I am certain Intel will remain competitive. Keep in mind that they are a giant with deep pockets, they may be feeling the burn now, but this has happened in the past when AMD released Athlon64. I am more worried for AMD, they are doing very well now and I hope they don't take their foot off the pedal like they did after Athlon64.
AMD should have taken off with Athlon64 and become a bigger competitor, with more money for research and development. Except that Intel gave OEMs deal where they got special pricing for only selling Intel processors. AMD didn't take their foot off the pedal, Intel cut the gas line. Even if the market isn't as susceptible to the practices Intel used in the pre-Core days, you are right that AMD will have to continue innovating to try and get market share...
ROFL. I hope you don't mean it cost Intel a lot...1.5B when you made 60B over the 10yrs you screwed them, means I'll do it to you AGAIN, AND AGAIN, AND AGAIN.
Do not expect Intel or MSFT to not act exactly as they have before because they are still monopolies. If you fine me 1.5b (heck 10B) every time I make 60B in 10yrs, I'll keep doing it as the fine means nothing but MORE profit for me while I bankrupt you (or whoever is next in line) over and over.
Business is WAR. Any bean-counter worth his salt should be able to figure out 60B NET is a lot higher than 1.5B after a 10yr legal war. I'll take that fine yearly for those 10yrs and it's STILL worth breaking the laws. 15B over 10yrs is still NOTHING compared to keeping the other 45B and your competition for getting ANY income (SEE AMD Q reports for 15yrs, they lost 8B over the life of the company!).
It appears Intel is doing the same thing as last time AMD was ahead. Announced 3B in payoffs. Kind of sounds EXACTLY like before right? "Don't use AMD you get X check". It will be interesting to see if we end up with ASUS boards in white boxes again with no ASUS silkscreen on boards at all like before. I truly hope AMD has figured out some defense for this that won't take 10yrs to get a verdict again and a piss poor 1.5B that some IDIOT at AMD accepted. They should have sued until it was 1/2 of that 60B or more. They chose to settle, instead of waiting for a jury to rule for a BIG hit. Continuing the suit is cheap compared to what they lost and what they taught Intel (do it again, it's a cheap fine).
You can coast off the brand loyalty from 'late adopters'. For example, my extended family would only ever buy Intel for the same reason they buy iPhones and Nike, they are the brand names they recognize. However, if you fast forward a decade this free lunch often comes to an end.
The sad thing about their record profits is that a nice part of this will probably go to stifling competition via various kickback / incentive schemes. This is what allowed them to stay competitive in the P3 / P4 days (until Core CPU were released) and I am really hoping that this time around it won't work.
It is different this time. All the big cloud providers want, no need, a second cpu provider. They will buy AMD just to make that happen. Hell Amazon was making their own CPU just to have another option. No intel cant buy there way out this time but frankly I don't think they will need to result to anything like that other than price cuts, if they execute properly going forward.
All will depend on 7nm. If 7nm come out in 2021, Intel can recover pretty quickly. We see it with the first Core CPU. If 7nm gets delayed, Itel will start having real problems.
7nm won't come in 2021 from Intel, except maybe initial laptop sales. There is a big question about any design improvements from Intel, because if they claim a 15 percent IPC boost due to fixing the security problems, then they may be at the same performance levels as they were in 2017 before the security mitigations first started showing up. If clock speeds are down by 20+ percent due to new process node, that's a net loss in performance between the new chips and the old 14nm++++ chips.
People often forget Intel and others have hands in MANY business ventures other than pure CPU. The jump from 14nm to 7nm is not as important as you think it is in the grand picture of a business.
First Intel 10nm process node with "aggressive" scaling was expected to yield around 100M tr/mm2, which is about 2.3x original 14nm and about 2.7x 14nm++ which has decreased density vs 14nm. OEMs have reported Intel shortages for more than 1 year now, so if you think making more than 3.5x the number or chips on a new 7nm node wafer vs 14nm is not a big deal you don't really understand how the semiconductor industry works. Keep this in mind as AMD pushed more cores each year and every 14nm Intel wafer makes less and less CPUs that need more cores to compete. Keep this in mind as AMD pushed more cores each year and every 14nm Intel wafer makes less and less CPUs that need more cores (and die space) to compete.
I gree. Intel planned fab capacity based off the fact they would be on 10nm well before now and they missed that mark hard. So now they simply cant make enough chips since they are not getting more chips per wafer with 10nm.
its not issues with 10 nm itself just overaggressive target. if they will build that 10 nm node on 7 nm process, its already will be enough to keep them afloat propably up to 65W, and at least competitive. AMD will be the king of top performers for next few years, or decade intel lost momentum in there.
It will certainly be impressive if they can keep it, though. TSMC Launched their first 7nm mobile processor in Q32018. HVM for 7nm was in Q2 2018. Their first 5nm HVM is coming in Q2 2020. So maybe we'll see a 5nm iPhone in 2020. Intel is launching a 7nm GPU in 2021. Sure, it's a year late, but they'll be launching with a much bigger die. I wouldn't be surprised if Intel's 7nm tech comes out at the same time as TSMC 5nm tech of a comparable size. That would be a hell of a quick turnaround for such a big ship - even parity with TSMC would be a big win for Intel, I think.
And here we have people who want BGA A series in their machines and Apple mass producing them to feed their cravings of a closed OS and closed ecosystem. While Apple itself is running on Linux for their services.
Amazing. AMD has arguably better and cheaper products in PC space (not mobile!) and data center space, yet not even a dent in Intels financials. Quite the opposite, they are growing. I don't get it... Really curious about AMD next quarterly financials.
I would say they have stagnated since last year, because the TAM has grown and revenue stayed roughly the same. Also "Gross margin was 58.8%, down from 64.5% a year ago. Operating income was down 12% to $6.4 billion, and net income was down 6% to $6.0 billion. This resulted in earnings-per-share of $1.35, down 2% from a year ago." Not much, but AMD has made a small dent in their margins and hence profits.
I mostly agree with you, esp. that margins are down somewhat - but by much less than I would have thought. Any source to back up the TAM growth statement? Did I miss something?
Intel has counted on OEMs and kickbacks to keep their position. You don't see nearly as many AMD Ryzen based offerings as there should be. You don't even see the 2019 batch of desktop APUs out there at decent prices(you still see the 2200g, barely any 3200g based desktops, I've not seen even the 3400g in generally low cost desktop machines from OEMs). Finding a big OEM desktop machine with a Ryzen 5 or 7 in it these days isn't common either. The only reason, when the new chips WILL work on previous generation motherboards, would be payments by Intel to keep these OEMs from offering decent AMD based machines.
Dont Forget that EPYC1 was just an intro, EPYC2 will hit Datacenter revenue hard. Not just because they will have reduced market share but they also had to decrease there pricing with 30%. On top of that they had to create a beast of a SP BGA 9000 platform that nobody wanted in the end in mass production. To close off in 2020 they are in such a bad shape with the withley platform that they will have to convince OEM and customers that they have 2 platforms in 1.... Cooper lake (a glued multidie) and Whiskey lake ( a failed 10nm platfrom) with different features and capabilities. THis is just the top of the iceberg, results will be very visible in the at least the next 5-6 quarters.
And dispite the fact that this is not good, it actually is, because remember as a customer it is good to have competition and to see companies have a drawback, when they believe they are invincible .Pricedrops - improved features and tech - higher investments back into the product.... in stead of just increasing margins.
"The good news for Intel is that despite the initial setbacks for 10 nm, they have stated that 10 nm yields are actually ahead of their internal expectations"
LOL, Bob Swan said the same during the q2 earnings release:
"Both yield and defect density are ahead of schedule for our 10-nanometer data center products."
Even in Q1 Bob said:
"And that played out as we expected, because I mentioned in my prepared remarks that our progress and our improvement on 10-nanometer yields was in line with what we've expected coming into the year, a little bit better, "
Brett, you should scale back your expectations that there was a sudden change and 10 nm is fine now.
Hasn't Intel been saying that 10nm is on track since the end of 2015? Why would anyone believe ANYTHING from Intel at this point? Why Intel didn't get a big fine for what is obviously false reporting for at least three years is obviously kickbacks and bribes. If you lie to your investors, that SHOULD result in penalties.
So, Q3 2018 to Q3 2019, Revenues are flat, Operating and Net Income are DOWN, and profits are down. Data Center is ALWAYS on a cycle, you don't think that businesses replace their servers every year, do you?
I don't see a lot of good news from this report, no matter that Intel is trying to spin it as good news. Intel is still making big profits, but I expect things will be more interesting when we see the numbers from AMD next week, purely because AMD has been recovering well with their new processor lines. Quarter to quarter...yea, third quarter always has a bump in income due to back to school, and preparation by OEMs for the holiday shopping season and needing to get inventory levels ready. Remember, before any business sells a product, they need to have product in hand. If they don't have product ready to ship 3-4 weeks ahead of Black Friday, they are running a risk of there being a shortage.
Whatever people say here, the fact that a company stuck on a 5 year+ old process can post such results is showing that they are managing things the right way. I would imagine that making a Core i9 9900K on an ancient process is much cheaper for Intel than what AMD pays for 3700X (one 7nm die and one 12nm). So given the fact that both CPUs sell at the same price, even if Intel sells less units compared to AMD, the end result is that Intel will still make more money. As for 10nm+ process, I think this time Bob is saying the actual truth. IceLake seems to be quite a good improvement over Whiskey Lake in terms of performance per watt. Don't know about density, but I would expect it to be quite a bit better. This whole 10nm fiasco came from changing lots of stuff (production wise) in one single node. The idea, if it worked, was very good and Intel could have said in 2017 that its 10nm process is as dense as TSMC 7nm. In any case, any mistake done in the development means you will be delayed. When you screw up the entire process premises, you first have to try to find solutions and that I expect it to be somewhere in 2017, followed by a hard decision to start from scratch from 14nm. This took two more years. I don't think Intel has forgot how to make chips. I think they just made the wrong bets. TSMC, Samsung would have the same delay if they would screw things up, because it takes a lot of time to develop, test, scale with profit a certain process. The fact that we see such a fast cadence from TSMC is just because they have teams working in parallel. All in all, I think Intel needs to push 7nm as early as possible. They also need to use chiplets like AMD. The fact that they announced Foveros and EMIB last year means that they intend to take this path also. I say we shoudn't jump to conclusions in regards to Intel/AMD since things can change very fast. And given how well Intel managed this whole 10nm situation, I think they have big plans for future once things start to settle on the manufacturing side of things.
" I say we shoudn't jump to conclusions in regards to Intel/AMD since things can change very fast." Things don't change fast - server guys are now jumping ship with Rome - EPYC they used basically just to validate. AMD is doing very well in the DIY market. I Agree INTC needs a chiplet approach and on 7nm but even at that point AMD will be competitive. The difference this time around with AMD is that they have a great MFG partner - in the past they would step on their own feet owning glofo. AMD set itself up to be competitive for the next many years.
Not surprising--after AMD's decisions during the A64/Opteron era moved the markets to DDR SDRAM and x86-64 (which AMD licensed to Intel years ago), at a time when Intel was pushing RDRAM and Itanium as "The only 64-bits you'll ever need"....Also, I see Intel's current Q3 report as more evidence of its heavy diversification into money/financial markets combined with the sort of clever accounting "tricks" that all companies use for a quarter or two when things are turning south. Intel cannot survive as a competitive force to AMD without new architectures (although Intel could easily survive as a different kind of company, probably), and we have seen none of that to date. Continuing to milk and heavily discount its old architectures will only buy them so much time. If we had a "performance watt" investment metric concerning how much a given company spends to develop per "performance watt," AMD would come out way ahead on that scale, as they to date have done much more than Intel with far less $$$ spent--so of the two companies in terms of their chips businesses, AMD clearly is the better-managed company today, imo.
"much cheaper for Intel than what AMD pays for 3700X" No, because AMD uses a chiplet design those 3700X core complexes are the worst of the stack. In the case of AMD's chiplet design, you have to take the entire stack into account to actually understand what the true cost of it is.
Intel's stacking design is also quite interesting, will see how that goes, it's really nice that CPU's have finally been moving forward again, just wish GPU was in the same market rather than Nvidia completely owning it.
What I find amusing these days is that while both Intel (now deeply discounting) and AMD show clear signs of their processor/chip businesses growing year over year, some people keep saying that the PC market is contracting. Instead of looking at the bottom line of the companies who make the hardware, they choose to look at various "estimates" like Peddie or even, bizarrely, the 100% voluntary Steam survey (you have to opt-in to the ss, it isn't automatic with a Steamworks account)...;)
Intel's main income is overcharging precipitously in the enterprise/server markets. AMD has only just begun to hammer Intel where it hurts in those markets. People accustomed to thinking that AMD will simply fizzle "inevitably" should adjust their thinking to incorporate the concept that AMD today is a different company entirely from what it was during its heyday A64/Opteron era. Today's AMD is far different and is already planning several years ahead. Right now, for Intel, the only way they can keep the revenue from their chips businesses flowing is to discount heavily, as there is little other reason *not* to go AMD these days. This may change when/if Intel is able to market competitive products at competitive prices looking ahead, certainly. AMD's server share will be appropriating ~10% per year of the entire enterprise market, I should think. Should be great fun to watch...;)
I don't think Intel is in a bind, not even close, they have a huge market outside of just pure CPU, and they're changing their design/product strategy to incorporate that.
This probably isn't going to continue if ROME has anything to say. Intel's Xeon brand might as well be irrelevant the way things are going. Which is good for the short/medium term, because a strong AMD means a healthier overal market for everyone. Businesses, consumers and human progress overall.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
39 Comments
Back to Article
Potato Power - Thursday, October 24, 2019 - link
Look at Intel has so much trouble with their 10nm. I won't bet on them to have 7nm ready in 2021. While TSMC will enter mass production for 5nm, ha.Cliff34 - Friday, October 25, 2019 - link
The bottom line is profit. Intel is still going strong despite AMD chipping some of their share.The question is how long will this last. Can Intel stay competitive in the next few years.
deepblue08 - Friday, October 25, 2019 - link
I am certain Intel will remain competitive. Keep in mind that they are a giant with deep pockets, they may be feeling the burn now, but this has happened in the past when AMD released Athlon64. I am more worried for AMD, they are doing very well now and I hope they don't take their foot off the pedal like they did after Athlon64.tiwi1391 - Friday, October 25, 2019 - link
AMD should have taken off with Athlon64 and become a bigger competitor, with more money for research and development. Except that Intel gave OEMs deal where they got special pricing for only selling Intel processors. AMD didn't take their foot off the pedal, Intel cut the gas line.Even if the market isn't as susceptible to the practices Intel used in the pre-Core days, you are right that AMD will have to continue innovating to try and get market share...
AshlayW - Saturday, October 26, 2019 - link
This. Intel is historically a scumbag company.Korguz - Saturday, October 26, 2019 - link
and it ended up costing them, a lotTheJian - Monday, October 28, 2019 - link
ROFL. I hope you don't mean it cost Intel a lot...1.5B when you made 60B over the 10yrs you screwed them, means I'll do it to you AGAIN, AND AGAIN, AND AGAIN.Do not expect Intel or MSFT to not act exactly as they have before because they are still monopolies. If you fine me 1.5b (heck 10B) every time I make 60B in 10yrs, I'll keep doing it as the fine means nothing but MORE profit for me while I bankrupt you (or whoever is next in line) over and over.
Business is WAR. Any bean-counter worth his salt should be able to figure out 60B NET is a lot higher than 1.5B after a 10yr legal war. I'll take that fine yearly for those 10yrs and it's STILL worth breaking the laws. 15B over 10yrs is still NOTHING compared to keeping the other 45B and your competition for getting ANY income (SEE AMD Q reports for 15yrs, they lost 8B over the life of the company!).
It appears Intel is doing the same thing as last time AMD was ahead. Announced 3B in payoffs. Kind of sounds EXACTLY like before right? "Don't use AMD you get X check". It will be interesting to see if we end up with ASUS boards in white boxes again with no ASUS silkscreen on boards at all like before. I truly hope AMD has figured out some defense for this that won't take 10yrs to get a verdict again and a piss poor 1.5B that some IDIOT at AMD accepted. They should have sued until it was 1/2 of that 60B or more. They chose to settle, instead of waiting for a jury to rule for a BIG hit. Continuing the suit is cheap compared to what they lost and what they taught Intel (do it again, it's a cheap fine).
Amandtec - Friday, October 25, 2019 - link
You can coast off the brand loyalty from 'late adopters'. For example, my extended family would only ever buy Intel for the same reason they buy iPhones and Nike, they are the brand names they recognize. However, if you fast forward a decade this free lunch often comes to an end.FreckledTrout - Friday, October 25, 2019 - link
Agree. I hope AMD realizes this and starts doing some rather profesional marketing. They need there "Intel Inside" moment of marketing.Irata - Friday, October 25, 2019 - link
The sad thing about their record profits is that a nice part of this will probably go to stifling competition via various kickback / incentive schemes. This is what allowed them to stay competitive in the P3 / P4 days (until Core CPU were released) and I am really hoping that this time around it won't work.FreckledTrout - Friday, October 25, 2019 - link
It is different this time. All the big cloud providers want, no need, a second cpu provider. They will buy AMD just to make that happen. Hell Amazon was making their own CPU just to have another option. No intel cant buy there way out this time but frankly I don't think they will need to result to anything like that other than price cuts, if they execute properly going forward.yannigr2 - Friday, October 25, 2019 - link
All will depend on 7nm. If 7nm come out in 2021, Intel can recover pretty quickly. We see it with the first Core CPU. If 7nm gets delayed, Itel will start having real problems.Targon - Friday, October 25, 2019 - link
7nm won't come in 2021 from Intel, except maybe initial laptop sales. There is a big question about any design improvements from Intel, because if they claim a 15 percent IPC boost due to fixing the security problems, then they may be at the same performance levels as they were in 2017 before the security mitigations first started showing up. If clock speeds are down by 20+ percent due to new process node, that's a net loss in performance between the new chips and the old 14nm++++ chips.imaheadcase - Friday, October 25, 2019 - link
People often forget Intel and others have hands in MANY business ventures other than pure CPU. The jump from 14nm to 7nm is not as important as you think it is in the grand picture of a business.sgeocla - Friday, October 25, 2019 - link
First Intel 10nm process node with "aggressive" scaling was expected to yield around 100M tr/mm2, which is about 2.3x original 14nm and about 2.7x 14nm++ which has decreased density vs 14nm.OEMs have reported Intel shortages for more than 1 year now, so if you think making more than 3.5x the number or chips on a new 7nm node wafer vs 14nm is not a big deal you don't really understand how the semiconductor industry works.
Keep this in mind as AMD pushed more cores each year and every 14nm Intel wafer makes less and less CPUs that need more cores to compete.
Keep this in mind as AMD pushed more cores each year and every 14nm Intel wafer makes less and less CPUs that need more cores (and die space) to compete.
FreckledTrout - Friday, October 25, 2019 - link
I gree. Intel planned fab capacity based off the fact they would be on 10nm well before now and they missed that mark hard. So now they simply cant make enough chips since they are not getting more chips per wafer with 10nm.tiwi1391 - Friday, October 25, 2019 - link
There are still issues though, like Intel promising FPGAs for 2017 to industry partners and then getting released last month...ilt24 - Friday, October 25, 2019 - link
@tiwi1391 ... "promising FPGAs for 2017 to industry partners"they fixed that problem by buying that industry partner.
deil - Friday, October 25, 2019 - link
its not issues with 10 nm itself just overaggressive target. if they will build that 10 nm node on 7 nm process, its already will be enough to keep them afloat propably up to 65W, and at least competitive. AMD will be the king of top performers for next few years, or decade intel lost momentum in there.Drumsticks - Friday, October 25, 2019 - link
It will certainly be impressive if they can keep it, though. TSMC Launched their first 7nm mobile processor in Q32018. HVM for 7nm was in Q2 2018. Their first 5nm HVM is coming in Q2 2020. So maybe we'll see a 5nm iPhone in 2020. Intel is launching a 7nm GPU in 2021. Sure, it's a year late, but they'll be launching with a much bigger die. I wouldn't be surprised if Intel's 7nm tech comes out at the same time as TSMC 5nm tech of a comparable size. That would be a hell of a quick turnaround for such a big ship - even parity with TSMC would be a big win for Intel, I think.Quantumz0d - Friday, October 25, 2019 - link
That's why the name Chipzilla. LOLAnd here we have people who want BGA A series in their machines and Apple mass producing them to feed their cravings of a closed OS and closed ecosystem. While Apple itself is running on Linux for their services.
jlp2097 - Friday, October 25, 2019 - link
Amazing. AMD has arguably better and cheaper products in PC space (not mobile!) and data center space, yet not even a dent in Intels financials. Quite the opposite, they are growing. I don't get it... Really curious about AMD next quarterly financials.klatscho - Friday, October 25, 2019 - link
I would say they have stagnated since last year, because the TAM has grown and revenue stayed roughly the same. Also "Gross margin was 58.8%, down from 64.5% a year ago. Operating income was down 12% to $6.4 billion, and net income was down 6% to $6.0 billion. This resulted in earnings-per-share of $1.35, down 2% from a year ago." Not much, but AMD has made a small dent in their margins and hence profits.jlp2097 - Friday, October 25, 2019 - link
I mostly agree with you, esp. that margins are down somewhat - but by much less than I would have thought. Any source to back up the TAM growth statement? Did I miss something?Targon - Friday, October 25, 2019 - link
Intel has counted on OEMs and kickbacks to keep their position. You don't see nearly as many AMD Ryzen based offerings as there should be. You don't even see the 2019 batch of desktop APUs out there at decent prices(you still see the 2200g, barely any 3200g based desktops, I've not seen even the 3400g in generally low cost desktop machines from OEMs). Finding a big OEM desktop machine with a Ryzen 5 or 7 in it these days isn't common either. The only reason, when the new chips WILL work on previous generation motherboards, would be payments by Intel to keep these OEMs from offering decent AMD based machines.duploxxx - Friday, October 25, 2019 - link
not a dent?Look at Gross Margin?
Dont Forget that EPYC1 was just an intro, EPYC2 will hit Datacenter revenue hard. Not just because they will have reduced market share but they also had to decrease there pricing with 30%.
On top of that they had to create a beast of a SP BGA 9000 platform that nobody wanted in the end in mass production.
To close off in 2020 they are in such a bad shape with the withley platform that they will have to convince OEM and customers that they have 2 platforms in 1.... Cooper lake (a glued multidie) and Whiskey lake ( a failed 10nm platfrom) with different features and capabilities.
THis is just the top of the iceberg, results will be very visible in the at least the next 5-6 quarters.
And dispite the fact that this is not good, it actually is, because remember as a customer it is good to have competition and to see companies have a drawback, when they believe they are invincible .Pricedrops - improved features and tech - higher investments back into the product.... in stead of just increasing margins.
Kishoreshack - Friday, October 25, 2019 - link
It will take time to show upthe AMD products were launched recently
it will take time to affect intels earnings
AshlayW - Saturday, October 26, 2019 - link
Consumer stupidity/Mindshare takes many years to erode.brakdoo - Friday, October 25, 2019 - link
"The good news for Intel is that despite the initial setbacks for 10 nm, they have stated that 10 nm yields are actually ahead of their internal expectations"LOL, Bob Swan said the same during the q2 earnings release:
"Both yield and defect density are ahead of schedule for our 10-nanometer data center products."
Even in Q1 Bob said:
"And that played out as we expected, because I mentioned in my prepared remarks that our progress and our improvement on 10-nanometer yields was in line with what we've expected coming into the year, a little bit better, "
Brett, you should scale back your expectations that there was a sudden change and 10 nm is fine now.
Targon - Friday, October 25, 2019 - link
Hasn't Intel been saying that 10nm is on track since the end of 2015? Why would anyone believe ANYTHING from Intel at this point? Why Intel didn't get a big fine for what is obviously false reporting for at least three years is obviously kickbacks and bribes. If you lie to your investors, that SHOULD result in penalties.Targon - Friday, October 25, 2019 - link
So, Q3 2018 to Q3 2019, Revenues are flat, Operating and Net Income are DOWN, and profits are down. Data Center is ALWAYS on a cycle, you don't think that businesses replace their servers every year, do you?I don't see a lot of good news from this report, no matter that Intel is trying to spin it as good news. Intel is still making big profits, but I expect things will be more interesting when we see the numbers from AMD next week, purely because AMD has been recovering well with their new processor lines. Quarter to quarter...yea, third quarter always has a bump in income due to back to school, and preparation by OEMs for the holiday shopping season and needing to get inventory levels ready. Remember, before any business sells a product, they need to have product in hand. If they don't have product ready to ship 3-4 weeks ahead of Black Friday, they are running a risk of there being a shortage.
yeeeeman - Friday, October 25, 2019 - link
Whatever people say here, the fact that a company stuck on a 5 year+ old process can post such results is showing that they are managing things the right way.I would imagine that making a Core i9 9900K on an ancient process is much cheaper for Intel than what AMD pays for 3700X (one 7nm die and one 12nm). So given the fact that both CPUs sell at the same price, even if Intel sells less units compared to AMD, the end result is that Intel will still make more money.
As for 10nm+ process, I think this time Bob is saying the actual truth. IceLake seems to be quite a good improvement over Whiskey Lake in terms of performance per watt. Don't know about density, but I would expect it to be quite a bit better. This whole 10nm fiasco came from changing lots of stuff (production wise) in one single node. The idea, if it worked, was very good and Intel could have said in 2017 that its 10nm process is as dense as TSMC 7nm.
In any case, any mistake done in the development means you will be delayed. When you screw up the entire process premises, you first have to try to find solutions and that I expect it to be somewhere in 2017, followed by a hard decision to start from scratch from 14nm. This took two more years. I don't think Intel has forgot how to make chips. I think they just made the wrong bets. TSMC, Samsung would have the same delay if they would screw things up, because it takes a lot of time to develop, test, scale with profit a certain process. The fact that we see such a fast cadence from TSMC is just because they have teams working in parallel.
All in all, I think Intel needs to push 7nm as early as possible. They also need to use chiplets like AMD. The fact that they announced Foveros and EMIB last year means that they intend to take this path also. I say we shoudn't jump to conclusions in regards to Intel/AMD since things can change very fast. And given how well Intel managed this whole 10nm situation, I think they have big plans for future once things start to settle on the manufacturing side of things.
stockolicious - Friday, October 25, 2019 - link
" I say we shoudn't jump to conclusions in regards to Intel/AMD since things can change very fast."Things don't change fast - server guys are now jumping ship with Rome - EPYC they used basically just to validate. AMD is doing very well in the DIY market. I Agree INTC needs a chiplet approach and on 7nm but even at that point AMD will be competitive. The difference this time around with AMD is that they have a great MFG partner - in the past they would step on their own feet owning glofo. AMD set itself up to be competitive for the next many years.
WaltC - Friday, October 25, 2019 - link
Not surprising--after AMD's decisions during the A64/Opteron era moved the markets to DDR SDRAM and x86-64 (which AMD licensed to Intel years ago), at a time when Intel was pushing RDRAM and Itanium as "The only 64-bits you'll ever need"....Also, I see Intel's current Q3 report as more evidence of its heavy diversification into money/financial markets combined with the sort of clever accounting "tricks" that all companies use for a quarter or two when things are turning south. Intel cannot survive as a competitive force to AMD without new architectures (although Intel could easily survive as a different kind of company, probably), and we have seen none of that to date. Continuing to milk and heavily discount its old architectures will only buy them so much time. If we had a "performance watt" investment metric concerning how much a given company spends to develop per "performance watt," AMD would come out way ahead on that scale, as they to date have done much more than Intel with far less $$$ spent--so of the two companies in terms of their chips businesses, AMD clearly is the better-managed company today, imo.RSAUser - Wednesday, October 30, 2019 - link
"much cheaper for Intel than what AMD pays for 3700X"No, because AMD uses a chiplet design those 3700X core complexes are the worst of the stack. In the case of AMD's chiplet design, you have to take the entire stack into account to actually understand what the true cost of it is.
Intel's stacking design is also quite interesting, will see how that goes, it's really nice that CPU's have finally been moving forward again, just wish GPU was in the same market rather than Nvidia completely owning it.
WaltC - Friday, October 25, 2019 - link
What I find amusing these days is that while both Intel (now deeply discounting) and AMD show clear signs of their processor/chip businesses growing year over year, some people keep saying that the PC market is contracting. Instead of looking at the bottom line of the companies who make the hardware, they choose to look at various "estimates" like Peddie or even, bizarrely, the 100% voluntary Steam survey (you have to opt-in to the ss, it isn't automatic with a Steamworks account)...;)Intel's main income is overcharging precipitously in the enterprise/server markets. AMD has only just begun to hammer Intel where it hurts in those markets. People accustomed to thinking that AMD will simply fizzle "inevitably" should adjust their thinking to incorporate the concept that AMD today is a different company entirely from what it was during its heyday A64/Opteron era. Today's AMD is far different and is already planning several years ahead. Right now, for Intel, the only way they can keep the revenue from their chips businesses flowing is to discount heavily, as there is little other reason *not* to go AMD these days. This may change when/if Intel is able to market competitive products at competitive prices looking ahead, certainly. AMD's server share will be appropriating ~10% per year of the entire enterprise market, I should think. Should be great fun to watch...;)
WaltC - Friday, October 25, 2019 - link
Wanted to add that AMD's Q3 report is due in four days, Oct. 29--that is going to be interesting!RSAUser - Wednesday, October 30, 2019 - link
I don't think Intel is in a bind, not even close, they have a huge market outside of just pure CPU, and they're changing their design/product strategy to incorporate that.AshlayW - Saturday, October 26, 2019 - link
This probably isn't going to continue if ROME has anything to say. Intel's Xeon brand might as well be irrelevant the way things are going. Which is good for the short/medium term, because a strong AMD means a healthier overal market for everyone. Businesses, consumers and human progress overall.