91916175
submission
Deathspawner writes:
After months (or even years) of anticipation, AMD has just launched its first EPYC server and datacenter processors, with 9 models set to launch over the next few months. AMD's top model, EPYC 7601, boasts 32 cores and 64 threads, while the smallest model, EPYC 7251, keeps things more modest at 8 cores and 16 threads. Fortunately for customers, every EPYC processor offers the same featureset, which includes AMD's entire security portfolio. That comes in addition to an eight-channel memory controller, and last but not least, 128 PCIe lanes per 1- or 2- socket system.
87325861
submission
Deathspawner writes:
Sony's PS4 Pro hasn't even been available for a week yet, but it's already flying off the shelves, helping to boost overall PS4 sales after a few months of tepid action. With its 4K and HDR capabilities, the allure of the Pro is high, but what about performance other than gaming? Techgage decided to find out, and discovered what most were expecting: that the Pro isn't that much faster outside of the improved gaming performance (or quality). The Pro includes a 1TB model of the same hard drive, so no gains there are seen, although the boosted CPU speed does seem to make improvements in load times in some rarer cases. What can really make a difference is upgrading the console to an SSD, which the site also explored.
87239641
submission
Deathspawner writes:
Sony's latest and greatest PlayStation 4 has just hit store shelves. Called Pro, Sony is targeting this particular model squarely at 4K and HDR televisions, although those with 1080p displays can still see some benefits. All told, the Pro sports a 31% faster CPU and 227% faster GPU, and as of the launch, most games will be targeting 4K resolution in lieu of increased detail of framerates. So far, reception to the console has been good, with Ars Technica saying that it will make games look crisper, especially when using PS VR. Techgage , meanwhile, believes the console leaves a lot to be desired — at least right now. One thing seems to be generally agreed-upon, though: the PS4 Pro is an ideal choice for those who don't own the original, or those who own PS VR, as both performance and fidelity can be improved significantly there.
85223551
submission
Deathspawner writes:
Intel's Basis has just sent an email to customers who own a Basis Peak smartwatch with some bad news: it's being recalled. In mid-June, Basis admitted that its flagship (and only) smartwatch had the chance to overheat, and then asked them to wait for a firmware update. Ultimately, a firmware update couldn't have been issued that wouldn't have compromised the user experience, and as such, the company is asking for every single Basis Peak to be returned for a full refund — it will even handle the shipping.
85013151
submission
Deathspawner writes:
At the ongoing SIGGRAPH 2016 conference, held in Anaheim, California, NVIDIA had a bevy of announcements to make, including a big one: Pascal-based Quadro professional workstation cards are en route. Similar to the latest TITAN X which was announced last week, the new top-end Quadro P6000 is based on the same GP102 architecture, but contains 256 more cores. This makes the P6000 an effective 12 TFLOPs (FP32) graphics card. Also announced was the 8.9 TFLOPs Quadro P5000, as well as updates to the company's Iray render (for VR), its DGX-1 deep-learning machine, and also its mental ray plugin for Autodesk Maya users.
84931091
submission
Deathspawner writes:
At a special artificial intelligence gathering at Stanford University on Thursday, NVIDIA's CEO Jen-Hsun Huang unveiled the world's fastest graphics card: the second-generation GeForce TITAN X. Based on the company's latest Pascal architecture, the new top-end card features 3,584 CUDA cores clocked at 1.53GHz, 12GB of GDDR5X, and is spec'd at 11 TFLOPs, which is at least 2 TFLOPs higher than the company's recently released GTX 1080. Jen-Hsun also touted for the first time a metric called TOPS (INT8), a deep-learning inferencing instruction. The new GTX TITAN X officially hits 44 TOPS. NVIDIA has said that its second-gen TITAN X will retail for $1,200, and will become available on August 2.
84771915
submission
Deathspawner writes:
NVIDIA has today released a Game Ready GeForce driver that introduces its interactive screenshot tool 'Ansel'. Named after famed photographer Ansel Adams, this new tool requires a developer to integrate up to a couple of hundred lines of code to give players the ability to pause their game, move around the environment, and then capture a more "artistic" image. To further that artistic value, users will have the ability to apply filters as well as capture an image in high-res 360 mode so that they can be viewed properly with a VR headset. Currently, Ansel supports only a single game, Mirror's Edge Catalyst, but NVIDIA promises that many more supported titles are on the way.
84769179
submission
Deathspawner writes:
NVIDIA has today released a Game Ready GeForce driver that introduces its interactive screenshot tool 'Ansel'. Named after famed photographer Ansel Adams, this new tool requires a developer to integrate up to a couple of hundred lines of code to give players the ability to pause their game, move around the environment, and then capture a more "artistic" image. To further that artistic value, users will have the ability to apply filters as well as capture an image in high-res 360 mode so that they can be viewed properly with a VR headset. Currently, Ansel supports only a single game, Mirror's Edge Catalyst, but NVIDIA promises that many more supported titles are on the way.
83673751
submission
Deathspawner writes:
Intel first teased their forthcoming NUC (Next Unit of Computing) mini PC, codenamed Skull Canyon, back at CES in January. However, systems have only just started shipping earlier this month. Styled with a new, thinner (but longer) all black chassis and Intel's classic Skull-branded logo, this NUC is targeted squarely at enthusiasts. The Intel Skull Canyon NUC6i7KYK not only boasts a Skylake quad-core CPU that boosts to 3.5GHz, but also Intel's fastest Iris Pro Graphics 580 integrated graphics core with 128MB of on-chip eDRAM (embedded DRAM). With the Skylake platform, Skull Canyon also sports DDR4-2133MHz memory, up to two M.2 NVMe Solid State Drives, four USB 3 ports, a Thunderbolt 3 port and a built-in SD card reader. It's a fair amount of computing horsepower in a roughly 8-inch by 4-inch, by 1-inch thick form factor. NUCs have been fairly popular in the market due to their size and convenience, though they've historically not be designed for higher-end workloads and gaming. However, in testing, this performance-built NUC, with its DX12 compatible graphic core, showed the ability to run most current-gen game titles at up to 1080p resolutions with medium image quality. The barebones kit is a bit on the pricey side at $650 street, but it's definitely one of the faster tiny PCs on the market.
83673591
submission
Deathspawner writes:
Intel first teased their forthcoming NUC (Next Unit of Computing) mini PC, codenamed Skull Canyon, back at CES in January. However, systems have only just started shipping earlier this month. Styled with a new, thinner (but longer) all black chassis and Intel's classic Skull-branded logo, this NUC is targeted squarely at enthusiasts. The Intel Skull Canyon NUC6i7KYK not only boasts a Skylake quad-core CPU that boosts to 3.5GHz, but also Intel's fastest Iris Pro Graphics 580 integrated graphics core with 128MB of on-chip eDRAM (embedded DRAM). With the Skylake platform, Skull Canyon also sports DDR4-2133MHz memory, up to two M.2 NVMe Solid State Drives, four USB 3 ports, a Thunderbolt 3 port and a built-in SD card reader. It's a fair amount of computing horsepower in a roughly 8-inch by 4-inch, by 1-inch thick form factor. NUCs have been fairly popular in the market due to their size and convenience, though they've historically not be designed for higher-end workloads and gaming. However, in testing, this performance-built NUC, with its DX12 compatible graphic core, showed the ability to run most current-gen game titles at up to 1080p resolutions with medium image quality. The barebones kit is a bit on the pricey side at $650 street, but it's definitely one of the faster tiny PCs on the market.
82682211
submission
Deathspawner writes:
At last week’s GPU Technology Conference, NVIDIA’s CEO Jen-Hsun Huang unveiled the Tesla P100 data center accelerator. The Pascal-based GPU at the heart of the Tesla P100 is codenamed the GP100 and it promises to be a very different animal versus the previous gen Maxwell. If NVIDIA’s past GPU naming convention rings-true, the GP100 will be the “big” version of Pascal, and presumably scaled down iterations of the chip will power consumer-class GPUs. Based on what we know so far about the GP100, it is an absolute beast of a GPU. It’s got roughly 3x the compute performance, 5x the GPU-to-GPU bandwidth and 3x the memory bandwidth of NVIDIA’s previous generation high-end products. The base clock is an impressive 1348MHz, with a boost clock of 1480MHz, and a 300 watt TDP. Considering how young TSMC’s 16nm FinFET process is, seeing clocks this high on such a big chip bodes well for NVIDIA. As configured, the GP100-powered Tesla P100 offers 5.3 TFLOPs of double-precision performance, 10.6 TFLOPs full-precision and 21.2 TFLOPs at half precision. Inside the GP100, 56 active SMs house a total of 3584 FP32 cores or 1792 FP 64 cores. The GPU links to its 16GB of HMB2 memory via a 4096-bit interface, which offers up 720GB/s of peak bandwidth. Finally, the GP100 will feature 16GB of Chip on Wafer on Substrate (CoWoS) HBM2 and support for NVLink. With NVLink, 160GB/s of serial bi-directional bandwidth is available between GP100 GPUs and up to eight GP100-based Tesla P100 boards can be interconnected. If this big iron version of Pascal is any indication, consumer GeForce versions will offer potent desktop graphics performance when they hit later this year.
82682001
submission
Deathspawner writes:
At last week’s GPU Technology Conference, NVIDIA’s CEO Jen-Hsun Huang unveiled the Tesla P100 data center accelerator. The Pascal-based GPU at the heart of the Tesla P100 is codenamed the GP100 and it promises to be a very different animal versus the previous gen Maxwell. If NVIDIA’s past GPU naming convention rings-true, the GP100 will be the “big” version of Pascal, and presumably scaled down iterations of the chip will power consumer-class GPUs. Based on http://hothardware.com/reviews...">what we know so far about the GP100, it is an absolute beast of a GPU. It’s got roughly 3x the compute performance, 5x the GPU-to-GPU bandwidth and 3x the memory bandwidth of NVIDIA’s previous generation high-end products. The base clock is an impressive 1348MHz, with a boost clock of 1480MHz, and a 300 watt TDP. Considering how young TSMC’s 16nm FinFET process is, seeing clocks this high on such a big chip bodes well for NVIDIA. As configured, the GP100-powered Tesla P100 offers 5.3 TFLOPs of double-precision performance, 10.6 TFLOPs full-precision and 21.2 TFLOPs at half precision. Inside the GP100, 56 active SMs house a total of 3584 FP32 cores or 1792 FP 64 cores. The GPU links to its 16GB of HMB2 memory via a 4096-bit interface, which offers up 720GB/s of peak bandwidth. Finally, the GP100 will feature 16GB of Chip on Wafer on Substrate (CoWoS) HBM2 and support for NVLink. With NVLink, 160GB/s of serial bi-directional bandwidth is available between GP100 GPUs and up to eight GP100-based Tesla P100 boards can be interconnected. If this big iron version of Pascal is any indication, consumer GeForce versions will offer potent desktop graphics performance when they hit later this year.
81755393
submission
Deathspawner writes:
Last March, Canada's regulatory agency for all things broadcasting, CRTC, ruled that cable TV providers would soon be forced to offer $25/mo packages. With enforcement having kicked-off on March 1, these inexpensive packages have now been made available. As Techgage has discovered, though, the first packages out-of-the-gate pack a number of caveats, and in some cases, are outright misleading. And, despite a simple framework to worth with, the two largest providers in the country, Rogers and Bell, offer vastly different packages, and ultimately vastly different values to the consumer.
81726575
submission
Deathspawner writes:
Last March, Canada's regulatory agency for all things broadcasting, CRTC, ruled that cable TV providers would soon be forced to offer $25/mo packages. With enforcement having kicked-off on March 1, these inexpensive packages have now been made available. As Techgage has discovered, though, the first packages out-of-the-gate pack a number of caveats, and in some cases, are outright misleading. And, despite a simple framework to worth with, the two largest providers in the country, Rogers and Bell, offer vastly different packages, and ultimately vastly different values to the consumer.
79204331
submission
Deathspawner writes:
There's never a lack of stuff to be impressed by in Minecraft, but rarely does that impressive stuff involve a corporation. Recently, Verizon teamed up with some prolific Minecraft streamers to design a mod that takes interactivity to a new level. After building an in-game smartphone and cellular tower, the gamer is not only able to browse the Web on the device, but also video call, all in a humorously low resolution. Verizon has created a GitHub page to explain how the magic is done.
