fast ethernet?
... single-core AMDs? What is this, the 90s?
Internet-grooming company Cloudflare has revealed that it was unable to put Intel inside its new home-brew servers, because they just used too much energy. A Tuesday post by platform operations engineer Chris Howells reveals that Cloudflare has been working on designs for an eleventh-generation server since mid-2020. "We …
This post has been deleted by its author
The "Intel Inside" circle logo, but with a big red circle-and-slash "NO" through it. Intel doesn't just need to step up the pace, they need to start running to catch up before the competition leaves them in the dust of irrelevance. "But our 10nm+++ is as good as or better than their 7nm stuff!" is just the whinings of a child that's unable to get with the program. It _might_ be nearly as good, but your 10nm is _not_ like their 7nm kit. While Intel is crowing about finally reaching a stable 7nm manufacturing capability, the competition is breaking ground on *three* nm plants. Intel's fix was not to get their manufacturing house in order, but to rebrand/remarket their stuff to sound the same as their competition. That's not an improvement, it's the magic smoke coming out your ears...
There is hope. The most recent rumours about Intel seem to suggest that they might actually have some interesting silicon inbound. I've always bought AMD and I'm not a big fan of the tactics of Intel in the past but we certainly do need stiff competition from Intel to keep the pressure on AMD to stop them becoming the next "Intel".
What a time to be alive though. I just wish we could get over the supply problems and return to sane prices though for both CPUs and Video Cards.
I've never been interested in the pissing contest between Intel and AMD over whose processor is the best, I was just interested in there being competition to force Intel to innovate and reduce prices - Intel are acknowledged experts at milking the purchasers of its products. (Also not too great at 'fessing-up to flaws and compensating!)
I don't want Intel to fail as AMD need the competition for the same reasons that keep Intel on their toes.
I honestly believe that if Intel does not get its act together soon, this will not end well for the rest of us.
I've also noted that few of the companies I deal with now would buy IBM because they want to...
Some edge servers serve as proxies for the x86_64 servers so need to be able handle the relevant code well. But the article also points out that they are also actively evaluating ARM chips. They may have simple decided that chips which provide the best current performance / power tradeoff are the Epycs and they are currently not at the scale large enough to warrant their own designs.
AMD thought ARM was the next big thing for servers too at one point
https://www.theregister.com/2016/01/14/amd_arm_seattle_launch/
Cloudflare and ARM
https://www.theregister.com/2017/11/08/qualcomm_centriq_2400/
seems every time ARM gets close to servers it's opportunity is killed. I guess one exception area is for those that are really building their own systems and designing their own chips(and such systems aren't available to purchase by anyone else). Don't really have much hope for ARM on servers in the general market.
It seemed to me as hard as it was for x86 to scale down in power(for mobile) it was as hard for ARM to scale up in performance/scale(for servers), to the point where the costs/power didn't really make a big selling point for ARM at that point. I'm guessing Qualcomm saw the margins Intel was getting and was hoping they could do similar with their server chips but pulled the plug when they realized they would not get those margins(most likely due to competition, and the cut throat nature that the big hyperscalers play by)
Companies tried to work around that to some degree for a while with micro server designs but those never really went anywhere(AMD bought one such vendor Seamicro which looked like a really neat system at the time, and HP still has their Moonshot system though it's been 4-5+ years since I heard them talk about it).
I like ur analysys - u sound like u have some inside knowledge.
I am inexpert, but I too am suspicious of the alleged threat of ARM to X86.
Yes it exists, but not to the extent, or the way in which it is portrayed.
I liked ur analogy of x86 vs mobile vs arm vs x86.
Given ARM's history, some points?
Yes cloud is a huge new force, but its not a complete compute solution (wan interconnects will ~always be relatively puny), & cloud companies with the massive apps to justify custom chips wont be vendors of them.
AFAICT, AMD's Fabric could integrate ARM processors for the best of both worlds
The sheer range of AMDs simplified core unit lego blocks - the same unit serves from massive servers down to laptops and embedded devices - spreading the massive development costs very thinly and at a power penalty which is almost imperceptible to the end user where it counts.
AMD'ss model allows it to cross subsidise and see off competitive threats with low pricing.
I doubt even Toyota have the volumes to justify serious custom ARM chips.
It seemed to me as hard as it was for x86 to scale down in power(for mobile) it was as hard for ARM to scale up in performance/scale(for servers),
Once the legacy x86 code is gone the difference is indeed small, though ARM now has more experience in asymmetric designs with big.Little now standard. But what ARM really does offer in addition is the ability for custom hardware: hardwiring certain encryption or codecs can make a huge difference overall.
The big problem Intel have being stuck on 10nm+++++++++ and 7nm is that the power consumption per flop is considerably higher that on the smaller architecture.
TSMC have perfected 5nm, and are sampling on 3 and working on 2 and 1.4
RISC-V FTW - why payARM for licencing if you're going to build your own.
the hyper-scalers will be moving to RISC just as the internet did to Linux