dresdenboy.blogspot.de escribió: AMD Zeppelin CPU codename confirmed by patch and perhaps 32 cores per socket for Zen based MPUs, too
The Zeppelin codename, first mentioned on a leaked slide shown by Fudzilla, has been identified as a "family 17h model 00h" CPU by a patch on LKML.org. The interesting parts of the patch are:
AMD Zeppelin (Family 17h, Model 00h) introduces an instructionsretired performance counter which indicated byCPUID.8000_0008H:EBX[1]. And dedicated Instructions Retired register(MSR 0xC000_000E9) increments on once for every instruction retired.
There might even be a meaning behind the similarity of parts of the "Zen" and "Zeppelin" codenames.
An older patch on the same mailing list also gives a little more info about Zen:
On AMD Fam17h systems, the last level cache is not resident in Northbridge. Therefore, we cannot assign cpu_llc_id to same value as Node ID (as we have been doing currently)
We should rather look at the ApicID bits of the core to provide us the last level cache ID info. Doing that here.
The most interesting part describes the way, how the last level cache (LLC) ID is being calculated for Zen based MPUs:
+ core_complex_id = (apicid & ((1 << c->x86_coreid_bits) - 1)) >> 3;
+ per_cpu(cpu_llc_id, cpu) = (socket_id << 3) | core_complex_id;
"Core complex" should be similar to "compute unit" and has been used in some AMD patents already. The expression marked in red means a shift right by 3, which equals a division by 8. So with two logical cores per physical core due to SMT, a core complex should contain four Zen cores and a shared LLC.
The next line shows the socket ID being shifted left by 3, leaving 3 bits for the core complex ID, which suggests a maximum number of eight core complexes per socket, or 32 physical cores. This number should first be seen as a placeholder, but we've already seen rumours mentioning that many cores.
Dfx escribió:Buf 8 modulos y 32 cores, imagino que sera para una version de servidor rebajada de frecuencias por que el consumo se puede ir por las nubes.
Lo de los 8 canales de memoria...., pasan de 2 directamente a 8 ?
TRASTARO escribió:Dfx escribió:Buf 8 modulos y 32 cores, imagino que sera para una version de servidor rebajada de frecuencias por que el consumo se puede ir por las nubes.
Lo de los 8 canales de memoria...., pasan de 2 directamente a 8 ?
Aqui dejamos el uso de modulos [1 modulo= 2 nucleos] pasando otra vez al concepto de nucleos independientes tradicional. En definitiva seria para servidores, si es que llegaran a contruir un procesador monolitico [un solo chip] de tantos nucleos. Pero para el mercado casero si podriamos pensar en uno de 10 o quiza 12 nucleos en alguin momento, pero con uno de 8 ya tendrias suficiente para una PC de render CGI.
Los 8 canales aun sera ver si es solo para servidores dealto desempeño o si llegara almenos 4 canales al mercado casero de gama alta, aunque lo mas probable para el gran mercado es que siga siendo de 2 canales.
AMD Details Bristol Ridge AM4 Performance
[..] the first entries under the Bristol Ridge umbrella. These “7th generation APUs” will feature an updated version of the Excavator architecture used within the Carrizo architecture with significantly higher single thread x86 performance and a DDR4 memory controller. According to AMD there will also be an optimized GPU core and more design changes to further refine performance per watt but what those entail isn't being discussed right now.
WCCTech escribió:Talking about Bristol Ridge, while the mobility platform has already been introduced, it’s time to talk about the desktop parts. The flagship A12-9800 processor comes with a core clock of 3.80 GHz. The chip features a quad core design along with 8 GCN 3.0 compute units, offering 512 stream processors. The processor is expected to feature a TDP around 65W (Configurable down to 45W). We can note in the following pictures that OEMs are readying desktop PCs with Bristol Ridge processors but the absence of AM4 boards from Computex may have indicated that Bristol Ridge could be an OEM only platform while Summit Ridge will bring AM4 to the masses.
Khalid Moammer, wccftech.com escribió:Rumor : AMD Zen & Bristol Ridge AM4 Motherboards Coming In October
AMD’s upcoming AM4 motherboards for Zen & Bristol Ridge are said to be debuting in October according to chatter by motherboard partners. Rumor has it that AMD’s motherboard partners will be showing off their next generation AM4 socket based motherboards in a couple of months.
Bristol Ridge APUs for the desktop are set to launch around the same time. AMD’s hotly anticipated high performance Zen FX processors will be available shortly afterwards. According to President & CEO Lisa Su, desktop Zen FX processors will see wide availability in Q1 2017 and could see limited availability in late 2016 depending on ramp up and yield.
One thing that AMD has hinted at during its recent performance demonstration of Zen which wowed the press, is that we’ll see Zen on display at CES in January of 2017. That’s when desktop Zen FX CPUs are also expected to launch.
techpowerup, btarunr escribió:
AMD today unveiled its 7th generation A-series desktop APUs. Unlike its predecessors, the new chips are full-fledged SoCs, built in the new socket AM4 package, on which the company plans to launch its "Zen" processors. The 7th gen A-series APUs are based on the "Bristol Ridge" silicon, and are the first fully-integrated SoCs (systems-on-chip) from the company, in that the APU completely integrates the functionality of a motherboard chipset, including its FCH or southbridge.
![[oki]](/images/smilies/net_thumbsup.gif "Ok!")
cada vez me gustan mas las APU AMD ZENSandungas escribió:..
Ahora que lo dices... creo que lo habia leido por ahi en algun lado... por eso en las conferencias que da AMD hablando de ZEN dicen que es una arquitectura "escalable" desde dispositivos portatiles (tablets, laptops, etc... de muy bajo consumo) hasta servidores de red (monstruos de alto consumo)
Lo que han hecho ha sido intentar estandarizar el socket AM4 y los chipsets para que todos ellos usen lo mismo
En la serie AMD ZEN "raven ridge" para portatiles era de esperar que estuvieran todos los chipsets integrados... creo que lei por ahi que han conseguido bajar el consumo hasta 15 o 10 watios (aunque esto hay gente que aun no se lo cree, es demasiado bajo)
Esos portatiles van a estar wapos... por eso el otro dia comente en otro mensaje en este subforo que a dia de hoy yo no recomendaria a nadie comprar un portatil nuevo ahora... es mejor esperar a esos portatiles nuevos con AMD ZEN a ver que rendimiento tienen (la bateria va a durar muchismo, no se van a calentar y no haran ruido de ventiladores)
Lo del southbridge externo para bristol bridge y servidores es como una "ampliacion" opcional... por poner un ejemplo... lo que va a pasar es que en las placas base que no lleven southbridge solo va a haber 4 conectores SATA para hdd's... pero las que lo lleven tendran muchos mas conectores y se podra hacer RAID de hdd's o cosas asi que se pueden considerar "extras" que no todo el mundo necesita. Con los conectores USB va a pasar lo mismo y con todos los otros perifericos que normalmente van conectados al southbridge en un pc "normal"
Mi interes en las APU AMD (a parte de porque lo usan los fabricantes de consolas) es porque me quiero montar una especia de "steam machine" o de "consola hecha en casa"... asi que lo que mas me interesa es el bajo consumo, poco ruido y que tenga el menor numero de componentes posible
Asi que intentare buscar una placa base mini-itx que no tenga southbridge... habra que esperar a ver que modelos de placas base sacan los fabricantes (es muy probable que asrock y gigabyte seran los primeros)
----------------------------------------------------------------------
Pongo un par de noticias relaccionadas:
http://www.guru3d.com/news-story/intel- ... -unix.html
Solo van a soportar sistema operativo windows 10 o superior.... o basados en unix (todos los linux, etc...)
A la gente no le va a hacer gracia, pero mi me parece bien porque no me gusta el windows10... soy de los que opinan que cada version nueva de windows es peor que la anterior, sigo usando windows pero en cualquier momento me pasare a debian y que le den a bill gates y su "coleccion de basura" a la que ellos llaman sistema operativo
http://techreport.com/review/30540/amd- ... ent-of-zen
Esto es una conferencia de hace unas semanas (un poco antigua) pero hay un video muy interesante donde se ve la potencia de ejecucion multihilo renderizando en un programa 3D
Ese test es muy importante porque usa drivers especificos para AMD ZEN y porque el procesador de intel con el que se esta comparando cuesta mas de 1000 euros !!! y quedan empatados
Lo de los drivers es muy importante, actualmente los programas que hacen tests de rendimiento grafico no pueden mostrar la potencia real de AMD ZEN... hay que esperar un tiempo a que el software se actualice, es necesario que aparezcan los drivers vulkan y ese tipo de cosas se hagan de uso comun en juegos y programas graficos
Anthony Garreffa, tweaktown.com escribió:AMD's high-end X370 chipset teased, arrives in Feb 2017 AMD's next-gen Zen CPUs will use the high-end X370 chipset, will arrive in February 2017
AMD will be launching its next-gen Zen architecture in February 2017, alongside a new high-end X370 series chipset. The news is coming from Benchlife, teasing the Bristol Ridge series of APUs based on the Excavator CPU architecture, and GCN 3.0-based graphics architecture. amds-high-end-x370-chipset-teased-arrives-feb-2017_02
The upcoming AM4 socket will be split into three different chipsets, with the high-end X370 leading the way, and behind it we have the B350 and A320 chipsets. The Zen FX-based CPUs will be part of the upcoming Summit Ridge family of processors, including PCIe 3.0 support, dual channel DDR4 memory controllers, lots of L3 cache, updated storage features including USB 3.1 and NVMe, and more. We can also expect to see CPUs arrive with between 4/8 cores with 8/16 threads, and consuming between 65-95W of power.
Motherboard makers are already reportedly building their inventories of X370-based motherboards, with companies preparing to unveil some of their AM4 motherboards in October. AMD will most likely show off their Zen CPUs and new X370 chipset during CES 2017 in January.
Wayne Manion, techreport.com escribió:[..]
The HP Willow microATX motherboard is built around AMD's Promontory Fusion Controller Hub, bringing with it support for PCIe 3.0 and DDR4 memory. HP's motherboard support page details support for Bristol Ridge A8-9600, A10-9700, and A12-9800 APUs, all of which are quad-core 28-nm chips with 65W TDP. Some future Summit Ridge CPUs have TDPs up to 95W, though this board appears to be limited to 65W units.
The board offers a single PCIe x16 slot and and A-keyed M.2 slot, which seems to imply support for PCIe x2 SSDs. DDR4 memory as fast as 2133 MT/s is supported in configurations up to 16GB spread across two slots. The Willow motherboard is suspiciously lacking USB 3.1 ports, Type-C or otherwise. Whether this is related to mid-summer rumors that AMD and Asmedia were dealing with signal integrity issues with USB 3.1-related circuitry remains unknown.
The specific PC available at Costco includes a 1TB magnetic hard disk, 16GB of DDR4 memory at an unspecified clock speed, a top-of-the-line A12-9800 APU, an 802.11ac Wi-Fi adapter, and a DVD recorder for just under $600.
New AMD x86 Zen Blender Benchmarks Surface – Show Equal Performance To A 10-Core Intel Xeon E5 Chip.
A benchmark of AMD’s upcoming x86 Zen architecture has been spotted on Blenchmark.com (a relatively unknown aggregate database for Blender tests). The test was spotted by this redditor and shows some very impressive performance if we take the result at face value. Unfortunately, the processor has no identifying information along with it so it is impossible to determine which variant of AMD’s Zen processor the results represent.
New blender benchmark of an AMD Zen CPU spotted in the wild – possible Naples variant
AMD’s philosophy with the upcoming Zen architecture is one of complete uniformity. The company will be shipping variants that scale (initially) from 8 cores for Summit Ridge and up to 32 cores for the Naples processor. While the result is undoubtedly legitimate in nature, the lack of identifying information makes it very hard to judge the performance. To put it simply, it is very impressive performance for an 8 core but only decent performance for a 16 core and above.
The processor is listed as “AMD Engineering Sample” and is several positions above the Core i7 6900k – the octa-core processor the Summit Ridge variant was originally tested against. It manages the Blender run in 69 seconds – which is exactly the same performance as an Intel Xeon E5 2680 v2 processor. This is a chip that has 10 cores! It is marginally faster than the Xeon E5 2650 v2 – which is an 8 core processor. Both of these feature relatively low base clocks (as is usually the case with Xeon chips) of 2.8 Ghz and 2.6 Ghz respectively.
The test date is approximately the same time that the official demo for Zen was shown but since we have evidence of Naples being ready for testing quite a while before that, this does not prove anything. If this is indeed the 8 core chip, then its performance exceeds all expectations and blows away the competition. The only thing that will remain will be the price point and as you can expect a significant premium over AMD’s usual value philosophy.
On the other, if this represent a Naples variant, it fits more or less within the performance we were expecting. Since Intel has had a significant process advantage over AMD for quite some time now I do not expect Zen to beat Intel core for core, clock for clock. However, we do believe that the Sunnyvale manufacturer will give Intel very real competition in terms of performance per dollar – which is all that gamers are usually interested in.
It is my belief however (caution: opinion) that this particular result represents a server class chip, i.e. Naples. This is because the benchmark was conducted in Windows 2008 Server R2 64 bit edition – an odd choice for a desktop variant. One would expect the Summit Ridge CPU to be tested on the latest OS: Windows 10. Since we do not know the clocks at this point and engineering samples are usually clocked low, this does go on to show that AMD can make a very big comeback in the server sector (if the price point is right). I am not going to go into any further speculation regarding the results since our readers can take care of that themselves and will end on the following note:
TRASTARO escribió:Supuestas pruebas bajo Blender para ver el rendimiento de los nucleos Zen en el render. Esto bajo una unidad de pueba de ingenieria ZEN.
http://wccftech.com/new-amd-zen-blender-benchmarksNew AMD x86 Zen Blender Benchmarks Surface – Show Equal Performance To A 10-Core Intel Xeon E5 Chip.
A benchmark of AMD’s upcoming x86 Zen architecture has been spotted on Blenchmark.com (a relatively unknown aggregate database for Blender tests). The test was spotted by this redditor and shows some very impressive performance if we take the result at face value. Unfortunately, the processor has no identifying information along with it so it is impossible to determine which variant of AMD’s Zen processor the results represent.
New blender benchmark of an AMD Zen CPU spotted in the wild – possible Naples variant
AMD’s philosophy with the upcoming Zen architecture is one of complete uniformity. The company will be shipping variants that scale (initially) from 8 cores for Summit Ridge and up to 32 cores for the Naples processor. While the result is undoubtedly legitimate in nature, the lack of identifying information makes it very hard to judge the performance. To put it simply, it is very impressive performance for an 8 core but only decent performance for a 16 core and above.
The processor is listed as “AMD Engineering Sample” and is several positions above the Core i7 6900k – the octa-core processor the Summit Ridge variant was originally tested against. It manages the Blender run in 69 seconds – which is exactly the same performance as an Intel Xeon E5 2680 v2 processor. This is a chip that has 10 cores! It is marginally faster than the Xeon E5 2650 v2 – which is an 8 core processor. Both of these feature relatively low base clocks (as is usually the case with Xeon chips) of 2.8 Ghz and 2.6 Ghz respectively.
The test date is approximately the same time that the official demo for Zen was shown but since we have evidence of Naples being ready for testing quite a while before that, this does not prove anything. If this is indeed the 8 core chip, then its performance exceeds all expectations and blows away the competition. The only thing that will remain will be the price point and as you can expect a significant premium over AMD’s usual value philosophy.
On the other, if this represent a Naples variant, it fits more or less within the performance we were expecting. Since Intel has had a significant process advantage over AMD for quite some time now I do not expect Zen to beat Intel core for core, clock for clock. However, we do believe that the Sunnyvale manufacturer will give Intel very real competition in terms of performance per dollar – which is all that gamers are usually interested in.
It is my belief however (caution: opinion) that this particular result represents a server class chip, i.e. Naples. This is because the benchmark was conducted in Windows 2008 Server R2 64 bit edition – an odd choice for a desktop variant. One would expect the Summit Ridge CPU to be tested on the latest OS: Windows 10. Since we do not know the clocks at this point and engineering samples are usually clocked low, this does go on to show that AMD can make a very big comeback in the server sector (if the price point is right). I am not going to go into any further speculation regarding the results since our readers can take care of that themselves and will end on the following note:
btarunr, techpowerup.com escribió:AMD Socket AM4 "Bristol Ridge" APU De-lidded
Here are some of the first pictures of an AMD socket AM4 APU being de-lidded. De-lidding is the process of removing the IHS (integrated heatspreader), the metal plate covering the CPU die. Some PC enthusiasts remove the IHS to improve heat-transfer between the CPU and extreme cooling solutions, such as LN2/dry-ice evaporators. Overclocker Nam Dae Won, with access to a couple of socket AM4 chips (most likely 7th generation A-series "Bristol Ridge" APUs), de-lidded the chips, revealing a large rectangular die. AMD is using high-quality TIM between the die and the IHS, which could either be solder or liquid metal. There's also a clear picture of the underside pin-grid of the AM4 chip, which has a central cutout that lacks any SMT components. Socket AM4 has 1,331 pins.
-page-002_678x452.jpg)
