Intel Larrabee to Support Both DirectX and OpenGL

Intel has been showing presentations on Larrabee for sometime now, but hasn’t openly talked about what instructions it will understand, clock speeds or even a hard launch date. For people who do not know, Larrabee is Intel’s attempt to integrate GPU with CPU and create GPGPU (General Purpose GPU) or computing through the use of graphics processing units. Intel today made it clear at Siggraph 2008 that Larrabee will support both DirectX and OpenGL for graphics processing.

Intel already has the highest market share in the GPU world, thanks to integrated graphics processors (IGP) in its chipset's sold with motherboards. But due to the dismal performance over the years of these IGP chipsets, no one ever considers these capable enough for mainstream gaming. Benchmarks have shown time and again that even the most basic discrete graphic cards can perform better than IGPs. With Larrabee, Intel has intentions to bring more computing/graphics power with lesser power requirements.

Larry Seiler, senior engineer in Intel's Visual Computing Group said that, Larrabee's consists of cores derived from Pentium processors and have added multithreading and 64-bit instructions. Each core consists of 256kb L2 cache and the first release of Larrabee will have 8 to 48 cores depending on the market segment. Larrabee also has advanced multi-core capabilities and uses a 1024 bits-wide, bi-directional ring network for fast, low latency communication between the cores.

The programming paradigm doesn’t change with Larrabee because it uses the same x86 instructions used by Intel’s current generation microprocessors. Also with DirectX and OpenGL support inside Larrabee it confirms that current generation games will also work fine. But without any mention of the performance improvement in graphics capabilities over current generation IGPs, I still have doubts whether Intel is actually competing with its own processors or with the GPU vendors like ATI (now AMD) and  NVidia. Larrabee does look solid on paper, but as we know papers do create a lot more hype than they can perform and Intel marketing does know how to play that game.

AMD also has been working on Fusion processor which will integrate AMD CPUs and ATI GPUs into one chip. NVidia on the other hand has create a new programming paradigm with CuDA, which tries to enable normal computer processing to NVidia’s GPU. Whichever company wins this GPGPU battle, we would like to see all of them fight on the same grounds and support existing standards. Today’s Larrabee announcement confirms that both the DirectX and OpenGL standards are followed and will allow current breed of games to run on the Larrabee.

Wimax Missed the Intel Centrino 2 Boat Again

Wimax has been Intel’s big push to wireless broadband for some years now. Intel has promised Wimax as the solution to fast wireless internet time and again, but has constantly failed to deliver real working chips for operation. Today, when Intel launched Centrino 2, it didn’t have much of a mention about the Wimax adapters 5150 and 5350, which was supposedly the main talking point of Centrino 2.

Centrino was probably one of the most talked about technology by Intel when it was first released. It standardized wireless access through Wifi and most people synonymously used Centrino for Wireless-capable laptops. That was back in 2003. After that Intel refreshed the platform with a lot of new processors/platforms, codenamed Yonah, Napa and Santa Rosa. Lots of new motherboards/chipsets came and went and few integrated graphics just changed. Centrino 2 after much delay was launched today and the platform was earlier known by codename Montevina.

The Montevina platform announced today covers new 45nm Core 2 Duo and Core 2 Extreme processors, the GM45 chipset and the WiFi Link 5300 adapter. You can look at all the information about Centrino 2 platform from the wide press coverage its got from all hardware sites. Look here, here and here for examples. But what most of the sites are not complaining about is the lack of Wimax at the launch. I mean what is special about Centrino 2, if its not got a new fast wireless medium, atleast that’s what the original Centrino was all about.

But like I have heard before, the Wimax adapters for laptops 5150 and 5350 are still not ready for release. The adapters have an issue with power consumption and if Intel would have today launched the adapters with Centrino 2 platform, all the promises about improved battery life would have gone down the drains. Intel says we have to wait till the end of the year to get one of those Wimax adapters, but didn’t mention any specific reason for the delay and also didn’t mention that it was due to these (and Atom’s marketing) that Intel has been delaying Centrino 2 for sometime. Lets hope by fall we do see a battery efficient Wimax adapter for my new laptop. For a change, this time the ISPs are not late!!

 

Few presentation slides from Centrino 2 launch:

 

 

The Car came, but Wimax didn’t want a ride!!

Power 7 Will Be IBM’s Supercomputing King

This weekend I heard a few guys from IBM India getting really excited about the next-generation microprocessor called Power 7 which will be released in H1 2010. The Power 7 processor is still not a final design but nonetheless, it is turning out to be an awesome processor. IBM’s latest plans show that Power 7 will be a 8-core processor, clocked at anywhere between 3.6Ghz-4.0Ghz.

Power 7 is a completely new architecture and is being developed at Austin, Böblingen, Rochester Research Centers. There are rumors that Opteron and Power 7 may share a common socket, but that doesn’t seem to be the case in the current design of the processor. But I got hints that things on the socket front could be changed during the final few weeks.

Power 7 is also a multi-thread per core CPU and current designs suggest it can execute about 4 threads per core. It still isn’t running optimally at the moment, but the target is to achieve full performance on all the 4 threads. This still makes Sun’s Rock the multi-threading king in the next-generation of processors. But IBM is high on the number crunching game. Each Power 7 core will be 32GFlops and that means on a 8-cores it going to output at 256GFlops. That’s no mean achievement!

Unlike the Power 6, which had its main aim towards improving single-core performance and max clocked at 6Ghz in prototypes, the Power 7 is working towards a multi-core, multi-threaded processor. From Intel to Sun, everyone has realized the potential efficiency of multi-core design and IBM hopes to lead the pack with Power 7. And to become the winner of the lot, Power 7 will compete against Sun’s Rock (UltraSPARC T2 Processors), Intel’s Itanium and Xeon (Nehelam-based) to some extent as well as its friendly technology partner AMD’s Opteron. IBM’s main goal with Power 7 is at the High-Performance Computing markets and supercomputers, where IBM already rules the roost.

If you look for stats at Top500 Supercomputers from around the world, you’ll realize what I mean by IBM ruling the roost. IBM has lots of BlueGene machines that use the old PowerPC processors, but not many newer generation processors. The reason to this is two-fold. IBM’s last generation processor Power 6 and to some extent Power 5 isn’t very power efficient. Power 6 heats up quite a bit and would require massive engineering skills to bring 1000s of these chips together inside a supercomputer. On the other hand, IBM has plans to mix-match Power 7 and Cell microprocessors in future supercomputers. Cell processors are known to be good specialized FP units and matching it with Power 7, IBM hopes to reach more than 10 petaflop from a supercomputer.

Niagara 3 Does 256-threads, But No Supercomputing ?

The Register claims to have learnt that the next version of Niagara processors from Sun Microsystems is going to be a 16-core and 16-thread/core processor. This means that the Niagara 3 will be doing 256 threads simultaneously. And we aren’t talking of some distant future plan, but Sun Microsystems will be releasing these processors in 2009.

If you did not know, Sun Microsystems has been selling the Niagara 2 processors which can run 64 threads simultaneously on 8-core processor. Niagara 2 is officially sold as UltraSPARC T2+ chips and has been the most successful processors from Sun’s lineup in recent years. Obviously with so many threads running simultaneously on a single chip means that the performance is screaming and is excellent value for money. There are 2 socket systems available as well as 4-socket systems with UltraSPARC T2+. Then there is also the Rock processors coming in 2009.

From what I heard 2 years back, UltraSPARC had the largest processor design team in the world. So it comes as no surprise that it has the best design and performance in multithreaded systems. They also open-sourced their older SPARC designs as part of OpenSPARC.net and have pledged to continue opensourcing the next-gen processors as well!!

But the sad thing is that UltraSPARC Tx processors haven’t been as big a hit in the HPC and supercomputer market. If you have a look at the list of top supercomputers for June 2008, you will realize that Sun Microsystems just figures once on the top 10. It features in the top 10 for Texas Advanced Computing Center at University of Texas and uses AMD’s Opteron processors and not UltraSPARC. In all of the 500 supercomputers, IBM rules the roost with 209 systems and HP comes second with 183 systems. IBM has been able to push POWER processors to build some awesome supercomputers whereas HP has taken Xeon and built powerful systems around it. Sun Microsystems on the other hand has just 4 supercomputer systems and all are with Opterons.

Most people would argue that single core performance proves to be more effective than multi-threaded performance in conventional supercomputer software and benchmarks. Processes are more common units of work than threads itself, but that’s where the overall system building plays its part. Like Seymour Cray has stated, “Anyone can build a fast CPU. The trick is to build a fast system.”

Intel Invests in Solar Cells

With the rising fuel costs and the dangers of global warming by the use of carbon fuels, everyone seems to be looking at alternate sources of energy. The last week has been  specially interesting because 3 major tech companies, specializing in semiconductor manufacturing have got into manufacturing solar cells. Intel today announced that its spin-off SpectraWatt will be manufacturing solar cells for panel makers starting in the middle of next year.

Just last week, IBM had made similar announcement of manufacturing solar cells with Japanese semiconductor firm Tokyo Ohka Kogyo. Earlier to that HP said that it has licensed some technology to a solar panel startup called Xtreme Energetics. All these companies are putting their expertise in semiconductor manufacturing to improve solar cell efficiencies and decrease the cost of manufacturing these cells.

Intel Capital has put some $50 million into SpectraWatt and Cogentrix Energy, PCG Clean Energy and Technology Fund and Solon AG are also investing into Intel’s latest spin-off. SpectraWatt is going to manufacture these photo-voltaic cells at a fabrication plant in Oregon. After the Intel’s plant at Oregon is fully-functional it plans to churn out 60 megawatts worth of cells sometime in Q2-2009.

Hopefully all this means that cheaper solar panels can be used in homes and sufficient power can be produced at a good value-for-money. Today the cost of solar panels are way too high and produce too little power to make them practically replace normal power lines.

AMD Working on Intel’s Physics Havok

AMD and Intel are rarely to be found in friendly talk. But this time, AMD is in talks with Intel to integrate “Havok FX” physics in its ATI graphics chips which is meant to improve realism in games and physics calculations in scientific applications. Intel last year acquired physics technology company Havok, and AMD plans to use its API inside GPUs.

AMD, like Intel has plans to have integrated CPU+GPU processors, but plans to use Havok physics in discrete graphics. The main competitor in the physics processing technology is Ageia, which was acquired by Nvidia. With two different technologies for physics, AMD decision to use Havok has important implications for game developers as well as scientific application developers. Nvidia has been pushing Ageia’s Physx technology to game developers and AMD was starting to feel lost. Today’s discussion about AMD using Havok gives AMD an advantage and its called Intel marketing. Intel will be using Havok inside its CPU+GPU Larabee and developers will be magically tilted towards using Havok because of Intel’s large market share.

The official announcement about how much AMD has to give Intel for the technology is still not out, but surely good money will change hand. Physics will definitely play an important role in the future of GPUs, either in games or general purpose computing. Along with ray-tracing, Intel is betting a lot on physics for the future of advanced computing and HPC. AMD has lots of room to maneuver, since it can switch between GPUs and CPUs with good ease. Lets wait and see how everything falls into place!!

NVidia Wishes It Could Live Like Intel

Nvidia has been kicking quite a lot of mud on Intel. Recently, it started with Nvidia claiming the GPU is more important than the CPU, then it was calling Larabee a powerpoint slide and then claiming that Intel is a monopoly that doesn't want anyone else in the market. But in reality, NVidia dreams to be at Intel's position and its been proved once again with NVidia buying Utah-based ray-tracing company RayScale.

If you've been following the discussion, or you read the above links, then you'll realize how pissed off the Nvidia CEO Jen-Hsun Huang has got with Intel. The reason for this anger is that Intel has been researching on architectures that are supposed to replace GPU (graphic cards) from computers. If Intel succeeds, it means Nvidia are without a market. Intel calls these next-gen processors that will be able to process graphics along with normal CPU operations as Larabee. AMD is also working on something similar called Fusion. NVidia on the other than doesn't have much to bet on. Although CUDA is starting to make GPUs do the work of CPUs, its way more steeper upgrade curve. Imagine porting all your normal Apps to CUDA vs Optimizing your games for a CPU that can show graphics.

With all that trouble knocking on Nvidia's doors, obviously you can't blame Jen-Hsun Huang for his anger. But today's buy of RayScale could point to what NVidia could do in the future. Intel has been pursuing a lot of research on Ray-tracing as a replacement for the vector processing that we see on today's graphic cards. Ray-tracing can be done on CPUs or Larabee/Fusion and could render images on screen, with more realistic lighting and shadow effects. Currently it takes a lot of processing, but it isn't too far from becoming mainstream once we have optimized processors. NVidia may be preparing for such a product and RayScale could be its software partner to move in that direction.

Nvidia also does not have the great fabs that Intel owns. Intel's strength is its production and NVidia really can't manufacture its own designs. Thus, if Nvidia even manages to create a killer GPU+CPU, how can it create enough of these for the entire world?? AMD on the other hand with their ATI buy can switch anywhere between a CPU+GPU or GPU+CPU... But haven't you heard: If you've got one leg in one boat and the other leg in another boat, you end up catching your groin!!

Intel HexaCore Dunnington at 2.66Ghz

Intel has taped out its six-core processor running at 2.66Ghz, which it plans to release sometime in Q4 2008. Intel's six-core processor is currently codenamed Dunnington has 3 dual-core processors stuck with each other and run as a single packaged processor. The processor will require a new socket to run. But the good news is that Intel is able to extract good speeds from it.

Dunnington is a reply to AMD's 3-core processors (TriCore), which AMD had initially thought to be value-for-money for the customers. Intel on the other hand has no plans to sell Dunnington's for cheap and will be sold only as server processors, under the Xeon brand. The Dunnington processors have a huge cache of 16MB, which should be very attractive for applications that run a lot of threads. Another news has it that, Dunnington will be running a single thread per core, and does not use HyperThreading (HT) technology.

The last time I met an Intel Design Engineer from Bangalore, I had heard that Intel was much impressed by Sun's designs in Rock or Niagara 2. Lots of concurrent threads on a core are very good on servers. Intel seems to going forward with the core-race at the moment, but I expect a switch to more threads on a core very soon!!

Will The "Sun" Shine With x86 ??

Latest news has it that Sun Microsystems has bought a struggling x86 chip manufacturing called Montalvo Systems. Montalvo is known to have been in talks with Sun for quite some time and there were long running rumours that Montavlo would be bought. When I heard the rumours, for the first-time I had thought it would be crazy for Sun to get into x86, but now it seems they have actually bought the company for a bargain buy of $5.5 million.

Monalvo Systems is a fabless startup funded through venture capital. It raised about $73 million from the venture capitals, but needed more $100 million lately. It still hasn't made any money and hence was denied the request for additional capital. In comes Sun Microsystems and gets a bargain buy!! Now if you are like me, then you would be wondering, why would Sun Microsystems buy an x86 chip design startup. Montalvo doesn't have the clients nor the SPARC designs. Sun Microsystems has over the years talked us into the greatness of the SPARC architecture and x86 manufacturing is really not its game. But so wasn't Sun's software business...

Popular belief has it that Sun Microsystem is going to use Montalvo's technology in its SPARC chips. Montalvo seems to have an excellent IP and chip design team which Sun was interested in. Sun Microsystems also has an amazing chip design team and probably its the human resources which has resulted in this deal.

If this buy is another completely new strategy from Jonathan, then it would be one to really watch out for. Lately, Sun has been selling x86 servers with AMD Opteron and you never know what Sun is upto after the ticker change... From JAVA to x86 can't be a more dramatic change than the last time.

Only Intel Marketing Can Sell At 1500% Margins

How many businesses can sell chips at 1500% margins?? I bet you won't believe to buy processors at that price even if you were in Zimbabwe's inflation... But that doesn't stop Intel to sell its upcoming Atom processors at those margins, if you were to believe TGDaily. TGDaily noted earlier that Intel produces these  small processors at $8 including packaging and shipping. And according to the latest pricelist, Intel hopes to sell these at $135!!

Atom is the upcoming generation of low-power processors meant for MID (Mobile Internet Devices) that come in really small packages. I noted earlier about how much noise Atom has been generating even though its not the best performing part out there. But according to Anand, it seems Atom is some special change in architecture that's gonna change the future of computing. May be Anand's seen that 'Crystal Ball' of the future, but it still got me wondering about the hype and the pricelist just added up to the hype.

The price list is as seen below:

- Atom Z500, 800 MHz clock speed, 512 KB L2 cache, FSB400, 0.65 watt TDP, $45 (including chipset)
- Atom Z510, 1.1 GHz, 512 KB, FSB400, 2 watt, $45 (including chipset)
- Atom Z520, 1.33 GHz, 512 KB, FSB533, Hyper-threading, 2 watt, $65 (including chipset)
- Atom Z530, 1.60 GHz, 512 KB, FSB533, Hyper-threading, 2 watt, $95 (including chipset)
- Atom Z540, 1.86 GHz, 512 KB, FSB533, Hyper-threading, 2.4 watt, $160 (including chipset)

Hyper-threading is back and just like James Reinders had told me to expect it in every upcoming Intel chip, it isn't really a surprise. Technologically, there are quite a few new things in Atom, but claiming all of them to be path-breaking doesn't make good sense. Its more like bringing the original Pentium back to a new fabrication and adding some lessons learnt from Pentium M. I think any ARM design without much dramatic changes would easily be able to beat Atom today. And if we had Intel giving x86 licenses away, then we would really have an 'apple vs apple' comparison (pun intended!!)... But I'll talk about the architecture some other day...

Is Intel Playing Bully Again ??

You must be thinking Intel has gone crazy with the pricing on this one, but there is some logic behind it. And the logic is called Intel marketing... Intel has been known to sell processors that are not the best performing in the market at higher street price during 'NetBurst' era. AMD alleges that Intel sold its processors at lower prices to PC vendors on the condition that they would sell Intel-only machines. It had major market share then, and I don't think it will be any different now!! If you go by the same logic, probably that margin is only for the lame buyer and not for Intel buddies.

AMD Working With Intel

As odd it may seem, the story holds ground and is one of the useful steps for the Centrino 2 platform of notebooks. Fuad Abazovic of Fudzilla claims that AMD is working hard with Intel to develop switchable graphics in future Centrino laptops. AMD is doing this in the notebooks with Puma and wants to have similar technology in notebooks with Intel chips and ATI/AMD graphic cards.

The thing about switchable graphics is that when your laptop is connected to the power unit or had good battery, it will use the graphics card. When the battery is low, it will switch to the onboard graphics. This will save power and make your battery life increase a little. Its generally close to 5-10% improvement, we heard the last time. But for the dying moments of laptop battery, I have experienced that you always remember the important work.

If you look at the image above, you'll see that the plus sign is red, just like red-cross... That's because it's one of the good things being done on the battlefield between Intel and AMD. This news makes you think how bitter rivals AMD and Intel are working together?? But actually its ATI which is working with Intel and I'm sure AMD has told its engineers not to look, talk, have lunch or discuss anything with Intel engineers. They just make faces when they meet and exchange some white documents with each other... Infact, news is that some even show the "middle-finger" to each other :))

nVidia To Buy Via To Make x86 Processors

Digitimes is reporting that nVidia may be in talks with Via for an acquisition, but no deal was made due to the high price named by Via. The war between nVidia, AMD and Intel is heating up and it seems the future holds a single chip for graphics and general purpose computing. It means very soon your computer will have a CPU or a GPU (graphics card), not both like we have in today's computers.

Intel recently talked about Larabee and its future platforms and made it clear that its moving through paces with its development of graphics in the CPU. AMD with ATI acquisition is also moving there and "Fusion", is AMD's plan for CPU+GPU processors. This leaves nVidia alone in its GPU world and it may well be the time when it wants to start making x86 processors as well. Via on the other hand, already has an x86 license and has been making processors for low-power devices. But Via isn't making enough money by selling these processors... This is what encouraged nVidia to make a bid for Via and sure is a good explanation for nVidia to buy Via.

Via also has good processor designs in the form of Isaiah which holds good future in the UMPC world. nVidia might be interested in acquiring the CPU design labs, Centaur because of some good designs that have come from there.

From Digitimes:

Nvidia and VIA discussed three possible scenarios, including a strategic alliance, acquisition of only VIA's processor division, or the acquisition of the entire company, according to the sources. Though no deal was agreed upon, since VIA continues to see losses, the company is expected to open up negotiations again soon with a lower price, noted the sources.

It'll be an interesting time in the CPU/GPU world in the coming few months... By next year, we would know if "GPU killed the CPUstar or CPU killed the GPUstar"

Intel To Release Hex-Core Processors

According to Maureen O'Gara from sys-con.com, it seems that Intel is all set to release a processor with Six cores. Yes, you heard me right...six. I think Intel understands that if AMD can start making tri-cores then may be six cores isn't a very bad idea.

The six core or hexacore processor goes by the codename "Dunnington" and has been designed at the Intel Design Center at Bangalore. It is a successor to the Harpertown processors of the Xeon-line of server processors from Intel. It is again supposed to be a sandwich of 3-dies, but this time its 3 dual-core processors that have been stitched together. The cores are sharing a 16MB L3 cache. Like other Penryns, Dunnington still uses a front-side bus. Dunnington is supposed to come at different clock speeds, with a total power consumption of less than 120W.

Intel will be releasing these just before the Nehalem are out and should be an interesting development. AMD on the other hand is just struggling with speed of the processors as well as the manufacturing and releasing of its products. Intel seems to be double of whatever AMD wants to do. AMD Tricore Vs Intel HexaCore.

Open-Source Hardware & Gadgets

I just read this article at the NYTimes about Neuros OSD, an Open Source Device (OSD) for digital video recording. Like the name suggests, the device is open-source in hardware and software. But when I look at all the hardware and gadgets I own, I can't find a single one which is open-source. Infact, most of the gadgets don't even have a proper manual. You have to open the box and look through all the chip ID and then hope that somewhere the chip manufacturer has left something for you to munch! Lets not talk about the software side of these gadgets... its even worse.

The NYTimes article is not the first time that I've heard about open-source hardware or read an article glorifying open-source. I have attended a few conferences on open-source evangelism and read quite a few books and articles on it. I have also looked closely at it as a business model as well. And that brings me to the question as to why isn't open-source hardware popular?

If you look at the word "open-source", its been applied at a lot of places. And not just in software or computing. Its been adopted in education, arts, automobiles and quite a few other places. But then its in these sectors as well where you can count open-source adoption on your fingers. If I have to say it bluntly, "Open-source is not human-like" and hence it's not popular. The open-source ideology in software may be gaining popularity, but its more so because open-source evangelism gets more media coverage. An evidence to this fact is that while I'm writing this post, I also read about McAfee not giving back code to open-source project as part of the GPL (GNU Public License).  And I'm not writing this piece as an open-source supporter. Instead, I'm here to tell you that open-source hardware has a few pitfalls as well.

Open-Source Starts Excitement

There could be two diametrically reasons for the excitement surrounding open-source. It could either be because its atypical (aberrant of  human tendency) or its very close to man being a "social animal". Whatever be the reason, open-source does provide excitement. This excitement means you get more eyes to your project and more hands to help you expand the project. All of this means, you have an opportunity to sell your product. Open-source gadgets allow more people to tweak and play with it and in-turn the company sells more of the gadget. The user of the product also gets additional enhancements made by other users. Open-source is educational for the geek as well as more usable for the novice.

Whose Opportunity

So all these are good for the product that is open-sourced...right? Not completely. Didn't I just say its an opportunity to sell your product? But it is obvious that anyone with better manufacturing capabilities can have an even better opportunity. For hardware manufacturing, we all know China, Taiwan etc. can make it cheaper than others. The hardware that you designed and manufactured in another country can be easily replicated and sold cheaper. Currently closed gadgets are copied and sold cheaper. If your hardware is open-sourced, its infact an open-permit for the other to copy.

This is probably one of the main reasons why open-sourcing hardware/gadgets is not popular and never will be. Other reasons could be security loopholes, not-so-great designs, upgrade markets, dumping old components in newer models...

But this hasn't stopped open-source hardware to be released. There are many examples of open-source hardware and the list seems to be growing every month. Some popular open-source hardware projects include OpenSPARC (CPU), LEON (CPU), S1 Core (CPU), Simputer (Handheld), PLAICE (flash programmer), OpenCellPhone (Mobile Phone), AstFin (Telephony hardware), OSCar (Automobile), gEDA (Electronic Design Automation tools), Daisy (mp3 player), OpenStim (Brain Stimulator) and many many more.

Bad Never Happens to Good People

I can't forget the argument a student of mine gave in support of open-source. He said that "Atleast my contribution is changing the world"... And then another said, "If that contribution can earn you your bread..."... and I just smiled...

Complete Nehalem Details Leaked

True to the earlier reports about Intel's next chips having HT (Hyper-Threading) and integrated-GPU, some more details about Nehalem appeared on the web, courtesy of Japan's PC Watch.

The architecture diagrams mention about 2 threads per core, which was what I had heard from James Reinders earlier. The PCWatch page also has pictures of the updated Intel roadmaps. It also has features of all the different variations of Nehalem that Intel is going to release. The code names of these are Beckton (Oct-core MP), Gainestown (Quad-core DP), Bloomfield (Quad-core, EE-Desktop), Lynnfield (Mainstream Quad-core) and Havendale (Mainstream Dualcore)

 

If you can understand Japanese, then go here and read all about it or use a translator and try to figure out what it means!!

Intel Releases Peryn To Make Life Tougher For AMD

Guest Blogger: Steven Parker
12th Nov, New York

Its been tough time for AMD already, since Core2 Duo was launched. Intel then stuck two of these together and created Quad Cores. These Quad Cores have  been selling like hot cakes, and Intel's making lots of profit. Intel today released its 45nm Peryn processors which are smaller in size and cheaper for Intel to manufacture. This means Intel can earn bigger margins on the same quad core processors. On the other hand, when AMD releases the Phenom its margins will go down because of the large size and competitive pricing as Sunny observed here...

The Intel Peryn is basically the same quad-core but built using a completing new process as well as substrate and gate transistors. Intel manufactures the Peryn using Hafnium-Based High-k Metal Gate Transistors, which seems to be reduce current leakages through gates. Infact, Intel's founder Gordon Moore called the technology "the biggest transistor advancements in 40 years". This technology would allow Intel to make faster processors in the short-term.

From Intel: (click here to read the press released)

New to the Intel line-up of server processors are 15 server dual-core and quad-core 45nm Hi-k Intel Xeon processors. The 12 new quad-core chips boast clock speeds ranging from 2GHz up to 3.20GHz, with front side bus speeds (FSB) up to 1600MHz, and cache sizes of 12MB. The three new dual-core chips feature clock speeds of up to 3.40GHz, an FSB of up to 1600MHz, and cache sizes of 6MB.

The 45nm Hi-k Intel Xeon processors are compatible with server platforms using the Intel® 5000 chipset family. In addition, Intel is launching three platform solutions to support 45nm processors, including:
   The Intel® 5400 chipset-based platform (previously codenamed "Stoakley") that is optimized for high-bandwidth applications such as high-performance computing (HPC).
   The Intel 5100 Memory Controller Hub chipset and Intel ICH-9R I/O controller (previously codenamed "Cranberry Lake"). These are cost-optimized solutions that support either one or two processors and also provide reduced power consumption using native DDR2 memory.
   The Intel 3200 chipset-based platform (previously codenamed "Garlow") that is specifically designed for single-processor entry servers.

Supercomputer On A Graphics Card (1 TFlops)

Not so long ago, who would have thought that a desktop computer would be able to process "trillion floating point operations per second (TeraFlops or TFlops). Most people not up-to-date with technology would say its impossible for a normal desktop computer to do it, but it isn't anymore!! Really..and I'm not marketing for any company...

Today AMD announced the FireStream 9170, a General Purpose Graphics Processing Unit (GPGPU) running at 775Mhz and performs 497 GFlops. That's about 0.5 TFlops!! The FireStream is based on the RV670 chip and has 320 double-precision floating-point (FP) units. The RV670 will also be available in desktop variants for gamers and will be known as the HD 3850 and 3870 as mentioned by me here.

But it is still just 0.5TFlops... right?? Yes, but in about a few months time, AMD (actually ATI) will be releasing R680 which will have 2 of these chips working together. Add a few more clocks to the chip and we'll have a chip that can perform 1TFlops!! Aint that great??

Consider this: The world's fastest supercomputer i.e. IBM's BluGene/L scores 280.6 TFlops and costs a few billion dollars to build whereas one GPGPU will cost some thousand dollars. Do the math and you'll know its some million times cheaper with the GPGPU!! But the entire comparison I did is comparing apple to oranges. And the reason is because BlueGene/L uses general purpose CPUs which are easily programmed, whereas GPUs require special programming. I don't say that programming GPU is tough, it's just that we are habituated to program CPUs and hence we currently don't have the same type of tools that are used to program CPUs. But as the tools come up and mature GPGPUs are going to be the next milestone in supercomputing.

Intel also about a year back showed a 80 core processor do some networking to process as much as 1.81TFlops at 5.7Ghz... But its still got a lot more tweaking to do and the yield I think won't be very good at the moment!!

All of this means that in the very near future, by the use of GPGPUs, each one of us may be having a supercomputer built into our computers. But where is the work that these fast mean machines would be doing??

How Intel Got Its Name

Have a friend of mine who's attending the IDF and got some news on Intel history, that I thought I should write down. Intel is a very big brand in itself and the name is synonymous with desktop computing. We all know how popular their "Intel Inside" branding was. But just a small history on how Intel got its name Intel was revealed by the founder Gordon Moore at IDF 2007.
Gordon Moore and Robert Noyce left Fairchild Semiconductor, since Fairchild didn't believe in the future of integrated circuits as much as Noyce or Moore. So, when they started this new company, they named it Moore & Noyce. They did the paper work with this name for the first 8 weeks, but didn't think this name was catchy or it gave an idea about the company. Then, a while later Noyce's daughter, Penny Noyce suggest the name Intel (INtegrated ELectronics). This name was liked by both Noyce and Moore and they changed the name to Intel. And now, as we all know its the biggest manufacturer of x86 processors.

So now you know, how Intel got its name!! Some history lesson ha!!

Does Barcelona / Phenom really help AMD ??

While reading Inq Charlie's post - having lunch with AMD, I got this idea that AMD is in some deep trouble!! AMD launched it's Barcelona processor (3rd generation server CPU - Opteron) on Sep 10th, 2007 with a lot of excitement and with a hope to become competitive with Intel's (Clovertown) Xeon processors. And frankly, it does bring a lot of performance improvement by increasing the IPC, and gives better power usage. An overall 124% improvement compared to the previous generation of Opterons. Clock-vs-Clock AMD beats Intel hands down!! Everything seems to be perfect for AMD... Doesn't it??

But on launch, Barcelona doesn't beat Clovertown in the benchmarks. Know why?? B'Coz the Barcelona is clocked at 2Ghz while the Clovertown is at 3Ghz. But that's truely not the bad news for AMD. By the end of 2007 AMD may have a processor clocked at 2.4Ghz and will then beat or equal Intel's new Xeons at 3.2Ghz. So really the clock-speed is a short-term issue!! So do AMD fanbois start dancing??... No, there is another problem and this one is bigger and a long-term problem. Its not related to the server processor (Opteron), but to the desktop-variant of the Barcelona called Phenom X4.

Since Intel released the Core 2 lineup in 2006, the Athlon64 X2 has failed miserably. The result is that AMD today sells the Athlon64 X2 at very low margins and Intel hasn't helped AMD's cause by creating a regular price war. This means AMD doesn't make profits as much as it should to run the R&D and fab maintainance/upgrade.

AMD's desktop-variant, PhenomX4 will be released by year-end and will have to compete with Intel's Peryn CPU. The Peryn is not much of an upgrade but is a die-shrink of current Core2 Quad to 45nm (basically means smaller processor size i.e. die size). On the other hand, if we look at PhenomX4, its actually quite large compared to the current Athlon64 X2. In the processor world, larger the die-size, its generally accepted that the cost to manufacture is higher. This means that the cost to manufacture PhenomX4 will be higher than the current Athlon64 X2, while the cost to manufacture Peryn will be lesser compared to Core2 Quad. Now ask AMD to get into a price war with Intel!! Let's try to sell PhenomX4 at the price of Peryn!! For AMD, it'll mean even smaller margins than today while Intel can still keep its current margins for selling the CPUs.

The reason why this is a long-term issue for AMD because Intel is moving really fast in upgrading it's fabrication plants (processor manufacturing plants) to a smaller process, while AMD has already mentioned it needs some more time to upgrade. AMD also has a cash crunch, which means for the next upgrade, it needs some more time until its coffers fill again. So AMD is really not helping itself by releasing a larger-sized processor, but it may be the only option to survive the Intel performance assault!! Goodluck AMD!!

Intel 4004 isn't the world's first microprocessor

Recently while I was giving an introductory lecture on microprocessor to Computer Engineering students, someone asked me about the evolution/history of microprocessors. I'm really not very good at remembering dates, but then I started speaking about it. After sometime, when I reached the year 1992, and the Intel Pentium processor at 66Mhz, a boy stood up and asked me which was the first microprocessor. I showed him the board where I had written the CADC (MP944), and he showed me the book which showed Intel 4004. I said, "I'm pretty sure it is MP944, and I've been teaching that to a few students", but I'm bad with dates, so thought I should check.

Looking at all the textbooks and reference books suggested by the University syllabus, I realized that all books were claiming the Intel 4004 to be the first microprocessor. I was a little shocked, coz I remembered reading a paper here, and discussions with my senior professors and colleagues, that the MOS-LSI based MP944 was the first microprocessor. I did some analysis with my students, who had never heard of that chip and drew on the board what I remembered... (Here's a better one from Ray Holt, the chip architect himself).



Looking at the diagram, the student argued that it was a 4-chip design, and I showed him that 4004 was also a part of a 4-chip chipset. The date for the chip's papers were older as old as 1968, which made everyone sure that the books had wrongly called the Intel 4004, the first microprocessor. We later found this from IBM which agreed to my observations.

I hope most textbooks and authors should acknowledge Ray Holt's designs from AI Garrett and give the facts correctly.