I sure hope AMD exploits every second of this situation. We needed proper competition in CPU space for a decade now.
Intel on Thursday reported $19.2bn in sales for its third quarter 2019 earnings, a quarterly record for the company. The chip maker's financial results [PDF] were significantly higher than the average analyst prediction of $18.05bn. Shortly after the results were announced, Intel stock was up almost 5 percent in after-hours …
Sadly, AMD remains largely a sideshow, almost just an alibi, in data centers where most CPUs are not sold. The only real competition for Intel remains ARM, which has taken much longer to come up with a plausible server CPU strategy than many of us hoped.
In the consumer space ARM chips have for a while provided sufficient oomph for devices but a lack of OS choice on devices with larger screens has hindered the market from development. The I-Pad Pro shows what could be possible but Google's decision to only allow ChromeOS on notebooks and can any Android for notebook development is the real blocker. Maybe they'll change their minds in Windows on ARM becomes properly usable.
>Sadly, AMD remains largely a sideshow, almost just an alibi, in data centers where most CPUs are not sold.
Funny that, the UK Supercomputer Archer is about to be replaced by Archer II using 11,696 AMD Epyc Rome CPUs which happen to be rather high margin high value processors rather than pile 'em high and sell 'em cheap ARMs.
12,000 CPUs keep the fabs busy for how long? OK, AMD no longer has its own fabs, but it still needs volume to cover the development costs.
I'm not dissing AMD but if you look at the numbers their earnings are tiny in comparison with Intel's which is why competition from other sources is welcome.
"12,000 CPUs keep the fabs busy for how long?"
Likely around 3 months from start to finish...
But that would be the case whether you run 12000 CPU's (likely 200 wafers @500 chiplets/wafer after defects and a few more 14nm wafers for the IO dies) or full capacity at 1500 wafers/day/fab (estimated TSMC 7FF production)
"Sadly, AMD remains largely a sideshow"
Be patient. It takes around 6 months after a server CPU launches before you get solid numbers on volume and then potentially 6 more months before you can easily judge market share.
The figures we see at the moment are largely AMD's Naples servers that are competitive with Intel but that doesn't make a great case for change in the server world. the Rome EPYC's released in August alter that significantly, particularly in light of minimal development on Intels side.
If a cloud provider (or 2 or 3 or 4) decide to use them, that is potentially 1-2 million CPU's a quarter which would equate to 10%-25% market share.
Yes, that is a very optimistic view of AMD's potential, but Intel finds itself in a very unusual position where it may have the single-threaded performance lead, but the majority of the market is after price/performance and Intel need to cut prices in half to even begin to compete, and may not have the capacity to do it.
This post has been deleted by its author
Biting the hand that feeds IT © 1998–2021