Vya DomusThat wasn't what I was talking about , notice I mentioned compute not ARM. I am talking about datacenters and more precisely , datacenters used in fields such as AI and Big Data (or even the automotive industry, though the situation there is a little bit different) where Nvidia has managed to garner an impressive amount of market share in just the last 2-3 years. These are multi-billion dollar industries that keep on growing by the year. Intel is at heart of the server market and here you have another company digging away what could have been their sales. They are trying to remain somewhat relevant with products such as Xeon Phi or by marketing their iGPUs as being capable of compute and heterogeneous computing. But their offerings still can't come even close to what Nvidia has.

Instead of having a customer buy 1000 Intel CPU nodes they now buy just 100 because they can spend the rest of the money on Nvidia GPUs and have a computer cluster that is many times more powerful and efficient and cheaper.

That hurts Intel badly , trust me. They aren't happy about that synergy at all. This is why Nvidia wont licence any of their GPU IPs , because it is the only thing that manages to keep them ahead by miles in these particular sectors.

Okay, I missed that you were taking about compute specifically, in which case I mostly agree with you. However I think you're overlooking how desperate Intel is - not having a decently-performing iGPU would IMO hurt them far more in the long run (consumer perception) than swallowing their pride and inking a deal with NVIDIA. Plus at this point, there isn't much Intel is able to do against the fact that NVIDIA is gobbling up the AI market.

Like I said, from NVIDIA's side, any licensing deal would come with very strict controls from them as to what is done with their GPU IP, precisely because of the lead in AI they currently have. If Intel was foolish enough to steal NVIDIA's IP, and NVIDIA found out, that would be the mother of all lawsuits that would not end well for Intel. And if NVIDIA were to loan the weakest design (GT 1030) or older-gen designs like Maxwell or even Kepler, there is much less chance of Intel being able to scale those up if they did decide to "borrow" concepts from them.

Just came across info about low end Ryzen chips. It's some older roadmap, but the chip I'm waiting for is called "Basilisk" and it's a Dual Core ZEN based APU with Vega based GPU (probably severely cut down version). ONly problem is, the roadmap states that for 2016 and we're near end of 2017. So, I'm not sure if that stands true anymore. On paper it sound sweet though. Ryzen IPC is super, so even 2 cores at up to 2GHz should deliver incredible performance on the budget. I mean, my current E-450 still "sufficient". I tried converting 4K phone recording to 1080p and it took forever. But it works fine for movies, music and browsing. Was looking at E2-9000e, while it's quite an improvement, it's not significant enough to be worthy of purchase of new device. But this "Basilisk" APU, based on Zen, it should provide quite significant boost. I hope it'll come soon.

RejZoRJust came across info about low end Ryzen chips. It's some older roadmap, but the chip I'm waiting for is called "Basilisk" and it's a Dual Core ZEN based APU with Vega based GPU (probably severely cut down version). ONly problem is, the roadmap states that for 2016 and we're near end of 2017. So, I'm not sure if that stands true anymore. On paper it sound sweet though. Ryzen IPC is super, so even 2 cores at up to 2GHz should deliver incredible performance on the budget. I mean, my current E-450 still "sufficient". I tried converting 4K phone recording to 1080p and it took forever. But it works fine for movies, music and browsing. Was looking at E2-9000e, while it's quite an improvement, it's not significant enough to be worthy of purchase of new device. But this "Basilisk" APU, based on Zen, it should provide quite significant boost. I hope it'll come soon.

The last news from AMD is that Raven Ridge is slated for a H2 2017 launch. Since we're already in Q4, I'm assuming we'll see a limited high-end mobile launch before Christmas, but probably low availability until Feb-March, and lower end SKUs rolling in as time goes by and production increases. Don't think AMD has much fab capacity to spare for the moment.

Since those roadmaps were for 2016 I'm assuming Basilisk will arive sometime in 2018. Was thinking of going Stoney Ridge when I spotted the info and decided E-450 (Zacate) will have to do till then.