Subject: Processors, Mobile | October 29, 2014 - 04:30 AM | Scott Michaud
Tagged: arm, mali-T800, mali
While some mobile SoC manufacturers have created their own graphics architectures, others license from ARM (and some even have a mixture of each within their product stack). There does not seem to be a specific push with this generation, rather just increases in the areas that make the most sense. Some comments tout increased energy efficiency, others higher performance, and even API support got a boost to OpenGL ES 3.1, which brings compute shaders to mobile graphics applications (without invoking OpenCL, etc.).
Three models are in the Mali-T800 series: the T820, the T830, and the T860. As you climb in the list, the products go from entry level to high-performance mobile. GPUs are often designed in modularized segments, which ARM calls cores. You see this frequently in desktop, discrete graphics cards where an entire product stack contains a handful of actual designs, but products are made by disabling whole modules. The T820 and T830 can scale between one to four "core" modules, each core containing four actual "shader cores", while the T860 can scale between one to sixteen "core" modules, each core with 16 "shader cores". Again "core modules" are groups that contain actual shader processors (and L2 cache, etc.). Cores in cores.
This is probably why NVIDIA calls them "Streaming Multiprocessors" that contain "CUDA Cores".
ARM does not (yet) provide an actual GFLOP rating for these processors, and it is up to manufacturers to some extent. It is normally a matter of multiplying the clock frequency by the number of ops per cycle and by the number of shader units available. I tried, but I assume my assumption of instructions per clock was off because the number I was getting did not match with known values from previous generations, so I assumed that I made a mistake. Also, again, ARM considers their performance figures to be conservative. Manufacturers should have no problem exceeding these, effortlessly.
As for a release timeline? Because these architectures are designed for manufacturers to implement, you should start seeing them within devices hitting retail in late 2015, early 2016.
Subject: General Tech, Processors, Chipsets | October 23, 2014 - 03:25 PM | Scott Michaud
Tagged: Intel, Broadwell, Broadwell-E, Haswell-E
VR-Zone China got hold of an alleged Intel leak, go figure, that talks about their next enthusiast processor platform, Broadwell-E. This architecture is mostly Haswell-E that has its (rated) feature size shrunk down to 14nm. Given an available BIOS, it is expected to support at least some existing LGA 2011-v3 motherboards with the X99 chipset. Like Haswell, they are sticking with a maximum of 40 PCIe lanes. We will need to wait for individual SKUs to see whether one or more models will be limited to 28 lanes, like the Haswell-E-based Core i7-5820K.
Image Credit: Chinese VR-Zone
Intel claims 140W TDP, which is identical to the current three offerings of Haswell-E, for all options. The slide claims six and eight core models will be available (also identical to Haswell-E).
One bullet-point that baffled me is, "Integrated Memory Controller: 4 Channels DDR4 2400, 1 DIMM per Channel". Double-checking with the other writers here, just to make sure sure, it seems like the slide claims that Broadwell-E will only support four sticks of DDR4. This makes zero sense for a couple of reasons. First, one of the main selling points of the enthusiast platform has been the obscene amount of RAM that workstation users demand. Second, and more importantly, if it is compatible with existing motherboards, what is it going to do? Fail to POST if you install a fifth stick? This has to be a typo or referring to something else entirely.
When will you be able to get it? A bit later than we were hoping. It is expected for Q1 2016, rather than late 2015.
Subject: Processors | October 22, 2014 - 10:02 PM | Josh Walrath
Tagged: Richland, Q3 results, lisa su, Kaveri, APU, amd, A10 7850K
While AMD made a small profit last quarter, the Q4 outlook from the company is not nearly as rosy. AMD estimates that Q4 revenues will be around 12% lower than Q3, making for a rare drop in what is typically a robust season for sales. Unlike Intel, AMD is seeing a very soft PC market for their products. Intel so far has been able to deliver parts that are as fast, if not faster than the latest APUs, but they also feature lower TDPs while at a comparable price. The one area that AMD has a significant advantage is in terms of 3D performance and better driver support.
To keep the chips selling during this very important quarter, AMD is cutting the prices on their entire lineup of FM2+ parts. This includes the entire Kaveri based lineup from the top end A10-7850K to the A6-7400K. AMD is also cutting the prices on the previous Richland based parts, which include the A10-6800K. Also of interest is that buyers of A10 APUs will be able to select one of three game titles (Murdered: Soul Suspect, Thief, or Sniper Elite 3) for free, or use the included code to purchase Corel’s Aftershot Pro 2 for only $5.
|Compute Cores||12 (4+8)||12 (4+8)||10 (4+6)||10 (4+6)||6 (2+4)|
|TDP (cTDP)||95 (65/45)||65 (45)||95 (65/45)||65 (45)||65 (45)|
The A10-7850K is a pretty good part overall, though of course it does suffer at the hands of Intel when it comes to pure CPU performance. It still is a pretty quick part that competes well with Intel’s 2 core/4 thread chips. 3D performance from the integrated graphics is class leading, and the potential for using that unit for HSA applications is another checkmark for AMD. We have yet to see widespread adoption of HSA, but we are seeing more and more software products coming out that support it. Having tested it out myself, the GPU portion of the APU can be enabled when using a standalone GPU from either AMD or NVIDIA. The Kaveri chips also support TrueAudio, which will show up in more titles throughout the next year.
One aspect of AMD’s latest FM2+ platform that cannot be ignored is the pretty robust selection of good and interesting motherboards that are offered at very low prices. Products such as the Gigabyte G1.Sniper.A88X and the MSI A88X-G45 Gaming motherboards are well rounded products that typically sell in the $90 to $110 range. Top end products like the Asus Crossblade Ranger are still quite affordable at around $160. Budget offerings are still pretty decent and they come in the $50 range.
One other product that has sparked interest is the Athlon X4 860K Black Edition. This product is clocked between 3.7 GHz and 4.0 GHz, features two Steamroller modules, and is priced at a very reasonable $90. The downside is that there is no GPU portion enabled, while the upside is that there is potentially more thermal headroom for the CPU portion to be clocked higher than previous A10-7850K parts. This will of course differ from individual chips, but the potential is there to have a pretty solid CPU for a very low price. Add in the low motherboard prices, and this has the making of a nice budget enthusiast system.
So why the cuts now? We can simply look at last week’s results for AMD’s previous quarter, as well as how the next quarter is stacking up. While AMD made a small profit last quarter, predictions for Q4 look grim. AMD is looking at around a 12% decrease in revenue, as stated above. AMD has a choice in that they can keep ASPs higher, but risk shipping less product in the very important 4th quarter; or they can sacrifice ASPs and potentially ship a lot more product. The end result of cutting the prices on their entire line of APUs will be of course lower ASPs, but a higher volume of parts being shipped and sold. In terms of cash flow, it is likely more important to see parts flowing rather than having higher inventories with a higher ASP. This also means that more APUs being sold will mean more motherboards from their partners moving through the channel.
Intel does have several huge advantages over AMD in that they have a very solid 22 nm process, a huge workforce that can hand tune their processors, and enough marketing money to make any company other than Apple squirm. AMD is at the mercy of the pure-play foundries in terms of process node tweaks and shrinks. AMD spent a long time at 32 nm PD-SOI before it was able to migrated to 28 nm HKMG. It looks to be 2015 before AMD sees anything below 28 nm for their desktop APUs, but it could be sooner for their smaller APUs and ARM based products on planar 20 nm HKMG processes. We don’t know a all of the specifics of the upcoming 16/14nm FinFET products from TSMC, Samsung, and GLOBALFOUNDRIES, so it will be hard to compare/contrast to Intel’s 2nd generation 14 nm TriGate line. All we know is that it will most assuredly be better than the current 28 nm HKMG that AMD is stuck at.
Subject: Graphics Cards, Processors | October 8, 2014 - 05:54 PM | Scott Michaud
In an abrupt announcement, Rory Read has stepped down from his positions at AMD, leaving them to Dr. Lisa Su. Until today, Mr. Read served as president and Chief Executive Officer (CEO) of the x86 chip designer and Dr. Su as Chief Operating Officer (COO). Today however, Dr. Su has become president and CEO, and Mr. Read will stay on for a couple of months as an adviser during the transition.
Josh Walrath, editor here at PC Perspective, tweeted that he was "Curious as to why Rory didn't stay on longer? He did some good things there [at AMD], but [it's] very much an unfinished job." I would have to agree. It feels like an odd time, hence the earlier use of the word "abrupt", to have a change in management. AMD restructured just four months ago, which was the occasion for Dr. Su to be promoted to COO. In fact, at least as far as I know, no-one is planned to fill her former position as COO.
These points suggest that she was planned to take over the company for at least several months.
I have been told that timing is everything. I guess this rings true, but only if you truly know the circumstances around any action. Today’s announcement by AMD was odd in its timing, but it was not exactly unexpected. As Scott mentioned above, I was confused by this happening now. I had expected Rory to be in charge for at least another year, if not two. Rory had hinted that he was not planning on being at AMD forever, but was aiming at creating a solid foundation for the company and to help shore up its finances and instill a new culture. While the culture is turning due to pressure from up top as well as a pretty significant personnel cuts, AMD is not quite as nimble yet as they want to be.
Rory’s term has seen the return of seasoned veterans like Jim Keller and Raja Koduri. These guys are helping to turn the ship around after some fairly mediocre architecturse on the CPU and GPU sides. While Raja had little to do with GCN, we are seeing some aggressive moves there in terms of features that are making their products much more competitive with NVIDIA. Keller has made some very significant changes to the overall roadmap on the CPU side and I think we will see some very solid improvements in design and execution over the next two years.
Lisa Su was brought in by Rory shortly after he was named CEO. Lisa has a pretty significant background in semiconductors and has made a name for herself in her work with IBM and Freescale. Lisa attained all three of her degrees from MIT. This is not unheard of, but it is uncommon to stay in one academic setting when gaining advanced degrees. Having said that, MIT certainly is the top engineering and science school in the nation (if not the world). I’m sure people from RPI, GT, and CalTech might argue that, but it certainly is an impressive school to have on your resume.
Dr. Su has seemingly been groomed for this transition for quite some time now. She went from a VP to COO rather quickly, and is now shouldering the burden of being CEO. Lisa has been on quite a few of the quarterly conference calls and taking questions. She also serves on the Board of Directors at Analog Devices.
I think that Lisa will continue along the same path that Rory set out, but she will likely bring a few new wrinkles due to her experience with semiconductor design and R&D at IBM. We can only hope that this won’t become a Dirk Meyer 2.0 type situation where a successful engineer and CPU architect could not change the course of the company after the disastrous reign of Hector Ruiz. I do not think that this will be the case, as Rory did not leave the mess that Hector did. I also believe that Lisa has more business sense and acumen than Dirk did.
This change, at this time, has provided some instability in the markets when regarding AMD. Some weeks ago AMD was at a near high for the year at around $4.66 per share. Right now it is hovering at $3.28. I was questioning why the stock price was going down, and it seems that my question was answered. One way or the other, rumors of Rory taking off reached investors’ ears and we saw a rapid decline in share price. We have yet to see what Q3 earnings look like now that Rory has rather abruptly left his position, but people are pessimistic as to what will be announced with such a sudden departure.
Subject: Processors | September 30, 2014 - 06:02 PM | Josh Walrath
Tagged: arm, cortex, Cortex-A, cortex-m, 90 nm, 40 nm, 28 nm, 32 bit
Last week ARM announced the latest member of their Cortex-M series of embedded parts. The new Cortex-M7 design is a 32 bit processor designed to have good performance while achieving excellent power consumption. The M7 is a fully superscalar design with 6 pipeline stages. This product should not be confused with the Cortex-A series of products, as the M series is aimed directly at embedded markets.
This product is not necessarily meant for multi-media rich applications, so it will not find its way into a modern smart phone. Products that it is leveraged at would be products like the latest generation of smart watches. Industrial control applications, automotive computing, low power and low heat applications, and countless IoT (Internet of Things) products can utilize this architecture.
The designs are being offered on a variety of process nodes from 90 nm down to 28 nm. These choices are made by the licensee depending on the specifics of their application. In the most energy efficient state, ARM claims that these products can see multiple years of running non-stop on a small lithium battery.
This obviously is not the most interesting ARM based product that we have seen lately, but it addresses a very important market. What is perhaps most interesting about this release not only is the pretty dramatic increase in per clock performance from the previous generation of part, but also how robust the support is in terms of design tools, software ecosystem, and 3rd party support.
Cortex-M7 can also be utilized in areas where a more complex DSP has traditionally been used. In comparison to some common DSPs, the Cortex-M7 is competitive in terms of specialized workload performance. It also has the advantage of being much more flexible than a DSP in a general computing environment.
ARM just keeps on moving along with products that address many different computing markets. ARM’s high end Cortex-A series of parts powers the majority of smart phones and tablets while the Cortex-M series have sold in the billions addressing the embedded market. The Cortex-M7 is the latest member of that family and will find more than its fair share of products to be integrated into.
Subject: Graphics Cards, Processors | September 30, 2014 - 03:33 AM | Scott Michaud
Tagged: iris, Intel, core m, broadwell-y, broadwell-u, Broadwell
Intel's upcoming 14nm product line, Broadwell, is expected to have six categories of increasing performance. Broadwell-Y, later branded Core M, is part of the soldered BGA family at expected TDPs of 3.5 to 4.5W. Above this is Broadwell-U, which are also BGA packages, and thus require soldering by the system builder. VR-Zone China has a list of seemingly every 15W SKU in that category. 28W TDP "U" products are expected to be available in the following quarter, but are not listed.
Image Credit: VR-Zone
As for those 15W parts though, there are seventeen (17!) of them, ranging from Celeron to Core i7. While each product is dual-core, the ones that are Core i3 and up have Hyper-Threading, increasing the parallelism to four tasks simultaneously. In terms of cache, Celerons and Pentiums will have 2MB, Core i7s will have 4MB, and everything in between will have 3MB. Otherwise, the products vary on the clock frequency they were binned (bin-sorted) at, and the integrated graphics that they contain.
Image Credit: VR-Zone
These integrated iGPUs range from "Intel HD Graphics" on the Celerons and Pentiums, to "Intel Iris Graphics 6100" on one Core i7, two Core i5s, and one Core i3. The rest pretty much alternate between Intel HD Graphics 5500 and Intel HD Graphics 6000. Maximum frequency of any given iGPU can vary within the same product, but only by about 100 MHz at the most. The exact spread is below.
- Intel HD Graphics: 300 MHz base clock, 800 MHz at load.
- Intel HD Graphics 5500: 300 MHz base clock, 850-950 MHz at load (depending on SKU).
- Intel HD Graphics 6000: 300 MHz base clock, 1000 MHz at load.
- Intel Iris Graphics 6100: 300 MHz base clock, 1000-1100 MHz at load (depending on SKU).
Unfortunately, without the number of shader units to go along with the core clock, we cannot derive a FLOP value yet. This is a very important metric for increasing resolution and shader complexity, and it would provide a relatively fair metric to compare the new parts against previous offerings for higher resolutions and quality settings, especialy in DirectX 12 I would assume.
Image Credit: VR-Zone
Probably the most interesting part to me is that "Intel HD Graphics" without a number meant GT1 with Haswell. Starting with Broadwell, it has been upgraded to GT2 (apparently). As we can see from even the 4.5W Core M processors, Intel is taking graphics seriously. It is unclear whether their intention is to respect gaming's influence on device purchases, or if they are believing that generalized GPU compute will be "a thing" very soon.
Subject: Graphics Cards, Processors, Mobile | September 29, 2014 - 01:53 AM | Scott Michaud
Tagged: apple, a8, a7, Imagination Technologies, PowerVR
First, Chipworks released a dieshot of the new Apple A8 SoC (stored at archive.org). It is based on the 20nm fabrication process from TSMC, which they allegedly bought the entire capacity for. From there, a bit of a debate arose regarding what each group of transistors represented. All sources claim that it is based around a dual-core CPU, but the GPU is a bit polarizing.
Image Credit: Chipworks via Ars Technica
Most sources, including Chipworks, Ars Technica, Anandtech, and so forth believe that it is a quad-core graphics processor from Imagination Technologies. Specifically, they expect that it is the GX6450 from the PowerVR Series 6XT. This is a narrow upgrade over the G6430 found in the Apple A7 processor, which is in line with the initial benchmarks that we saw (and not in line with the 50% GPU performance increase that Apple claims). For programmability, the GX6450 is equivalent to a DirectX 10-level feature set, unless it was extended by Apple, which I doubt.
Image Source: DailyTech
DailyTech has their own theory, suggesting that it is a GX6650 that is horizontally-aligned. From my observation, their "Cluster 2" and "Cluster 5" do not look identical at all to the other four, so I doubt their claims. I expect that they heard Apple's 50% claims, expected six GPU cores as the rumors originally indicated, and saw cores that were not there.
Which brings us back to the question of, "So what is the 50% increase in performance that Apple claims?" Unless they had a significant increase in clock rate, I still wonder if Apple is claiming that their increase in graphics performance will come from the Metal API even though it is not exclusive to new hardware.
But from everything we saw so far, it is just a handful of percent better.
Subject: General Tech, Processors, Mobile | September 27, 2014 - 02:38 PM | Scott Michaud
Tagged: Intel, spreadtrum, rda, Rockchip, SoC
A few months ago, Intel partnered with Rockchip to develop low-cost SoCs for Android. The companies would work together on a design that could be fabricated at TSMC. This time Intel is partnering with Tsinghua Unigroup Ltd. and, unlike Rockchip, also investing in them. The deal will be up to $1.5 billion USD in exchange for a 20% share (approximately) of a division of Tsinghua.
Image Credit: Wikipedia
Intel is hoping to use this partnership to develop mobile SoCs, for smart (and "feature") phones, tablets, and other devices, and get significant presence in the Chinese mobile market. Tsinghua acquired Spreadtrum Communications and RDA Microelectronics within the last two years. The "holding group" that owns these division is apparently the part of Tsinghua which Intel is investing in, specifically.
Spreadtrum will produce SoCs based on Intel's "Intel Architecture". This sounds like they are referring to the 32-bit IA-32, which means that Spreadtrum would be developing 32-bit SoCs, but it is possible that they could be talking about Intel 64. These products are expected for 2H'15.
Subject: Processors | September 25, 2014 - 02:56 PM | Jeremy Hellstrom
Tagged: linux, X99, core i7-5960x, Haswell-E
After the smoke from their previous attempt at testing the i7 5960X CPU Phoronix picked up a Gigabyte X99-UD4-CF and have now had a chance to test Haswell-E performance on Linux. The new processor is compared to over a dozen others on machines running Ubuntu and really showed up the competition on benchmarks that took advantage of the 8 cores. Single threaded applications that depended on a higher clock speed proved to be a weakness as the 4790K's higher frequency allowed it to outperform the new Haswell-E processor. Check out the very impressive results of Phoronix's testing right here.
"With the X99 burned-up motherboard problem of last week appearing to be behind us with no further issues when using a completely different X99 motherboard, here's the first extensive look at the Core i7 5960X Haswell-E processor running on Ubuntu Linux."
Here are some more Processor articles from around the web:
- Intel's Xeon E5-2687W v3 @ The Tech Report
- Intel Core i7-5960X Extreme @ Benchmark Reviews
- Intel Core i7-4790K and Core i5-4690K @ X-bit Labs
- Return of the Athlon: AMD Brings Kabini to the desktop @ Bjorn3d
- AMD FX8370E @ Kitguru
- AMD FX-8370E @ eTeknix
Subject: General Tech, Motherboards, Processors | September 20, 2014 - 06:51 PM | Scott Michaud
Tagged: xeon, Haswell-EP, ddr4, ddr3, Intel
Well this is interesting and, while not new, is news to me.
The upper-tier Haswell processors ushered DDR4 into the desktops for enthusiasts and servers, but DIMMs are quite expensive and incompatible with the DDR3 sticks that your organization might have been stocking up on. Despite the memory controller being placed on the processor, ASRock has a few motherboards which claim DDR3 support. ASRock, responding to Anandtech's inquiry, confirmed that this is not an error and Intel will launch three SKUs, one eight-core, one ten-core, and one twelve-core, with a DDR3-supporting memory controller.
The three models are:
|E5-2629 v3||E5-2649 v3||E5-2669 v3|
|Cores (Threads)||8 (16)||10 (20)||12 (24)|
|Clock Rate||2.4 GHz||2.3 GHz||2.3 Ghz|
The processors, themselves, might not be cheap or easily attainable, though. There are rumors that Intel will require customers purchase at least a minimum amount. It might not be worth buying these processors unless you have a significant server farm (or similar situation).