TweakTown NewsRefine News by Category:
Samsung have teased their latest 32nm quad-core Exynos chip at the ISSCC event, which has been the first example shown of a quad-core Exynos design. The new chip still sports an ARM Cortex-A9, but is now built upon a 32nm process, which provides not only more speed, but better battery life when compared to 2011-era chips.
A new graphics core is featured with four pixel effects processors, and is expected to be around 26-percent faster, even though it uses 34- to 50-percent less power. Just like NVIDIA's Tegra 3 chip, it sports the ability to shut cores off independently to save power, and thus battery life, when a task isn't needed. Samsung's new Exynos chip can support up to dual-channel memory, with either low-power DDR2 or DDR2, to keep the chip working at full speed.
Graphics-wise, it's fast enough to render a whopping 57 million polygons per second. Samsung hasn't made made the new quad-core Exynos chip public yet, but the Mobile World Congress is next week where we're sure they'll want to show it off.
Get ready, people. Intel are ready to unveil a breakthrough System-on-a-Chip (SoC), which is said to be a new dual-core Atom-based chip, that sports a digital 2.4GHz Wi-Fi receiver, get this, built into the same die.
This new chip is dubbed "Rosepoint", and is built upon a 32nm architecture and will be shown off for the very first time as a research project at ISSCC 2012 in San Francisco. We should see devices sporting this technology to significantly improve power consumption and produce a more reliable Wi-Fi connection than current analogue-based receivers.
Intel Chief Technology Officer, Justin Rattner, says:
With a digital approach to radio, you can bring the benefits of Moore's law to RF and radio circuits.
In order to achieve this, Intel needed to find a way of stopping the radio waves from interfering with the CPU. Rosepoint steps in and stops this by employing noise-cancelling and radiation-shielding methods to enable the concept to work. At the moment, Rosepoint is just a working prototype, but Intel believes that they can develop the technology into a shipping product by 2015.
AMD have updated their Q1 2012 line-up to include three brand-spankin' new Bulldozer/Zambezi-based chips. This is all from a leaked roadmap that Donanimhaber published, which shows the there new chips as the FX-4150, FX-8140 and FX-6120.
These chips are said to hit anytime between now and April. All of the new chips sport TDPs of just 95W. AMD also have plans to launch quad-core-based Trinity APU models with A10 and A8 processors in the following quarter (Q2), which is said to be followed by A6 and A4 dual-core versions in Q3. I've listed the chip details below:
- Cores: 4
- Clock speed: 3.9GHz
- Turbo speed: 4.1GHz
- Cache: 12MB
- TDP: 95W
- Cores: 6
- Clock speed: 3.3GHz
- Turbo speed: 3.9GHz
- Cache: 14MB
- TDP: 95W
- Cores: 8
- Clock speed: 3.1GHz
- Turbo speed: 4.0GHz
- Cache: 16MB
- TDP: 95W
Intel Engineering Samples, back in my Celeron 300A days, those things were the thing to have. You were pretty hectic to own one of those bad boys and these days, with the likes of eBay's and mainstream (and more accepted) overclocking, it's not as cool, but it still gets my nerd juices flowing.
The latest ES sample comes in the form of Intel's Ivy Bridge-based Core i5-3570K. Henry posted two CPU pictures on the Expreview forums, where we can see the Core i5-3570K is pretty similar to the Core i5-2500K, but it's process has been shrunk to 22nm, clock speeds are up and TDP is down. We have clock speeds of 3.4GHz, and just a 77W TDP.
The IGP featured on the Core i5-3570K is Intel HD Graphics 4000, which is stamped into the K-series of processors, as well as some Ivy Bridge processors. HD Graphics reportedly sports 16 EUs, whereas the current HD Graphics 3000-series has 12 EUs.
Expreview's article on it goes into some incredible detail with some benchmarks with the 4000-series vs. the 3000-series, but an overall test featuring 3DMark Vantage, Left 4 Dead 2, Street Fighter IV, Starcraft 2, DiRT 3 and Far Cry 2 see the 4000-series come out on top to the tune of 67.25-percent faster. This is not bad, not bad at all, considering it's integrated graphics.
Intel have introduced the third member of the LGA2011 Sandy Bridge-E family in the form of the Core i7-3820. The chip joins the ranks with the current chips, i7-3960X and i7-3930K.
The new Core i7-3820 has a clock speed of 3.6GHz, has just four cores with four hyper-threaded cores (versus the six-of-each on the two other SB-E chips), 10MB of L3 cache, Turbo Speeds of 3.9GHz, a max overclock multiplier of 45x, a TDP of 130W and a price of just $294.
This new chip is the Sandy Bridge-E chip, on a budget. AnandTech give three reasons why you'll want the Core i7-3820:
- You need PCIe 3.0 today and/or you need more PCIe lanes than a Core i7-2600K can provide.
- You need tons of memory bandwidth for a particular application.
- You want a 2600K but you need a platform that can support more memory (32GB+).
Yep, that's enough convincing for me. Looks tasty indeed.
Intel's new processor architecture codenamed "Haswell" won't see the light of day until 2013, but that doesn't stop information from leaking out onto the Internet about it. The successor to Ivy Bridge now has some more details on its embedded graphics processor, with DonanimHaber detailing the integrated GPU (iGPU) plans.
Haswell's iGPU will be DirectX 11.1 compliant, which means it will take advantage of API optimizations that improve performance, for "typical desktop usage scenarios". It also supports OpenCL 1.2, which will help in GPGPU-optimized applications, and is designed around a new stereoscopic 3D standard called Auto-Stereoscopic 3D (AS3D), which will benefit Blu-ray 3D acceleration, stereo 3D photos, and more.
It would be compared to entry-level GeForce or Radeon GPUs for acceptable performance with stereo 3D.
Something else to look forward to is a little thing Intel refer to as "digital display repartition". At the moment, current Intel processors with embedded graphics have relied on the PCH to perform all display I/O functions, the iCPU communicates to the PCH over the Flexible Display Interface (FDI), a special data link dedicated to this exact function.
AMD's client roadmap has revealed Hondo, a 4.5W APU with 1-2 low voltage Bobcat cores as well as an on-die DX11 GPU built on a 40nm process. Windows 8 will be the OS to receive some of AMD's Hondo lovin'. AMD is wanting to get serious of entering the sub-2W market, with mobile being a very important path for AMD.
AMD has talked about being "ambidextrous" when it comes to architectures, implying that AMD's future products will include chips with ARM-based CPU cores for markets where it makes sense. AMD have also said they will take advantage of all sorts of ecosystems, including Windows 8 and Android.
The AMD 2013 Client Roadmap unveils some interesting things indeed, with 2013 mobile chips "Kaveri", "Kabini", and "Tamesh" being three new APUs to be introduced. Kaveri will sport 2-4 "Steamroller" CPU cores, as well as a GCN-based GPU with HSA Application support. Kabini will feature 2-4 "Jaguar" cores backed up by a GCN-based GPU, and finally, Tamesh sports 2 "Jaguar" CPU cores, and a GCN-based GPU.
Desktop wise from AMD, we should expect the second-generation FX CPUs codenamed "Vishera" which will include 4-8 "Piledriver" CPU cores, "Kaveri" APU which will sport 2-4 "Steamroller" CPU cores, GCN-based GPU, HSA Application support and finally, "Kabini" which is identical to its mobile counterpart, sporting 2-4 "Jaguar" cores, and a GCN-based GPU.
IBM, hard at work obviously, have developed the smallest carbon nanotubule transistor, measuring in at just 9 nanometers (nm) across. Currently, the smallest transistors possible using silicon is 10nm across, so while it doesn't sound like a huge achievement, when talking in nm, it's actually quite the achievement.
IBM claims that the new transistor consumes less power, all while being able to carry more current than today's technology. John Rogers, Professor of Materials Science at the University of Illinois at Urbana-Champaign, says:
The results really highlight the value of nanotubes in the most sophisticated type of transistors. They suggest very clearly, that nanotubes have the potential for doing something truly competitive with, or complementary to, silicon.
The smallest production-grade transistors are currently limited to 22nm.
Ivy Bridge is just weeks away, but this hasn't stopped Intel from releasing some new parts. Intel updated their processor price list over the weekend, debuting three new Core i5 desktop parts, and four single- and dual-core Celeron mobile parts.
The new Core i5 models are: Core i5-2550K, Core i5-2450P, and Core i5-2380P, with prices of $225, $195, and $177, respectively. All three of the new Core i5 models are quad-core models, which handle four concurrent threads, include 6MB of L3 cache, and a TDP of 95W. All three models do not include integrated graphics.
Clock speeds on the new models are 3.4GHz with Turbo Boost speeds of 3.8GHz, 3.2 and 3.5GHz, 3.1 and 3.4GHz for normal and Turbo Boost for the Core i5-2550K, Core i5-2450P, and Core i5-2380P, respectively.
The new Celeron mobile parts are the Mobile Celeron B815 (1.6 GHz, 512KB L2, 2MB L3, two cores, $86), B720 (1.7GHz, 512KB L2, 1MB L3, two cores, $70), ultra-low voltage 867 (512KB L2, 2MB L3, 2 cores, $134) and 797 (1.4GHz, 512KB L2, 1MB L3, single core, $107).
Globalfoundries CEO has called last quarter a "remarkable quarter," stating that Globalfoundries is back on track to "keep the momentum going," after a rough 12 months.
Globalfoundries' new fab in upstate New York is expected to ramp up 20nm tech in June, with Globalfoundries expected to spend more than $3 billion in capital expenditures this year alone. Last year, GloFo spent $5 billion on capex, so $3 billion doesn't seem as much now.
GloFo's Dresden-based facility is said to continue with its 32 and 28nm manufacturing, with plans to shrink all the way down to an insane 14nm. GloFo's CEO said during an interview with EE Times, that their competitors have only shipped "a few thousand wafers" when compared to GloFo's 700,000-plus.