strafejump
Power Member
Não percebo como a intel vai usar um processo de fabrico, que se tem mostrado um fracasso nos cpus, em gpus que ocupam uma área maior que se traduz em 'yields' ainda piores.
- We have seen random kernel patches previously that cite Intel multi-GPU systems where there may be Intel integrated graphics but also the possibility of discrete Intel graphics, which obviously isn't possible right now but will soon be the case with the "Xe Graphics" adapters expected to launch in 2020. The latest on the Intel multi-GPU front is a perf PMU (Processor Monitoring Unit) bit of code for handling the Intel iGPU+discrete use-case. Nothing too exciting by itself, but just another example showing Intel getting ready for their multi-GPU future capabilities.
https://videocardz.com/newz/intels-first-xe-graphics-processor-is-called-ponte-vecchioIntel Ponte Vecchio is not a gaming GPU. The first Xe graphics are for exascale computing. On November 17th Intel will share details on the project “Aurora”. This exascale computer features Sapphire Rapids Xeon CPUs, Ponte Vecchio GPUs and Intel’s new initiative called oneAPI (unified programming model).
In the new press deck (that we have totally not seen) Intel claims that Ponte Vecchio will use Foveros packing technology and will utilize CLX interconnection (this was already known). What’s new is that they confirmed that the Xe graphics feature: ultra-high cache and high memory bandwidth. Intel Ponte Vecchio will also have high double-precision FP throughput.
06:56PM EST - Rambo cache, high footprint, very large
06:56PM EST - unified cache, accessible to CPUs and GPUs
06:56PM EST - and other GPUs connected to this GPU
Isto é muito curioso atendendo a que o responsável de gpu's da Intel estava na AMD, e que AMD vai agora fazer o caminho inverso: tinha arquitectura unificada para gpgpu e para gpus gaming/de consumidor (pelo menos os de performance mais elevada, Vega) e que agora a AMD vai deliberadamente ter arquitecturas diferentes para gpu's de consumidor/gaming e gpu's gpgpu.XeHPC, XeHP e XeLP. A mesma arquitectura, desde o GPU integrado, até HPC.
IC: Turning to gaming solutions, because there is a lot of interest in how Intel is going to attack the gaming space: what we’ve seen today is a compute GPU based on chiplets. Moving from a monolithic graphics chip to a chiplet design is a tough paradigm to solve, so does working on chiplets help solve the ‘multi-GPU’ issue on graphics? Is the future of graphics still consigned to single GPU, or should we expect multiple GPU scaling easier to manage?
RK: That’s a great question. As you know, solving the multi-GPU problem is tough – it has been part of my pursuits for almost 15 years. I’m excited, especially now, because multiple things are happening. As you know, the software aspect of multi-GPU was the biggest problem, and getting compatibility across applications was tough. So things like chiplets, and the amount of bandwidth now going on between GPUs, and other things makes it a more exciting task for the industry to take a second attempt. I think due to these continual advances, as well as new paradigms, we are getting closer to solving this problem. Chiplets and advancement of interconnect will be a great boost on the hardware side. The other big problem is software architecture. With many interesting cloud-based GPU efforts, I am optimistic that we will solve the software problems as well.
XeHPC, XeHP e XeLP. A mesma arquitectura, desde o GPU integrado, até HPC.
A escalabilidade baseada em chiplets não irá passar por SLI, CF ou múltiplas placas gráficas, mas por GPUs "grandes" compostos por múltiplos chiplets. Isso permite poupanças enormes em termos de fabrico e que beneficiam toda a gama. Tal como disse, o melhor exemplo disso é o Zen 2: onde um único chiplet é usado na gama toda, do CPU mais barato ao mais caro.facilmente compra a 3x gráficas se isso lhes garantir a antiga promessa que o SLI e CF nunca foi.