Processors - Page 111

During the 2015 IEEE international Solid-State Circuits Conference (ISSCC) in February 2015, Intel will reveal more technical details about its upcoming Haswell-EX processor. This processor will be the most complex CPU the company has ever made.

Intel's Xeon E7 v3 "Haswell-EX" processor will work in motherboards with up to eight sockets, packing 18 cores (and 18 Hyper-Threaded cores) for a total of 36 threads per CPU. We will also have 45MB of last-level cache (LLC), quad-channel DDR4 support, PCI Express 3.0 and much more. The new Intel Xeon E7 platform will usher in new reliability, availability, scalability (RAS) capabilities, something that will bring Intel closer to the older Itanium-based servers.

The Haswell-EX processor will be using Intel's 22nm Tri-Gate technology, packing in an insane 5.56 billion transistors, making it one of the most complex x86-based processors ever made.

As a result of a class-action lawsuit against HP and Intel, the chipmaker will have to give $15 to those who have purchased an Intel Pentium 4 processor about 15 years ago.

As of now, this applies to those who are residing in the United States and have purchased computers for personal/general use with Intel Pentium 4 processors between November 20, 2000 and June 30, 2002. The lawsuit points out that Intel and HP have deliberately manipulated benchmark scores for the Intel Pentium 4 processors at the time it was facing tough rivalry from AMD. There were also allegations that Intel Pentium III and AMD Athlon line ups at the time performed better in comparison to Pentium 4 line ups.

The lawsuit states that Intel secretly wrote benchmarks which would favour Pentium 4 processors. The company also paid software companies to make changes to favour Pentium 4's performance scores for third-party benchmark software, so that it will stand out against AMD. The benchmarks that were in question were WebMark2001 and SysMark 2001. Both companies have denied these allegations but said that they were willing to settle the matter via compensation. You do not require to show a purchase invoice that you've bought a Pentium 4 powered PC during that time frame, but you will need to present some proof such as the retailer's details and date of purchase.

Intel may have just launched its Haswell-E and X99 platform, but that doesn't stop the chipmaker from teasing its upcoming HEDT (high-end desktop) processor train from slowing down. We're now hearing about the next-gen HEDT tech, Broadwell-E, which will be based on Intel's 14nm technology, using the same LGA2011v3 package.

The new CPU will not be an architectural change, but it will provide smaller changes over what we have with the current Haswell-E processors. The new Core i7 Broadwell-E will be built on Intel's 14nm process, and will feature between 6 and 8 cores based on their, you guessed it, Broadwell microarchitecture. These cores will feature up to 20MB of L3 cache, and is pin-compatible with current Haswell-E, meaning we have quad-channel DDR4, too.

Intel could provide the full 40-lane PCIe interface, instead of the cut down 28-lane PCIe interface that the entry-level HEDT currently has. We should expect a 140W TDP, even with the die shrink, when the Broadwell-E processors launch in 2016.

AMD would be announcing its next generation notebook APU Carrizo-L in December. The processor is based on a 28nm quad-core architecture based on their Excavator core.

The notebook APU will have support for 2133MHz DDR3 memory. The news reported pointed out that Carrizo-L will succeed AMD's Beem and Mullins APU which is currently positioned for entry-level notebooks and tablets. It was also pointed out that this will also be 'officially' compatible with Windows 10. The rest of the operating systems, such as Windows 8.1, Ubuntu and SLED operating systems were added on the list.

The APU is designed for entry-level notebooks and will be placed to compete against Intel Pentium and Celeron series processors. But as far as mainstream segment is concerned, AMD would not be releasing the full-fledged Carrizo APU before March 2015 which will be succeeding after the long running Kaveri APUs. According to another report, Carrizo will be supporting both DDR3 and DDR4 along with an on-package memory die. It is also speculated that Carrizo series will have a desktop APU variant, which will be using the existing FM2+ socket motherboards. The expected timeframe for the desktop variant is also assumed for March 2015 launch.

AMD has announced a new multi-year agreement with Synopsys IP that will see the chipmaker receiving a slew of Synopsys DesignWare intellectual property on its advanced 16/14nm technologies, as well as its upcoming 10nm FinFET technology. AMD will be handing over specific IP and engineering resources to the company. Considering NVIDIA just catapulted it's more-than-impressive GeForce GTX 900 series, there's never been a better time for AMD to partner up with someone who can handle the move to smaller processes.

The agreement sees AMD securing interface, memory compiler, logic library and analog IP from Synopsys, where it will use these technologies to create future generations of its chips on the 14nm and 16nm FinFET manufacturing process, eventually moving onto the 10nm process down the track. Synopsys will reportedly hire around 150 of AMD's IP & R&D engineers and receive access to AMD's leading interface and foundation IP. AMD will be saving money with this deal, but provides some holes in its resources, while Synopsys is only gaining from this deal.

If you've never heard of Synopsys, they are a leading power in silicon-proven IP for advance process technologies, with the company helping chip designers on a broad range of high-end IP for integration into system-on-chips, or SoCs, as well as delivering expert technical support. This power allows companies like AMD to come to them, in order to save money on pumping into their own R&D. But, AMD still packs a punch when it comes to the complex IP used in advanced microprocessors and GPUs. AMD will gain silicon-proven IP for its chips over the coming years, while handing over interface and foundation IP, as well as engineers to Synopsys, something the company explains will give it the ability to "focus its valuable engineering resources on its ongoing product differentiation and IP reuse strategy".

We found out not too long ago that NVIDIA was suing Samsung and Qualcomm, without going after any other companies, even if those companies used chips and parts from Samsung and Qualcomm, but now we might have found out why: Samsung is rumored to be working on its own GPU.

The news is coming from Fudzilla, and is just a rumor right now, but the company has been reportedly hiring people from the likes of AMD, NVIDIA and Intel. If Samsung were to be building its own GPU, it would be competing directly against Qualcomm and NVIDIA, with the latter having a very capable SoC with its Tegra K1 processor.

If Samsung did build its own GPU, it would save itself from having to license one from another company, as it would have nearly all of the components it needs for a flagship device built-in-house, from the screen, right down to the GPU.

Intel has just launched its new high-end Haswell-E platform, but what is AMD doing? Well, according to a recent interview with Bloomberg, AMD won't be releasing a new micro-architecture until 2016, with any CPU or APU products released between now and then based on current architecture.

AMD CEO, Rory Read, talked with Bloomberg, but didn't reveal any information on future microarchitecture, but he did say that the hardware coming out next year will be based on existing architecture, and won't be much better than what AMD has on the market now. Read said: "AMD engineers are now proving they can deliver new designs on time, something that didn't happen in the past."

In 2015, we can expect AMD to release new APUS that will be based on the low-power Puma+ and high-performance Steamroller architecture. Both of these architectures aren't expected to deliver much additional performance, but we should expect lowered power consumption and heat output.

Intel will be launching its new Haswell-E based Core processors tomorrow, but some leaked benchmarks are already surfacing over at Videocardz and WCCFTech. When it comes to games, the new Core i7-5960X is around 14% faster than its predecessor, the Core i7-4960X.

The new Core i7-5960X is Intel's first 8-core processor for the consumer market, with a stock frequency of 3GHz, and Boost frequency of 3.5GHz. We have 20MB of L3 cache, 140W TDP and support for DDR4 memory. We should expect a price of $999, which isn't too bad for a processor of this calibre.

When it comes to 4K video editing the new Core i7-5960X is around 20% faster than the 4960X, and around 32% faster in 3D rending. 'Thayn3' in the coolaler forums was able to overclock the Core i7-5960X to 4GHz using just 1.2V, but there has been an insane overclock found online, with the new 16-thread CPU clocked up to 6.2GHz on LN2.

It shouldn't be long until Intel officially launches its new X99 chipset along with a slew of new high-end processors, with the star of the Haswell-E show being the upcoming Core i7-5960X processor. This new CPU has been spotted in some newly leaked photos that Hermitage Akihabara got its hands-on.

Intel's new LGA 2011-based Haswell-E processors are expected to be released on August 29, with three models to be unveiled: the Core i7-5960X, the Core i7-5930K and the Core i7-5820K. The top-of-the-line Core i7-5960X will have eight physical cores and eight provided through Hyper-Threading for a total of 16 threads - a monster of a consumer CPU.

The new Core i7-5960X will also feature 20MB of L3 cache, quad-channel DDR4 RAM support, and 40 PCIe 3.0 lanes in total. The default clock speed on the Extreme CPU will be 3GHz, and it'll be built on Intel's 22nm process.

NVIDIA's Tegra K1 processor is quite the performance powerhouse, with a quad-core processor with four A15 CPUs, up to 2.3GHz clock speed, and a 192 Kepler-based GPU cores for the graphics side of things. We've seen the Tegra K1 power NVIDIA's cheap, but very powerful Shield Tablet, but the company is already showing off the next version of its SoC.

At HOT CHIPS, a technical conference in the world of high-performance chips, NVIDIA has unveiled more details on the 64-bit version of its Tegra K1 processor. The 64-bit Tegra K1 is powered by the 192-core Kepler GPU, with NVIDIA's own custom-designed 64-bit, dual-core "Project Denver" CPU, which is fully ARMv8 architecture compatible. The big shift here is that the Denver part of the Tegra K1 is a dual-core variant, with a clock speed of up to 2.5GHz, but is 64-bit capable. The current Tegra K1 is a quad-core chip, with 32-bit capabilities. This makes the 64-bit Tegra K1 the world's first 64-bit ARM processor for Android, demolishing the competition when it comes to performance.

NVIDIA has used some clever optimizations, as well as its advanced technology in its Denver CPU cores, to deliver performance from its dual-core Denver-based Tegra K1 that rivals even four or eight-core CPUs that we find in our mobile devices today. Better yet, The 64-bit Tegra K1 processor offers PC-class performance, extended battery life, better gaming and multi-tasking, and much more. NVIDIA will see its 64-bit Denver-based Tegra K1 processor baked into mobile devices later this year, with the company also teasing that it is already working on support for the upcoming release of Android L on its 64-bit Tegra K1.

Intel is facing troubles with its schedule of its 14nm manufacturing process, however the chipmaker said that this won't affect 10nm fabrication's schedule. Intel may be under the pressure to reassure its investors as its postponed its 14nm processor production plans that was supposed to roll out from its Fab 42 plant in Arizona, USA. 10nm is scheduled for mass-production for 2016.

Intel's CEO Brian Krzanich said during its quarterly conference call with financial analysts and investors,"We have done no changes or shift to our 10nm schedule, but we will not really talk about 10nm schedules until next year". However, Intel didn't reveal details about the production of these chips.

This might be the reason why Intel may show-off its first 10nm wafer during the upcoming Intel Developer Forum 2014. The demonstration of these wafers should reinvigorate investor's faith in Intel's schedule and in its tick-tock strategy, despite 14nm delays. It is also rumoured that Taiwan-based semi-conductor maker TSMC is also making plans to fabricate 10nm chips, which may also pressure Intel to go ahead of schedule with its 10nm roadmap.

It looks like things could get quite good at the Intel Developer Forum (IDF) in September, according to DigiTimes' sources. These sources have said that Intel will show off its 14nm processors in September, but it will also be teasing its 10nm wafers at the event, too.

DigiTimes' sources said: "Intel will release its 14nm Core M-series processors in the fourth quarter and 14nm Broadwell-based processors in January 2015". Intel is expected weaker-than-expected yields, and has a lot of 22nm-based processors in its inventory, and mixed with poor PC demand right now, Intel has reportedly "postponed 14nm processor production, which is planned to be conducted at its Fab 42 in Arizona, the US", according to these sources.

According to these sources, we should expect TSMC to pump up the mass production of its 20nm process in Q3 2014, where it will announce its 16nm FinFET process in 2015, followed by a 10nm process that will enter mass production in 2016.

It seems that AMD is working on a new APU using 28nm process and stacked DRAM, codenamed 'Carrizo'. It is said that these APUs will benefit from HBM (Higher Bandwidth Memory) implementation compared to current DIMM slot counterparts.

Though the reports are unconfirmed, it is known that AMD is collaborating with Hynix to make stacked DRAMs. The HBM provides higher bandwidth which will benefitted by the APU especially by the onboard graphics core. The APU will be made with 28nm process, but the onboard HBM die will be based on 20nm process. Its speculated that Carrizo's APU core die size is smaller than Kaveri.

HBM can provide maximum bandwidth of 128-256GB/s, which will prove to be a better implementation over DDR4 support. These APUs will most likely use the FM2+ socket and maintains 65w TDP envelope. If AMD incorporates on package DRAM solution, it will allow higher speeds for the memory and have lesser latency even compared to DDR4 implementation and it would cost lesser than integrating L3 cache. Whether the stacked DRAM be implemented in all of Carrizo APU lineups and feasibility especially for low-cost APUs is currently unknown.

IBM thinks that the days of silicon are numbered, as it spends $3 billion over the next five years on finding ways to create the future generations of microprocessors. Senior VP of IBM Systems & Technology Group, Tom Rosamilia, says: "We really do see the clock ticking on silicon".

Right now, IBM's very latest silicon components are baked onto a 22nm process, but the company is looking five years into the future where parts will become so small that it will be hard to maintain a reliable on and off state. Rosamilia adds: "As we get into the 7 nanometer timeframe, things really begin to taper off".

This has IBM looking at new ways of making components work, funding this new set of research. The company has faith in an alternative to silicon, something known as carbon nanotubes. The concept of this technology still needs considerable work if it hopes to see the fabrication of carbon nanotube-based processors as an alternative to silicon. Another route that IBM could go into is silicon nanophotonics, which uses light instead of electrical signals to blast data around the chip.

TSMC is reportedly increasing the development of its upcoming 10nm process so that it can better prepare itself against Samsung, which has reportedly received an order from Qualcomm to build 14nm FinFET chips, reports DigiTimes.

DigiTimes writes: "TSMC and Samsung are currently competing fiercely in the development of FinFET process, with the Korea-based foundry house utilizing a 14nm process and TSMC a 16nm node. Both the 14nm and 16nm processes are scheduled to enter volume production in early 2015". TSMC has been at the forefront of FinFET development, with plans to begin producing 16nm FinFET chips in Q4 2014.

DigiTimes' sources have said that TSMC has rescheduled its commercial production for the 16nm FinFET process, pushing forward with the more advanced 16nm FinFET Plus process. This process will consume less power, and shrink die sizes even more. TSMC is running scared at the moment, as it didn't anticipate Samsung to develop its 14nm process so quickly, so now the Taiwanese company is accelerating its development of the 10nm process, to continue staying out ahead of its competitors.

If new reports are to be believed, VIA is working on the next-generation version of its x86-based CPU architecture, Isaiah. The company's last try at the x86 CPU game was back in 2006, but the latest version looks like it'll be released in just a few weeks time.

German site 3DCenter is reporting on some of the performance comparisons with other low-power chips, such as AMD's Kabini APU which has a 25W TDP, and a mobile Bay Trail processor from Intel. Isaiah II looks like it can keep its own, but we don't know how genuine these benchmarks are, if they are genuine at all. Centaur is the name of VIA's CPU division, with its website teasing an upcoming refresh which will happen on September 1. This could be the date when we learn about the new CPU, so we'll be keeping our eyes open.

We can also confirm hearing stories from certain VIA employees at Computex that its next-gen CPU may end up being able to compete with Intel Core i5 level parts at low power consuming levels.

The Russian Industry and Trade Ministry has announced plans to replace the US-made processors from companies like Intel and AMD, with its own x86-based processors. These new processors will run on a new Linux-based system, with a CPU built-in Russia called Baikal.

Baikal is being made by an electronics division of T-Platforms, a supercomputer maker, and looks to have some serious funding being pumped into it. Rosnano, a technology firm, and Rostec, a rather large defense contractor, are both chipping into the project. The first processors off the production line will feature an ARM Cortex A-57 at 2GHz, and will run both PCs and servers.

Each and every year, the Russian government reportedly purchases 700,000 PCs which costs around $500 million. On top of this, the government spends a further $300 million acquiring 300,000 servers per year. The new Baikal processors should begin replacing the Intel- and AMD-powered machines starting in early 2015.

Intel is about to launch its 16-threaded (but 8-core) processor in September, and while that is for consumers, what is being played with behind closed doors in experiments is incredibly exciting - with a new 36-core processor teased by researchers at the International Symposium on Computer Architecture.

Li-Shiuan Peh, the Singapore Research Professor of Electrical Engineering and Computer Science at MIT, has said that the future of massively multi-core processors will be more like little Internets, where every core packs a router, with data travelling between cores in packets of fixed size. Peh's group unveiled a titanic 36-core processor that features this "network-on-chip" at the event.

Today's processors are connected by a single wire, and feature between 2 and 6 cores, with the multiple cores needing to talk to each other through exclusive access to the bus. But, this way won't work as the core count increases, as the other cores will be waiting for the bus to free up, rather than performing the duties you've set it out to do. With the network-on-chip, each and every CPU core is connected only to those that are directly next to it. Bhavya Daya, an MIT graduate student in electrical engineering and computer science explains: "You can reach your neighbors really quickly. You can also have multiple paths to your destination. So if you're going way across, rather than having one congested path, you could have multiple ones".

After watching season three of Game of Thrones, Intel's Knights Landing just reads and sounds like Kings Landing... but, onto the news. Intel has just announced its next-gen Xeon Phi chips, codenamed Knights Landing, at the International Supercomputing Conference being held in Leipzig, Germany.

The new processors will be capable of delivering close to three times the peak performance power of its predecessor, Knights Corner. The new Xeon Phi CPUs will use a new high-speed fabric technology - Intel's own Omni Scale fabric - that should help with performance, scalability, reliability, power and density requirements to speed up the rate of scientific discovery.

Intel's new Knights Landing processors will use 16GB of stacked memory, something that will be based on Micron's incredibly fast Hybrid Memory Cube technology. This technology is capable of 15x the bandwidth provided by DDR3, and 5x the bandwidth of DDR4, all while being 5x more power efficient, and only requiring 1/3 of the space.

Intel will reportedly launch its next-generation high-end desktop (HEDT) processors in mid-September, with the new Haswell-E chips arriving with the new X99 desktop platform too. We should see Intel officially launch its Core i7-5960X, the Core i7-5930K and the Core i7-5820K on the 14th of September.

On the same day, we should expect the chipmaker to launch its new X99 chipset, designed to power the new CPUs. One of the exciting new things about Haswell-E and X99 is that it supports DDR4 RAM, and that the top of the line Core i7-5960X is a 16-theaded CPU (8 cores, 8 HT threads = 16 threads total). The Core i7-5960X will arrive with a base clock of 3GHz, 20MB of L3 cache, support quad-channel DDR4 RAM up to 2133MHz, and will be an unlocked part for overclockers to have some fun with.