Update: The actual interview can be heard as part of the Special Edition Podcast.
I just got off a phone call with Bob Pette, Vice President of Corporate Marketing at SGI, regarding today’s announcement of bankruptcy and acquisition of a merger with Rackable. The phone call will be in an upcoming Special Edition podcast to come later tonight, but some highlights for you:
- SGI was saddled with a large debt coming out of the last bankruptcy, and was doing fine until the recent economic downturn.
- The new company’s name is still to be determined, although the “SGI” and “Silicon Graphics” names bring alot of brand equity.
- No significant reduction in personnel is expected, although some overlap will obviously happen in HR & Operations that will result in slight reductions
- No overlap of high-end products, a little overlap in the lower-end x86 space may result in a few products being scrapped.
- SGI expects to no impact to current support contracts, or outstanding purchases (like the DOD’s recent $40mil acquisition) (Corrected 10:09)
Check back later for the audio of the phone call, and more!
CNet News has an article up about the upcoming fight between Intel & Nvidia about multi-core processors, and the different directions that the two firms are going. Intel is betting on Larrabee, which puts a few generic-X86 cores on a chip, while NVidia has it’s CUDA and GPGPU solutions with hundreds of very specialized chips. What which solution does industry want?
Keith Gray manager high performance and technical computing at oil giant BP spelled out why he has hesitated to use GPUs to date while expressing interest in adopting them in the future. “Our business is about accelerating our development of new seismic imaging research algorithms. At this point we actually believe the level of programming difficulty and lack of standardization of application development tools make the move to accelerated computing a bit risky ” he said.
via Needs of big firms foretell Intel, Nvidia battle | Nanotech – The Circuits Blog – CNET News.
Update #2: Hear the actual interview with Bob Pette, along with a better explanation of the acquisition from John West of InsideHPC in our Special Edition Podcast.
Update: See a summary of our Interview with SGI VP Bob Pette about the acquisition.
John West at InsideHPC has gathered some more information on the SGI/Rackable deal, and it seems the deal isn’t quite as solid as originally reported.
The sale is actually being made via a section 363 “stalking horse” procedure, which means there is a 25-day window for other bidders to come in and buy SGI instead. If none come, however, Rackable gets SGI without the debt. It’s pretty much a done deal because, as John points out:
But the word “short” is part of the problem from the perspective of those other bidders. Rackable has (likely) had a much longer amount of time for due diligence on the sale and, according to the Focus article probably has stacked up some other benefits as well.
I’m still attempting to reach people inside SGI for more information. If anyone inside wants to speak about the impacts of this on SGI business (VUE, Hardware, sales, support, etc) contact me!
SGI, Rackable, and the stalking horse | insideHPC.com.
Once the industry standard in Computer Graphics, father of the famed Onyx, Infinite Reality graphics pipes, and Large Shared Memory architectures, SGI has filed for bankruptcy and is now to be acquired by Rackable Systems.
According to the press release, Rackable has agreed to acquire “substantially all the assets of Silicon Graphics, Inc. for approximately $25 million in cash.” The agreement includes Silicon Graphics filing for Chapter 11 bankruptcy protection in order to reduce their overall debt load.
So what does this mean for the VUE product line? As information comes, we’ll let you know.
via Rackable Systems to Acquire Silicon Graphics | insideHPC.com.
One of Nvidia’s big announcements yesterday was “SLI Multi-OS”, but alot of people are still uncertain what that really means. InfoWorld has an article on the latest version of Parallels which uses the feature, and gives a nice description of how it will be useful to people performing large visualizations (the Oil & Gas industry specifically).
“Until now it has been impossible for them to enjoy the benefits of virtualization as the impact on graphics performance has been significant. (…) The solution also offers up to 16 CPU cores and 64GB of RAM for guest OSs creating a superior customer experience compared to today’s normal computing speed in a virtual machine. (…) .” Rather than taking the typical path of emulating the graphics adapter in a virtual machine Parallels Workstation Extreme employs what it calls the Parallels FastLane architecture which takes full advantage of the intelligent performance features in the new Intel Xeon processor 5500 series and the Intel X58 Express Chipset.
via Parallels gives 3D graphics a boost within virtualization |Virtualization Report | David Marshall | InfoWorld.
Yesterday the internets were alight with talk of NVidia’s latest QuadroFX card offerings. However, a few astute readers saw something fishy in the announcement. In particular, the following comment about sums it all up:
What is NVidia announcing today?
Certainly it is not the Quadro FX 5800 with 4GB of video ram on-board. That was announced on November 10, 2008.
Certainly it is not the Quadro FX 4800. That was announced on December 18, 2008.
So really it appears that the new cards are the NVIDIA Quadro FX 3800, NVIDIA Quadro FX 1800, and the NVIDIA Quadro FX 580/380.
The Quadro FX 3800 should sell for between $900 to $1200. This card has 1GB GDDR3 memory, 192 shader processors, 256-bit memory bus, and 51.2GB/s memory bandwidth. The maximum power consumption is about 107 watts.
The Quadro FX 1800 should sell for about $600. This card has 768 MB GDDR3 memory, 64 shader processors, 192-bit memory bus, and 38.4GB/s memory bandwidth. The maximum power consumption is about 59 watts. It is based on the G94 core (a.k.a. the GeForce 9600).
The Quadro FX 580 should sell for about $150. This card has 512 MB GDDR3 memory, 32 shader processors, 128-bit memory bus, and 25.6GB/s memory bandwidth. The maximum power consumption is about 40 watts. It is based on the G96 core (a.k.a. the GeForce 9500).
At the low end you have the Quadro FX 380 with 256 MB of memory, and 22.4 GB/s memory bandwidth. That is, it is a Quadro FX 580 with slower clocks and less memory. It is based on the G96 core (a.k.a. the GeForce 9500).
The Quadro NVS 295 is really multi-display graphics for businesses.
In other words, even these new cards are nothing really new. The only new thing that I can see is the SLI Multi-OS. Now that is worth paying attention to.
Seems NVidia is trying to gain some press with another “Paper Release”.
expreview has the specs on the new Radeon HD 4770, based on the RV740 chip. Of interesting note: They ship it with high-speed GDDR5 memory, but they’ve cut the memory interface in half (from 256bit to 128bit).
Slated to lauch May4th, price tag is at $99.
AMD Radeon HD 4770 Specs Unveiled – Expreview.com.
BOXX is breaking ground again with their new 3DBOXX8500, a new workstation based on the ‘Nehalem’ Xeon processor from Intel.
Capable of reaching 3.46 GHz using Intel Turbo Boost Technology, the 3DBOXX 8500 provides digital artists with 16 virtual cores of high-powered, multitasking performance for 3D design, animation, rendering, and visualization applications, as well as VFX compositing, video editing and digital intermediate workflows. Designed to accommodate multi-threading, multiple applications, and complex production pipelines, the 8500 features six Intel Xeon Processor 5500 Series options, up to 192GB advanced DDR3 ECC memory, 12 terabytes of storage, and up to six hard drives.
Time to upgrade?
via CGSociety – 3DBOXX 8500.
expreview.com has news that NVidia, presumably in response to AMD’s latest move to reschedule the Radeon HD4890 release to April 2nd, has done the same with the GTX275.
Unfortunately, NVidia has a reputation for “paper launches” (eg, Issuing press releases saying that it’s released, but without any inventory to actually sell). So even with the card “released” on Thursday, it’ll probably be a few months before anyone is actually selling them. That should give AMD a good lead in the short-term.
Specs on the GTX275 can be found on Hardware.Info
Update: Tom’s Hardware has more on this story.
According to Expreview, Nvidia will release the GeForce GTX 275 on April 2nd. Naturally, without some kind of source, such a date would be chalked up as simply rumor. However, that may not be the case, as DollarShops actually had the card listed for around $327 USD (€249) and promised to have the card in stock by April 4. Manufactured by Sparkle, the GeForce GTX 275 offers the 55nm GT200 GPU clocking in at 633 MHz and a 1164 MHz memory clock, 896 MB of DDR3 (448-bit), 240 stream processors, 2 DVI outputs and HDCP. However, now the card is gone, ripped from the online retail shop altogether.
Will it be released this week or not? That’s the big question right now.. Oh the Drama..
NVIDIA Brings GeForce GTX 275 Launch Forward to April 2nd – Expreview.com.
Update: Read our follow-up article, “What is NVidia Really Releasing?“.
NVidia has issued a press release announcing a new line of QuadroFX cards.
- NVIDIA Quadro FX 5800 – the industry’s first and only 4GB, ultra high-end solution delivers the highest visualization performance and capabilities, enabling professionals to work with large-scale models and datasets.
- NVIDIA Quadro FX 4800 – ultra high-end solution provides professionals with the most complete toolset to dramatically push the boundaries of realism, performance, and quality.
- NVIDIA Quadro FX 3800 – single slot solution with a robust feature set and added capabilities such as SLI Multi-OS and SDI support to enable a no-compromise, high-performance, interactive visualization experience.
- NVIDIA Quadro FX 1800 – mid-range solution offers the best price performance for workstation graphics and provides the optimal blend of quality, precision, performance, and programmability.
- NVIDIA Quadro FX 580 – best-in-class entry-level solution boosts productivity for a variety of industry-leading volume CAD and digital content applications.
- NVIDIA Quadro FX 380 – an affordable professional-class graphics solution extends energy efficiency while delivering 50-percent faster performance.
- The Quadro NVS 295 – capable of supporting up to two 30-inch digital displays at maximum resolutions, taking business graphics to a new level visual perfection.
Another interesting piece of this is their new “SLI Multi-OS” system.
Built into the new Quadro FX 3800, Quadro FX 4800, and Quadro FX 5800, SLI Multi-OS allows users to tap into the advanced visualization and compute capabilities of the Quadro GPUs to experience full graphics performance within a virtualized system. SLI Multi-OS works in association with Parallels Workstation Extreme virtualization software and Intel’s VT-d technology, assigning both the host and guest virtual machine its own dedicated GPU. Available in the new HP Z800 workstation, the combination of these innovative technologies delivers application performance nearly identical to systems configured with a dedicated operating system and GPU.
This means you can now run OpenGL applications with full hardware acceleration inside a Virtualized environment. With virtualization becoming more and more attractive to larger organizations, and virtualization being used more on cluster-configurations to simulate one single large machine across a cluster of smaller machines, a good working virtualized OpenGL accelerator will be a huge win for NVidia in the upcoming thin-client space created by companies like OnLive & Gaikai
via NVidia Press Releases: SLI Multi OS and new QuadroFX’s.