Hardware stub

HW News - AMD GPUs in Samsung Phones, Fake X499, Apple Monitor Stand

Posted on June 8, 2019

As we board another plane, just five days since landing home from Taipei, we're recapping news leading into next week's E3 event, positioned exhaustingly close to Computex. This recap talks AMD and Samsung partnerships on GPUs, Apple's $1000 monitor stand and accompanying cheese grater, and the Radeon Vega II dual-GPUs located therein. We also talk tariff impact on pricing in PC hardware and, as an exclusive story for the video version, we talk about the fake "X499" motherboard at Computex 2019.

Show notes below the video embed.

AMD & Samsung Announce Radeon Partnership

AMD and Samsung announced a joint, multi-year partnership that will see AMD licensing its Radeon graphics IP to Samsung. Samsung will leverage AMD’s graphics technology -- specifically its recently announced RDNA architecture powering new Navi cards -- for use with its own ARM-based Exynos chips inside mobile devices.

This will likely allow Samsung to replace the Mali GPUs, which is more ARM-based IP. It also gives AMD’s Radeon graphics yet another platform and revenue stream. AMD has made inroads with its graphics IP outside of the PC market, with a successful semi-custom strategy that will see AMD deliver solutions for next-gen consoles from both Sony and Microsoft, in addition to Google Stadia using Radeon powered graphics.

Indeed, partnering with one of the biggest phone manufacturers will help AMD carve out a new path to market for Radeon and expand its ecosystem.    

Source: https://www.amd.com/en/press-releases/2019-06-03-amd-and-samsung-announce-strategic-partnership-ultra-low-power-high

AMD Pro Vega II 7nm GPUs

In other AMD news, it seems the announcement of the new Mac Pro also served as the launch pad for AMD’s Pro Vega II GPUs. AMD’s new workstation-class hardware will come in two flavors: the AMD Pro Vega II and the AMD Pro Vega II Duo. For the time being, it seems AMD’s newest cards will be solely available inside Apple’s newest Mac Pro workstation.

The new AMD Pro Vega II cards are based on AMD’s second-generation Vega architecture, while using 7nm Vega 20 silicon. The AMD Pro Vega II offers 64 compute units, or 4,096 stream processors. The card’s clock rate tops out at 1700 MHz, and is capable of 14.2 TFLOPS of FP32 compute performance. The Pro Vega II card will also make use of 32GB of HBM2 memory across a 4096-bit memory bus, delivering up to 1 TB/s of memory bandwidth.

The Pro Vega II Duo returns AMD to its dual-GPU days, offering two Vega 20 dies atop the same PCB. The Pro Vega II effectively doubles the previous specifications: 128 compute units, equal to 8,192 stream processor, 28.4 TFLOPS of FP32 performance, and 64GB of HBM2 memory. Of course, AMD’s Infinity Fabric is the interconnect the GPUs will communicate through, with 84GB/s of bandwidth per direction.

No word on price, but these GPUs likely cost as much as a good used car.        

Source: https://www.tomshardware.com/news/amd-radeon-pro-vega-ii-7nm-gpus-apple-specs,39571.html

Samsung Could Oust TSMC as NVIDIA’s 7nm Partner

According to reports, one being from industry soothsayers Digitimes, Nvidia is intending to tap Samsung as its fab partner for the 7nm Ampere. As far as we know, Ampere is intended as the successor to Turing, although there are precious little details about the architecture other than a 2020 release window.

It’s been expected Nvidia would continue its long partnership with TSMC, but another report by EE Times claims Samsung is “aggressively undercutting” TSMC with its own 7nm EUV node. Samsung landing Nvidia as a 7nm launch partner would certainly take some shine away from TSMC, who’s been leading the pack in terms of foundry processes.

“Separately, one source said that Samsung is aggressively undercutting prices for its 7-nm node with EUV, offering some startups a full mask set for less than a multi-layer mask (MLM) set at its rival. TSMC introduced the MLM mask sets in 2007 to lower costs for small-volume runs. They are said to be about 60% of the cost of a full mask set,” the EE Times report reads.

As Tom’s Hardware notes, capacity could also be a consideration. TSMC’s 7nm node has been seeing a burgeoning demand as of late, with AMD’s Ryzen 3000 and EPYC Rome, as well as Apple’s A13 chip coming later in the year.  

Source: https://www.pcgamesn.com/nvidia/ampere-7nm-gpu-samsung-discount



U.S.-China Trade: MSI ‘Prepares for the Worst,’ DRAM Decline

The US-China trade conflict continues to have adverse effects on many facets of the tech industry. Tensions between the US and China recently culminated with the blacklisting of Huawei, of which the ramifications haven't yet been fully realized. On top of many American tech companies being forced to drop their business with the company, the newest revelation is the impact it will likely have on the DRAM market.

“DRAMeXchange, a division of TrendForce, points out that, as ripples from the US ban continue to spread, Huawei's shipments of smartphone and server products are feared to face heavy obstacles for the next two to three quarters, impacting peak-season-demand for DRAM products 2H and the time of price precipitation. TrendForce officially adjusts its outlook for 3Q DRAM prices from its original prediction of a 10% decline to a widened 10-15%,” says DRAMeXchange. The report also notes that DRAM could rebound in 2020, behind bottoming prices and limited bit growth. Until then, it’s unlikely anyone empathizes with DRAM makers much after several quarters of inflated pricing and murmurings of price fixing.

Meanwhile, PC vendors like MSI are caught directly in the crossfire of the trade war, and Tom’s Hardware was able to catch up with MSI CEO Charles Chiang at Computex 2019, where he stated MSI was preparing for the worst.

"I've been very, very busy lately; you always need to prepare for the worst, right? Right now, the $200 billion [in trade] is already at 25% [tariff], and then we are worried the next wave will be $325 billion in goods will be at 25% tariff. That's why we need to prepare for the worst," Chiang said. "So what I'm doing right now, we will move more and more capacity back to Taiwan. It is our short- and mid-term strategy. In the long term we will go to Vietnam or somewhere, but Taiwan is the short-term solution. That is the only way we can do it right now."

Some vendors have been able to absorb the first wave of tariffs without sacrificing pricing; others are trying to wait it out. Some, like MSI, are moving product lines out of China altogether, with little to no incentive to return. The tariffs have recently affected our products here at GN, most notably our modmats. We have played a lot of chess with the few pieces we have and, with a slight ding to our margin, we think we’ve found some ways to preserve pricing without needing to increase it. We don’t have many more cards to play, but at least for this round, we can keep the pricing where it is. We’ll talk about our story more in a future news piece.

Finally, the server market continues to see uncertainties as a result of the US-China trade war.

“As suppliers for the cloud computing datacenter segment are still clearing their inventory, and US-China trade tensions have created uncertainties, demand for datacenter servers has been decreasing since early 2019 and may cause Intel's datacenter business group to suffer its first on-year revenue decline in 10 years in 2019,” explains Digitimes.

The problem that the server market faces appears to be two-pronged: Huawei’s blacklisting will cause major server outfitters like Intel and Nvidia serious revenue losses. More to the point, China could retaliate by boycotting US goods, cutting server companies off from the Chinese market period.      

Source: https://www.digitimes.com/news/a20190604PD206.html



AMD Will No Longer License x86 Chip Designs to China

AMD’s 2016 technology licensing deal it struck with China will come to an end, as Lisa Su confirmed with Tom’s Hardware. China will still be able to develop chips based on AMD’s first generation of Zen, which encompasses Ryzen and EYPC Naples. However, AMD will not extend its more recent x86 IP to China, including the recent Zen 2 architecture powering Ryzen 3000 and EPYC Rome.

Known as the Tianjin Haiguang Advanced Technology Investment Co. Ltd. (THATIC), the deal between AMD and China served as both a path to the lucrative Chinese market for AMD, and a chance for China to develop its own x86 technology. Regarding the future of THATIC, Lisa Su stated "we are not discussing any additional technology transfers," and that "THATIC was a single-generation technology license, and there are no additional technology licenses."

Source: https://www.techpowerup.com/256253/amd-halts-further-x86-technology-licensing-to-china

Windows: Patch BlueKeep Exploit; Windows 10 Variable Refresh

Not so long ago, we reported on a wormable exploit that warranted enough caution for Microsoft to deliver a critical update to the ancient Windows XP. The exploit has since earned the name BlueKeep, and Microsoft has doubled down on its warning to update and patch almost all Windows versions, and even the NSA has added its voice to the chorus.

“The National Security Agency is urging Microsoft Windows administrators and users to ensure they are using a patched and updated system in the face of growing threats. Recent warnings by Microsoft stressed the importance of installing patches to address a protocol vulnerability in older versions of Windows. Microsoft has warned that this flaw is potentially “wormable,” meaning it could spread without user interaction across the internet. We have seen devastating computer worms inflict damage on unpatched systems with wide-ranging impact, and are seeking to motivate increased protections against this flaw,” warns the NSA.

TL;DR: Despite Microsoft’s botched series of updates in recent months, given the circumstances and the importance of heading off any chance of a WannaCry clone, it’s worth getting patched up.

In less troublesome Microsoft Windows news, the Windows 10 May 2019 Update brought with it new variable refresh rate support for DX11 games. The new graphics setting has to be enabled manually, and is intended to supplement the adaptive refresh rate technology found in FreeSync and G-Sync monitors.  

Source: https://devblogs.microsoft.com/directx/os-variable-refresh-rate/



May 2019 Steam Hardware Survey: RTX Is A Slow Burn

Despite the aspiring competition in the Epic Games Store, Steam is the foremost PC gaming platform, and will likely remain so for the foreseeable future. Steam regularly uses its massive platform to poll users’ choice of hardware, and while the results are never that shocking month-over-month, they tend to highlight emerging trends and growths.

For May 2019, the most notable observation is that RTX card adoption is on the rise -- slowly, but surely. The install base for Nvidia’s RTX cards rose roughly 0.31% between April and May, with the RTX 2060 accounting for the bulk of that growth, unsurprisingly. Nvidia’s 1060 and 1050 Ti remain the most popular cards.

Outside of GPUs, most gamers are still using quad-core CPUs, with hexa-core and octa-core processors being a distant second and third place. Windows 10 is still the OS of choice, and 1920 x 1080 is still the dominant resolution for the majority of gamers.