Acceleware cuts half its staff
ACCELEWARE JUST LET GO half it's staff, starting with President/CEO, Sean Krakiwsky and 39 others including the COO, VP of Business development and the VP of sales. It looks like the CUDA based GPGPU benefits are not quite a bountiful as Nvidia PR says.
The operative quote from the press release here, is "To preserve capital and limit expenditures, the Corporation will reduce the number of personnel to approximately 40 employees", down from about 80. From this point on, "The restructured Corporation will focus its efforts and resources on its more mature vertical markets, including the electronics and oil & gas markets, until additional capital is obtained."
While it is sad to see a promising company like Acceleware implode, it does show exactly what we already knew, CUDA and GPGPU is on it's last legs. That Nvidia has been pouring money into it is a rather transparently desperate attempt to make people believe that it is still worthy. It isn't.
If Acceleware, the big name in the industry, with solid technology, couldn't make the paradigm work, can anyone? Doubtful, but even if they can, it is now quite clear that GPGPU is not the path to riches. It may be an interesting aside, but that is about the limit.
The most interesting part of this is at the top of the release. It says "NOT FOR DISTRIBUTION TO US NEWS WIRE SERVICES OR DISSEMINATION IN THE US", I wonder why? Almost like someone is trying to keep something out of the press, to use the vernacular, eh? I wonder who would want to keep the biggest proponent of a core technology imploding quiet a few weeks ahead of Nvision?
In the end, there is not much more to say, Google bought Peakstream likely for a song, and now someone will pick up the distressed pieces of Acceleware. They have solid technology, the real question is will anyone actually pay for it when it doesn't make financial sense to use the underlying paradigm? µ
Share this:
Business Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
I seriously wonder what nVidia did to Charlie to make him such an ATI-fanboi.

He relishes in the fact (misattributed by himself to CUDA) that a bunch of people just lost their jobs... and see's it as another way to prop up him self and bag out nVidia.

Well, I'm not an undereducated hack, I'm an engineer that uses AMD/ATI and nVidias GPU's for acceleration (and no, I have naught to do with anyone mentioned in the article).

Do you know what? CUDA is easier to write in than AMD/ATI's CTM. It gets results quicker and is more mature. But what do I know, I'm just an engineer who does it for a living, not some hack who has a chip on his shoulder...

Acceleware's problems mean Cuda is a failure...how comes I don't see the connection.
The statement says "While it is both disappointing and unfortunate that the significant downturn in Canadian and U.S. equity markets has restricted Acceleware's access to the capital necessary to fund growth opportunities".
Understood? Ever heard of credit crunch? You will if you didn't.

Now since I trust Carmack more than this author, Carmack here http://www.pcper.com/article.php?aid=532
said that he's going to implement a new rendering algorithm using CUDA.
Now why don't you say that Carmack is wrong... good luck...
Yes, they are so rotten that Apple will use nVidia chipsets and mobile GPUs.
Here: http://www.pcper.com/article.php?aid=598
You have overlooked that one, eh Charlie, didn't you?
ClusterInABox are getting the AXE!
Why So Serious?
I get a feeling deep down inside
Something just ain't right

Horrors! The sham of it all! Will the Nvillians get their chips up?
Itune in tomorrow. Same Bat Time. Same Bat Channel.
The problem was that they didn't have a product, they were selling R&D.

Companies go with what they know works. Processing too slow - get faster hardware. It is never, "Invest in rewriting software". Think of all the companies that have gone belly up while re-writing / retooling / refactoring their code. 

Now, if Acceleware actually had product that companies wanted, they might have succeeded. The problem is that SIMD processing is pretty niche. And GPUs are pretty limited in what sort of problems they can tackle.
Not even constrained by elementary logic anymore Charlie? Some small start up cuts staff, therefore a different companies larger effort is not lucrative. That's great. We should apply your groundbreaking logic to other applications.

AMD is cutting staff, therefore all software makers are doomed. AMD processors are designed to run software, after all. If software was so great, AMD wouldn't have to cut staff, right?
Ah, but Charlie, if the actual technical talents behind Acceleware are saved, they are potentially valuable, not because of their GPGPU experience per se, but their experience in developing compilers and structuring applications to take advantage of SIMD/MIMD architectures. Like Intel Larabee, or IBM/Sony/Toshiba CellEngine.

Doubtless, they uncovered bottlenecks in the GPGPU architecture that could guide others in developing better solutions.
classic case of small startup hired too many suits and burned all the $$. nothing interesting to see here, move along.

otoh, nvda can expand the access they have given to Acceleware, for example, by creating a partnership program. sooner or later some shops would create interesting applications, and voila, nvda get to sell a lot more chips. win win for everyone.


I wonder if Intel paid you to write an article like this! Why dont you substantitate what you say?

Just merely holding CUDA responsible for Acceleware's failure is NOT fair-play!

Kindly investigate and post a detailed report -- we'll agree!!

To me, CUDA rocks! I have programmed! I have seen unbelievable speedups!! 

And, btw, what is that "CUDA and GPGPU are in its last legs"?? So, Intel is making Larrabee just for fun, huh?

It is the reporter's responsibility to report things in a fair manner. Dont be so biased!
Looks like Larrabee has won the massively parallel wars before they even began. x86 is the instruction set that simply won't die. I have to hand it to Intel - they are a very well run company when arrogance doesn't go to their head.
As a site to offer "news/reviews/facts & friction" it is apparent that articles such as this one discredits Inq. It is neither fact or opinion, in fact it is intentionally misleading.

It *would* have been a good editorial if Charlie had backed up his claim (that CUDA/Tesla lines are failing to pick up) by
*thorough analysis across verticals* and *named sources*. None of which exists. Furthermore, Inq's editorial policy suggests that "opinions" should be clearly marked as such. Well, this is clearly marked as an article which it equally clearly isn't.

--

While I find it interesting that Acceleware is cutting its staff by 50% the rest of the article is pure rubbish. In fact, if Charlie had bothered (actually, if he were allowed access) to check with larger industry vendors in Tesla targeted verticals (oil and gas, financials, medical) he would have found that the majority of companies are *increasing* their investments in the GPGPU domain. How do I know? Because we deliver services & software to these clients, directly related to Nvidia GPGPU technology.
Furthermore, to expect that the Tesla boards would roll directly into production environments in thousands when CUDA 1.0 was released only a short time back shows Charlie's total neglect of market dynamics. Vendors *do* rewrite important codes when the reward is good enough, which in the case of Tesla & CUDA - it will be.

I suggest that Charlie should attend NVISION 2008 to gain some knowledge about GPGPU, discuss with vendors (such as ourselves) that rely on the GPGPU technologies (we do have choices, you know) and then produce a proper analysis.

-CEO
With quad off the shelf dual GPU's, (8 GPU's) the proof is in the pudding:
http://fastra.ua.ac.be/

NVIDIA has had GPGPU in the planning for a long time, and Microsoft is finally going to be building in support. This means NVIDIA's IGP strategy will continue, as it means their chipset business will continue.