CPU, APU & Chipsets News - Page 138

AMD has been teasing its Zen architecture for a little while now, but the new Zen-based APUs will be packing quite the performance punch when they drop in 2017.

It's said that we're looking at Xbox One and PS4 level performance, even when the Zen APU is found inside of notebooks - and personally, I think we might see it in VR headsets, too. Another exciting part of this news is that there's a rumor that AMD is working on a Zen APU that would feature High Bandwidth Memory, or HBM.

If AMD did this, it could really begin to make a dent in CPU sales for Intel, as people could choose a semi-powerful APU with HBM over a traditional CPU from Intel and DDR4 RAM.

We knew they were coming, but the 10-core/20-thread Broadwell-E CPUs have been confirmed through a leaked roadmap from Intel. We should expect the chipmaker to launch an entire family of Broadwell-E processors throughout 2016.

Intel is expected to launch 6-, 8-, and 10-core offerings - with the assistance of Hyper-Threading boosting those up to 12-, 16-, and 20-threaded processors. Not only that, but we should expect new Broadwell, Skylake-S, and Skylake-U processors throughout 2016, where in Q4 2016 it looks like we should see Skylake-S 4+4e and Kaby Lake-S offerings unveiled.

The Broadwell-E offerings from Intel will arrive as the Core i7-6950X, with this CPU being the flagship 20-threaded beast. Under that, we have the Core i7-6900K, Core i7-6850K, and Core i7-6800K.

Intel has provided more details on its second-generation Xeon Phi co-processor at the annual SC conference, with the new Knights Landing beast nearly ready to be shipped.

The company has said they've already sent out various pre-production systems to clients through its early ship program, with Cray enjoying the new Xeon Phi co-processor, with it handling multiple customer applications in preparation for supercomputer deployments at Los Alamos and NERSC. Intel has said that the new Knights Landing chip has various innovative features that provide unmatched performance.

For example, Xeon Phi co-processor powered systems will be capable of delivering single-precision performance that is over 8TFlops with double-precision performance sitting at around 3TFlops. Inside, the Xeon Phi co-processor features 16GB of stacked memory that Intel says has 5x the bandwidth of DDR5, and is also 5x more power efficient and 3x denser than GDDR5 that is used on today's video cards.

While we just found out some details on the upcoming Broadwell-E platform, including the flagship Core i7-6950X and Core i7-6900K SKUs, there is some more exciting stuff for the Xeon family of processors.

The Broadwell-EP update will arrive in Q1 2016, with the new Xeon E5-2600 v4 processor featuring up to 22 CPU cores, with 44 threads thanks to HyperThreading technology. Intel's new Xeon E5-2600 v4 processors will boast up to 55MB of L3 cache, and have TDPs of up to 145W. The new Xeon processors will be slotted into the Grantley family, with quad-channel DDR4 support to boot.

Over the next few years, we'll see the new Skylake-EP and Skylake-EX processors from Intel, with the Xeon CPUs featuring up to 56 and 64 CPU threads. From there, in 2018 we can expect the Cannonlake-EP and Cannonlake-EX processors to really hit their stride, being made on the incredibly small 10nm process on the new Socket P platform. These processors should rock up to 68 or more CPU threads and have between 80-85MB of L3 cache. Exciting times.

Intel's next generation Broadwell-E platform is on its way, where we're learning about the next-gen HEDT family that will still work on existing X99 platform, with an entire new SKU lineup of exciting processors.

The flagship CPU in the new Broadwell-E family will be the Core i7-6950X, which will feature 10 x CPU cores and thanks to Hyper-Threading, 20 threads. The Core i7-6950X should be clocked at around 3GHz, but this will change between now and its release which is expected in the first half of 2016. But can you imagine 20 x 4GHz processors being inside of your PC?

There will be plenty of other CPUs released in the Broadwell-E family, including the new Core i7-6900K. The new Core i7-6900K will also be manufactured on Intel's impressive 14nm technology, with 16 threads of power behind it. The Core i7-6900K should rock a frequency of 3.3GHz, and 20MB of L3 cache. It will reportedly land on the X99 platform, which will confuse most people as the Core i7-6700K fits onto the Z170 chipset.

GlobalFoundries has achieved its first success in producing AMD products using the performance-enhanced version of its 14nm FinFET process technology known as 14nm Low Power Plus (14LPP).

With that, the tech will be officially integrated into a new set of AMD compute and graphics products, which it's said will be simultaneously high-performing and power-efficient.

Mark Papermaster, senior vice president and chief technology officer at AMD, said of the milestone, "FinFET technology is expected to play a critical foundational role across multiple AMD product lines, starting in 2016. GLOBALFOUNDRIES has worked tirelessly to reach this key milestone on its 14LPP process. We look forward to GLOBALFOUNDRIES' continued progress towards full production readiness and expect to leverage the advanced 14LPP process technology across a broad set of our CPU, APU, and GPU products."

AMD has a new Embedded single-chip system-on-chip (SOC) processor family on the scene with its new R-Series. It's aimed at the high-end market, and packs its latest 64-bit CPU core "Excavator", as well as a third-gen GCN GPU architecture. It's also the first of its kind to support dual-channel 64-bit DDR4 (with speeds as high as DDR4-2400; DDR3 is also supported with speeds up to 2133).

On the graphics and video front, the R-Series supports High Efficiency Video Coding (HEVC)1 for full 4K decode and DirectX 12; the specifications note up to eight compute units and two rendering blocks and GPU clock speeds up to 800MHz resulting in 819 GFLOPS. The company claims 22 percent better performance over the 2nd Generation AMD Embedded R-Series APU2 and 58 percent better performance over the Intel Broadwell Core i7 in graphics-intensive benches.

As for efficiency, AMD boasts full HSA implementation and hUMA utilization for what they feel results in great CPU, GPU, and memory performance. Additionally, thermal design power is configurable, and power usage is said to be reduced 35 percent compared to the 2nd Generation AMD Embedded R-Series APU.

AMD CEO and President Lisa Su affirmed her company's commitment to high performance x86 CPUs in its earnings call last week, for anyone in doubt following the recent news of financial troubles.

"In terms of long-term roadmap we are extremely committed to high performance x86 CPUs," she says. "And there should be no confusion on that point. [Chief Technology Officer] Mark Papermaster is currently directly engaging with the team on that execution and we'll have more details to come. But [we're] overall pleased with the execution and it continues to be our number one priority for the company."

More, she states Zen's impressive target 40% IPC performance is "on track" and that multiple FinFET designs have been taped out (meaning design is finalized, and the chips need only be produced).

It looks like AMD is preparing to launch a new 8-core server SoC codenamed "Hierofalcon", based on a 64-bit ARM design using eight A57 cores clocked at 2GHz each. AMD's new Hierofalcon SoC has a TDP of 30W, so we should expect some decent performance from the new chip.

An engineering sample of the new 8-core "Hierofalcon" processor has been spotted, where it has been put up against a bunch of desktop processors from both Intel and AMD. We have those results above. These results are looking at the performance on a single core, at a normalized clock speed. The new Hierofalcon processor does quite well, taking swings at some of AMD's desktop chips.

When it comes to "Whetstone" performance, the desktop processors kick some serious ass, but that's to be expected. The same thing goes for the Execl benchmark and the kCopy and kPipe results.

Qualcomm has just unveiled its plans to enter the server CPU market with a new custom processor based on an ARM design. Qualcomm's latest chip is being aimed at the likes of hyper-scale customers like Google and Facebook, as well as many service providers and large enterprises.

The new chip was shown off last week, with the CPU being a purpose-built system-on-chip that is different to other Snapdragon processors. The first version of Qualcomm's new chip is a 24-core processor, combining PCIe, storage and other features into the SoC. According to Qualcomm Senior Vice President Anand CHandrasekher, the first pre-production processor has 24 cores, but future chips will have many more.

The chipmaker is now shipping its processor to big customers for testing, but Qualcomm isn't naming names. The company said it would provide a release window for the chip, but it would update the world in 2016.