Robert Hallock y Frank Azor, ingenieros de AMD, debaten más en detalle sobre los nuevos Ryzen 7000

Durante el último episodio de la serie de webcasts ‘The Full Nerd’ de PCWorld , los invitados Robert Hallock (director de marketing técnico de AMD) y Frank Azor (arquitecto jefe de soluciones de juegos) respondieron a una serie de preguntas y detallaron más sobre las CPU de escritorio Ryzen 7000 y sus características.

Para empezar, confirmaron las ya conocidas especificaciones de los nuevos Ryzen 7000

Algunas de las cosas que se confirmaron en la entrevista incluyen:

  • Ryzen 7000 is 125W TDP / 170W Package Power
  • Ryzen 7000 5.5 GHz Demo Was In Stock-Spec (No Overclock)
  • Ryzen 7000 Double L2 Cache is IPC-Benefit
  • Ryzen 7000 CPUs Have 28 PCIe Gen 5 lanes (24 Usable)
  • 1:1 Infinity Fabric Clock (No Frequencies Mentioned)
  • B650 Motherboards will support overclocking (like B550 series)
  • Integrated RDNA 2 GPU supports both Video encode/decode
  • Integrated RDNA 2 GPU For Commercial/Diagnostic Purposes
  • Smart Access Storage Details (Requirements Highlighted)

En primer lugar, las cifras de TDP (Potencia de disipación térmica) para las CPU AMD Ryzen 7000 serán de 125 W, mientras que el PPT (Seguimiento de energía del paquete) será de 170 W. AMD proporcionó esta información en respuesta a si la cifra de 170 W era un TDP real para los próximos chips o un límite superior para la alineación completa.

Según AMD, este es un aumento de alrededor de 28 W sobre el límite de potencia del paquete AM4 (PPT), que era de 142 W, mientras que las CPU tenían un TDP de 105 W. Según AMD, los fabricantes de placas base ahora podrán implementar características de potencia más premium en sus placas base, lo que debería permitir mejores oportunidades de overclocking.

So what we want to clarify is that it's a 170 Watt socket power which with AMD, that spec is PPT (Package Power) for us. That doesn't mean that every CPU is going to go up to 170 Watts but it's 30 (Watt) higher than the socket AM4 power cap which was a 142 (watts). And we did this to mainly improve multi-thread performance as many of the core count chips were actually held back in overall compute performance by relatively modest socket power.

The other point that I want to make is that by raising the minimum required socket power or minimum spec, you also raise the power delivery with every motherboard built to that spec so you get more robust power characteristics on all the boards which we are pretty excited about as well, It should be good for people who want to experiment with overclocking, people who appreciate premium board designs.

Robert Hallock (AMD Director of Technical Marketing)

Además, Hallock confirmó que las CPU AMD Ryzen 7000 cuentan con un total de 28 carriles PCIe Gen 5, lo que confirma todas las filtraciones anteriores. La cuestión es que de esos 28 carriles, solo 24 son utilizables por la GPU dedicada y los dispositivos de almacenamiento M.2, ya que el resto de los cuatro carriles están conectados a la PCH de la serie 600. En cuanto a los conjuntos de chips, Robert Hallock también confirmó que las placas base B650 serán totalmente compatibles con el overclocking como sus predecesoras de clase B*50.

En cuanto a la demostración de juegos con velocidad de reloj de 5,5 GHz, Robert aseguró que las frecuencias estaban completamente dentro de las especificaciones de stock. La placa base utilizada fue un diseño de referencia X670 y el disipador fue un AIO estándar ASETEK de 280 mm. También es obvio que no hubo overclocking ya que los relojes variaban entre 5,1 y 5,5 GHz.

AMD mostró algunas frecuencias increíblemente rápidas con la misma muestra de CPU Ryzen 7000 alcanzando hasta 5,52 GHz, pero se pudieron ver algunas variaciones en las velocidades de reloj que comenzaron en 5,1 GHz y subieron a la velocidad máxima de 5,52 GHz de la que todos hablan. Curiosamente, Robert afirma que en la demostración del juego, vieron que la mayoría de los subprocesos registraban hasta 5,5 GHz (32 subprocesos). El prototipo Ryzen 7000 de 16 núcleos se produjo a finales de abril o principios de mayo, por lo que AMD aún podría obtener más espacio para la mejora de este chip si así lo desea o simplemente dejar que los overclockers hagan el trabajo.

We used a 280mm (ASETEK) watercooler so nothing exotic it's just a dual 140 loop that you can buy from Amazon or Newegg. It was running an AMD reference motherboard so one of our internal AM5 socket motherboards, a 16 core prototype part built in late April or early May and we just plugged it in and ran it. It was not an overclocked part, its just the natural frequency of that particular prototype.

So in the game, we were running most of the threads around 5.5 (GHz), it depends on the game load, depends on the scene, of course, clock speed fluctuates up and down so somewhere between 5.2 and 5.5 is pretty common on all the threads playing that game. So I wanna reassure people that this was nothing exotic in terms of cooling and nothing exotic in terms of parts selection or config or overclocking. It is exactly what you see, you plug in a Ryzen 7000 series part and play the game, & that's the frequency!

Robert Hallock (AMD Director of Technical Marketing)

Hallock también habló sobre cómo la duplicación de la memoria caché L2 puede ayudar a mejorar el rendimiento del núcleo relacionado con IPC. El caché L2 no se incrementó dramáticamente ya que más caché L2 da como resultado más aciertos, pero también significa una latencia deficiente y tamaños de troqueles cada vez más grandes, lo que aumenta el coste. Así que AMD encontró un terreno en el que la memoria caché L2 aumenta lo suficiente como para aumentar los beneficios de rendimiento de las CPU de escritorio Ryzen 7000.

So L2 cache, we did double it in Zen 4, it is an IPC benefit, it helps the core contain internally more information, overall it improves throughput and so it will be beneficial to multiple categories of workloads, be it integers like gaming or more floating-point like stuff like scientific and creative workloads.

Robert Hallock (AMD Director of Technical Marketing)

Cuando se le preguntó si la nueva plataforma AM5 ejecutará la memoria DDR5 en una proporción de 1: 1 con Infinity Fabric, Robert respondió «Sí», pero no mencionó a qué frecuencias se ejecutaría. Ese es otro tema de discusión cuando nos acerquemos al lanzamiento oficial de las CPU de escritorio Ryzen 7000 en las placas base AM5.

En cuanto a lo que aportan las nuevas iGPU RDNA 2 en las CPU de escritorio Ryzen 7000, AMD afirma que con gráficos integrados en toda su cartera de CPU, pueden expandir su negocio al segmento comercial y eso tiene mucho sentido ya que muchos consumidores no requieren una tarjeta gráfica dedicada y quieren algo que sea tan simple como plug-and-play.

Para los DIY, la iGPU RDNA 2 puede proporcionar capacidades de diagnóstico y resolución de problemas donde los usuarios con tarjetas gráficas pueden depurar si su tarjeta gráfica está defectuosa o no, o para otros fines.

Frank Azor llevó el tema a un espacio más interesante al afirmar que, si bien las CPU AMD Ryzen 7000 tendrán núcleos de «gráficos pequeños» en comparación con los núcleos de «gráficos más grandes» en las APU, seguirán albergando algunas de las tecnologías Smart Eco que se jactan de los portátiles. Si bien las iGPU RDNA 2 de AMD pueden permitir un uso de energía inferior a 50 W en modo inactivo, las CPU Smart Shift on Desktop con iGPU RDNA 2 pueden cambiar de gráficos dedciados a gráficos integrados para cargas de trabajo livianas y ofrecer una potencia inferior a 5 W o incluso mW.

La otra cosa es que, a diferencia de las GPU Navi 24 de AMD, la iGPU Ryzen 7000 basada en la misma arquitectura de núcleo RDNA 2 (pero en una revisión Rembrandt de 6 nm) vendrá con un motor VCN que admite tanto la codificación como la decodificación de video AV1.

We still think of the Ryzen 7000 series as a CPU. The graphics cores in that IO die are not many, the purpose of adding graphics is three-fold. One, it greatly expands these products in the commercial market where they don't buy discrete at all, they just want to turn it on, have video encode/decode and light up some displays for office work and that's what the GPU in the IO die will offer so that's a huge business opportunity for us on the Ryzen PRO side as we start migrating these components over to that business.

The second is for diagnostic purposes, how do you know that you have a bad graphics card? Well, you have to swap in another graphics card but with the graphics core we have, you can do a little bit of troubleshooting thirdly, we were thinking about users who are planning to buy a discrete graphics, and it's still in transit in the mail but all the other hardware has arrived first so it's all sat there, looking at a pile of components and don't have a GPU to actually set that all up. That would go away with the Ryzen 7000 series.

We are still going to do APUs with big graphics so APUs 'BIG GRAPHICS', CPUs 'little graphics'. That would be our strategy going forward.

Robert Hallock (AMD Director of Technical Marketing)

We are developing a lot of technologies that make use of integrated graphics in many ways and there are things that we are able to do with technologies such as Smart Shift ECO where we can turn off the discrete graphics and we can run the notebook off of the iGPU and say you want that because you want less heat, longer battery life (even when you are playing a game) or you want less fan noise or lower power consumption, there's all these benefits to it. Because we have that thin integrated graphics in Ryzen 7000 series, it's going to allow us to bring more of these types of smart technologies over to the desktops aswell so those customers can get some of these benefits.

Frank Azor (Chief Architect of Gaming Solutions)

En cuanto a si veremos una variante iGPU deshabilitada de las CPU de escritorio Ryzen 7000, Robert afirmó que todos los chips Zen 4 tendrán gráficos RDNA 2 integrados, por lo que aquellos que esperan ver una variante KF (como en el caso de Intel) deberían sentirse un poco decepcionados.

Pasando al lado del almacenamiento directo, AMD introdujo su propia tecnología conocida como Smart Access Storage SAS. Los usuarios lo han estado confundiendo como un reemplazo de Direct Storage cuando, de hecho, SAS se basa en los mismos algoritmos y API que Direct Storage y permite que el SSD NVMe se comunique directamente con los gráficos dedicados. Frank proporciona una explicación detallada de lo que es SAS a continuación:

I have seen Smart Access Storage in action and it works pretty well! The primary difference between Direct Storage and Smart Access Storage is that, think of, first of all, Smart Access Storage uses Direct Storage so that's really important to know. It's not a replacement for it is something that we built that takes advantage of Direct Storage, it uses the compression algorithms and APIs that Direct Storage promotes, supports, and endorses and asks ISVs to design so it's fully compatible with them.

What we do a little differently with Smart Access Storage is that because we own an entire platform from an architecture perspective, end to end, from CPU to all the way to storage subsystem to all the way to graphics, that allows us to create more efficient streamlines between the game content and GPU itself. We can avoid some parts of the PC subsystem in order to reduce latency and increase performance to reduce bottlenecks when it comes to decompressing game assets and streaming game assets on the GPU.

Smart Access Storage is more of a ubiquitous solution that has to work across a lot of different combinations of hardware that are out there and those combinations may include mixed configurations, let's say AMD's CPU and somebody else's graphics card. and that's getting more and more complex, you are gonna have three graphics card partners, you have two CPU partners, that may even get more complex, so because it's more of a hiding compatible ubiquitous solution, there's more overhead built in order for it to be compatible so there are more CPU calls than more system memory that it has to do in order for it to work so there's more decompression and management that is done in the CPU versus in our solution, it's a direct pipe from NVMe to the GPU with those bottlenecks and overheads removed.

We have designed Smart Access Storage to be as open as possible from a compatibility perspective so that's why we are using Direct Storage we are going to work with Direct Storage titles and we are going to optimize to work with Direct Storage. We don't have any intention of developing our own algorithm or API since we don't want to create a walled garden or another walled garden in the industry, so ideally what we would like to have to happen and this is how we designed Smart Access Storage is that you have a Direct Storage compatible game and you have an all AMD desktop or laptop, we'll have a foundational infrastructure in place so that so we can avoid those routes that add latency and some of those bottlenecks because of compatibility and we kinda short cut and allow the NVMe to direct send data over to the GPU for decompression. So that's our architecture and our approach is fully compatible but when we can take a shortcut because we have the end-to-end platform, we are gonna enable that shortcut to be taken.

Frank Azor (Chief Architect of Gaming Solutions)

Frank Azor también dice que las CPU AMD Ryzen 7000 y la plataforma AM5 contarán con soporte SAS en todos los ámbitos, pero habrá una lista de SSD y componentes que cumplan específicamente con sus requisitos SAS. No todos los SSD Gen 4 y Gen 5 serán compatibles con Smart Access Storage:

We have an authorized vendor list of parts that we are developing right now with NVME drives and controllers that are meeting our performance requirements around Smart Access Storage. Just because you have a PCIe Gen 4 or the upcoming Gen 5 drive, it doesn't mean that those drives are going to be able to keep up with the performance demands and capabilities of Smart Access Storage so something folks should understand is that today a lot of the reason why your NVMe drive isn't able to fully capitalize on its theoretical performance capability is that there are bottlenecks in the data fabric and those bottlenecks exist in different parts of the PC so that often affects your theoretical versus real performance.

What will happen with Smart Access Storage is that once you remove all those paths from point A to point Z that are your obstacles (bottlenecks), suddenly, your NVMe drive will become the bottleneck because the dGPU can receive the data extremely fast but if your NVMe drive was designed from the get-go counting on these bottlenecks along the data path, then you knew you didn't put much performance on it because you knew it was going to get wasted. You may have used a slower controller or not used the most premium parts or not optimized your product for the highest theoretical limit that it can reach.

Frank Azor (Chief Architect of Gaming Solutions)

En cuanto al lanzamiento, se dice que las CPU de escritorio AMD Ryzen 7000 se lanzarán este otoño, lo que significa que lo más pronto que veremos los chips en acción es septiembre de 2022.

Fuente Wccftech PcWorld – Full Nerd

Publicaciones relacionadas

Jonsbo V8. Análisis

admin

Los resultados del i9-13900K muestran buen rendimiento multiproceso y ligeras mejorías en juegos

admin

Nuevas noticias sobre el lanzamiento de las RTX 4000

admin

Deja un comentario