Here is all of the news we’ve collected about Intel Corporation. This mostly includes their (now defunct) Larrabee chipset, as well as their many whitepapers and benchmarks regarding parallelism and their various platforms.
You might also want to check out
Earlier this week, Nvidia’s Steve Scott took on the hype around the new Intel MIC system. Acknowledging the power of a hybrid solution, he then picked apart their design and claims of “port-free performance boosts” from running your x86 code on MIC.
The recent news and industry reaction regarding Intel’s forthcoming “Many Integrated Core” (MIC) accelerator has been interesting to watch. It appears Intel, like NVIDIA and AMD, has now concluded that hybrid architectures are the proper response to the growing power constraints of high performance computing.
While I agree with this, some of the discussions around programming the upcoming MIC chips leave me scratching my head – particularly the notion that, because MIC runs the x86 instruction set, there’s no need to change your existing code, and your port will come for free.
It’s been covered widely, including over at HPCWire.
Scott is not arguing against the MIC as an accelerator, per se. He and most of the community are convinced that HPC needs a hybrid (or heterogeneous) computing to move performance forward without consuming unreasonable amounts of energy. Traditional CPUs, whose cores are optimized for single-threaded performance, are not designed for work requiring lots of throughput. For that type of computing, much better energy efficiency can be delivered using simpler, slower, but more numerous cores. Both GPUs and the MIC adhere to this paradigm; they just come at the problem from different architectural pedigrees
It’s an interesting design, but I think the real power isn’t necessarily in the Intel MIC hardware, but rather in the libraries and tools that Intel is finally developing to assist in the upcoming hybrid computing surge. While you’ll always get better results from custom hand-tuning your code for these systems, it’s a huge boost to productivity to find that your existing libraries have already been ported, and you get a nice boost (but not the best) just from using their libraries.
In the end, I don’t think either NVidia’s or Intel’s MIC are the true path forward, but each will do their own part to push the technology forward into the right answer.
Intel has a pretty new infographic online showing the explosion of internet connected devices since the first computer came online back 1960.
The time was when the internet was people on their PCs sending email and surfing web content dished up by servers. Sure, it was around before that in its academic/military guise, but as far as the public consciousness was concerned, PCs, laptops, email and the web, was all the internet meant. A few years ago that changed, imperceptibly at first, such that now the recurrent understanding of the internet is far more diverse, feral even. Myriad devices and people creating content in multiple forms, for consumption on an increasing array of devices. And it’s changing industry too- internet-connected combine harvesters, anyone? It has become known as the internet of things, guided by the fact that if something has an on/off switch, it should probably compute, and if it is going to compute, it should also connect to the internet.
I agree with Randy Krum’s take on it, it’s realliy pretty but lacking any true information. OF course, their point was to show address space explosion, but all of the points twist and tangle for apparently no real reason other than to look pretty.
This is one of those graphs that has the illusion of far more information than is actually there.
In a surprising move, Intel’s Havok division (the physics SDK developer that Intel bought a few years back) has just acquired Trinigy and their 3D Vision Engine. It’s a surprising move for Intel, who dropped their Larrabee and Project Offset projects long ago. For Havok tho, this may be an attempt at growing their physics SDK to include cross-platform gaming.
“The team at Trinigy is very excited by the opportunity to join Havok.” said Felix Roeken, General Manager of Trinigy. “Havok and Trinigy have been partners for a number of years and both companies share very similar philosophies about how technology should be built and delivered to customers. We are confident that this acquisition will be very positive for Trinigy’s customers and employees.”
Intel Acquires Trinigy, Adds 3D Engine To Gaming Portfolio – HotHardware.
Intel has just released a collection of Code Kernels to do photo-realistic ray-tracing entirely using instruction accelerators (SSE and AVX) on the CPU.
Embree is a collection of high-performance ray tracing kernels, developed at Intel Labs. The kernels are optimized for photo-realistic rendering on the latest Intel® processors with support for the SSE and AVX instruction sets. In addition to the ray tracing kernels, Embree provides an example photo-realistic rendering engine. Embree is designed for Monte Carlo ray tracing algorithms, where the vast majority of rays are incoherent. The specific single-ray traversal kernels in Embree provide the best performance in this scenario and they are very easy to integrate into existing applications.
The results are beautiful I must admit, and a stab against GPU-accelerated raytracers like iRay. However, they haven’t published any details on the time required for rendering other than to say “a perfect result in about a minute” on one particular model. You can get a few more details in this short video from the creators:
via Embree – Photo-Realistic Ray Tracing Kernels – Intel® Software Network – Intel® Software Network.
The video compares an Intel Core i7-2600 CPU with an AMD 3510MX. The Intel CPU is a Sandy Bridge processor and has Intel HD 3000 graphics. The AMD quad-core processor A8-3510MX has Radeon HD 6620M graphics. Of course it shows that AMD’s Llano performs better.
Intel has posted an article on using the Cloud to perform ray tracing in games. Normally, I am not a fan of the term Cloud, because of its overuse and hype. However, in this case, I will make an exception because it uses Intel’s Knights Ferry. What is Knight’s Ferry? Remember Larrabee? Larrabee was the codename for a GPGPU chip that Intel was developing, that was canceled. The follow-on project is called Intel’s Many Integrated Core (MIC) project, with the first product being codenamed Knight’s Corner. Knight’s Ferry falls up under the MIC project. The article describes Knight’s Ferry as:
Intel code name Knights Ferry is the first-generation development platform for the Intel MIC Architecture. It includes a PCIe card that has a 32-core chip on it that is clocked at 1.2 GHz. The development platform is programmable with the regular tools and programming languages that developers regularly use. A bit further out there are plans for an Intel MIC Architecture-based product, code named Knights Corner, that will use 22nm manufacturing technology and will therefore be able to even have more than 50 cores on the chip.
The researchers used four server machines, each with a Knights Ferry PCIe card (32 cores, 4 threads per core). Each server also had a i7-980X processor, which has 6 cores running at 3.33 GHz. The thin client was a a small laptop, running on a Core2 Duo processor P9600 and with a 1280×800 screen.
Everything was connected over a Gigabit Ethernet LAN. Now, not everybody has Gigabit Ethernet, but this is a proof of concept, not a finished product. I suspect that Gigabit Ethernet is not necessary anyway. Sure, it gives you a lot of bandwidth, but from my own testing of remote visualization services cloud-based applications, it is the latency that is most important. If you are getting a 100 ms ping, then that means you can only get 10 frames per second, at best. You really need a connection with less than 33 ms ping time to get 30 frames per second, and it would be best if it was even lower.
Now if all of this sounds familiar, it is because we covered it last year in Wolfenstein Gets Ray Traced, On a Laptop. At the time, we did not know that they were using Knight’s Ferry. From the conclusion to the article:
Over the last sections it has been shown that ray tracing can offer a variety of new and interesting effects to games. Through this research using a cloud-based gaming setup with machines that utilize the Intel code name Knights Ferry development platform, ray-traced games with a high frame rate can already be achieved today.
Further progress could be made by optimizing the video codec used in order to be able to use it for even smaller devices such as netbooks and tablets. Instead of assuming a Gigabit Ethernet setup, optimizations for wireless networks could be investigated to bring the technology to handheld devices like smartphones. In order to cut down on the number of servers needed, it should be possible to develop support for using multiple Knights Ferry PCIe cards within a single machine. To increase image quality, several well-known post-processing techniques like HDR bloom and depth of field could be added. A smart solution on how to do anti-aliasing for ray tracing with high performance on the Knights Ferry platform could also be investigated.
AMD has posted a video demonstrating its Fusion Accelerated Processing Unit (APU), which has been code named Llano. AMD compares it against an Intel Core i7-2630QM, which is based on the Sandy Bridge architecture. As one might expect, since it is an AMD demonstration, the Fusion APU beats the Intel Sandy Bridge chip. Then again, AMD does have a history of making great graphics cards through their purchase of ATI. Intel’s graphics have always been less than stellar.
Lots of people heard that Nvidia wasn’t going to join the X86 race but rather “go it alone” in the ARM market, and were baffled that NVidia would drop major computer architecture design and go after a market that’s proven over and over again it’s good for not much more than embedded systems. A new article over at ArsTechnica does a good job breaking down why NVidia will have a tough road ahead of them in “beating” Intel, but that they honestly may not even have to to win.
If it turns out that the ARM ecosystem can get within a factor of two of x86 in terms of performance and performance per watt as ARM chips move to higher levels of size and complexity, and if that ecosystem can simultaneously keep the cost of ARM chips much lower than that of x86 chips, then ARM could do to x86 what x86 did to Alpha, MIPS, SPARC, PowerPC, and the other RISC workstation architectures. Intel used its process strength to get close enough to RISC workstation and server performance that the PC’s status as a low-cost commodity machine gave the PC an edge. We all know how the story played out: the high-end RISC vendors were marginalized, as the lower volumes on their now-boutique chips kept their prices up; meanwhile, Intel relentlessly narrowed the performance gap and moved the PC into new markets because it was dramatically cheaper and almost as good.
I think it’s a smart move on NVidia’s part to let Intel run with the Server space (where NVidia can always have an “in” with Tesla and Quadro GPGPU addons), and instead target the growing smartphone, tablet, and portable computer spaces. It’s slimmer margins for sure, but a vastly larger market that’s already used to upgrades every 6 months.
All this has happened before: NVIDIA 3.0, ARM, and the fate of x86.
Apple has just refreshed the MacBook Pro line of hardware with some impressive new offerings, dumping the previously used NVidia chips for Intel Sandy Bridge and AMD Radeon chips. Just check out what you get for $1800:
15-inch MacBook Pro – No more Core i5 options for the middle child — these two strictly get the Core i7 and AMD Radeon HD 6000M goods. The $1,799 model packs a 2.0GHz quad-core Core i7 processor, AMD Radeon HD6490M graphics with 1GB of VRAM, and a 500GB hard drive. On the higher end, there’s a $2,199 model which buys you a faster 2.2GHz Core i7 chip, AMD Radeon 6750M graphics, and 750GB hard drive. There are also 128GB, 256GB, and 512GB SSD options. Also, as far as we understand, the graphics switching will work exactly like the previous models — it automatically switches between the GPU and IGP depending on what you’re doing.
Quad Core CPU, Radeon 6490 video card, and the new Intel Light Peak technology now called Thunderbolt. I sell a few more ads (or Amazon affiliate commisions!) here on VizWorld, and I might finally upgrade my old Black MacBook.
A new whitepaper from Intel brings in some statistics and stories from Luxology, Luxion, and Modo on the power of CPU’s for ray-tracing and how they can smoke any GPU on the market with CPU-only solutions.
“Modern GPUs offer a brute force solution to ray tracing, but the memory available to GPUs is relatively limited compared to the system memory available to 64-bit CPUs such as Intel Core i7 and Xeon processors. That means that GPUs typically can’t handle the huge scene files required in full-scale production rendering, which may involve tens of millions of polygons and hundreds of high-resolution texture maps. And CPUs offer greater flexibility in terms of shading complexity and plug-in shaders, which may or may not have been ported to run on a GPU.”
These are the same arguments I’ve been hearing for the last year or so. And I have to admit they’re right, if not a bit short-sighted. It’s my belief that most of the arguments they use are going to fall apart soon.
In fact, I think within the next 5 years we may see the distinction between CPU and GPU disappear almost entirely, as they both wind up on the same die (similar to how Processor and Math Co-Processor eventually merged several years ago).
It’s a good whitepaper tho, full of some concrete numbers on attempts to GPU-ize code unsuccessfully and benefits achieved from using some of Intel’s newest CPU-optimization technology.
Check it out, and see what you think?
VizWorld.com Got an idea for a story? Drop us a line at [email protected] or other means.
Comments