Renowned overclocker Elmor has set a new world record by pushing Intel's i7 6950X an insane 91% past its base clock speed.
At this year's Intel IDF 2016 "Technical Innovation" event, Elmor was able to overclock Intel's Broadwell-E i7 6950X CPU to 5731 MHz, which is 91.06% past the CPUs base clock speed. Elmor achieved this speed by using liquid nitrogen cooling on an ASUS Rampage V Edition 10 X99 motherboard, an ASUS GeForce GTX 980 Ti Matrix video card, and G.Skill's Trident Z DDR4 RAM.
For more info on Intel's Broadwell-E i7 6950X Extreme Edition CPU, check out our official review.
Intel used a benchmark slide during its presentation at ISC 2016 to detail their Xeon Phi processors being 2.3x faster than NVIDIA GPUs in neural network training. According to Intel's data and comparisons, the Xeon Phi is also 38% better at scaling. Well, NVIDIA isn't happy and it's ready to fight, taking to a new blog post saying "We'd like to address these claims and correct some misperceptions that may arise".
First off, NVIDIA claims that Intel used Caffe AlexNet data that is 18 months old, with four Xeon Phi servers against four Maxwell-based GPUs. The latest version which is publicly available, and NVIDIA providing a link to it in their blog, sees NVIDIA's four Maxwell GPUs delivering 30% faster training time than the four Xeon Phi servers. Using four of the new Pascal-based Titan X GPUs, training is a whopping 90% faster than four of Intel's Xeon Phi servers, and if you want to put the boot heel on Intel's neck, just a single DGX-1 machine from NVIDIA is 5x faster than four Xeon Phi servers.
NVIDIA goes into great detail on its blog so be sure to check it out, with the post finishing with NVIDIA saying "It's great that Intel is now working on deep learning. This is the most important computing revolution with the era of AI upon us and deep learning is too big to ignore. But they should get their facts straight".
AMD is working towards Zen as we speak, but more leaked details have arrived on the 4-core and 8-core CPUs which will feature 8 and 16 CPU threads in total, respectively. The new details are on the clock speeds and TDP, with the 16-threaded Zen CPU clocking in at 2.8Ghz base frequency and up to 3.2GHz boost. The new 16-thread CPU will be a 95W part according to the leaks.
If we compare it to Intel's Broadwell-E based, 16-threaded Core i7-6900K processor, which is also made on the 14nm process, Intel has a higher 130W TDP. AMD's 8-core/16-thread Zen-based CPU has a TDP of just 95W in comparison, 35W less than Intel's offering with the 6900K. AMD has some great efficiency tricks up its sleeve with Zen, but the 14nm process used by Intel is different to the 14nm FinFET process that Samsung and Globalfoundries use for AMD's new CPUs.
Performance wise, we already expect over 40% more instructions per cycle (IPC) over the previous generation CPU architecture from AMD, codenamed Excavator. The older FX-8350 from AMD was an 8-threaded CPU (4 physical CPU cores) codenamed Orochi, while the Zen-based Summit Ridge CPU is twice as fast in Cinebench 15, according to AMD. If we're looking at twice the efficiency per core, the new 4- and 8-core processors are going to be rather large improvements for AMD, putting them on a more level playing field against Intel.
Originally planned to be exclusively supported by Windows 10, Microsoft has reverted its policy and decided Skylake will be supported by Windows 7 and Windows 8.1 as well.
"Enterprise customers are moving to Windows 10 faster than any version of Windows. At the same time, we recognize that, in some instances, customers have a few systems that require longer deployment timeframes," explains Shad Larsen, Director of Windows Business Planning. "We listened to this feedback and today are sharing an update to our 6th Gen Intel Core (Skylake) support policy. We have extended the support period from July 17, 2018 to the end of support dates for Windows 7 and Windows 8.1; and we will provide all applicable security updates."
The change means Windows 7 support until 2020 and 8.1 until 2023. It does not apply to Kaby Lake and future Intel processors which will not be supported by Windows versions older than Windows 10.
Intel has finally flicked the power button on its 10nm fabrication facility, with the prototyping phase to begin very soon. Intel will reportedly enter the trial production stage in the coming weeks, with Intel shipping Kaby Lake-based CPUs to customers on time.
The company will be starting its 10nm factory during Q3 2016, confirming that it's upcoming Cannonlake processors are still on track. This means we should see trial productions begin later this year, where volume production will begin in the first half of 2017. Intel's current 14nm process was delayed into the mainstream market by 6-9 months or so, but in that time it still managed to keep a node lead over competitors in Samsung and TSMC.
AMD is set to get back into the CPU game in a big way with its next generation Zen architecture, which will be arriving to consumers in early 2017 - but reports have surfaced stating that an early run of Zen CPUs will hit later this year. We've already reported on the 32-core server variants of Zen, with the consumer CPUs hitting 4- and 8-core CPU markets.
During AMD's recent earnings call, company CEO Lisa Su said: "We have been very focused on the server launch for first half of 2017. Desktop should launch before that. In terms of true volume ability, I believe it will be in the first quarter of 2017. We may ship some limited volume towards the end of the fourth quarter, based on how bring up goes and the customer readiness".
Su said that AMD should have a limited run of Zen processors for their partners to use in testing in order to prepare for the desktop launch of the Zen-powered CPUs. The limited shipment should kick off in Q4 2016 (so, not long now) with the consumer launch in early 2017 aimed at high-end desktops and enthusiasts - where AMD has been losing to Intel for a number of years now.
Right now it's all eyes and ears on Polaris and graphics cards with the Radeon RX 480, and the soon-to-be-released Radeon RX 470 and RX 460, but the upcoming Zen architecture is beginning to rear its head, and will rock CPU sockets in the very near future.
According to the latest leaks, engineering samples of the Zen CPU are around the place, with 4-, 8-, 24-, and 32-core variants in the wild. The first two SKUs will slot into the new AM4 socket, while the 24- and 32-core variants will be used in servers.
The quad-core Zen CPU has a 65W TDP, while the 8-core model rocks a 95W TDP - teasing the efficiency tweaks that AMD will deliver with Zen. The 24- and 32-core server SKUs ramp the TDP up to 150W and 180W, respectively. When it comes to tech specs, we should see the quad-core Zen CPU with 8 CPU threads 2MB of L2 cache and 8MB of L3 cache while the 8-core Zen CPU will rock 16 threads with double the cache.
AMD's next-gen Radeon RX 480 is finally here, and it rocks quite a lot of performance for $199 - but what about the CPU side of the business? Zen isn't going to be here for a little while yet, but the company has reached a milestone in its production.
AMD has started the rollout of its enterprise-class Naples processor, which is a 32-core beast with 64 threads of CPU power. Naples is the codename of AMD's next Opteron platform, which will support the 32-core CPU. Naples will reportedly be built into an 8-core base design, with 16- and 32-core chips being made in an MCM (Multi-Chip Module) format.
The new Naples processor was spotted on the Zauba shipping database, where it was noted to be shipped on June 20, listed as the Naples Test Board. AMD will be tapping the 14nm FinFET process for its Naples CPU, just like the Polaris architecture on the GPU side of things. It will support 8 memory channels, and up to 128 PCIe 3.0 lanes, and up to 32 x SATA/NVMe devices.
While there might be a 1000-core processor in the wild, we won't see it in consumer or prosumer machines in that form, ever - at least not until the likes of Intel or AMD can hit 1000 processors on a single chip. Well, Intel is getting closer with its new Xeon Phi 7290 processor, which has 72 CPU cores.
Intel's new Xeon Phi 7290 is the fastest processor the company has released, and it's also their most expensive. The new Xeon Phi 7290 has 72 processing cores at 1.5GHz each, with Intel sending them out in September, with a huge price tag of $6294. Intel's next-gen Xeon Phi 7290 doesn't rely on the PCIe bus, allowing it to handle a much higher wider variety of workloads, and configurations that aren't supported by usual accelerators.
Intel has included 16GB of HBM with 490GB/sec of memory bandwidth for applications that are memory bound, while the dual-port Intel Omni-Path Fabric makes an appearance, reducing the cost of Intel's new Xeon Phi 7290, in terms of cost, power consumption and space required. The new Xeon Phi 7290 is one of four new processors, with the new chips packaged very similar to video cards, functioning as primary processors, or co-processors - depending on the use.
This isn't something I thought I'd be writing in 2016, but a team at the University of California, Davis, Department of Electrical and Computer Engineering have hit a world's first. The team have designed the first "KiloCore" processor, which has 1000 independent programmable processors, with a maximum computation rate of 1.78 trillion instructions per second.
The new 1000-core processor has 621 million transistors and was shown off during the recent 2016 Symposium on VLSI Technology and Circuits in Honolulu on June 16. Professor of Electrical and Computer Engineering, Bevan Baas, who led the team that designed KiloCore, said: "To the best of our knowledge, it is the world's first 1,000-processor chip and it is the highest clock-rate processor ever designed in a university".
Up until now, there has been no other chips made with over 300 processors, with most of them being made for research purposes and only a handful are sold commercially. The new 1000-processor KiloCore was fabricated by IBM on their 32nm CMOS technology, so we're not even talking anything radically advanced like the 14nm and 10nm nodes that are being used and played with now.