CPU's Article (28)
How to Overclock an Intel Pentium G3258 Guide
» January 13, 2015 05:00PM
AMD FX 8370 & FX 8370E Review
» September 1, 2014 05:00PM
Intel Core i7 5960X Extreme Edition Review
» August 28, 2014 05:00PM
AMD FX-9590 & FX-9370 Review
» July 7, 2014 05:00PM
Intel Fourth Generation Core i7 4790K Review
» June 18, 2014 05:00PM
CPU's News (451)
Posted: May 15, 2015 10:03AM
A roadmap concerning Intel's Skylake plans has recently leaked, and while the roadmap starts from back in 2013, it carries us into next year. We're nearing the end of the Haswell series, obviously, with Broadwell desktop CPUs expected to arrive on June 2. The Broadwell processors are already here on the M, U, and Y series, but then Skylake will replace those right around the start of quarter four this year. The M processors will stay with Skylake until the very end of 2016, when Cannonlake arrives, but the U and Y series will get the Cannonlake replacements in quarter two of 2016. That doesn't give those series a long time on Skylake, but Cannonlake is expected to be 10nm, so reducing the power consumption makes sense with those.
Moving back to the desktop side, the Skylake processors figure to arrive this September, with the Core i5-6500K and Core i7-6700K leading the pack. Those two will be 95W models, but the Skylake series will have 65W and even 35W models. Next year, during the third quarter of 2016, the Skylake processors will receive a refresh, much like the Haswells did, and maybe even a Devil's Canyon-style refresh. On the high-end side, the Haswell-E series will continue a little while longer, with the Skylake-E chips not expected until the third quarter of 2016. It isn't known what motherboard chipset or even socket those will use, but considering the Skylakes are moving to Z170 and LGA 1151, the Skylake-Es will probably move to something else, too. We'll just have to wait and see what those bring, but first, people have the traditional Skylake CPUs to look forward to.
Intel has plenty in the works for the desktop and mobile side for processors, and I imagine many people are planning some upgrades when Skylake arrives in the next few months.
Posted: May 14, 2015 03:36PM
Author: Nick Harezga
Information about the upcoming Broadwell processors from Intel has become available in bits and pieces over the past few weeks. The Core i5-5675C and i7-5775C started to arrive at OEMs about a week ago with an anticipated public release of "soon." Benchmarks using an engineering sample of the i7-5775C were released yesterday by Chinese site HKEPC showing some impressive overclocking numbers. It has now been confirmed that the release date will match earlier speculation of a near Computex arrival, with the official release coming on June 2.
Source: WCCF Tech
Posted: May 13, 2015 04:21PM
Author: Nick Harezga
Information about the first engineering samples of the upcoming Intel Broadwell and Skylake CPUs has been released by Chinese site HKEPC. The site has obtained the i7-5775C Broadwell and i7-6700K Skylake processors and has been busy putting them through some testing. The 5775C was paired with an ASRock Z97 OC Formula motherboard using the L1.52 BIOS, indicating that current motherboards can support the upcoming chips. It was overclocked to a frequency of 5GHz at a voltage of 1.419V using air cooling, completing 32 million calculations in WPrime in 4.399 seconds.
Posted: May 11, 2015 05:35AM
Author: Brentt Moore
Late last month, it was reported that the Godavari refresh platform was spotted in the wild. Over the weekend, AMD decided to officially launch its series of Godavari APUs, which refresh the company’s highly successful Kaveri lineup. The latest chips by AMD continue to offer features found in Kaveri products, such as "Steamroller" x86 cores, a dual-channel DDR3 memory controller, heterogeneous system architecture capabilities, and a built-in Radeon graphics engine based on the GCN 1.1 architecture. The Godavari lineup does bring some improvements to the table, as it offers consumers Windows 10 readiness including DirectX 12 Multiengine and Multiadapter features.
The Godavari lineup is available immediately for purchase and features the same 7000 series nomenclature that the Kaveri lineup boasts.
Posted: May 7, 2015 11:56AM
The server and enterprise market is going to get a big boost, as the Intel Haswell-EX Xeon E7 v3 processor line has arrived. The Xeon E7 v3 processors feature 12 new chips, with the flagship Xeon E7-8890 v3 packing 18 cores, 36 threads, and 45MB of L3 cache. The other processors are all derived from that flagship beast, with the Xeon E7-8800 v3, Xeon E7-8870 v3, and Xeon E7-8860 v3 considered the advanced models (along with the flagship, of course). All of those have a 9.6 GT/s interconnect speed, DDR4/DDR3 support, and S8S scalability, with all featuring 18 cores, except for the E7-8860 V3, which has 16.
Moving down to S4S scalability are the 14-core Xeon E7-4850 v3 and 12-core E7-4830 v3, which have an 8 GT/s interconnect speed and DDR4/DDR3 support. Still on the S4S are the 10-core Xeon E7-4820 v3 and 8-core Xeon E7-4809 v3, with an interconnect speed of 6.4 GT/s and DDR4/DDR3 support. There are also the 10-core Xeon E7-8891 v3 and quad-core Xeon E7-8893 v3 for the enterprise and database markets, the 18-core Xeon E7-8880L v3 for low power servers, and the 16-core Xeon E7-8867 v3 for HPC.
Clock speeds range from 1.9GHz on the E7-4820 v3 to 3.2GHz on the E7-8893 v3, with Turbo speeds ranging from 2.7GHz on the E7-4830 v3 to 3.5GHz on the E7-8891 v3 and E7-8893 v3. The E7-4820 v3 and E7-4809 v3 do not have a Turbo frequency. TDPs range from 115W to 165W, a grand total of 1536GB of RAM is supported (although some S8S servers can handle up to 12TB), and all are 22nm parts. Prices go from $1,224 for the E7-4809 v3 all the way up to $7,175 for the flagship E7-8890 v3. All these Xeon E7 v3 CPUs are available now from selected manufacturers.
Posted: May 7, 2015 11:23AM
Recently some OEMs have been listing pre-built PCs with new Intel Broadwell chips, which is a bit of a surprise considering the fifth generation parts had not received anything definitive about a launch date. This relatively quiet launch of the 14nm Broadwell chips means some people may get a surprise when browsing for all-in-one and mini PCs. The Intel Core i5-5675C and Core i7-5775C CPUs are the unlocked models, with that "C" a bit of a change from the familiar "K" SKUs. Broadwell parts will be available in Broadwell Y (Core M), Broadwell-HQ, and Broadwell-U lines as well to better fit what kind of needs fit the consumer, whether it's low-power laptops, Ultrabooks, all-in-ones, or desktop PCs. Desktop models will be available in BGA or LGA versions, with BGA being the soldered-in and LGA running on LGA 1150 sockets.
The Core i5-5675C and Core i7-5775C both have a 65W TDP, which is expected to increase once the Skylake series arrives. Broadwell is the last generation of chips with a fully integrated voltage regulator, as that will be changed on Skylake, too. What Broadwell does different compared to Haswell and other previous models is that it integrates eDRAM directly on the die to help boost the Iris Pro 6200 graphics. Anyone using the integrated graphics should see quite a nice boost in performance, so those on Ultrabooks or mini PCs should be happy. Broadwell is not seen as the direct successor to the current Haswell parts, as that is Skylake's job, but rather a sidestep for anyone in need of a new PC right now.
The Intel Broadwell processors will be arriving soon at your favorite (r)etail outlets, although an exact date is not known. Currently there are just some OEMs using them, although it won't be for long considering Skylake is set to arrive later this fall. Skylake also uses the new LGA 1151 socket and a new motherboard line, so Broadwell is the last of LGA 1150 processors to still use in your current motherboard.
Posted: May 6, 2015 03:47PM
Author: Nick Harezga
A wave of details about the upcoming Intel Skylake based processors, including the Core i7 6700K and Core i5 6600K, were revealed in April. We learned about the CPUs starting with CPU speed and thermal characteristics and moving to benchmarks that compared the chips to the i7 4790K and other current generation CPUs. Also revealed were some motherboards using the Intel Z170 chipset from Chinese company Colorful, with an anticipated release of September 2015. The recently released Intel roadmap confirms that the new processors will be released in the third quarter of this year, indicating that initial estimates of a September release were accurate. In addition to confirming the release date of the processors, it was revealed that there will be a total of ten CPUs in the Skylake family. Full details for all chips can be seen by following the link.
Source: WCCF Tech
Posted: May 6, 2015 01:15PM
After plenty of leaks and rumors, and even a core block diagram, AMD has officially unveiled its Zen CPU core. During the company's Financial Analyst Day, the wraps were finally taken off the next-generation x86 core, and things are looking quite good. Zen is the successor to AMD's Bulldozer line and is the first processor from the company to make use of Simulatenous Multithreading (SMT). SMT allows for a processor to use multiple, independent threads to better execute tasks, which should speed things up considerably from AMD's past use of Clustered Multithreading. Zen also makes use of a high-bandwidth, low latency cache system to help improve cache performance compared to Bulldozer. AMD announced the Zen cores will make use of FinFET technology, but did not specifiy if it would be built on a 14nm or 16nm process. The 14nm chips sound most likely, especially if yesterday's roadmaps are any indication.
Compared to AMD's upcoming Excavator CPUs, the last of the Bulldozer line due out later this year, the Zen cores have a 40% increase in instructions per clock. It would put Zen performance on the same level as Intel Haswell, at least according to single threaded benchmarks. It's something AMD needs to aim for, especially with the financial woes.
The first AMD Zen products won't be available until next year when the Summit Ridge CPUs, part of the FX line, appear. Those FX CPUs will have DDR4 support and a "high core count with multi-threading," which could mean up to eight cores, or potentially more. There will also be Bristol Ridge and Basilisk APUs utilizing Zen cores for those who don't need quite that FX level performance. AMD also seems to be pushing the desktop Zen products over to the AM4 socket and not FM3 like yesterday's roadmap leak showed. The mobile APUs will be on the FP4 socket.
Zen right now sounds extremely promising, but we need a lot more information before we form any definitive thoughts on the matter. Hopefully that and more arrives during Computex next month.
Posted: May 5, 2015 09:32AM
A recent roadmap leak shows what AMD has in store for 2016, and it looks like desktop and mobile parts will get some Zen and K12 love. The remainder of 2015 will bring the second generation FX CPUs based on up to eight Piledriver cores, the Godavari APUs with up to four Steamroller cores, and Beema APUs with up to four Puma core. Both APUs will be built on the 28nm process, with the Beema APUs also in an SoC form. Pretty familiar stuff, just with some tweaks here and there. Once the calendar turns to 2016, however, we get some new toys, starting with the Summit Ridge performance CPU. It will use up to eight Zen cores, DDR4 support, and be on the FM3 socket instead of AM3+, like the current FX processors. Performance of this CPU should be far better than the current ones, and maybe even the current 16 core Opteron CPUs, if the Zen core block diagram is anything to go by.
Moving on to the APU side, there's the Bristol Ridge APU with up to four Zen cores, DDR4 support, and the next generation GCN core. It has full HSA 1.0 support and could have the first APU with stacked HBM, although right now that remains to be seen. There are also the Basilisk APUs, essentially a lower-powered Bristol Ridge, which have up to two Zen cores, DDR4 support, and a less powerful next generation GCN core. Both APUs are built on the 14nm process, but while the Bristol Ridge APUs use the FM3 socket, just like the high-end Summit Ridge CPUs, the Basilisk APUs use the FT4 BGA socket. There is even an SoC model for both with an integrated southbridge.
On the mobile line, which covers both laptops and tablets/smartphones, the 2015 lineup has the Carrizo and Carrizo-L APUs, as well as the Amur APUs. The Carrizo APUs feature up to four Excavator cores, the Carrizo-L APUs have up to four Puma+ cores, with both on the 28nm process. The Amur APUs have up to four ARM Cortex-A57 cores and are built on the 20nm process. Heading into 2016 brings the Bristol Ridge and Basilisk APUs, like the desktop side, but also the ultra-low power Styx APUs. The Bristol Ridge and Basilisk APUs are just like their desktop counterparts, but the Styx APUs change things up. These use up to two K12 cores, which are the company's custom made ARM cores.
Posted: April 29, 2015 03:51PM
Author: Nick Harezga
Details about the Intel 6th Generation Core processors, currently know as Skylake, have begun to leak over the past week starting with the processors and followed by compatible motherboards. The first rumored benchmarks for the quad core i7 6700K have been obtained by PCFRM that pit the new CPU against current offerings from Intel and AMD. Benchmarks in the report include PCMark, 3DMark, Cinebench, Crysis 3, GTAV, and Battlefield 4. The 6700K outperformed the comparable i7 4790K in all tests and also beat the six core 5820K in the PCMark test.
Source: Legit Reviews
Posted: April 27, 2015 08:35PM
We have had some news lately about AMD's Zen CPU line, with today bringing one that is especially interesting. First things first, this information is from an unverified slide, but if true sounds extremely promising. It seems a block diagram of the Zen CPU has found its way online, which compares Zen to AMD's final Bulldozer part, the Excavator. The upcoming AMD Carrizo APUs will be the last of the Bulldozer line and should launch later this year. A big difference between the Zen and Excavator lines is that Zen only has one integer cluster instead of two, which puts it more in line with older Phenom and Athlon K chips in terms of internal layout. This should greatly help out in floating point performance, since instead of two cores sharing one floating point unit, each core has its own.
Both floating point and integer performance should see a boost this way, with single threaded performance likely surpassing the Bulldozer line. The floating point for Zen has two FMAC 256-bit units that could work together for 512-bit AVX floating point instructions. Less complex process instructions should be handled twice as fast as Bulldozer, too. A 50% wider integer pipeline compared to Excavator is also present, which again will help with single threaded performance.
Again, all of this is based on an unverified slide, but it certainly sounds like the Zen CPUs will be something to keep an eye on when they launch next year.
Posted: April 27, 2015 07:53PM
Last week some leaked details appeared for the Intel Skylake processors, and today we have some information on the motherboards. Chinese company Colorful recently held an event to show off some of the Intel Z170 chipset motherboards it's working on, which use the new LGA 1151 socket for the Skylake processors. None of the motherboards have a finalized color or name scheme, but all show off the new features Intel is implementing in the chipset. Three different motherboards are seen, with support for up to 64GB of DDR4 2133MHz RAM. All three feature VRM cooling by way of thermal armor, and one even has that armor down near the PCIe slots. Three PCIe 3.0 x16 slots are seen on all the motherboards, as are three PCIe 3.0 x1 slots and even an M.2 port. There are also USB 3.0 ports, USB 2.0 ports, onboard audio powered by Gamer Voice, Dual Gigabit LAN, a Killer E2201 chip, and Intel I211-AT to help with connectivity.
All of these features and more will be seen on Intel Z170 motherboards likely to be shown off during Computex. That's still a little over a month away, but we should at least get an idea of what Intel has in store for the future. The Intel Z170 motherboards and Skylake processors are expected to arrive this September.
Posted: April 27, 2015 07:10AM
Author: Brentt Moore
While the Kaveri lineup by AMD has continued to impress gamers with its built-in Radeon graphics engine based on the GCN 1.1 architecture, a platform refresh is just on the horizon. Computerbase.de has announced that it has spotted the Kaveri-Refresh platform, which is officially known as Godavari, boasting the newer JC stepping. The upcoming platform, which is expected to use the FM2+ form-factor, will feature the same design as Kaveri, such as "Steamroller" x86 cores, a dual-channel DDR3 memory controller, heterogeneous system architecture capabilities, and the aforementioned Radeon graphics engine.
While the Godavari lineup was originally expected to feature a new 8000 series nomenclature, it looks like AMD will be sticking with the 7000 series naming scheme for unknown reasons. Below is a table that features the leaked SKU specifications of the Godavari platform before it was known that AMD was sticking with the current 7000 series nomenclature.
|Godavari||CPU Cores||CPU Clock Base/Boost||L2 Cache||GPU/Cores||GPU Clock||TDP|
|AMD A10-Pro 8850B||4||3.7/4.1GHz||4MB||R7/512||800MHz||95W|
|AMD A10 PRO-8750B||4||3.0/3.5GHz||4MB||R7/512||TBD||65W|
|AMD A8 PRO-8650B||4||2.9/3.2GHz||4MB||R7/384||760MHz||65W|
|AMD A6 PRO-8550B||2||3.1/3.5GHz||2MB||R5/256||TBD||65W|
|AMD A4 PRO-8350B||2||3.0/3.5GHz||2MB||R5/256||757MHz||65W|
|AMD Athlon X4 870K||4||3.5/3.7GHz||2MB||N/A||N/A||95W|
|AMD Athlon X4 850||4||2.9/3.2GHz||2MB||N/A||N/A||65W|
Posted: April 24, 2015 04:04PM
AMD has been previously rumored to be working on a 16 core APU, but a new rumor is doubling that core count for the server market. The Opteron CPU would pack 32 Zen cores and, being a server part, would not need a Greenland GPU along for the ride. It seems that stripping out the GPU component allows for double the cores. Each core apparently will have 512KB of L2 cache and every four cores share 8MB of L3 cache. The CPU would have eight clusters of four cores each, with support for secure boot, cryptographic co-processing, and eight DDR4 memory channels. Each channel would support 256GB of RAM, so that means 2TB in total is possible. The 32 core model should sit at the top, with fewer cores available as you get further down the line, based on what your server needs.
Again, this is all just rumor, with the 32 core Opteron not likely to arrive until sometime next year. Once we have some confirmation on this and any other AMD part in the works, you can be sure to find it here.
Posted: April 21, 2015 02:01PM
Intel is still riding the wave of Haswell processors on the desktop side, but before long the Skylake series will launch. Today we have some new information by way of a leak concering the Intel Core i5 6600K and the Core i7 6700K. Both processors sit atop their respective lines, and both are built on the 14nm process. The Core i5 6600K apparently has a core clock of 3.5GHz with a boost clock of 3.9GHz, while the Core i7 6700K has a core clock of 4.0GHz and a boost clock of 4.2GHz. That is a fairly high core clock for the 6700K, which may limit its overclocking capabilities. Intel's Devil's Canyon processors were the same, and the i7 4790K managed to hit 4.6GHz during OCC's test, so that may be the limit on the 6700K, too.
However, both the Core i5 6600K and Core i7 6700K have a 95W TDP, which puts them above the Devil's Canyon and could help out in the overclock landscape. The 6600K is a quad-core processor without Hyper Threading, while the 6700K is a quad-core with Hyper Threading, so that distinction between the two lines remains the same. Each one also has Intel HD 5000 Graphics integrated directly on the die. The Core i5 has 6MB of L3 cache and the Core i7 has 8MB, and interestingly, both have a DDR3/DDR4 memory controller capable of 1600/2133MHz RAM. The motherboard you pick will decide which RAM it uses, so it will be interesting to see what that holds. One thing those motherboards have is a new socket, as the Skylake processors are on LGA 1151. So yes, you will need a new motherboard to take advantage of the new processors.
Now, keep in mind all of this is just rumor based on information from a leak. Intel is keeping basically everything about the Skylake processors under wraps until its big reveal date, so take the above with a grain of salt. It all sounds quite interesting, so we'll have to see how much pans out.
Posted: April 20, 2015 10:57AM
In the middle of this year, Intel is planning to release the 5th generation Broadwell Socketed CPUs for desktops. Broadwell CPUs are already available and used in devices, but this will be their first appearance for the desktop platform. They will come in two series, with one being soldered BGA chips and the other being the unlocked C-Series. The two socketed CPUs will have 65 W TDPs and feature Iris Pro graphics.
Posted: April 14, 2015 10:05AM
A couple weeks back a rumor appeared that AMD was working on a new APU featuring 16 Zen cores and a Greenland GPU. It would also feature HBM and DDR4 support, but not a whole lot else was known. Now some new rumors have turned up on this APU, and it seems even more fantastical. This APU would have up to 16 Zen x86 cores with 32 threads, with 512KB of L2 cache per core and 32MB of L3 cache. It also has AMD Crypto Co-Processor and Secure Boot, which would make this more of a server chip than desktop. On the graphics side, a Greenland GPU is along for the ride, as is 16GB of HBM with 512GB/s bandwidth for just the graphics. A double precision compute rate of 1/2 is also expected, which again points to this being a server chip, especially when you factor in the possibility of ECC, RAS, and true HSA support.
DDR4 support of up to 1TB would be possible, with speeds at 3200MHz. An impressive 64 lanes of PCIe 3.0 would be available, with up to 16 lanes being split between SATA and SATA Express. Both the CPU and GPU would be able to share the cache, HBM, and system RAM, which would help boost performance to high levels.
Now, all of this does sound rather impressive and would be one amazing APU from AMD, but all of this is just rumor. It's based on a slide that may or may not be a total fabrication, and if it is true, odds are we won't see this in its full capacity on the desktop market. Servers could, maybe in 2017 or later, but desktops would probably have a stripped down version to fit in an affordable price range. Either way, whether it is all true or not, take it with a grain of salt and hope we get official news at some point.
Posted: April 12, 2015 02:47PM
Author: Nick Harezga
The first details for the Knights Landing family of processors and co-processors from Intel were released at the end of last month and additional details have now been made available. Knights Landing will have 72 cores with the capacity for four threads each. Up to 36MB of shared L2 cache and 16GB of High Bandwidth Memory will be included on each processor, with support for six DDR4 channels and up to 384GB of RAM. The cores used will be 14nm Silvermont cores that Intel claims will have "up to 3 times the single threaded performance over the last generation of Xeon Phi co-processors and up to 300% gains in power efficiency."
Source: WCCF Tech
Posted: April 3, 2015 12:59PM
Intel recently did a fairly quiet launch of several new processors, with some Braswell SoCs and Broadwell CPUs rolling out. The Braswell SoCs have either one or two CPU modules, with two Airmont cores per module. They also use the 14nm manufacturing process and have some fairly low TDPs that would be a replacement for a Bay Trail-M or Bay Trail-D SoC. The Celeron N3000 and N3050 are both dual-core models with 1MB of L2 cache, with the N3000 featuring a 1.04GHz to 2.08GHz speed and a 4W TDP, while the N3050 has a 1.6GHz to 2.16GHz speed and a 6W TDP. The Celeron N3150 and Pentium N3700 are both quad-core SoCs with 2MB L2 cache and a 6W TDP, with the N3150 running between 1.6 and 2.08GHz, and the N3700 at 1.6 to 2.4GHz.
On the Broadwell side, all three are mobile CPUs with Hyper Threading support, and could likely appear in Ultrabooks or even tablets. The Pentium 3825U, Core i3 5015U, and Core i3 5020U are all dual-core CPUs with 3MB of L3 cache and a 15W TDP. The Pentium runs at 1.9GHz and has Intel HD Graphics (the exact type wasn't listed), the Core i3 5015U at 2.1GHz with an HD 5500 providing the graphics, and the i3 5020U at 2.2GHz with the HD 5500.
All the SoCs and CPUs range in price between $107 and $281, but as these are most likely for trays of 1,000 units, actual retail prices will be a bit higher. Considering these will undoubtedly appear in pre-built devices, hopefully things aren't too outrageous. Exactly when Intel will have these parts ready for retail is not known, but expect word on that before long.
Source: Maximum PC
Posted: April 1, 2015 09:41AM
The AMD news has been coming out fairly often of late, and today is no exception. A new APU is supposedly in the works that will use up to 16 Zen cores and a Greenland GPU to bring the gaming potential. Greenland was mentioned yesterday as the rumored codename of part of the RX 400 series, maybe the high-end line. However, with this new APU, Greenland and maybe the entire Arctic Islands line will be reserved for APUs, as this new APU is said to use HBM, just like Greenland. As for the capability of 16 Zen cores, this doesn't mean it will use all 16, but could. The new APU may be that rumored HPC APU, but it could also be in the running for a server product. Zen is the codename for AMD's APU line launching sometime next year or in 2017, and is said to have PCIe 3.0 and quad-channel DDR4 support.
It all sounds like the makings of a high-powered APU, but remember all of this is just rumor, and this 16 core APU could just be for the server market. The desktop market would probably top out at 8 cores, but you never know what could change in the next year or so.
Posted: March 30, 2015 04:18PM
Author: Nick Harezga
The Knights Landing Xeon Phi co-processor is the latest entry from Intel into the high performance computing market, taking on established products such as the NVIDIA Tesla and IBM Power8. Knights Landing will be available in three products, a co-processor, a standalone processor, and a standalone processor with integrated fabric. All three units will be able to achieve double precision floating point performance in excess of 3 teraflops using a Many Integrated Core design built with up to eight billion transistors. The 14nm chips are from the Silvermont family and will have access to up to 16GB of on-board DDR4 memory with a maximum bandwidth of 400GB/s. The Knights Landing products are expected to be available in the second half of this year.
Source: WCCF Tech
Posted: March 30, 2015 09:32AM
AMD was recently at the PC Cluster Constorium in Osaka, Japan, and revealed its CPU, APU, and GPU roadmap for the next five years. It is a rather nice look at what exactly AMD has in store, and while the pictures aren't of the greatest quality, it does provide a look into what the company has up its sleeves. A pair of CPUs has been planned to arrive early next year, with Zen being an x86 64-bit part, and the K12 a 64-bit ARMv8 part. Both products will be 14nm FinFET CPUs aimed at the server, embedded, semi-custom, and client markets.
The interesting thing about these two products is how they'll support "many threads," which should mean AMD is adding in simultaneous multi-threading (SMT) instead of clustered multi-threading (CMT) like on the Bulldozer line. SMT essentially allows for one fast thread and one slower thread, while CMT allows for two equally fast threads. For single-thread applications, SMT is the way to go, while CMT is best for mutli-threaded apps. With the new CPUs apparently supporting many threads, maybe AMD is using SMT to support more than just that one extra thread, which could put it above Intel.
On the GPU side, it sadly does not cover the discrete cards, but rather how the GPUs will be updated in the APUs. AMD is planning to update its GPU architecture in the APUs every two years, so don't freak out by that because it does mean a new discrete card line will launch every two years; this is strictly for APUs. Arriving in 2017 will be something called a High Performance Computing APU, or HPC as seen on the roadmap. The HPC APUs will evidently have a TDP between 200 and 300 watts, which definitely puts them up there in terms of powerful CPUs. A high-powered APU was not attempted before due to memory limitations, but with the likes of stacked High Bandwidth Memory (HBM), like expected on the company's upcoming graphics cards, an HPC APU is certainly possible.
Roadmaps with greater detail should be coming in May during AMD's Financial Analyst Day event, so keep it tuned for that.
Posted: March 27, 2015 02:47PM
The 20nm manufacturing process is supposed to be the next step, with AMD, NVIDIA, and Intel releasing parts with it. So far only Intel has done anything, as it has used the similar 22nm process on Ivy Bridge and Sandy Bridge, and even the new 14nm process on Broadwell. However, GPUs from both NVIDIA and AMD, and CPUs/APUs from AMD, have only dropped down to 28nm. The mobile processor from NVIDIA, the Tegra X1, is 20nm, however. AMD was supposed to have 20nm parts out by now, and even NVIDIA's Titan X card is still on the 28nm process (hopefully 22/20nm for the Pascal line), but so far nothing. Recently some news has come out from the AMD side, and it looks like we'll have 20nm AMD APUs and SoCs, part of the company's Project Skybridge initiative, in the second half of this year. It may even extend to the AMD video card stack, but first will be the APU/SoC products.
AMD's first 20nm parts are projected to be the Amur line, which will come in both x86 and ARM flavors. The Amur line is aimed at the low power market, like the Intel Bay Trail and upcoming Cherry Trail, but will be the followup to the Nolan line, which AMD has repurposed as the Carrizo-L APUs on the 28nm process. Unfortunately, one wrinkle in the whole thing is that AMD has apparently not even begun using TSMC's 20nm process for the chips. Production usually kicks off months before release anyway, but the earliest the 20nm chips can launch will be anytime from July to the end of this year. AMD does plan for these chips to be in notebooks, tablets, and even Android devices thanks to HSA support and the GCN architecture; just not anytime soon.
Posted: February 25, 2015 04:45PM
Author: Nick Harezga
The next iteration of AMD A-Series Accelerated Processing Units (APU) is codenamed "Carrizo" and targets the low-power system market. The APU is built on the x86 architecture using a 28 nm manufacturing process and will be "optimized for both power and area efficiency." AMD has optimized how voltage is used by the chip resulting in power savings of up to 10% for the GPU and 19% for the CPU. AMD has also included adaptive voltage and frequency scaling (AVFS) which will "enable each individual APU to adapt to its particular silicon characteristics, platform behavior, and operating environment." AVFS technology can contribute to power savings of up to 30%. AMD Corporate Fellow Sam Naffziger described the benefits of Carrizo stating, "As a part of our continued focus on building great products, the advanced power and performance optimizations we have designed into our upcoming Carrizo APU will deliver the largest generational performance-per-watt gain ever for a mainstream AMD APU."
Posted: February 9, 2015 05:28AM
Author: Brentt Moore
According to Taha Khalifa, the general manager for Intel in the Middle East and North Africa region, the company plans to launch its 10nm chips in early 2017. The first batch of 10nm processors, which will be code named Cannonlake and are the "Tick" in Intel’s "Tick Tock" cadence, are expected to offer SoC interconnect improvements such as shared coherent physical memory addresses and a more capable set of heterogeneous functionality. The 10nm chips will initially be integrated into tablet offerings and mobile devices, allowing these products to take on slimmer, lighter, and more elegant designs, all while using energy more efficiently. Mainstream and high performance 10nm chips will launch sometime afterwards, catering to other markets and consumers.
Time will tell if Intel is able to follow its current projection of launching 10nm chips in 2017, as Moore’s law continues to be difficult for semiconductor companies to keep up with.
Posted: December 2, 2014 06:09AM
Author: Brentt Moore
An Intel roadmap that has recently been made available to the public shows that Intel is expecting technology innovation to continue in regards to shrinking its processors. Desktop processors that make use of 14nm technology are expected to be available sometime next year, while 10nm processors are likely coming late next year or early 2016. In order to make 10nm processors a reality, the semiconductor industry will have to go through the process of upgrading to extreme ultraviolet lithography, or EUVL, technology. Intel is also planning on offering 7nm processors sometime in 2017, which is amazing considering that, in 2007, the chip manufacturer was offering 45nm processors to consumers.
Posted: November 21, 2014 07:36AM
Author: Brentt Moore
Intel is teasing its upcoming Haswell-EX processor, which will feature a wealth of various features and technologies aimed directly at server environments. The Intel Xeon E7 v3 microprocessor, codenamed Haswell-EX, is set to offer up to 18 cores with the Hyper-Threading technology, 45MB of last-level cache, PCI Express 3.0 links, a quad-channel DDR4 memory controller, and more. The processor will be made using 22nm tri-gate technology and will feature 5.56 billion transistors, making it one of the most complex x86-based processors ever manufactured. New reliability, availability, and scalability capabilities are expected to be introduced thanks to the new Xeon E7 platform, making Xeon platforms as a whole more comparable with Itanium-based servers in terms of included features.
Intel is set to reveal technical details about its Xeon E7 v3 microprocessor at the 2015 IEEE international Solid-State Circuits Conference, which will take place in February, 2015.
Posted: November 6, 2014 06:20AM
Author: Brentt Moore
Due to a class action lawsuit against Intel and HP, which claimed that Intel manipulated certain benchmarks to perform better on Pentium 4 processors than AMD Athlon chips, Intel has agreed to give $15 to customers who purchased a personal computer powered by a Pentium 4 CPU. The compensation only applies to United States residents who purchased a personal or general computer between November 20, 2000 and June 30, 2002 that contained an Intel Pentium 4 processer. Individuals who did in fact purchase a Pentium 4 powered computer during this timeframe are required to provide the retailer's details for the computer and date of purchase; a purchase invoice is not required.
It is worth noting that Intel and HP have denied the allegations contained with the class action lawsuit that claim that WebMark2001 and SysMark 2001 were modified to make Pentium 4 CPUs perform faster than competing AMD processors.
More details surrounding the settlement, as well as resources to file a claim, can be found on the Intel Pentium 4 Settlement website.
Posted: October 24, 2014 08:40AM
Author: Brentt Moore
AMD has significantly reduced the cost for the majority of its APU lineup, making the CPU and GPU combination more attractive than ever for value orientated consumers. Newegg previously sold the A6-7400K for $85, the A8-6600K for $99, the A8-7600 for $110, the A10-6800K for $140, the A10-7700K for $159, and the A10-7800 for $165, but is now selling the AMD APUs for $58, $92, $92, $112, $123, and $133, respectively. The top of the line Kaveri APU from AMD, the A10-7850K, is now retailing for $143, a notable decrease from its previous $189 price tag.
Along with reduced pricing, AMD is offering APU customers a coupon code for one of three games for the rest of the month. Consumers who purchase an APU between now and the end of October will qualify Murdered: Soul Suspect, Thief, or Sniper Elite 3, with Thief and Sniper Elite 3 supporting the Mantle API from AMD.
Posted: September 2, 2014 07:12AM
Author: Brentt Moore
AMD has just introduced three new eight-core processors into its FX lineup, the FX-8370, FX-8370E, and FX 8320E. While performance is certainly a marketing component of these new processors, as they are geared directly towards content creation and gaming, their price points are what make them attractive for potential buyers. The 4GHz FX-8370 and the 3.3GHz FX-8370E are priced at $199.99, while the 3.2GHz FX 8320E is priced at $146.99. AMD is hoping that the value of its three new FX-series processors wins over consumers who are looking at low-level and mid-level offerings by Intel. Unfortunately, the three new FX processors are still built upon the Piledriver microarchitecture that AMD launched in 2012 and are also manufactured with a 32nm process.