Intel Unveils 7nm Ponte Vecchio GPU Architecture For Supercomputers and AI (hothardware.com) 28
MojoKid writes: Intel has unveiled its first discrete GPU solution that will hit the market in 2020, code name Ponte Vecchio. Based on 7nm silicon manufacturing and stack chiplet design with Intel's Foveros tech, Ponte Vecchio will target HPC markets for supercomputers and AI training in the datacenter. According to HotHardware, Ponte Vecchio will employ a combination of both its Foveros 3D packaging and EMIB (Embedded Multi-die Interconnect Bridge) technologies, along with High Bandwidth Memory (HBM) and Compute Express Link (CXL), which will operate over the newly ratified PCIe 5.0 interface and serve as Ponte Vecchio's high-speed switch fabric connecting all GPU resources. Intel is billing Ponte Vecchio as its first exascale GPU, proving its meddle in the U.S. Department of Energy's (DOE) Aurora supercomputer. Aurora will employ a topology of six Ponte Vecchio GPUs and two Intel Xeon Scalable processors based on Intel's next generation Sapphire Rapids architecture, along with Optane DC Persistent Memory on a single blade. The new supercomputer is schedule to arrive sometime in 2021.
Does anyone trust Intel at this point? (Score:2, Troll)
Do you really want their engineering “skill” inside your GPU as well?
Re: (Score:2)
Tough to side-load a GPU. OTOH, stacking them with 7mm wafers might requires syrup to be correctly digested.
Proving its "meddle"? (Score:2)
Damn. Mojokid: where did you go to skool?
Re: (Score:2)
Proving that this chip will mess with you.
Re: (Score:2)
Apparently someone went to a school where they don't meddle with dictionaries. But you can't blame MojoKid, that mess was in TFA.
Re: (Score:2)
I'm pretty sure Mojokid IS Hothardware.com. He always submits this hothardware stuff. I am assuming this is more of the same (didnt look)
Re: (Score:2)
Probably the skool of hard cocks. He's a legacy, after all.
Re: (Score:1)
That's called a Freudian slip.
Slashvertisement? (Score:2)
Re: (Score:2)
How is this a Slashvertisement? What next, discussion on kernel releases with new features are Slashvertisements for linux?
No thanks. I don't want to live in your Slashdot where the only thing left to do is discuss petty politics because a summary better not dare mention a product or vendor by name.
Re: (Score:2)
Promote an specific product is not an AD?
No Blockchain mentioned? (Score:2)
If we are already past that hype, then I need to string-replace Blockchain with AI...
Re: No Blockchain mentioned? (Score:1)
Intel should release a Pinocchio CPU (Score:2)
Re: (Score:2)
Re: Intel should release a Pinocchio CPU (Score:1)
Re: (Score:1)
There are lots of low end laptops at Walmart with AMD processors in them.
Re: Intel should release a Pinocchio CPU (Score:2)
They should release a 7nm CPU period.
A bloody fast one?
Re: (Score:2)
A lot of the new supercomputers are AMD too.
Not the first discrete GPU (Score:2)
When will people learn that Intel did have discrete GPUs [wikipedia.org] in the past?
The article gets it right though by saying that it is the first discrete GPU for supercomputers.
Not 2020 (Score:2)
The article says 2021. And even that timeframe is optimistic. Still cool. But get facts right please.
How's it's TLA generation & processing? (Score:2)
Intel "unveils" a PowerPoint slide deck (Score:3)
Intel "unveils" a PowerPoint slide deck. Holy shit NVIDIA is totally in trouble now. The walls are closing in! What I want to know is how the hell did they get a supercomputer contract for this vaporware that doesn't yet exist?
Re: (Score:1)
I'm guessing the usual way: bribes and blackmail. They certainly have the expertise.
Re: (Score:2)
Reputation. That's one of the key Intel selling points, and one of the reasons why it could dominate sales even back during the early K7 days, when they literally had nothing on AMD in CPU market. Later they had to go with the monopoly abuse route.
Like it or not, "Intel" logo matter to many people.
Ponte Vecchio, Great name choice (Score:2)
A charming architecture regularly in danger of being destroyed by a flood.
Mettle. (Score:2)
meddle - (v) To interfere in or with; to concern oneself with unduly.
mettle - (n) A quality of endurance and courage.
Software Support is Key (Score:1)
Much like AMD's, Intel won't be making inroads in the Deep Learning field as long as TensorFlow, PyTorch and other libraries only really support CUDA and cuDNN.
Sorry AMD, but maintaining a separate fork of TF is not my idea of compatibility. Hopefully Intel's effort is more serious, because NVIDIA could use some competition.