3 ms·
I wonder how are they going to compete with Intel? I don’t see how they’re going to match raw performance against those 1tflops+ 20-cores xeons. I don’t see ho
by Const-me 11y ago
I wonder how are they going to compete with Intel?
I don’t see how they’re going to match raw performance against those 1tflops+ 20-cores xeons.
I don’t see how they’re going to match performance/watt against 7w quad-core C2338.
I don’t see how they’re going to match price against $20 x5-Z8300.
If none of the above, then why should Google/HP/Oracle/anyone else buy that?
- mtgx 11y agoMaybe they think the more Intel conquers the markets it's in, the less competitive it will be performance/price-wise - like say how it replaced its $110 Core-based Celeron chips for Atom chips with half the performance and almost no extra "features" that also cost $110, because it knows people will keep buying "Intel" in the PC market and are not even giving AMD a second look now. It surely helped when Microsoft killed Windows RT and eliminated the ARM competition from the market for the foreseeable future. Maybe these RISC-V chips will never get out of the labs of Google, but they could still serve to force Intel to keep prices the same on its new generations in the future, unless they want Google to really get serious about making its own chips.
- Const-me 11y ago“how it replaced its $110 Core-based Celeron chips for Atom chips” — they already reverted that decision. Latest Celerons 3855U, 3955U are Skylake i.e. 6th generation Core chips.
- jeffbush 11y agoI don't think they need to compete with Intel. The numbers I've seen suggests Intel ships around 300-400 Million processors per year[1]. In 2014, 12 Billon (with a B) devices with ARM cores shipped[2]. In 2011, around 500 Million MIPS based cores shipped[3]. These are obviously very different businesses. But it's clear there is an enormous market for licensable low-end IP cores that are used in everything from cars, cable boxes, TVs, cellphones (many phones have multiple cores internally to operate functions like the cellular baseband), etc. Most of the licensees are just gluing together licensed cores and contracting foundries like TSMC or Global Foundries to build them. The margins tend to be pretty thin (we're talking chips that are in the single digits for cost), so saving a little money on IP licensing is attractive. Since these are often custom applications, binary compatibility is less of an issue. That seems like a more interesting market for an open ISA and open core processors. [1] http://www.ft.com/cms/s/0/beff6e56-53ed-11e4-80db-00144feab7de.html#axzz3wFGbKjgu http://www.ft.com/cms/s/0/beff6e56-53ed-11e4-80db-00144feab7... [2] https://atlas.qz.com/charts/Ek18VmbP https://atlas.qz.com/charts/Ek18VmbP [3] http://www.forbes.com/forbes/2011/0509/technology-mips-sandeep-vij-dvd-micro-chips-less-more.html http://www.forbes.com/forbes/2011/0509/technology-mips-sande...
- Const-me 11y agoThanks, that sounds reasonable.
- nickpsecurity 11y agoCounter subversion concerns of various black boxes for SaaS operations. Have customizable cores to add HW accelerators to like Cavium's Octeon III's do. Things might get cheaper over time as big vendors buy in volume and that money goes back into enhancements.
- Const-me 11y agoVendors won’t be buying from RISC-V, they’ll be buying from someone else who’ll use that open source design. By not paying for R&D and enhancements, that OEMs will be able to offer lower prices.
- nickpsecurity 11y agoYou might be replying to the wrong comment by mistake as I listed a number of reasons people might like an open ISA and reference implementation. That OEM's will repackage it for those purposes doesn't contradict what I said.
- igmor 11y agothey will be competing with Apple's ARM product