Goldman Sachs Upgrades Stock Ratings for AMD, Downgrades Intel to "Sell"

[Raevenlord]

Goldman Sachs, citing problems with Intel's 10 nm manufacturing process delivery - which was supposed to be available in the market for years now - has reduced the blue giant's stock rating. Previously at a "neutral" stance - already downgraded in the face of Intel's manufacturing woes - the stock is now at a "Sell" level. Even though Intel is still outperforming the S&P 500's 6.7 percent return with their (current) 8.6 percent this year through Thursday, the outlook isn't good for the company.

"We see Intel's struggles with 10 nm process technology having ramifications in terms of its competitive position - across a broad set of products. While the 10nm push is well-publicized at this point, we believe Intel's manufacturing issues could potentially be deeper-rooted than what most think and could have a sustained impact on market share and/or spending levels as Intel competes with a growing/stronger TSMC eco-system."

Goldman Sachs Analyst Toshiya Hari

As to AMD, Goldman Sachs has a much more positive outlook, moving the company's shares from the "sell" rating it was in to a "neutral" stance. Citing belief in increased server and consumer market penetration with AMD's products in the next two years: AMD's server chip market share will rise to 5.1 percent in 2019 and increase to 9.4 percent in 2020 from an estimated 2.2 percent this year.

"We find it increasingly harder to argue our prior bear thesis - even following the recent stock price move - given Intel's struggles with 10nm process technology. The delay in Intel's new products will allow AMD to gain share in not only client (i.e. desktop PC, notebook PC) CPUs, but also in the lucrative server CPU market."

Goldman Sachs Analyst Toshiya Hari

AMD's stock was - and likely still is - terribly undervalued for the company's IP and expertise - the underdog terminology has served to put the company's outlook in a much more negative view than it needed to be - this still is the only second company that's capable of designing x86 chips, and let's not forget how it's been delivering on that front in no shy means in the last two years. Owing to that same undervaluation, though, AMD's stock has risen 85.8 percent this year through Thursday versus the S&P 500's 6.7 percent return. The chip stock's performance ranks as the second-best return in the S&P 500 this year.

View at TechPowerUp Main Site

 

[FordGT90Concept]

Oh how the tides have turned.

 

[Xzibit]

The real issue for Intel is what does it do after they sort their 10nm. Their last years roadmap were stil showing moving to 7nm then 5nm. Are those going to experience the same delays.

TSMC will be producing 5nm by 2020 with a new fab. Their also looking at 3nm by 2023.

Intel has to keep pace or start transitioning to a fabless model

 

[Aquinus]

This is what happens when you become complacent.

The real issue for Intel is what does it do after they sort their 10nm. They are moving to 7nm then 5nm. Are those going to experience the same delays.

TSMC will be producing 5nm by 2020 with a new fab. Their also looking at 3nm by 2023.

Intel has to keep pace or start transitioning to a fabless model

No way. The issue with intel is that they need to rely on large monolithic dies. AMD going with the MCM approach has some big advantages when it comes to quality control and yields. It's easier to build and control characteristics of a smaller chip than a bigger one. It also spreads the heat out over a larger area. The cost is latency when it comes to communication to segments of memory and cache that aren't local to the chip but, with the right tweaks to the CPU scheduler, a lot of that can be mitigated and situations where more cores are needed tends to suffer less from latency.

Simply put, AMD found a sweet spot and exploited it and it's going to take Intel time shift towards a different (MCM,) design because you know it's going to happen. When push comes to shove, what's important are yields and AMD has plowed ahead on that front. I honestly think MCM is the future and that's the wind behind AMD's sails right now.

 

[Fluffmeister]

AMD have played a blinder, or more specifically... Intel have loved playing doupoly with an underdog.

The silver lining for Intel is tech communties love underdogs, so when AMD becomes the big bad wolf they no doubt will get all the love and fight back against the evil Dr Sue.... sorry Lisa Lu.

 

[crazyeyesreaper]

The real issue for Intel is what does it do after they sort their 10nm. Their last years roadmap were stil showing moving to 7nm then 5nm. Are those going to experience the same delays.

TSMC will be producing 5nm by 2020 with a new fab. Their also looking at 3nm by 2023.

Intel has to keep pace or start transitioning to a fabless model

There not complacent what people forget is how foundries label the new nodes is kinda bs.

Intel's 10nm is likely similar to TSMC etc at 7nm. Meanwhile Intels fabrication will likely still be slightly better since its tailored to their own designs.

Finfet and other design aspects has kinda made the 14 / 10 / 7 nm node aspect moot since it is kind of a lie at this point. Its smaller yes but not in the way the average user would think.

That said Intel 10nm will likely = TSMC etc at 7nm. Both will likely be available on a similar time table. Still it does mean Intel is no longer ahead and therefore has actual competition again.

 

[R0H1T]

The real issue for Intel is what does it do after they sort their 10nm. Their last years roadmap were stil showing moving to 7nm then 5nm. Are those going to experience the same delays.

TSMC will be producing 5nm by 2020 with a new fab. Their also looking at 3nm by 2023.

Intel has to keep pace or start transitioning to a fabless model

According to the latest rumors 10 nm will be less dense than initially planned, there's also the bigger possibility that it'll perform worse than initially thought.

 

[Arjai]

 

[StrayKAT]

There not complacent what people forget is how foundries label the new nodes is kinda bs.

Intel's 10nm is likely similar to TSMC etc at 7nm. Meanwhile Intels fabrication will likely still be slightly better since its tailored to their own designs.

Finfet and other design aspects has kinda made the 14 / 10 / 7 nm node aspect moot since it is kind of a lie at this point. Its smaller yes but not in the way the average user would think.

That said Intel 10nm will likely = TSMC etc at 7nm. Both will likely be available on a similar time table. Still it does mean Intel is no longer ahead and therefore has actual competition again.

Exactly. Sad that people get their knowledge on cpu engineering from bankers. Heh. But I predicted this would happen last week in another thread. These kind of subtle points never did well on Wall Street or headlines.

 

[Xzibit]

According to the latest rumors 10 nm will be less dense than initially planned, there's also the bigger possibility that it'll perform worse than initially thought.

Yes, The rumors were they will come out with an easier to manufacture 10nm and those initial uplifts they reported aren't going to be meet and the people who were working on the next node 7nm had to be pulled to help out on the 14++nm improvements. If all those "sources" turn out to be true which over the past years have been right. Once Intel 10nm gets ironed out. Will they be stuck optimizing 10nm like 14nm (++). How long will it take to get 10nm to the initial state they wanted. Will their 7nm take longer to transition to then planned ?

Exactly. Sad that people get their knowledge on cpu engineering from bankers. Heh. But I predicted this would happen last week in another thread. These kind of subtle points never did well on Wall Street or headlines.

Didnt know David Kanter was a banker.

 

[RejZoR]

The real issue for Intel is what does it do after they sort their 10nm. Their last years roadmap were stil showing moving to 7nm then 5nm. Are those going to experience the same delays.

TSMC will be producing 5nm by 2020 with a new fab. Their also looking at 3nm by 2023.

Intel has to keep pace or start transitioning to a fabless model

Well, Intel can still manufacture CPU's at TSMC if they wish so even with their own fabs. The problem that creates is perception of their competence if their own fabs can't produce the latest and greatest they design chips for, one may start questioning if their latest and greatest node design isn't all that good. Again, that would just be a perception to an outsider, not necessarily an inferior product or node design.

 

[HTC]

Well, Intel can still manufacture CPU's at TSMC if they wish so even with their own fabs. The problem that creates is perception of their competence if their own fabs can't produce the latest and greatest they design chips for, one may start questioning if their latest and greatest node design isn't all that good. Again, that would just be a perception to an outsider, not necessarily an inferior product or node design.

Something that would likely be punished, from a stock market point of view.

Not to mention the elephant in the room: the amount of chips suddenly required from TSMC / GLOFLOW would potentially put quite a strain on capacity from both foundries.

 

[StrayKAT]

Something that would likely be punished, from a stock market point of view.

Not to mention the elephant in the room: the amount of chips suddenly required from TSMC / GLOFLOW would potentially put quite a strain on capacity from both foundries.

Don't forget the inevitable viruses dragging a whole new component market down.

 

[FordGT90Concept]

Intel could still pull several rabbits out of their hat. Probably not within the next year but Intel could surprise the market just like AMD did. Examples:
1) Intel is working on discreet GPUs but we don't know to what end.
2) Intel knew Moore's Law was falling apart all of the way back in 2014. I don't believe their myopic enough to have not put contingency plans in place if 10 nm wasn't working. Four years later, those contingency plans could easily be bearing fruit behind closed doors.
3) Intel has all kinds of prototype working going on in elastic computing (neural, AI, deep learning, etc.).

Intel may be down but it definitely isn't out.

AMD needed a breather to get their Radeon house in order and it looks like they have one. I hope they don't squander it.

 

[Xzibit]

Well, Intel can still manufacture CPU's at TSMC if they wish so even with their own fabs. The problem that creates is perception of their competence if their own fabs can't produce the latest and greatest they design chips for, one may start questioning if their latest and greatest node design isn't all that good. Again, that would just be a perception to an outsider, not necessarily an inferior product or node design.

They would have to re-design because they are using different methods. Reallocating resources.

Fab 42 was repurposed to 7nm and isnt going to come online until 2021 "if" its on schedule. TSMC is scheduled for 5nm in 2020. The perception is going to be there not only now at 10nm but going foward.

 

[Space Lynx]

AMD needed a breather to get their Radeon house in order and it looks like they have one. I hope they don't squander it.

I really hope they don't squander it as well. However, I have a gut feeling GTX 2080 is going to decimate whatever GDDR6 cards they can produce amongst all tier levels, AMD will probably win at budget as they always do, but for men like you and me... 90% chance we are forced to roll with Nvidia again, just from sheer number gains. Such as the 1080 ti beating vega in a maojrity of games by over 40% 50%+ when both cards were overclocked, at 1440p high refresh. For budget gamers though, AMD next year = a great time to be alive.

GTX 1070 is not overkill at 1080p like so many claim, if you like high refresh that is, which most of us here do these days.

 

[HD64G]

Point is that blue giant is at least 4 years behind their 10nm manufacturing process schedule. And their most possible solution is to abandon their original plan and redesign that 10nm for a close to 14nm one just to not let all know their total failure. My bet is that they will go to 7nm very fast after they release some 10nm-ish products. They are in deep problem atm.

 

[Space Lynx]

Point is that blue giant is at least 4 years behind their 10nm manufacturing process schedule. And their most possible solution is to abandon their original plan and redesign that 10nm for a close to 14nm one just to not let all know their total failure. My bet is that they will go to 7nm very fast after they release some 10nm-ish products. They are in deep problem atm.

It's also possible that none of this matters, Intel chips are still best for gaming according to all the numbers, especially in minimum FPS, also a lot of productivity software I know yuotubers use seem to have odd issues once in awhile with Ryzen. If the 8 core 16 thread 9900k is priced right and it really can do 5ghz overclocked.... even Ryzen x570 next year might not be able to match it... AMD is struggling hardcore with its ability to increase clock speed. I have my eye on the 9900k next year, unless its way too pricey, or AMD x570 lineup in summer 2019 is still stuck at 4.4ghz or less... I need IPC, nothing else. IPC is king for what I do.

 

[StrayKAT]

Point is that blue giant is at least 4 years behind their 10nm manufacturing process schedule. And their most possible solution is to abandon their original plan and redesign that 10nm for a close to 14nm one just to not let all know their total failure. My bet is that they will go to 7nm very fast after they release some 10nm-ish products. They are in deep problem atm.

Their 10nm is already similar to AMD's 7nm. I know people want AMD to just completely decimate Intel with 4 years in advancement, but it's not that simple... or possible. As if they're some hobbyists working out of a Chinese flea market electronics shop.

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

 

[InVasMani]

Wonder if Intel will begin to think about bringing back eDRAM like skylake predecessor had to desktop CPU's that might help slow the bleeding since it has it's uses and they could probably double the density by now. Perhaps the cost of adding it was just too high relative to the gains though. It looked promising so shame it just got tossed aside. Oddly enough though eDRAM would likely be of much higher use to AMD.

 

[HD64G]

It's also possible that none of this matters, Intel chips are still best for gaming according to all the numbers, especially in minimum FPS, also a lot of productivity software I know yuotubers use seem to have odd issues once in awhile with Ryzen. If the 8 core 16 thread 9900k is priced right and it really can do 5ghz overclocked.... even Ryzen x570 next year might not be able to match it... AMD is struggling hardcore with its ability to increase clock speed. I have my eye on the 9900k next year, unless its way too pricey, or AMD x570 lineup in summer 2019 is still stuck at 4.4ghz or less... I need IPC, nothing else. IPC is king for what I do.

Most pc owners need vfm and since 2017 with ryzen 1st gen and the am4 platform being upgradeable till at least 2020 with all new cpus up to that point, amd has a well-paved road in front of them. High-refresh rate@1080P gamers aren't that many. And for the 1440P or 4K ones, cpu doesn't botlleneck their performance, so...

Their 10nm is already similar to AMD's 7nm. I know people want AMD to just completely decimate Intel with 4 years in advancement, but it's not that simple... or possible. As if they're some hobbyists working out of a Chinese flea market electronics shop.

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

I didn't even imply that 7nm will be so much better than intel's 10nm. Read again carefully. I only referred to intel 's problems that force them to alter their plans since they are at least 4 years behind their schedule. Amd's 7nm cpus are unknown quantity for now. When we have more and reliable info, only then we will start making estimations. Now, only predictions are possible, and those not safe at all.

 

[chimonow]

Loving the turn around! Amd just bitch slapped Intel! That's what they get for charging up to five times the price per core. At this point there's really no reason to even buy an Intel CPU unless you have a 1080 Ti and game at 1080p, which is just stupid to begin with. and absolutely no reason to go with any of their HEDT CPU's. I'll gladly take 32 cores/64 threads for $200 less that their 18 core/36 thread part.

Their 10nm is already similar to AMD's 7nm. I know people want AMD to just completely decimate Intel with 4 years in advancement, but it's not that simple... or possible. As if they're some hobbyists working out of a Chinese flea market electronics shop.

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

It's totally possible! Intel's not going to have mainstream 10nm until 2nd half of 2019! Amd may very well be on their second iteration of their 7nm CPU's by then. It all depends on the memory latency and clock frequency improvements that 7nm brings and we'll find out in a relatively short 6-8 months.

 

[sepheronx]

Meh.

As long as it does what it needs to do, and it does so with decent wattage and performance, I don't care how many nm it is.

 

[dorsetknob]

Sad that people get their knowledge on cpu engineering from bankers. Heh. But I predicted this would happen last week in another thread. These kind of subtle points never did well on Wall Street or headlines.

This is not so Much about technology (INTEL v AMD) its more about leveraging a Stock market for Profit

 

[DeathtoGnomes]

This is not so Much about technology (INTEL v AMD) its more about leveraging a Stock market for Profit

everyone disregards that little fact and it turns into another thread into intel vs amd cpu performance thread.