TweakTown NewsRefine News by Category:
AMD has just announced its new Opteron A1100 processor, a chip built on a 64-bit capable ARM-based server SoC. This chip has been branded as an Opteron processor, which is the first processor the two companies have worked together on.
The new Opteron A1100, codenamed "Seattle", comes in two flavors: four or eight CPU cores. These cores are based on ARM's Cortex-A57 design, which can be clocked up to 2GHz. Each core shares 1MB of L2 cache, as well as a total of 8MB of L3 cache that all shares can access. The SoC is built at GlobalFoundaries on a 28nm process.
Opteron A1100's memory controller is capable of supporting both DDR3 and DDR4 memory through its 128-bit bus, and AMD's reference platform will support up to 128GB of Registered DDR3 DIMMs. The SoC features an 8 lane PCIe 3.0 controller that supports one x8 or two x4 slot configurations, plus an 8-port 6Gbps SATA controller, and two 10GbE ports.
AMD's new Opteron A1100 chip will sample in March, with the company predicting that ARM-based solutions will occupy 25% of the server market by 2019.
Qualcomm is reportedly hard at work on a new processor, something that should blow most other chips away when it comes out. The chipmaker is rumored to be working on a new SoC that would feature 8 processing cores, and would be 64-bit compatible.
The new SoC would feature eight Krait64 general-purpose cores designed in-house at Qualcomm, and would be compatible with ARMv8 architecture with up to 4MB of cache and a 2.5GHz clock speed. We should also expect Adreno 430 graphics on-board (up from the 330 series) with a 500MHz frequency, LPDDR3/LPDDR4 memory controller, and various special purpose hardware/accelerators.
Qualcomm's new SoC would be made on the 20nm process over at TSMC. We should expect it to arrive as the Snapdragon 810 series, with two versions: MSM8994 with integrated baseband and APQ8094 that will require an external telecommunications chip.
Today AMD announced the launch of its new 12-core and 16-core Opteron 6300 series server processors. The CPU's code named Warsaw have been highly anticipated over the last few months and now that they are here, AMD says that they have been designed with heavy emphasis on Virtualization deployments.
The new 12-core 633P processor features a Piledriver-based 12-core CPU that is clocked at 2.3GHz, and peaks at 2.8GHz during heavy loads. The 16-core 6370P is also Piledriver based and comes clocked at a base frequency of 2.0GHz and when needed can boost to 2.5GHz. Both CPUs are fully compatible with all hardware running existing 6300 series chips, and have a TDP of 99 Watts.
We reported over the weekend on the 16-core processor from AMD, but if you want to get in early, you can buy the Opteron variant of it... right now. The AMD Operton 6300-series, or "Warsaw" processor, is designed for 2- and 4-way goodness in servers.
It uses the socket G34, and are based on the Piledriver architecture. The new chips feature between 12 and 16 cores, 16MB of L3 cache, quad-channel DDR3 memory and much more. They should have a much better performance-per-watt ratio over today's Opteron 6300 series, too. The company were poised to release these new CPUs later in Q1 2014, but ShopBLT is offering a pre-order on the 12-core version for $421, and the 16-core Operton for $663.
The details on the processors are as follows: the 12-core version is the Opteron 6338 HE, which has a 99W TDP. The 16-core processor, is the Opteron 6370 HE, and also has a 99W TDP. The HE stands for high efficiency, and should feature moderate clock speeds to keep the power numbers down.
Intel should fully unveil its 9-series chipset as we get closer to Computex 2014 in Taipei in June, but between now and then we'll have to tease you a little on what to expect from Intel's upcoming Haswell refresh-capable chipset.
The new 9-series chipset won't bring many new technologies to the consumer, but it will feature six SATA 6Gbps ports, 14 USB ports, PCI Express 3.0, and the usual array of other technologies. Where things are new, we have support for SSDs in the PCIe M.2 form factor - capable of driving 1GB/sec bandwidth. Intel device protection with boot guard technology, as well as Intel Rapid Storage technology which will support PCI Express-based drives.
There shouldn't be SATA Express support, which is a disappointment, as it is capable of driving 2GB/sec maximum bandwidth. We can't complain about 1GB/sec, as it is much higher than SATA 6Gbps' maximum rate of 600MB/sec.
According to a report from DigiTimes, we should expect Intel to refresh its Haswell lineup of CPUs sometime in May - just before Computex kicks off in Taipei, Taiwan.
Intel will begin shipping its new Z97 and H97 chipsets in April, so that motherboard makers can start selling some new motherboards moving into May. We should expect 20 or so new processors for the Haswell refresh, which will include the Core i7-4790, Core i5-4690, Core i5-4590, Core i3-4360, Pentium G3450 and Celeron G1840 and low-power Core i7-4790S, Core i5-4590S and Core i3-4150T.
After which the chipmaker will release the more interesting processors, the Haswell Refresh K series, and Haswell-E processors in Q3 2014.
I was pretty impressed what what I read about NVIDIA's Tegra K1 processor at CES 2014, but it looks like the new SoC will be kicking some serious benchmark ass when it hits.
Some leaked benchmarks involving Lenovo's ThinkVision 28, which is powered by the Tegra K1 SoC, have leaked. These leaked benchmarks are using Futuremark's 3DMark, where the Tegra K1 SoC just dominates the competition. Keep in mind that the ThinkVision 28 was actually clocked at just 2GHz, and not the 2.3GHz that NVIDIA claims the Tegra K1 can be operated at.
Other benchmarks suggest that NVIDIA's Tegra K1 processor can also beat an Intel Haswell-powered notebook with integrated graphics. This is a big deal, as a Haswell-powered notebook will chew up much more power than an SoC. It looks like NVIDIA is onto a big, mobile winner here, folks.
The last we heard on Intel's Broadwell CPUs is that they would feature 18 cores, but the release date on them was Q4 2014, now we're hearing dates of Q3 2014.
The new Broadwell CPUs will usher in a refined 14nm process, and three-gate transistors, the first CPUs to be made on the new tech. The new Broadwell architecture will be close to the company's on-the-market Haswell processors, but will see reduced power consumption, as well as increased performance. We should also see more advanced integrated graphics on Broadwell-based CPUs, too.
We should expect Intel's Broadwell CPUs to be released in Q3 2014, but the big problem here is that Intel will be competing with its Haswell-E CPU. Unleashing two new CPUs, that will overlap each other in some markets, might be hard for Intel. We should see some more concrete information, specs and prices toward Computex in June.
CES 2014 - We've already seen the Consumer Electronics Show 2014 start off, unofficially, with a bunch of 4K goodies being announced, with Qualcomm not being left out on its own - with the company announcing its new SoC, the Snapdragon 802.
Qualcomm's Snapdragon 802 SoC is destined for TVs, where it is a TV-specific quad-core SoC that borrows features from the Snapdragon 800 SoC, and the "Ultra HD" 805 chip for mobile video. The Snapdragon 802 is all about powering 4K content in your living room, with the processor capable of multi-tasking, where it can easily handle browsing the Internet, or switching things up and streaming a movie or playing four HD video streams on your TV.
The Snapdragon 802 also has HQV, or Hollywood Quality Video, a video-processing engine which upconverts 1080p content to a level "approaching Ultra HD." The actual specs of the Qualcomm Snapdragon 802 are: a quad-core 1.8GHz Krait CPU, Adreno 330 GPU, two-stream dual-band 802.11ac, AllJoyn integration and Snapdragon Studio Access content protection.
Qualcomm will begin pushing these new Snapdragon 802 SoCs to OEMs in the near future, where consumers will begin to see the SoCs powering their future TVs and set-top boxes. This is expected in late 2014.
CES 2014 - NVIDIA unveiled its impressive Tegra K1 processor yesterday, but in the weeks leading up to the event held in Las Vegas for CES 2014, the GPU maker plowed through a farm to drum up marketing for its new mobile processor which features 192 GPU cores.
The crop circle NVIDIA used has some keen-eyed readers notice the numbers 1, 9 and 2 in the crop circle - teasing '192' - the amount of GPU cores on its just-announced Tegra K1 processor. It seems like a long shot, but if this is as far as NVIDIA wants to go with marketing its Tegra K1 processor. -