TweakTown NewsRefine News by Category:
Deep neural-networks require a tremendous amount of power to actually be as effective as the human brain. Sure your Tesla Model S might ave a small DNN powered by NVIDIA inside, but it isn't nearly complex enough to to provide a full true-to-life aritifical intelligence experience. New breakthroughs from MIT might be able to provide full-on human-brain inspired AI experience on your phone.
Researchers at MIT presented a new chip mobile chip that's designed specifically for neural networks and it happens to be 10 times more efficient than any mobile GPU currently in production. They're calling it the "Eyeriss" and the researchers are hoping that it can potentially change small-device computing. Just imagine having Siri or Cortana being that much more useful because they've got the processing power local to them.
And beyond that this innovation could help to further develop the idea of the Internet of Things, where powerful AI programs can communicate with other devices and coordinate tasks to get things done nearly invisible to the user. The possiblities are endless with the way that individual small-machines needn't be connected to the Internet itself for the compute power itself, but instead merely for communication.
Announced in a recent press release, AMD has launched its new desktop processor cooler called Wraith alongside a slew of new near-silent operating CPUs and stock cooler packages.
The AMD Wraith cooler is designed to work in a whisper quiet way, further providing a white LED illuminated AMD logo on the black plastic shroud. Promising to operate at whisper-quiet sound levels, the Wraith is rated by AMD to clock only 39 decibels when operation.
Announced alongside the Wraith cooler are the all-new AMD 10-7860K and the Athlon X4 845 processors, both coming packed with a new 95W stock fan cooler. These two new processors are not the only models to share the updated cooler, with AMD A8 processors such as the 7670K and 7650K joining in the action alongside the Athlon X4 870K and X4 860K.
AMD has been in the headlines quite a bit this year, and we're only a month into 2016. According to the latest information, AMD is working on a new EHP (or Exascale Heterogeneous Processor) that has been confirmed by an AMD engineer, who has also confirmed the name of the processor, too: Zeppelin.
Zeppelin is a new Multi Chip Module, that features 32 Zen cores tied into AMD's impressive Coherent Fabric - which offers 100GB/sec of bandwidth - compared to PCIe that offers around 15GB/sec. The latency has also been reduced from 500ns to something much smaller, but the bigger question is whether Zeppelin refers to the Zen processor cluster or the MCM as a whole.
Digging into Zeppelin, we find the Vega 10-based MCM. Vega 10 will sport up to 4 GMIs (Global Memory Interconnects), which allow the CPU to talk to the GPU at 100GB/sec, courtesy of the Coherent Fabric. The MCM is also capable of communicating with the RAM at 100GB/sec too, which allows for a super-fast system with magnitudes more bandwidth than we have right now.
What we do know, is that the Zeppelin-based EHP variant will feature next-gen Vega 10 graphics, but we don't know how powerful it will be. We should expect AMD to push more details out on Zeppelin, Vega 10 and more as the months fly past.
During TSMC's earning's call, they revealed that they're expecting to begin the process of switching over to 10nm production, ramping it up fully into 2017.
16nm and FinFet are relatively new processes with GPU manufacturers just now taking advantage of a more complete production process and higher yields. Now that those are mature processes they can concentrate on perfecting 10nm, which will likely be somewhat of a challenge, though the market should transition to the new process as demand for 20nm and even 16nm slowly drops as they ramp up. They plan on capturing a large part of the market share with their early introduction of the process.
They've already taped out working devices based on the node, but aren't quite ready for full production. They did invest heavily into research into 10nm back in 2014. We should see actual hardware using this process, likely mobile hardware first, in 2017. Intel is also slated to have their 10nm node ready in 2017.
A couple of weeks ago it was made known a bug with Intel Skylake CPUs was causing systems to freeze up. Very shortly after, MSI released a fix for it, and now ASRock has followed suit.
While Intel has made it clear the issue only occurs when running intensive applications like Prime95, for what it's worth we've heard user reports that say otherwise. These same users claim updating the BIOS does indeed cure the instability.
We now know that AMD will be launching its Zen-powered Summit Ridge FX CPUs late this year, after being in development for the last two years.
AMD will launch its new processor families; Summit Ridge and Bristol Ridge, later this year - both hitting the enthusiast and mainstream markets. The news came from AMD's recent earnings call - where they reported a 28% decline in revenue. The Zen architecture is super-exicting, as it will be a completely revised design, more CPU cores, Simultaneous Multithreading Support (SMT), high-bandwidth, low-latency cache on the FinFET process, and so much more.
Next year, Intel will launch its new 10nm processor family with the Cannonlake CPUs, replacing the current 14nm processors. The CPU family to replace Cannonlake in 2018 will be Ice Lake, which is also based on the 10nm node, which will be followed by Tiger Lake in 2019.
Intel has been having issues in getting their 10nm-based Cannonlake CPUs to the market, but we had these issues with the 14nm-based Broadwell processors - where they were delayed by months in both the desktop, and mobile markets. Intel will launch the Cannonlake processors in 2017, a year after the upcoming Kaby Lake CPUs debut. After that, we'll see Intel introduce its 10nm-based Ice Lake processors in 2018, followed by the Tiger Lake processors in 2019 - both based on 10nm.
But now Intel is facing competition in TSMC, which has just announced their 7nm node for 2017, which will be followed through with 5nm in 2020.
Last week we reported on a discovered bug with Intel Skylake CPUs causing freezing in specific usage scenarios. If you're running Skylake in combination with an MSI motherboard, you'll be pleased to know MSI has released a fix for its Z170A XPower Gaming Titanium and Z170A Gaming M7 boards and will be releasing it for the rest of its Skylake lineup later on.
The company has confirmed the bug and says it will very rarely affect typical users. But, as they say, it's better to be safe than sorry.
You can grab the update here. Note it is a beta BIOS update, so if everything has been running smoothly for you, we recommend waiting for a non-beta update, given how delicate the BIOS can be.
The latest rumor for Intel is that the chipmaker will launch its new Broadwell-EP processors later this year, with a new Xeon E5-2602 V4 and a huge clock speed of 5.1GHz.
The Intel Xeon E5-2602 V4 processor will not only sport a clock speed of 5.1GHz, but 10MB of L3 cache, with each core featuring 2.5MB of shared cache, with a TDP of 165W. This would make the Xeon E5-2602 V4 processor have the fastest default frequency of any processor on the market, surpassing the AMD FX-9590 processor and its huge 5GHz frequency.
The problem is, Intel's super-quick Xeon E5-2602 V4 processor will only be made through specific OEMs, and will not be pushed out into the world for public consumption.
Samsung not long ago elected to use its in-house Exynos chip over Qualcomm's Snapdragon 810 in its Galaxy S6 and S6 Edge; today it confirms it has begun manufacturing Snapdragon 820 mobile processors utilizing its new 14nm LPP process for the semiconductor company. The new process is said to deliver up to 15 percent faster speeds and 15 percent less power consumption over its 14mn LPE process (found in the Exynos 7 Octa chip).
Samsung wouldn't say whether the new chip would find its way into its future phones, but rumors indicate a Galaxy S7 is on the horizon and will indeed contain the 820. Patrick Moorhead, president and principal analyst at Moor Insights & Strategy (and formerly of AMD), says, "I do believe Samsung must use the 820 to be more competitive with Apple. 820 hits performance per watt levels Exynos just can't hit."