Articles

404 ERROR: REQUEST COULD NOT BE FOUND

The page that you have requested could not be found at this time. We have provided you a list of related content below or you can use our site search to find the information that you are looking for.

Samsung Introduces Next Generation HBM2E

At NVIDIA's GPU Technology Conference, Samsung unveiled an even faster version of their HBM2 memory. A stack of "Flashbolt," as they call it, can deliver up to 410 GBps of bandwidth, which they claim is 33% faster than previous offerings, and a single package can hold up to 16GB of memory. Samsung says the new memory is aimed at "supercomputers, graphics systems, and artificial intelligence (AI)" applications, though they didn't mention just what GPUs or accelerators will make use of it in the near future.

"Flashbolt's industry-leading performance will enable enhanced solutions for next-generation data centers, artificial intelligence, machine learning, and graphics applications," said Jinman Han, senior vice president of Memory Product Planning and Application Engineering Team at Samsung Electronics. "We will continue to expand our premium DRAM offering, and improve our 'high-performance, high capacity, and low power' memory segment to meet market demand."

Discussion
Posted by alphaatlas March 20, 2019 9:35 AM (CDT)

Google Pay Equity Analysis Results in $9.7 Million in Adjustments

In 2018, Google included 91% of Google employees in its annual pay equity analysis that covers all job groups that meet minimum n-count thresholds for statistical analysis. Google's annual compensation planning process includes modeling algorithmically each employee's compensation based on work-related inputs like the market rate for their job, their location, level and performance rating. This allows the company to equally compensate employees equitably across gender and racial lines. Any significant discrepancies are handled with a pay raise as compensation. Google made $9.7 million in adjustments to 10,677 Google employees after the 2018 analysis. The main group that was flagged to receive adjustments were men in the Level 4 Software Engineer career. Google said men received less discretionary funds than women in that career path. 49% of total dollars spent on adjustments went to new hires. Google's next goal is to create a leveling equity analysis to assess how employees are leveled when they are hired and improve how they level. The Washington Post says Google is embroiled in a lawsuit that alleges the tech giant discriminates against women.

Our pay equity analysis ensures that compensation is fair for employees in the same job, at the same level, location and performance. But we know that's only part of the story. Because leveling, performance ratings, and promotion impact pay, this year, we are undertaking a comprehensive review of these processes to make sure the outcomes are fair and equitable for all employees.

Discussion
Posted by cageymaru March 05, 2019 10:55 AM (CST)

The 16GB of HBM2 on the Radeon VII Is Needed for Real World 4K Video Production

Digital Foundry's Richard Leadbetter brought to light an important real-world use case for the 16GB of HBM2 on the AMD Radeon VII. 4K video editing in Adobe Premiere Pro needs lots of GPU memory to export video when mixing several 4K streams and using taxing transitions. He said that Digital Foundry had upgraded all of their 6GB and 8GB GPUs to 12GB cards in the past as video cards like the RTX 2080, RTX 2080 Ti, GTX 1080 Ti, etc, lack the amount of GPU memory necessary to export 4K video in their everyday workflow. He documented an example of this at the 5:00 mark of the video below. In this real-world use case, the 12GB NVIDIA Titan X Pascal would crash during the export phase of head-to-head 4K video production when mixing six 4K video clips on a timeline and using 3 taxing transitions. This "Accelerated Renderer Error" was 100% due to a lack of GPU memory according to Mr. Leadbetter. But when the exact same task was presented to the AMD Radeon VII; there were no problems. The pro-level, 16GB of HBM2 video memory on the AMD Radeon VII powered through the task without breaking a sweat! These GPU related "out of memory" crashes have dogged his production team in the past and has even made some of their content late. He goes on to explain that producing 1080p video with 4K assets will cause the same crashing issue if the production team tries to use too many fancy transitions. Special thanks to Jason Evangelho of Forbes.

But for those of you producing hi-res video now or entertaining it in the near future, AMD may be onto something here -- like opening more doors for amateur producers on a budget. I also suspect that certain compute workloads may see considerable benefit from the hefty amount of HBM2 and the corresponding 1TB/second of memory bandwidth.

Discussion
Posted by cageymaru February 07, 2019 7:32 PM (CST)

An Analysis of GDDR6 and HBM2 Technologies

Frank Ferro, senior director of product management at Rambus, discusses the differences in GDDR6 and HBM2 memory technologies; with a focus on the type of applications that each would be best suited for. Around the 8:54 mark he discusses how to combine older process nodes with 7nm process nodes and chiplets. Thanks to TheRetiredEngineer for the links.

Tech Talk: Frank Ferro, senior director of product management at Rambus, talks with Semiconductor Engineering about memory bottlenecks and why both GDDR6 and high-bandwidth memory are gaining steam and for which markets.

Discussion
Posted by cageymaru January 21, 2019 10:48 AM (CST)

GDDR6 Memory Costs 70% More than GDDR5

Pricing lists from two major electronics components distributors are showing that GDDR6 memory carries a whopping premium over GDDR5: "14 Gbps GDDR6 memory chips from Micron Technology cost over 70 percent more than common 8 Gbps GDDR5 chips of the same density, from the same manufacturer." TechPowerUp suggests that graphics card manufacturers can save around $22 per card by using six GDDR5 chips instead of GDDR6.

Although GDDR6 is available in marginally cheaper 13 Gbps and 12 Gbps trims, NVIDIA has only been sourcing 14 Gbps chips. Even the company's upcoming RTX 2060 performance-segment graphics card is rumored to implement 14 Gbps chips in variants that feature GDDR6. The sheer disparity in pricing between GDDR6 and GDDR5 could explain why NVIDIA is developing cheaper GDDR5 variants of the RTX 2060.

Discussion
Posted by Megalith January 05, 2019 11:10 AM (CST)

An Analysis of the Z390 Socket Proves the Extra Pins Aren't Necessary

Der8auer on YouTube (Roman Hartung) has performed an in-depth analysis of the Z390 socket where he physically extracted the pins from a dead Z370 board and measured how much of an load each pin can take. After applying up to 5 amperes of power to the pin, he concluded that it can easily withstand the rigors of a daily 1.01 amperes load. Then he taped off 18 pins on an i9-9900K to simulate it running in a Z270 motherboard. This increased the load on each pin, but no adverse changes were observed in the socket or the board. He then taped off varying amounts of pins and tested the i9-9900K with up to 69 pins taped off which created a 1.92 ampere per pin load. Again, no changes were observed during testing that exceeded 6 hours. The conclusion is that the LGA-1151v2 is absolutely unnecessary.

Roman Hartung has proven very clearly that the LGA-1151v2 socket is basically completely unnecessary. The pins withstand even a very limited power supply without any problems, there is no damage to the mainboard, the socket or the processor. This also proves once more that Intel probably didn't allow compatibility with the old motherboards for sales reasons.

Discussion
Posted by cageymaru January 03, 2019 2:54 PM (CST)

An In-Depth Analysis of the 7nm 64-Core AMD EPYC "Rome" Server Processor

Charlie Demerjian of SemiAccurate has written an in-depth analysis of the 7nm 64-core 9-die AMD EPYC "Rome" server processor. He hypothesizes over the real world performance potential for the processor versus benchmarks by analyzing the chip's design features. He explains why the 14nm IOX chip didn't need a shrink, and how the design choice will speed up production of the processor. He articulates how doubling the instruction widths internally for the new Zen 2 cores leads to a 2X FP performance increase per core and a 4X FP increase per socket. Many other revelations about the new "Intel killers" can be found in the article.

AMD has full transparent memory encryption, can secure VMs from the host, and many other useful things that protect users in the real world. Better yet they are enabled with a BIOS switch and have a tiny overhead. Officially there is now more space for keys so you can support more encrypted memory VMs per box. The other big box to check is Spectre mitigations are now rolled in to the core but Meltdown and L1TF/Foreshadow are not. Why? Because AMD wasn't affected by either one and never will be. Patching can work but to be immune from the start is always a better choice.

Discussion
Posted by cageymaru November 09, 2018 11:57 AM (CST)

An Analysis of the Dark Souls 3 Alpha: Time of Day System

Lance McDonald has created a video series of cut features from the Dark Souls 3 alpha that never made it into the retail game. The features in this particular video were called "Ceremonies," but in reality they were time of day changes that a player could experience in the alpha client. Some of them add flying creatures to the already beautiful, but ominous and foreboding skies. If you're a fan of the series of games, his channel has plenty of videos detailing cut lore, unseen items and more from the alphas of the Souls games.
Discussion
Posted by cageymaru September 28, 2018 7:15 PM (CDT)

Adored's RTX 2080 Ti Review Roundup and Analysis

Jim "The Scottish Hammer" from AdoredTV gives us his thoughts about what all just happened with NVIDIA and its new RTX GPUS.
Discussion
Posted by Kyle September 20, 2018 5:49 PM (CDT)

Nvidia DLSS Analysis

Eurogamer posted an analysis of Nvidia's DLSS tech found in their Turing graphics cards, and the results are interesting. DLSS does boost game performance significantly in some cases, while still providing good image quality. The author points out that Final Fantasy 15 ships with a particularly blurry temporal AA implementation, and running DLSS instead provides a image quality boost in some areas. Check on the screenshots comparisons on Eurogamer's website, or watch video below:

The Infiltrator demo also serves to highlight that the performance boost offered by DLSS is not always uniform - it's not a straight 35-40 per cent uplift throughout. The demo features a number of close-up scenes that stress the GPU via an insanely expensive depth of field effect that almost certainly causes extreme bandwidth issues for the hardware. However, because the base resolution is so much lower, the bandwidth 'crash' via DLSS is far less pronounced. On a particular close-up, the DLSS result sees the scene play out over three times faster than the standard, native resolution TAA version of the same content. But has Nvidia truly managed to equal native quality? For the most part, it passes muster and inaccuracies and detail issues are only truly noticeable when conducting direct side-by-side comparisons - but we did note that in the demo's climactic zoom-out to show the high detail city, the lower base resolution does have an impact on the quality of the final image. Is it likely to distract a user playing the game and not conducting detailed side-by-side comparisons? Highly unlikely.

Discussion
Posted by alphaatlas September 20, 2018 11:52 AM (CDT)

An Analysis of the Take-Two Interactive Lawsuit Against a GTA V Cheat Creator

YouTuber Law has created a video analysis of the recent Take Two Interactive win against David Zipperer who created cheats for the online game Grand Theft Auto V. He goes into how cheats are a derivative product of the original copyrighted work; thus it is the same as creating a sequel to the original copyrighted work. He relates previous court rulings against derivative works to the Take Two Interactive case and then explains why commercializing a derivative of another company's copyrighted work is frowned upon by the courts.

We've seen many game developers suing individuals who develop software to cheat in their games. Most lawsuits against video game cheats settle quickly. Today, we have one lawsuit that the court has rules on. In a lawsuit by take two against a developer of cheat or grief software for gta v (grand theft auto 5), the court issued a preliminary injunction. In the process the court told us why is cheating in a video game considered copyright infringement.

Discussion
Posted by cageymaru September 17, 2018 3:40 PM (CDT)

Rumored NVIDIA RTX 2070 Specs: 2304 Cores, 8GB GDDR6 at ~$400

Now that the RTX 2080 Ti and 2080 are less of a mystery, interest is ramping up for their less powerful siblings, the 2070, 2060, and 2050. Rumors from VideoCardz and AdoredTV suggest the former (possibly priced at $300 to $500) will be 40% faster than the 1070, carrying a total of 2304 CUDA cores and 8GB of GDDR6 memory. Wccftech speculates a 256-bit GDDR6 memory interface, for a total of 448GB/s of bandwidth.

The GeForce RTX 2070 is set to launch sometime in September, following the debut of the RTX 2080 Ti and RTX 2080 next Monday, at an estimated suggested price of ~$400 give or take a hundred bucks. The card is said to slightly outperform NVIDIA’s current GeForce GTX 1080 by roughly 8%, but not quite match the GTX 1080 Ti.

Discussion
Posted by Megalith August 19, 2018 9:50 AM (CDT)