AMD on the road to recovery.

[humbug]

Instead of bumping years old doom and gloom AMD financials threads over and over again i thought a new thread was needed, a thread that was more relevant to AMD's current situation which for the past year or so has been a story of not only recovery but now also starting to move into a position of strength.

With that said since Ryzen AMD's revenue it up about 70% from just under $1bn a quarter to about $1.7bn a quarter, their R&D spend is up 25% and over this year for the first time in years AMD are in Net Profit, to the tune of about $200m so far this year.

It is fair to say that as far as Intel are concerned they have lived in a bubble with their only rival (AMD) of no consequence what-so-ever.
As a result we have had a serious lack of competition, Intel have spent the last 5 years or so re-hashing what are basically the same 4 core CPU's over and over again, with one exception of adding a DDR4 memory controller.
In my mind Intel would not have launched 6 core CPU's when they did in the form of the 8700K, if at all, certainly i do not believe the upcoming 8 core Coffeelake was on their radar at all and is as a direct result of Ryzen.

Intel's bubble has burst

This feeling is now also starting to creep in amongst experts, i'm not saying its significant beyond the fact that for the first time in probably more than a decade Intel's stock value has been down graded for reasons Citing a better outlook for AMD, there are many other agencies and this feeling is not yet wide spread but it does mark an end to Intel's infallibility perception.

BofAML Downgrades Intel from $62 to $56, Intel shares down 10% https://seekingalpha.com/news/3374975-bofaml-downgrades-intel-another-10nm-chip-delay#email_link

AMD's shares on the rise now approaching $20

https://seekingalpha.com/symbol/AMD

 

[humbug]

*Reserved*

 

[smilingcrow]

Thank God the decade of pain has ended and not with bankruptcy.
Now to sort out their GPU woes and then sky is the limit for the immediate future.

 

[gavinh87]

Saw title, looked a author. Smiled

 

[Gibbo]

The Ryzen CPU's are selling incredibly well, but so they should be, it actually feels like were giving them away with 8/16 CPU's from £169.99, seems too cheap but if AMD can do it and still make money, then great, just hope they are making money whilst been so aggressive.

 

[Glanza]

Good news, competition is good for consumers especially when they can start turning a profit. Considering all the issues happening with Intel currently it doesn't surprise me that AMD are on the up.

 

[humbug]

Good news, competition is good for consumers especially when they can start turning a profit.

Exactly

 

[4K8KW10]

How far can they go, though? Can they become the single, largest CPU and GPU manufacturer, overshadowing Intel and nvidia?
For now, it seems Intel is an easier target.

 

[drunkenmaster]

Intel just straight up said they expect first 10nm chips (when production restarts, the initial run that happened doesn't really count) to now ship in Holiday 2019 season, meaning very end of 2019 for client.

IE they are going to release dual core stuff again end of 2019. That puts server and probably higher power desktop way into 2020. For big server chips it's going to be late 2020 or even 2021.

AMD can make a ~700mm^2 effective die using 4x 200mm^2 dies (some of it is duplicated and some in interconnect that a monolithic die wouldn't require so you 'waste' some transistors to achieve it), that can double core count as soon as they can make 200mm^2 dies with even poor yields, Intel won't be able to double the core count of their 700mm^2 14nm chips till they can make something ~700mm^2 on 10nm and currently they can't make a 70mm^2 core fully working.

This is also a date Intel are giving, but they've given a half dozen dates for 10nm already. If they are saying end 2019 that is the earliest possible time but could be pushed back again. Funny as when they announced another 10nm delay they said they identified the problem and it would be a quick fix but in reality they've just stated their announcement (which was what, Feb/March) is a 1.5 year delay.

AMD is going to rake it in next year, server parts will be utterly unmatched in core count, in performance per watt (the most important metric) and quite possibly straight up higher IPC. With the ability to use TSMC and Glofo for 7nm they've increased their effective capacity massively and given themselves a few months headstart on 7nm by being able to use TSMC as well.

2019 is going to be utterly insane for AMD. A full process node advantage over Intel across pretty much every segment for the best part of a year with an updated architecture when the current architecture has made significant headway while using an inferior node...

 

[smilingcrow]

IE they are going to release dual core stuff again end of 2019. That puts server and probably higher power desktop way into 2020.

You might as well speculate wildly as that also appears to be what Intel are doing.

 

[drunkenmaster]

hehe, not too much speculation though. intel's usual thing is to go dual core/mobile first, often even that Core M thing that is even smaller and lower power than their dual core laptop chips.

So say Core M, then 4-6 months later a dual core low power laptop stuff, then 4-6 months later desktop, then 6 months later smaller server chips not too much bigger.

Looking at 14nm you have Sept 2014 for Core M stuff, dual core Broadwell for mobile was Jan 2015, 4 core was June 2015 then Skylake was August 2015.

Broadwell server is a fairly bad sign also, the first 14nm server stuff ranged from ~ 250-450mm^2 die sizes and only came out June 2016, so a full year to go from yielding desktop parts to server sized stuff and that is small compared to the next lot of server chips. It in fact took another 6 months before they released a fully working 24 core part, before then it was salvaged parts only. It was mid 2017 before they hit those 28 core parts that were ~700mm^2 in die size and in doing so the top end 22 core parts were $7000 and moved to $13000 for the top end 28 core stuff, largely because the yields are still poor.

Intel took close to to 2 years to go from 50mm^2 mobile parts to not fully working 436mm^2 parts and another year to get insanely low yield high cost 700mm^2 parts. The problem in being so long in 14nm meant instead of moving to 10nm and making a ~400mm^2 die with twice the cores as their Broadwell server parts which makes sense, they went HUGE on 14nm. But that means a ~400mm^2 10nm part will have broadly speaking a similar transistor count to their current big 14nm parts. IE not an awful lot of performance improvement.

IF <150mm^2 desktop stuff is coming out for next xmas then on the same schedule as 14nm the first ~400mm^2 server parts wouldn't launch till xmas 2020 and a huge part with maybe double the core count wouldn't be coming till xmas 2021.

Where as AMD will launch seemingly 64 core EPYC chips by fairly early next year.

 

[mmj_uk]

tbh in light of the success of Ryzen and recent security fixes hurting perfermance on Intel, I'm surprised Intel are still charging as much as they are for their processors. It's as though they don't really care about the consumer market if people don't pay way over the odds and are happy to just make their money selling to industry with their Xeon's.

 

[smilingcrow]

hehe, not too much speculation though. intel's usual thing is to go dual core/mobile first...

Your choice of the word Usual is what makes your whole spiel speculation. It varies what they release first so time will tell.
It's not that I have much confidence with Intel's 10nm process but due to the time scale involved it's too early to even speculate not that I'd bet on them.

 

[smilingcrow]

tbh in light of the success of Ryzen and recent security fixes hurting perfermance on Intel, I'm surprised Intel are still charging as much as they are for their processors. It's as though they don't really care about the consumer market if people don't pay way over the odds and are happy to just make their money selling to industry with their Xeon's.

Most sales to consumers are via OEM complete systems and also Intel has a very strong brand name so you won’t see many AMD systems in PC World or on Dell.com etc.
The large OEMs don’t pay retail prices or even tray pricing per 1K so retail pricing has little impact on Intel’s overall profit.
Even at retail chip pricing if you look at the psychology of consumers Intel are the premium brand so can afford to charge more for the same performance like Nvidia.
On top of that, only a small minority are paying above ~£200 for a mainstream CPU partly because they have no use for such performance so it’s a rarefied market.
I think Intel know what they are doing pricing wise as their profits clearly show.
But when Zen 2 is available in a Ryzen chip next year things might well get interesting and then Intel might need to adjust pricing noticeably especially if 10nm is still MIA.
In other words, life is more complicated than it appears on the surface as Brexit is making very clear.

 

[Drollic]

How come you don't post much praise for AMD in the graphics forum but really, this pro AMD anti Intel stuff gets old quick.

 

[kitfit1]

Your choice of the word Usual is what makes your whole spiel speculation. It varies what they release first so time will tell.
It's not that I have much confidence with Intel's 10nm process but due to the time scale involved it's too early to even speculate not that I'd bet on them.

Yep, The word USUAL is the big one there.

Let me give my take. Intel have gained server and consumer space by deliberately gimping the security of ther own cpu's to gain an advantage over AMD. This has now come come home to roost on them. I bet Intel will not be able to produce another competitive secure cpu for the next 3 to 4 years. Simply because they have to start from the ground up, because they have been lying to the world for years. They can lie to the world as much they like, the server and data farm world know's different.......................we as clockers should know different as well.

 

[smilingcrow]

Intel have gained server and consumer space by deliberately gimping the security of ther own cpu's to gain an advantage over AMD.

I’m going to ignore the conspiracy theory stuff as it doesn’t interest me but if you look at the consumer space and the performance hit of the security updates, Intel’s CPUs would surely still annihilate AMD’s CPUs for the 10 years prior to Ryzen even with the patches in place?
On a pragmatic level, as those exploits weren’t known so not exploited the net result is that we had faster consumer CPUs with no downside during that period so all good.
The situation since the exploits were revealed is a whole other ball game though.

.. the server and data farm world know's different…

Yeah, that is the real issue for Intel and with Zen 2 due next year and Intel’s 10nm delayed until who really knows when that is a massive opportunity for AMD even without Intel's security issues.

 

[Panos]

How come you don't post much praise for AMD in the graphics forum but really, this pro AMD anti Intel stuff gets old quick.

https://forums.overclockers.co.uk/threads/dark-days-amd-share-price-at-lowest-ever.18682064/

 

[drunkenmaster]

Your choice of the word Usual is what makes your whole spiel speculation. It varies what they release first so time will tell.
It's not that I have much confidence with Intel's 10nm process but due to the time scale involved it's too early to even speculate not that I'd bet on them.

Yeah, it's usual because that is how yields work on processes, they start off low on an immature process, in the very early days before mass production such that you could get literally no working chips off a wafer if you went with a 400mm^2 die but you might get 100 dies if they are 70mm^2.

The entire industry makes smaller chips first and brings in bigger chips as process maturity leads to less defects and the ability to make larger dies in a financially viable manor.

Rather than attack a perfectly normal usage of the word usual, it would be sensible to suggest a reason why Intel will all of a sudden, despite horrendous issues with 10nm to date, will suddenly throw out every single thing known about the foundry business and be able to make 700mm^2 dies at the same point in time 70mm^2 dies become viable. It would literally change everything known about how chips are currently made, change decades of how chips are usually produced and fly in the face of all logic but why not at least suggest a reason why smaller chips won't come first before having a pop at how I'm wildly speculating.

I mean... which chip did Intel actually already make on 10nm? Which chip did they already try to get into full production but find yields so bad they pushed the node back another 1.5 years... right, a 70mm^2 dual core. But I'm wildly speculating that Intel's plan is to make smaller chips first.........

 

[smilingcrow]

Rather than attack a perfectly normal usage of the word usual, it would be sensible to suggest a reason why Intel will all of a sudden, despite horrendous issues with 10nm to date, will suddenly throw out every single thing known about the foundry business and be able to make 700mm^2 dies at the same point in time 70mm^2 dies become viable.

Intel have already clearly stated that they are releasing 10nm server chips well beyond the time scale they have stated for their first 10nm chips so your whole 70 v 700 mm^2 argument is completely bogus as nobody has suggested they are doing that. ^10.