AMD and Nvidia both appear to be certain to get their “14 nm” out next year.
According to TweakTown Nvidia is apparently dotting the “I” and working out where to put in the semi-colons for its Pascal GPU using TSMC’s 16nm FinFet node. AMD rumored has been wining and dining its old chums at GlobalFoundries to use its 14nm process for its Greenland GPU.
Although these sound like different technologies the “14nm and 16nm” is difference how you measure a transistor. The outcome of both 14 and 16 should be a fairly same sized transistor with similar power features. TSMC calls its process 16nm FinFet, while Samsung and GloFo insist on calling it 14nm FinFet.
The dark satanic rumor mill suggests that the Greenland GPU, which has new Arctic Islands family micro-architecture, will have HBM2 memory. There will be up to 32GB of memory available for enthusiast and professional users. Consumer-oriented cards will have eight to 16GB of HBM2 memory. It will also have a new ISA (instruction set architecture).
It makes sense, AMD moved to HBM with its Fury line this year. Nvidia is expected to follow suit in 2016 with cards offering up to 32GB HBM2 as well.
Both Nvidia and AMD are drawn to FinFET which offers 90 percent more density than 28nm. Both will boost the transistors on offer with their next-generation GPUs, with 17 to 18 billion transistors currently being rumored.
AMD is continuing to lose market share to Nvidia, despite the fact that its new best video card, the Fury is out.
AMD always had a get out of jail card when the last GPU market share numbers were out on the basis of it not having released anything. At the time NVidia had 76% of the discrete GPU market. This was when Nvidia’s best card was the GeForce GTX 980.
A lot happened in that time. There was the release of the Titan X in March, and before the GTX 980 Ti in June. AMD had its Hawaii architecture inside of the R9 290X, and the dual-GPU in the form of the R9 295X2. It was expected that the R9 390X might turn AMD’s luck around but that turned out to be another rebrand. Then there was the arrival of the R9 Fury X.
AMD has new products on the market: the R9 Fury X, R9 Fury, R9 390X and a bunch of rebranded 300 series video cards. But according to Mercury Research’s latest data, NVIDIA has jumped from 76% of the discrete GPU market in Q4 2014 to 82 per cent in Q2 2015.
AMD has 18 per cent of the dGPU market share, even after the release of multiple new products.
It is not that the Fury X isn’t selling well, but because of yield problems there will only 30,000 units made over the entire of the year.
AMD also rebranded nearly its entire product stack thus making no reason to buy a R9 390X if you own an R9 290X.
Sure there is 8GB of GDDR5 on board compared to the 4GB offered on most R9 290X cards, but that’s not enough to push someone to upgrade their card.
Tweaktown noted that there was a big issue of the HBM-powered R9 Fury X not really offering any form of performance benefits over the GDDR5-powered GeForce GTX 980 Ti from NVIDIA. The 980 Ti beating the Fury X in some tests which it should not have.
Nvidia has plenty of GM200 GPUs to go around, with countless GTX 980 Ti models from a bunch of AIB partners. There is absolutely no shortage of GTX 980 Ti cards. Even if you wanted to get your paws on a Fury X, AMD has made it difficult.
Now it seems that next year could be a lot worse for AMD. Nvidia will have its GP100 and GP104 out next year powered by Pascal. This will cane AMD’s Fiji architecture. Then Nvidia will swap to 16nm process when its Maxwell architecture is already power efficient. Then there is the move HBM2, where be should see around 1TB/sec memory bandwidth.
All up the future does not look that great for AMD.
TSMC has mentioned that it will be looking to commence production of its 16nm process in the third quarter now it looks like it is happening.
The outfit told the Taiwan Stock Exchange that its 16nm chips “smoothly entered” volume production as expected. It sounds like someone in TSMC’s PR department has been reading 40 Shades of Grey again.
Production of a 16nm node was underway, with higher yields expected later this quarter. Nvidia wants to get its GP100 Pascal GPU manufactured on this node as it switches to 16nm next year.
TSMC’s 16nm FinFET+ node will deliver twice the density and 65 percent higher speed at 70 percent less power than the current 28HPM process.
So far there has been a shortage of dates mentioned. We think that it will be the first quarter of 2016, with the hardware becoming available in the second quarter.
Both will still be behind Samsung which is going be 14nm fairly soon.
Our well informed industry sources have shared a few more details about the AMD’s 2016 Zen cores and now it appears that the architecture won’t use the shared FPU like Bulldozer.
The new Zen uses a SMT Hyperthreading just like Intel. They can process two threads at once with a Hyperthreaded core. AMD has told a special few that they are dropping the “core pair” approach that was a foundation of Bulldozer. This means that there will not be a shared FPU anymore.
Zen will use a scheduling model that is similar to Intel’s and it will use competitive hardware and simulation to define any needed scheduling or NUMA changes.
Two cores will still share the L3 cache but not the FPU. This because in 14nm there is enough space for the FPU inside of the Zen core and this approach might be faster.
We mentioned this in late April where we released a few details about the 16 core, 32 thread Zen based processor with Greenland based graphics stream processor.
Zen will apparently be ISA compatible with Haswell/Broadwell style of compute and the existing software will be compatible without requiring any programming changes.
Zen also focuses on a various compiler optimisation including GCC with target of SPECint v6 based score at common compiler settings and Microsoft Visual studio with target of parity of supported ISA features with Intel.
Benchmarking and performance compiler LLVM targets SPECint v6 rate score at performance compiler settings.
We cannot predict any instruction per clock (IPC improvement) over Intel Skylake, but it helps that Intel replaced Skylake with another 14nm processor in later part of 2016. If Zen makes to the market in 2016 AMD might have a fighting chance to narrow the performance gap between Intel greatest offerings.
After a lot of asking around, we can give you some actual numbers about the AMD’s coherent fabric.
The inter-connecting technology already sounded very promising, but now we have the actual number. The HSA, Heterogeneous System Architecture MCM (Multi Chip Module) that AMD is working on can give you almost seven times faster score than the traditional PCIe interface.
Our industry sources have confirmed that with 4 GMI (Global Memory Interconnect) links AMD’s CPU and GPU can talk at 100GB/s. the traditional PCIe 16X provides 15GB/s at about 500 ns latency. Data Fabric eliminates PCIe latency too.
AMD will be using this technology with the next gen Multi Chip module that packs a Zeppelin CPU (most likely packed with a bunch of ZEN cores) and a Greenland GPU that of course comes with super fast HBM (High Bandwidth Memory). The Greenland and HBM can communicate at 500 GB/s and can provide highest performance GPU with 4+ teraflops.
This new MCM package based chip will also talk with DDR4 3200 memory at 100GB/s speed making it quite attractive for the HSA computation oriented customers.
It was rumored back in April about AMD’s upcoming Exascale Heterogeneous Processor (EHP) with 16 cores and a Greenland APU, and now it seems that the rest of the world has caught up to the news.
A paper was submitted at IEEE and it was the first time AMD mentioned sixteen Zen cores wrapped around the GPU and powered by HBM 2 memory. We believe that this is a 16-core processor with 32 thread support and not 32 core as many reported. We will know soon enough and then can have another “we told you so” headline.
We would not be surprised if we hear more about this AMD processor at the Hot Chips conference on August 23rd. The EHP computing solution uses a silicon interposer and an APU chip that, almost as a raison d’être for AMD over the past several years, packs a GPU and CPU into a well-tuned band. All this will be surrounded by die-stacked DRAM.The Italian website that brought this news back to life claims that AMD expects to ship the product between 2016 and 2017. That is the sort of timing you can expect with the rest of the ZEN based cores on the market. One can only hope that it will happen sooner rather than later. AMD needs to get more of the high performance compute market and earn some profits.
The IEEE article gives a bit of light on AMD exascale computer strategy:
Exascale computing requires very high levels of performance capabilities while staying within very stringent power budgets. Hardware optimized for specific functions is much more energy efficient than implementing those functions with general purpose cores. However, there is a strong desire for supercomputer customers to not have to pay for custom components designed only for high-end HPC systems, and therefore high-volume GPU technology becomes a natural choice for energy-efficient data-parallel computing. To fully realize the capabilities of the GPU, we envision exascale compute nodes comprised of integrated CPUs and GPUs (i.e., accelerated processing units or APUs) along with the hardware and software support to enable scientists to effectively run their scientific experiments on an exascale system. [In the paper submitted to IEEE...] We discuss the hardware and software challenges in building a heterogeneous exascale system, and we describe on-going research efforts at AMD to realize our exascale vision.
AMD has come up with a new interface, interconnect technology it calls coherent fabric.
It is a new inter-chip connection that will be used in the Exascale Heterogeneous Processor (EHP). Coherent fabric is primary interconnect for IP on AMD SOCs, APUs, CPUs and GPUs beginning with the next generation CPU SOC and followed by the Greenland GPU ASIC. Incoherant fabric, on the other hand, is something worn by loud American tourists and is sometimes called a Hawaiian shirt.
AMD’s GMI – Global Memory Interconnect, extends the coherent fabric between chips on the same multi-chip module (MCM) package. The coherent fabric extends between packages using combo phys which also support PCIe interconnection.
All this sounds rather complicated, but is basically a faster way for Zen SoC to talk to the Greenland Die on future processors.
Coherent fabric will speed up the communication in Zen CPUs between the cores and cache, host controllers, eg. USB, SATA or GbE, memory controllers, PSP, timer, counter, ACPI or legacy interface, combo phys and GMI layers. GMI phys of Zen die will talk directly with GMI phys of Greenland HBM 2.0 graphics die and uses coherent fabric too.
Our industry sources did not want to go into the specifics of achievable speeds of this interconnect, but the promise of a significant improvement is there. And of course the upcoming AMD’s 16 core Heterogenous EHP Processor will benefit from the fast coherent fabric interconnect.
AMD’s coherent fabric is the direct competitor to Nvidia’s NVlink interconnection that is expected in 2016 with the Pascal GPUs.
ARM has acquired Sansa Security to bolster protection against cyber threats in its Internet of Things (IoT) offerings.
Sansa is an Israel-based company that provides hardware security intellectual property (IP) and software for system-on-chip components that end up in around 150 million devices every year.
The technology makes it easier for manufacturers to build secure products by offering a complete hardware subsystem that adds additional isolation of security operations from the main application processor.
ARM’s decision to snap up the firm will deliver the technology across the ARM security portfolio, including TrustZone and SecurCore processor IP.
ARM said that the acquisition, the terms of which have not yet been disclosed, creates “extra protection against malware and malicious software”.
“It is a system-wide approach that underpins security-related chipset and trusted software needs. This enables the protection of any connected device and management of sensitive data and content,” said the firm.
ARM CTO Mike Muller added: “Any connected device could be a target for a malicious attack so we must embed security at every potential attack point.
“Protection against hackers works best when it is multi-layered, so we are extending our security technology capability into hardware subsystems and trusted software. This means our partners will be able to license a comprehensive security suite from a single source.”
ARM is definitely taking this IoT thing seriously. In April, the firm announced another acquisition in a bid to expand its presence in the IoT arena, creating a new portfolio dubbed ARM Cordio in the process.
The UK semiconductor designer picked up Wicentric, a Bluetooth smart stack and profile provider, and Sunrise Micro Devices, a provider of sub-one volt Bluetooth radio IP.
It could be said that the firm is pushing its stance in the IoT market in a bid to capitalise on what is essentially the next big thing in tech before it becomes ubiquitous.
For instance, ARM joined forces with IBM in February to launch its mbed Device Platform as a starter kit with cloud support, offering developer tools with cloud-based analytics.
The mbed tool was announced last year and is primarily an operating system built around open standards to “bring internet protocols, security and standards-based manageability into one integrated tool” and make IoT deployment faster and easier and thus speed up the creation of IoT-powered devices.
Launching the mbed IoT Starter Kit Ethernet Edition with IBM means that the company can channel data from internet-connected devices directly into IBM’s Bluemix cloud platform.
The IoT Starter Kit consists of an ARM mbed-enabled development board from Freescale, powered by an ARM Cortex-M4-based processor, together with a sensor IO application shield.
The PlayStation business has had another phenomenal quarter in the first four months of 2015, selling three million PS4 units and turning in an operating income of $160 million from revenues of $2.365 billion. There are now 25.3 million PlayStation 4 units in the hands of players worldwide – a number achieved in less than two full years.
The console continues to be the company’s fastest seller – outpacing the PS2, which took two years and eight months to reach the 20 million mark. Furthermore, thanks to dropping production costs for PS4 hardware, a 12 per cent increase in sales from the same quarter last year translated to a massive 350 per cent rise in operating income.
A strengthening dollar again hurt Sony’s bottom line, having an estimated impact of 15.6 billion Yen on the revenue total of 288.6 billion Yen, but this was massively outweighed by the increase in sales and the efficiency gains of Sony’s operation. On the strength of the results, Sony has added another 20 billion Yen in operating income to the sector’s full year forecast.
The sales rate of PS4 shows a healthily steady growth in player base, returning to a gradual upswing after a huge blip in Q3, 2014. Sony has upgraded it full year forecast from 16 million units to 16.5 as a result – a figure which would show a substantial increase on 2014′s 14.8 million total. By Sony’s own reckoning, the end of Q1 2016 will see nigh on 40 million of the consoles in homes. Vita sales once again went unmentioned in the report, whilst the gradual decline of PS3 continued.
Hardware wasn’t the only success story. Network, (“Network includes network services relating to game, video, and music content provided by Sony Network Entertainment Inc.”) mad almost as much in revenues, netting around 105.8 billion Yen compared to Hardware’s 129.5 billion. The Other category (Other includes packaged software and peripheral devices) brought in 30.6 billion.
Overall, the corporation turned a healthy profit, banking $676 million in net from sales of nearly $15 billion. Whilst the PlayStation business is very healthy indeed, it’s far from Sony’s only, or even biggest, success story: Devices, Imaging, Financial Services and Music all continue to return a higher operating income.
Samsung has put 10nm FinFET in its roadmap to stop its customers migrating to TSMC.
There were some rumours that Samsung may alter its schedule in order to prevent clients that might consider switching to 10nm chips from TSMC as that outfit is expected to skip the 14nm process and go straight to 10nm
Kelvin Low from Samsung Foundry confirmed in a video posted on YouTube that Samsung has formally added 10nm FinFET into the process roadmap, for chip designers working in mobile, consumer or networking market segment the new chips will provide significant performance and power consumption improvements.
Samsung LSI division has already shown off its first 10nm wafers which was a symbolic message to major clients that Samsung is more than capable of getting its 10nm production lines up and running without much hassle. Low expected 10nm products to appear at the end of 2016
Investors in ARM are deeply worried about its close relationship to the fruity cargo cult Apple.
ARM released its results which looked great, but investors were looking at its close ties to Jobs’ Mob which posted results which were disappointing.
Shares dropped 3.1 per cent on the back of Apple’s results. Apple uses ARM’s processor designs in its range of iThings.
It seems odd as ARMs Revenues rose 22 per cent to $17.5m for its second quarter, while pre-tax profits increased 32 per cent to $90.9m, compared with the same period last year.
The chip designer signed 54 processor licences for the three months, a “record” number.
Simon Segars, ARM chief exec, said a diverse range of companies chose to license ARM’s latest processors in the second quarter and physical IP for future product developments.
“ARM has been investing in advanced technology products for mobile devices, automotive applications and enterprise infrastructure, and in Q2 ARM signed licences for many of these new products. This licensing activity will help to grow the royalty revenue opportunity for years to come,” he said.
TSMC president and co-CEO Mark Liu has announced his outfit has begun volume shipment of chips based on its 16-nm FinFET manufacturing process.
He added that the ramping of the 16-nm process will be even more aggressive than that of its 20-nm process and he wants to gain foundry market share over the remainder of 2015 and well into 2016 on the back of the technology.
The foundry expects 16-nm processor shipments to begin contributing to revenue in the fourth quarter of 2015, since the process will ramp up during much of the third quarter.
TSMC moved quickly from 20-nm to 16-nm manufacturing claiming that 16 nanometer shared a similar metal backend process with 20 nanometer. In other words its 16 FinFET benefited from what it learnt doing 20-nanometer.
Liu also talked about the foundry’s 10-nm and 7-nm processes, saying that the recent product-like validation vehicle milestone was encouraging and that its plans are on-track.
TSMC plans to make 7-nm validation samples in the second quarter of 2017, just fifteen months after 10-nm validation.
Fabless chipmaker AMD has come up with a mixed set of results for the second quarter. The company managed to make as much cash as the cocaine nose jobs of Wall Street expected, but missed revenue expectations.
In fact its revenues were below the psychologically important billion figure at $942 million.
We knew it was going to be bad. Last week we were warned that the results would be flat. The actual figure was $942m, an 8.5 per cent sequential decline and a 34.6 per cent drop from the same period a year ago.
As you might expect, there are some measures of this not being AMD’s fault. The company is almost entirely dependent on PC sales. Not only have these fallen but don’t look like they are going to pick up for a while.
AMD’s Computing and Graphics division reported revenue of $379m, which was down 54.2 per cent, year-on-year. Its operating loss was $147m, compared to a $6m operating loss for last year’s quarter.
Lisa Su, AMD president and CEO, in a statement said that strong sequential revenue growth in AMD’s enterprise, embedded, and semi-custom segment and channel business was not enough to offset near-term problems in its PC processor business. This was due to lower than expected consumer demand that impacted sales to OEMs, she said.
“We continue to execute our long-term strategy while we navigate the current market environment. Our focus is on developing leadership computing and graphics products capable of driving profitable share growth across our target markets,” she added.
In the semi-custom segment, AMD makes chips for video game consoles such as the Nintendo Wii U, Microsoft Xbox One, and Sony PlayStation 4 consoles. That segment did reasonably well, up 13 percent from the previous quarter but down 8 percent from a year ago.
But AMD’s core business of processors and graphics chips fell 29 percent from the previous quarter and 54 percent from a year ago. AMD said it had decreased sales to manufacturers of laptop computers.
Figures like this strap a large target on AMD’s back with a sign saying “take me over” but AMD is not predicting total doom yet.
For the third quarter, AMD expects revenue to increase 6 percent, plus or minus 3 percent, sequentially, which is a fairly conservative outlook given the fact that Windows 10 is expected to push a few sales its way.
AMD supplies chips to the Nintendo Wii U, Microsoft Xbox One, and Sony PlayStation 4 consoles and these seem to be going rather well.
The maker of chips and bits, Intel is doing better than the cocaine nose jobs of Wall Street expected.
Intel issued its quarterly results today and saw growth in its data centers and Internet-of-Things businesses offset weak demand for personal computers that use the company’s chips.
Intel said that it was expanding its line-up of higher-margin chips used in data centers to counter slowing demand from the PC industry. Its cunning plan to buy Altera for $16.7 billion in April was all about trying to do this.
Revenue from the data centres grew 9.7 percent to $3.85 billion in the second quarter from a year earlier, helped by cloud services companies and demand for data analytics.
Chief Financial Officer Stacy Smith was predicting robust growth rates of the data center group, Internet of Things group and NAND businesses.
Revenue from the PC business, which is still Intel’s largest, fell 13.5 percent to $7.54 billion in the quarter ended June 27.
However there was more doom about the PC market which Smith said was going to be weaker than previously expected.
Research firm Gartner thinks global PC shipments will fall 4.5 percent to 300 million units in 2015, and life is going to be pretty pants until 2016.
Intel forecast current-quarter revenue of $14.3 billion, plus or minus $500 million. Wall Street predicted a revenue of $14.08 billion.
The company’s net income fell to $2.71 billion from $2.80 billion a year earlier.
Net revenue fell 4.6 percent to $13.19 billion, but edged past the average analyst estimate of $13.04 billion. Intel’s stock fell about 18 percent this year.
AMD’s Project Quantum PC system, with graphics powered by two of the new Fiji GPUs may have got the pundits moist but it has been discovered that the beast has Intel inside
KitGuru confirmed that the powerful tiny system, as shown at AMD’s own event, was based upon an Asrock Z97E-ITX/ac motherboard with an Intel Core i7-4790K ‘Devil’s Canyon’ processor.
Now AMD has made a statement to explain why it chose to employ a CPU from one of its competitor in what is a flagship pioneering gaming PC.
It told Tom’s Hardware that users wanted the Devil’s Canyon chip in the Project Quantum machine.
Customers “want to pick and choose the balance of components that they want,” and the machine shown off at the E3 was considered to be the height of tech sexiness right now.
AMD said Quantum PCs will feature both AMD and Intel CPUs to address the entire market, but did you see that nice Radeon Fury… think about that right now.
IT is going to be ages before we see the first Project Quantum PCs will be released and the CPU options might change. We would have thought that AMD might want to put its FinFET process ZEN CPUs in Project Quantum with up to 16 cores and 32 threads. We will not see that until next year.