Breaking News
News & Analysis

Intel further retreats from discrete GPUs

5/26/2010 11:00 PM EDT
13 comments
NO RATINGS
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
KB3001
User Rank
CEO
re: Intel further retreats from discrete GPUs
KB3001   6/30/2010 8:31:51 PM
NO RATINGS
It seems to me this is an admission of failure from the part of Intel in conquering the Graphics market. Rather than do that, they are trying to keep their market share in HPC, which could be eroded with the rise of GPGPU.

sandhu233
User Rank
Rookie
re: Intel further retreats from discrete GPUs
sandhu233   6/18/2010 10:39:48 PM
NO RATINGS
I love this reply to a post concept !!

Warren3
User Rank
CEO
re: Intel further retreats from discrete GPUs
Warren3   6/18/2010 9:39:07 PM
NO RATINGS
In "Bright Side Of News" on May 27th, 2010, Andrew Richards wrote a 3,400 word analysis regarding Larrabee that concludes: "So, what matters in GPU and CPU design is how do we reduce power. So, when thinking about graphics, we should think, what is the most power-efficient way to implement each part of the graphics pipeline. If that means lots of different bits of custom hardware, then so be it. Implementing it all in software will take longer, run slower, and use more power. I don’t know how many of us would have predicted that at the start of all of this, but that’s where we are now. Intel will, of course, have to produce a GPU. But for the meantime, a Larrabee without graphics, but with data-parallel cores, will be a formidable HPC accelerator, possibly even supplanting NVIDIA GPUs for HPC applications." I cannot say if he's right or wrong but he raises some interesting points either way.

BigJoJo
User Rank
Rookie
re: Intel further retreats from discrete GPUs
BigJoJo   6/18/2010 1:31:51 AM
NO RATINGS
another great one!

alan.varghese
User Rank
Rookie
re: Intel further retreats from discrete GPUs
alan.varghese   6/16/2010 3:02:01 PM
NO RATINGS
Larrabee was a hybrid of both a multicore CPU and a GPU. Thus the architecture had to meet the requirements of both high-end computing, as well as those of high-end graphics. This made the software architecture complex too. So I am not surprised that key product milestones were missed. The business case was not clear either - in some segments the Larrabee would have cannibalized Intel's main CPU business. The only reason in favor of Intel staying with the program was to be on top of high-end Graphics requirements before they become mainstream, but that by itself is not sufficient reason to invest in the high costs of producing a chip. Altogether, it makes sense that the project was stopped.

Robotics Developer
User Rank
Rookie
re: Intel further retreats from discrete GPUs
Robotics Developer   6/15/2010 5:16:40 PM
NO RATINGS
I would be curious to know what are the overall market opportunities for discrete GPUs? I have read a few months ago about the use of GPUs for general processing of code. The highly pipe-lined GPUs were being used as high speed general processors running user code without the overhead of operating systems. While there were a number of constraints involved in general purpose GPU processing, it seems to be an interesting path worth pursuing. I would like to see further development in this arena as a means to provide very high throughput using mass produced GPUs (read here lower cost) for more general purpose computing.

marktwain
User Rank
Rookie
re: Intel further retreats from discrete GPUs
marktwain   6/14/2010 11:55:40 PM
NO RATINGS
#nic - do you have a link to speediest supercomputer story?

SteveSaunders
User Rank
Rookie
re: Intel further retreats from discrete GPUs
SteveSaunders   6/8/2010 7:46:21 PM
NO RATINGS
In Tuesday's blog posting, Intel spokesperson Bill Kircos expanded on the status of Larabee, saying that Intel, "will not bring a discrete graphics product to market, at least in the short-term." Kircos said the company missed some key product milestones and determined that media and high-defintiion video, as well as mobile computing, are the important areas to focus on moving forward. Nathan Brookwood, principal of Insight64 (Saratoga, Calif.), said Intel originally had two goals when it started down the Larrabee path: develop a multi-core HPC processor and to deploy the same type of hardware as a high-performance graphics processor. Now it appears Intel is going to focus on the HPC market and never going to use Larrabee to build high-performance, discrete graphics GPUs, he said "The graphics part turns out to be a distraction," said Brookwood. "The market for high-performance discrete GPUs is a very narrow market and, by any measure, it is shrinking. Why would Intel spend a lot of money to target a market like that?" Intel remains focused on developing integrated graphics technologies for its processors, Kircos wrote. The blog also states that Intel intends to introduce a multi-core extension of its server product line, derived from Larrabee, at the International Supercomputing Conference in Germany next week. According to Brookwood, efforts to continue building high-end GPUs with more and more power now appears to be a losing proposition. Even ATI Technologies, the graphics chip unit of AMD, has backed away from building high-end graphics processors over the past two years, he said. ATI has been targeting the mid-range graphics market and serving the high-end with multiple GPUs, Brookwood said. Only Nvidia is still trying to build "the most humungous GPUs," Brookwood said, suggesting it is "one of the reasons they have struggled" over the past 18 months.

nicolas.mokhoff
User Rank
Rookie
re: Intel further retreats from discrete GPUs
nicolas.mokhoff   6/4/2010 3:26:57 PM
NO RATINGS
What does Gartner think about this?

nicolas.mokhoff
User Rank
Rookie
re: Intel further retreats from discrete GPUs
nicolas.mokhoff   6/3/2010 10:16:28 PM
NO RATINGS
China claims speediest supercomputer, made from graphics chips?

Page 1 / 2   >   >>
Most Recent Comments
August Cartoon Caption Winner!
August Cartoon Caption Winner!
"All the King's horses and all the KIng's men gave up on Humpty, so they handed the problem off to Engineering."
5 comments
Top Comments of the Week
Like Us on Facebook

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
Flash Poll
Radio
LATEST ARCHIVED BROADCAST
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.