP. Nevertheless, as we found out after the fact with the launch of GCN, major users weren’t interested in moving to VLIW4, almost certainly having early knowledge that VLIW4 would be a dead-end architecture to be replaced by GCN in 2012.

In any case, with the release of GCN and its significant compute enhancements we have been expecting a major update to the FireStream product family. But as it turns out AMD has other plans.

Starting with the launch of the FirePro W series the FireStream family of products is being discontinued entirely. From here on the FirePro family will officially be pulling double-duty as both AMD’s professional graphics product and AMD’s compute product.

So why is AMD choosing now to discontinue the FireStream family at this point in time? Officially, AMD believes the FireStream family to be redundant. AMD’s FireStream cards were nearly identical to their FirePro cards in both build and performance, with the only practical difference being that FireStream cards had most of their display connectors removed. And AMD is right – by all accounts the FireStream 9300 series did very little to differentiate itself from the equivalent FirePro cards.

Meanwhile FireStream as a brand hasn’t kept up with NVIDIA’s Tesla business in the dedicated compute market. NVIDIA’s Tesla business is still a fledging business – it’s grown by leaps and bounds since 2010, but not as much as NVIDIA would like – but even so the company has created several distinctions between Tesla and Quadro that AMD never did replicate with FireStream. Chief among these was a compute-focused driver for Windows (NVIDIA calls it TCC), which stripped away all of the graphics capabilities of the card in order improve compute performance by freeing it from the control of the Windows display driver subsystem. Furthermore NVIDIA went and developed a couple different lines of Tesla cards, branching out into both traditional self-cooled cards for workstations and servers, and purely passive cards meant for specialized rackmount servers. For the FirePro V series AMD does have both active and passive cooled FirePro cards, however the FireStream cards were only available with passive cooling.


FirePro V9800P(assive)

The other limitation for AMD in this arena was of course their GPUs. GCN gives AMD a very potent compute architecture that is unquestionably competitive with Fermi and little Kepler, but there’s one thing NVIDIA will do that AMD won’t: build it big. AMD doesn’t strictly adhere to a small-die strategy (300-400mm2 is now their sweet spot), but they also don’t build 500mm2+ behemoths like NVIDIA. There’s a great deal more to compute performance than die size of course (especially when NVIDIA has disabled functional units on most Fermi Tesla parts for yield and power reasons), but it does lock AMD out of certain markets that desire as much individual GPU performance as possible. GPUs excel at parallel problems, however not every problem scales well across multiple GPUs (mostly due to memory sharing needs), which means there’s still a need for a very large GPU.

As a result of these factors, AMD’s RV870 FireStream cards never did gather a great following. GCN fixes the fundamental compute performance problem that was the chief factor holding back FireStream, but this would appear to be a case of too little, too late for FireStream. Of course AMD will continue to have a very important presence in the overall GPU computing market, but at least for now they’re seceding from producing a dedicated compute product.

In lieu of having a dedicated compute product FirePro will be serving both markets. In fact AMD tells us that there are already plans to build compute clusters out of FirePro W series cards (though they can’t name names at this time), reinforcing the notion that AMD is still active in the compute market even without a dedicated product. The big question of course is how potential customers will respond to this; customers may not be interested in a mixed-function part even if the performance was to be the same. Throwing a further wrench in AMD’s plans will be pricing – they may be underpricing NVIDIA’s best Quadro, but right now they’re going to be charging well more than NVIDIA’s best Tesla card. So there’s a real risk right now that FirePro for compute may be a complete non-starter once Tesla K20 arrives at the end of the year.

Perhaps for that very reason AMD’s promotional focus with FirePro is largely focused on professional graphics right now. Certainly the company is proud of their achievements in compute performance, but going by AMD’s marketing materials AMD is definitely focused on graphics first and foremost with FirePro.

The Rest of the FirePro W Series Feature Set More to Come, So Stay Tuned
POST A COMMENT

35 Comments

View All Comments

  • johnthacker - Tuesday, August 14, 2012 - link

    The W7000 has some uses in specific situations, but that's because it's a single slot card. Single slot Radeon HD 7850s (much less 7870s), which also use Pitcairn, are difficult to find; there was one OEM that showed off a design IIRC. Other than that it's hard to see exactly when someone would want these cards.

    The same generally holds for NVIDIA (the Fermi Quadro cards are cut down GF100 based, so they can be better at compute than their gaming numbers suggest, and the old Quadro 4000 is a single slot card.) Interesting that NVIDIA so far is trying to reserve GK110 for Quadro and Kepler only. We'll see if that works.
    Reply
  • Dribble - Tuesday, August 14, 2012 - link

    AMD doesn't need to provide compatible, they need to provide better.

    Bottom line is companies won't change gpu manufacturer. Nvidia works well, has traditionally worked better then AMD and they still have much better driver support (team is much bigger).

    There are no AMD fanboys routing for the underdog, you have to provide a clear business reason to change, and "we're almost as good as nvidia in sometimes" isn't it.
    Reply
  • CeriseCogburn - Wednesday, August 29, 2012 - link

    Compute baby ! amd compute ! compute ! GPGPU ! amd wins! amd wins!
    (that's all I've heard since 7000 series)
    Hey, winzip man.
    Reply
  • Pixelpusher6 - Wednesday, August 15, 2012 - link

    I really have to question AMD's move here to kill off Firestream and have the FirePro line serve both markets. At the present time this is where they have an advantage on Nvidia. 1TB double precision performance is pretty good even though its only 1/4 of single precision, even clocked low enough to be passively cooled it should still beat Nvidia's best compute card. K10 is not really a compute card at all and to me it seems like they just wanted to get something, anything out until K20. And K20 is by no means a certainty for Q4 2012, my guess is it will be delayed. I just don't have confidence in Nvidia's mastering of the 28nm process yet, especially given the enormous die size of this chip which I've heard presents some unique challenges. And when K20 does come out it will probably be more expensive than their current compute cards.

    If I were AMD I would re-brand the compute card, drop the Firestream name because of it's association with VLIW, and come out with a new brand to highlight what really makes up GCN...Compute. Does anyone know if HPC clusters use actively cooled cards or only passively cooled? I was under the impression that compute cards generally were clocked a little lower but passively cooled. If that is the case then that rules out using the FirePro W9000 and W8000 in these server clusters. It seems like AMD just conceded this market completely when they finally have a competitive compute GPU to gain a foothold. As someone else noted this market will only be expanding. If AMD wants to only focus on professional graphics I sure hope their drivers will be better than the consumer counterparts.
    Reply
  • dtolios - Wednesday, August 15, 2012 - link

    When will AMD start improving compatibility with VRay RT and other similar OpenCL apps? All this computational potential remains unused outside benchmarks - at least for the CG world.

    Radeons are vastly better in OpenGL than GeForce cards, so the switch to FirePro's is way less "mandatory" for such apps. But even if those driver issues were solved, AMD would secure a huge increase of share in the professional CG market which now uses nVidia (yes, mostly gaming cards) almost exclusively.
    Reply
  • AG@FirePro - Monday, August 27, 2012 - link

    You might imagine that it's in AMD's best interest to work very closely with all the important ISVs in this space - and you would be right! :) Helping our technology partners and the broader software development community implement open-standards-based GPU acceleration in their applications is an area of heavy onging investment for us.

    Of course, not all apps are written the same. Some applications -especially those written in years past- are architected in a way that makes it challenging to enable the best performance across all the modern GPU options on the market. Proprietary or "hybrid" codebases often make full cross-compatibility quite difficult. can assure you that neither ISVs nor end-users want their toolsets be tied to a particular hardware vendor or proprietary technology base. Unfortunately, it's not always as easy as flipping a switch and sometimes this takes a while. This said, I think it's fair to say that our aim is that very soon, everybody will have the option to run the hardware of their choice in conjunction with their favorite realtime raytracer, physics solver or any other hardware-accelerated toolset.

    AMD FirePro cards fully support OpenCL in both hardware and software. Our devices offer certified and acknowledged compatibility and killer performance for a broad range of OpenCL-based applications. The same is true for tons of applications accelerated under OpenGL, DirectX and DirectCompute APIs. Compatibility and reliability are crucial. Nobody understands this more than us.

    To this end, we continue to be closely aligned with all the key ISVs in the M&E, DCC and CAD space to help them provide maximum flexibility, choice and value for their end-users.We also continue to refine and expand our range of developer tools (profilers, compliers, debuggers, etc) while at the same time contributing heavily to the open-source community in the form of optimized libraries and other free developer resources.

    The OpenCL story gets better every day. Every day, there are new and better OpenCL libraries being written and shared. There are new compiler optimizations being made all the time which allow for faster andmore flexible implementations. More and more software devs are liberating their code and their customers from proprietary APIs. While CUDA-bound apps still provide lots of value for many end-users, the writing is clearly on the wall. The age of proprietary GPU acceleration has begun to yield to a new reality of flexibility and choice for consumers.

    This is a good thing, no?

    *PS* You may have noticed an announcement about certain new server-side GCN-based FirePro GPU offerings today. Stay tuned. Things are about to get seriously fun up in here.

    Adam G.
    AMD FirePro Team
    Reply
  • CeriseCogburn - Wednesday, August 29, 2012 - link

    It's not a good thing because it has not happened, and it doesn't appear it will even in the next decade.
    It's still proprietary, and is not cross card company compatible, so it's all crap talk.
    As we saw amd do after their years of PR bashing, WinZip PROPRIETARY.

    It's gonna be all seriously vendor specific up in there for a long, long time.
    Reply
  • warpuck - Sunday, August 19, 2012 - link

    Does this mean I wont need 2 PCs? one for games and another for graphics. I did notice what appears to be a crossfire connector. I know most companies would not go for PC configuration like that, unless it was in the boss's office. I am one of those independants. I like taking a break when I feel like it. Not having 2 PCs would simplify things for me. Reply
  • peevee - Friday, August 24, 2012 - link

    OK, "later this week"? In the review written 8/14. "This" week ended, then "next" week ends today... Reply
  • Death666Angel - Tuesday, August 28, 2012 - link

    Hey!
    No problem about the inconsistent data, but maybe you can present it in a more accurate way? Currently the interval of the X axis is not to scale and the line through the data points makes it seem as though you know the way it progressed in between the data points. I'd rather make a simple bar chart with the intervals showing correctly. It would be a more honest and easy to read diagram. :)
    Great article though! :D
    Reply

Log in

Don't have an account? Sign up now