They need their own CPU architecture now if they want to continue to grow. X86 is so inefficient and high-end ARM server CPUs are coming online, and the market in 5 years might be completely different. Maybe they could build off of RISC-V, but likely they'll need something completely custom.
Intel is going to face the same problems, with competition from ARM servers. But at least they have a broad line of support IP - Optane (great for databases), FPGAs (great for EDA accelerators), Xe (for AI), QuickAssist Technology (web servers), QuickSync video (server video transcoding), AVX-512, etc..
AMD will need a broad line of server IP to compete in the long term. An efficient CPU isn't going to be enough.
AMD is in no position to make a new architecture. Even Intel failed with IA64. Maintaining x86 compatibility is the cornerstone to these companies chip business. This is specifically why AMD built on top of x86 when they developed AMD64 (or as Intel would call it when they implemented it, x86-64.)
Fortunately AMD also has more diversity with their GPU business (which is significantly larger than Intel's other ventures such as networking, embedded, AI, etc.) But realistically neither of these companies are venturing away from x86 in the consumer space for a long time.
Data centers might be another story. Who knows, I'm no expert, but Anapurna\Amazon has spent billions and apparently it paid off for them, but that combined with other ARM competition means the market has a lot of players. Sometimes it's nice to be in a market when you only have one other player.
Intel has failed with IA64 exactly because "x86 is inefficient" is a myth. Nowadays and for years, CISC > RISC since we can afford cramming insane number of transistors into tiny chips.
AVX-512 is barely alive and "quick sync video" is a funny mention, as it is basically an optimized codec.
If x86 were efficient, Atom would have stood a chance against ARM (it even launched on a better manufacturing node). The fact that ARM has done in a half decade what took Intel 20+ years also speaks to an architecture that is much easier to optimize and iterate on.
IA64 had *very* different issues. They argued that a "sufficiently smart compiler" could optimize their code. That essentially required solving the halting problem which is provably unsolvable and as expected, despite using VLIW GPUs for decades, the problem was never really solved (the big reason for moving to less-fixed architectures).
Intel already knew what a disaster EPIC would be from their i860 failure. That raises the interesting question about why they'd push tech they already knew was bad. As a result of all the Itanium marketing, all their upcoming RISC competitors were shut down (PA-RISC, Alpha, etc) or moved on to specialty markets (MIPS with network switches, PowerPC or SPARC very-parallel supercomputers, etc). Who knows if this was intentional, but in practice, Intel managed to make hundreds of billions more off its monopoly for only a couple billion in investments (probably nowhere near as much as they spent on their illegal anti-AMD campaigns over the years).
The Atom is and was plenty efficient. It was the ecosystem surrounding it that doomed it to failure. For example, with the N270 they had a 2.5 watt CPU paired with a 15 watt chipset. The wonkiness doesn't end there either, the other CPUs for even more mobile devices had Imagination chipsets which weren't supported worth a hill of beans. As far as available devices, that was an uphill battle from the start in mobile because ARM was already heavily entrenched with near full market dominance. This meant compatibility was always going to be difficult for Intel on mobile where operating systems like Android, iOS, and various embedded Linuxes are dominant.
As far as how ARM caught up in performance so quickly, it is because there was never actually any reason why it couldn't be other than it wasn't the target market. There is also a matter of hindsight and poaching. The engineers at ARM weren't working in the dark. They already knew what works and what doesn't. Intel and AMD were pathfinding in the high performance space so of course it was going to take longer.
There is nothing about ARM arch that makes it better or special in comparison to x86. You'd be foolish to believe there is.
ARM's success in the low power, low price market has much more to do with the ARM business model and Intel's refusal to compete than it does to any functional realities of the silicon CPU market.
"Fortunately AMD also has more diversity with their GPU business (which is significantly larger than Intel's other ventures such as networking, embedded, AI, etc.) But realistically neither of these companies are venturing away from x86 in the consumer space for a long time."
This is simply not true... from Intel's latest quarter results: IoT is $816M, Mobileye $146M, Programmable Solutions (former Altera business) $501M, Non-volatile memory $1.7B... only these are bigger than all AMD combined. And I am not even sure if Intel puts their networking products in these categories....
That's incorrect, Intel called its implementation IA32e, EM64T, then settled on Intel 64. The term x86-64 or x64 are vendor neutral terms. As a matter of fact, when the extension was introduced, AMD called it x86-64, later renamed AMD64.
Good increase in DC side for Q2 but slightly less than year ago. Client side has enormous growth from year ago but slight dip from Q1. Overall excellent and Q3 guidance is excellent. These are good times for AMD.
"finally captured a double-digit share of the server processor market, a major goal for the company" - really ? Intel server CPU sales are 7.2B, so AMD sales are 8%. After excluding chips for consoles, this share drop below 5% probably.
I can believe it. EPYC is entering the mix at a price and with scalability that makes it a great choice for cloud and high-density VM providers, and they've clearly been snapping them up. I use three AMD VMs myself - one is on a 3900X, which probably isn't even included on the server side.
We're still rely mainly on Intel VMs for our CDN, but that's because old Intel CPUs are still out there. It's really rare for us to use current-gen hardware, but with AMD's price point we suddenly are.
AMD's market share predictions on the server side were very conservative. Lisa has been trying to provide consistency to wall street and meetable goals.
One of the problems with AMD stock (and the management) in the past was always the unpredictability and the constantly missing goals and Lisa is trying to change that whole culture at AMD. It didn't just hurt their stock either, constantly missing goals and failing to meet releases damaged them greatly with the OEM's who have to invest lots of money to prepare products that don't show up or dramatically under perform the predictions.
Lisa has made AMD a much better company, one of the reasons I invested several years ago when Ryzen showed she was making these changes successfully.
The forward PE is only 48, still high but not nearly as exorbitant as the 170 you claim. Looking at the same PE from what was predicted the growth has far exceeded the prior predicted which means a higher PE than average is justified.
My own investment in 2018 had similar predicted PE's, and I've seen a 400% increase in value. As long as AMD exceeds their sale goals the growth in stock value will continue, especially since intel has stumbled again on process tech which should make hitting those sale goals even easier.
Look at it this way, AMD is still pushing Intel for the top on CPU's, has them licked on the server side (which practically guarantees server sales increases) and now has a guaranteed 6 months (and probably longer if you go by history) more time to hit those profits.
ROFL. He said, PE, then you change topic to forward PE.
At some point you have to grow NET INCOME too. You can get all the revenue you want, but if you make NOTHING on it, what is the point? At some point people will ask you to PROFIT or dump your shares like I did ;) At this point AMD has to prove more than "we can sell it cheap, get revenue and ACT like we make money"...Not good enough after 3 years of this crap. If you're not making .5B GROSS PROFIT on 1B revenue, you're doing it WRONG in this business and the best do far better than that (see Intel, while losing share, losing fab war, etc, STILL understand how to MAKE MORE NET INCOME). Wake me when AMD figures it out.
Making 150mil on 2B is FAILING. PERIOD. There are NO GUARANTEES when the other guy can simply PAY people with 23B NET INCOME he earned to people YOU need to survive. IE, a simple extra 10% maybe (use your imagination, it's called contra revenue etc) for the same fab wafers as AMD, and promise even more crap every year you ignore AMD, what do you think TSMC does? They don't care who wins, only who makes them more money. They just make chips, they don't care WHO they are from, only who pays more to get them.
4.1B+ per year for 4yrs+ was BLOWN just to stay in mobile by Intel. In case you missed it, Intel was pissing away MORE than 1/2 of AMD's revenue...YEARLY...ROFL. You don't think they will allow the enemy to buy wafers they need do you? They already said ~7 new divisions now, each can decide what fab to go with, OTHER than Intel's.
Translation: All your fabs belong to US (Intel, well, ok, an apple...ROFL). WAFERS to the highest bidders. DUH. Guarantee...LOL. Until person X writes a larger check than person Y. If AMD doesn't have a contract for years, the party is OVER. Intel and even NV can pay MUCH more per wafer start than AMD. Rumors are abound about the fighting over them already, and Intel supposedly already has an agreement with TSMC now. Huawei gone sept, so I'm guessing Intel outbids anyone but Apple for that stuff, and since apple is on 5nm, It's Intel vs. everyone else for those wafers. I'm guessing Intel's 23B NET INCOME per year matters here. I guess you missed the part where AMD SOLD their fabs (and everything else they owned but IP, due to stupid LOW PRICING for decades). IF you don't OWN the fab, anyone can ruin your day with a check (use your imagination, 100 ways to pay you without a check too, easily made legal). How do you stop someone from paying MORE than you can afford for a wafer?
What is AMD's option here, take or pay agreement again on wafers? How many years did it take to get out of that from GF? It cost them billions for that mistake. Intel put the fear of god in everyone (board makers couldn't put their name on mboards, Asus sold boards in whiteboxes etc, PC makers threatened, stores, etc etc), and AMD got stuck with years of wafers they couldn't use. Read The Art of War (or heck, Trump's Art of the Deal), and maybe you'll get how warfare works (business is ruthless, China reads trump's book...LOL, odd USA isn't using it in all schools).
The best product doesn't always win. Just ask Netscape, Beta, etc etc :) Intel can piss away another 4B a year just to stop you from getting wafers easily and still make nearly 20B NET. What happens with 5nm TSMC server chips from Intel? You are not listening to INTEL telling you what is coming. OTHER FABS making INTEL chips. Do the math.
Price to Earnings ratio has two equally valid calculations. Trailing P/E is based on the previous years earnings while forward P/E is based on the projected earnings. The trailing P/E is the one typically calculated because it's not based on projections. But in the case of AMD the trialing PE is distorted because of the cyrto-coin GPU market distortion and as a result the trialing P/E IMO is not a reliable indicator until that distortion is no longer in the calculation. Good investors investigate the basis of numbers before accepting them.
You then spend a couple paragraphs scattershoting various topics that appear to be a stream of consciousness that makes little sense. If you'd like a reply please post something coherent.
AMD remains a very good investment, they have significant potential to not only hit their growth targets but exceed them, as they've done for the last several years, especially given that Intel continues to miss their goals and struggle with manufacturing.
There's a much, much longer lag between good hardware arriving and it being put into play on the server side of the equation.
They have to make the CPUs, then the OEMs have to integrate them, then the resellers have to change their ideas about what products to sell, then the IT managers have to convince the people holding the purse strings to change architecture, then the solutions have to be designed... it's a slow process, end-to-end.
Compared to "shiny new product released, build self new PC" it's positively glacial.
It's not just those reasons, your typical enterprise that purchases large volumes likes to test new systems before deploying them. I'd argue your average enterprise server purchase is at least a year after commercial availability because that's as fast as they can test and deploy. But the cheaper cost of AMD is definitely having an impact.
Just reminding everyone that contrary to what many people say on the internet, Ryzen 4000 series doesn't beat Intel 10th gen in performance both single-core and multi-core.
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
35 Comments
Back to Article
Memo.Ray - Tuesday, July 28, 2020 - link
Good to see them continue to grow!vFunct - Wednesday, July 29, 2020 - link
They need their own CPU architecture now if they want to continue to grow. X86 is so inefficient and high-end ARM server CPUs are coming online, and the market in 5 years might be completely different. Maybe they could build off of RISC-V, but likely they'll need something completely custom.Intel is going to face the same problems, with competition from ARM servers. But at least they have a broad line of support IP - Optane (great for databases), FPGAs (great for EDA accelerators), Xe (for AI), QuickAssist Technology (web servers), QuickSync video (server video transcoding), AVX-512, etc..
AMD will need a broad line of server IP to compete in the long term. An efficient CPU isn't going to be enough.
Samus - Wednesday, July 29, 2020 - link
AMD is in no position to make a new architecture. Even Intel failed with IA64. Maintaining x86 compatibility is the cornerstone to these companies chip business. This is specifically why AMD built on top of x86 when they developed AMD64 (or as Intel would call it when they implemented it, x86-64.)Fortunately AMD also has more diversity with their GPU business (which is significantly larger than Intel's other ventures such as networking, embedded, AI, etc.) But realistically neither of these companies are venturing away from x86 in the consumer space for a long time.
Data centers might be another story. Who knows, I'm no expert, but Anapurna\Amazon has spent billions and apparently it paid off for them, but that combined with other ARM competition means the market has a lot of players. Sometimes it's nice to be in a market when you only have one other player.
medi05 - Wednesday, July 29, 2020 - link
Intel has failed with IA64 exactly because "x86 is inefficient" is a myth.Nowadays and for years, CISC > RISC since we can afford cramming insane number of transistors into tiny chips.
AVX-512 is barely alive and "quick sync video" is a funny mention, as it is basically an optimized codec.
quadrivial - Wednesday, July 29, 2020 - link
If x86 were efficient, Atom would have stood a chance against ARM (it even launched on a better manufacturing node). The fact that ARM has done in a half decade what took Intel 20+ years also speaks to an architecture that is much easier to optimize and iterate on.IA64 had *very* different issues. They argued that a "sufficiently smart compiler" could optimize their code. That essentially required solving the halting problem which is provably unsolvable and as expected, despite using VLIW GPUs for decades, the problem was never really solved (the big reason for moving to less-fixed architectures).
Intel already knew what a disaster EPIC would be from their i860 failure. That raises the interesting question about why they'd push tech they already knew was bad. As a result of all the Itanium marketing, all their upcoming RISC competitors were shut down (PA-RISC, Alpha, etc) or moved on to specialty markets (MIPS with network switches, PowerPC or SPARC very-parallel supercomputers, etc). Who knows if this was intentional, but in practice, Intel managed to make hundreds of billions more off its monopoly for only a couple billion in investments (probably nowhere near as much as they spent on their illegal anti-AMD campaigns over the years).
linuxid10t - Thursday, July 30, 2020 - link
The Atom is and was plenty efficient. It was the ecosystem surrounding it that doomed it to failure. For example, with the N270 they had a 2.5 watt CPU paired with a 15 watt chipset. The wonkiness doesn't end there either, the other CPUs for even more mobile devices had Imagination chipsets which weren't supported worth a hill of beans. As far as available devices, that was an uphill battle from the start in mobile because ARM was already heavily entrenched with near full market dominance. This meant compatibility was always going to be difficult for Intel on mobile where operating systems like Android, iOS, and various embedded Linuxes are dominant.As far as how ARM caught up in performance so quickly, it is because there was never actually any reason why it couldn't be other than it wasn't the target market. There is also a matter of hindsight and poaching. The engineers at ARM weren't working in the dark. They already knew what works and what doesn't. Intel and AMD were pathfinding in the high performance space so of course it was going to take longer.
rahvin - Sunday, August 2, 2020 - link
There is nothing about ARM arch that makes it better or special in comparison to x86. You'd be foolish to believe there is.ARM's success in the low power, low price market has much more to do with the ARM business model and Intel's refusal to compete than it does to any functional realities of the silicon CPU market.
kingmouf - Wednesday, July 29, 2020 - link
"Fortunately AMD also has more diversity with their GPU business (which is significantly larger than Intel's other ventures such as networking, embedded, AI, etc.) But realistically neither of these companies are venturing away from x86 in the consumer space for a long time."This is simply not true... from Intel's latest quarter results: IoT is $816M, Mobileye $146M, Programmable Solutions (former Altera business) $501M, Non-volatile memory $1.7B... only these are bigger than all AMD combined. And I am not even sure if Intel puts their networking products in these categories....
dotjaz - Wednesday, July 29, 2020 - link
And surely Intel's empire can rely on what you listed to survive then. I mean losing x86-64 related products is merely 90% revenue, so diverse.dotjaz - Wednesday, July 29, 2020 - link
Diversity is measured against its own size, not the competition.Spunjji - Wednesday, July 29, 2020 - link
Maybe Samus meant "larger as a proportion of the company"? Not sure that's a particularly worthwhile point to make either, though 😅dotjaz - Wednesday, July 29, 2020 - link
That's incorrect, Intel called its implementation IA32e, EM64T, then settled on Intel 64. The term x86-64 or x64 are vendor neutral terms. As a matter of fact, when the extension was introduced, AMD called it x86-64, later renamed AMD64.https://software.intel.com/content/www/us/en/devel...
rahvin - Sunday, August 2, 2020 - link
It's correctly called AMD64 on the linux side because it was created, specified by and submitted to the kernel by AMD.Spunjji - Wednesday, July 29, 2020 - link
I see some variant of this nonsense posted every year, and every year it fails to become true.It's the hardware variant of "this year will be the year of Linux on desktop".
trivik12 - Tuesday, July 28, 2020 - link
Good increase in DC side for Q2 but slightly less than year ago. Client side has enormous growth from year ago but slight dip from Q1. Overall excellent and Q3 guidance is excellent. These are good times for AMD.TristanSDX - Tuesday, July 28, 2020 - link
"finally captured a double-digit share of the server processor market, a major goal for the company" - really ? Intel server CPU sales are 7.2B, so AMD sales are 8%. After excluding chips for consoles, this share drop below 5% probably.martinw - Tuesday, July 28, 2020 - link
I believe it is market share by units, not by revenue.Ryan Smith - Tuesday, July 28, 2020 - link
Correct. AMD now at at least 10% of the server CPU market in units sold.GreenReaper - Tuesday, July 28, 2020 - link
I can believe it. EPYC is entering the mix at a price and with scalability that makes it a great choice for cloud and high-density VM providers, and they've clearly been snapping them up. I use three AMD VMs myself - one is on a 3900X, which probably isn't even included on the server side.We're still rely mainly on Intel VMs for our CDN, but that's because old Intel CPUs are still out there. It's really rare for us to use current-gen hardware, but with AMD's price point we suddenly are.
rahvin - Sunday, August 2, 2020 - link
AMD's market share predictions on the server side were very conservative. Lisa has been trying to provide consistency to wall street and meetable goals.One of the problems with AMD stock (and the management) in the past was always the unpredictability and the constantly missing goals and Lisa is trying to change that whole culture at AMD. It didn't just hurt their stock either, constantly missing goals and failing to meet releases damaged them greatly with the OEM's who have to invest lots of money to prepare products that don't show up or dramatically under perform the predictions.
Lisa has made AMD a much better company, one of the reasons I invested several years ago when Ryzen showed she was making these changes successfully.
brucethemoose - Tuesday, July 28, 2020 - link
Good for them!The stock price is still insane though. A P/E of 170... holy moly.
ads295 - Wednesday, July 29, 2020 - link
P/E is dependent on price as well as earnings. Earnings for AMD have been poor enough to now lead to ludicrous P/E ratios.brucethemoose - Wednesday, July 29, 2020 - link
Well yes, exactly. Its like theres an expectation that earnings will jump 10x, and do it soon.rahvin - Sunday, August 2, 2020 - link
The forward PE is only 48, still high but not nearly as exorbitant as the 170 you claim. Looking at the same PE from what was predicted the growth has far exceeded the prior predicted which means a higher PE than average is justified.My own investment in 2018 had similar predicted PE's, and I've seen a 400% increase in value. As long as AMD exceeds their sale goals the growth in stock value will continue, especially since intel has stumbled again on process tech which should make hitting those sale goals even easier.
Look at it this way, AMD is still pushing Intel for the top on CPU's, has them licked on the server side (which practically guarantees server sales increases) and now has a guaranteed 6 months (and probably longer if you go by history) more time to hit those profits.
TheJian - Monday, August 3, 2020 - link
ROFL. He said, PE, then you change topic to forward PE.At some point you have to grow NET INCOME too. You can get all the revenue you want, but if you make NOTHING on it, what is the point? At some point people will ask you to PROFIT or dump your shares like I did ;) At this point AMD has to prove more than "we can sell it cheap, get revenue and ACT like we make money"...Not good enough after 3 years of this crap. If you're not making .5B GROSS PROFIT on 1B revenue, you're doing it WRONG in this business and the best do far better than that (see Intel, while losing share, losing fab war, etc, STILL understand how to MAKE MORE NET INCOME). Wake me when AMD figures it out.
Making 150mil on 2B is FAILING. PERIOD. There are NO GUARANTEES when the other guy can simply PAY people with 23B NET INCOME he earned to people YOU need to survive. IE, a simple extra 10% maybe (use your imagination, it's called contra revenue etc) for the same fab wafers as AMD, and promise even more crap every year you ignore AMD, what do you think TSMC does? They don't care who wins, only who makes them more money. They just make chips, they don't care WHO they are from, only who pays more to get them.
4.1B+ per year for 4yrs+ was BLOWN just to stay in mobile by Intel. In case you missed it, Intel was pissing away MORE than 1/2 of AMD's revenue...YEARLY...ROFL. You don't think they will allow the enemy to buy wafers they need do you? They already said ~7 new divisions now, each can decide what fab to go with, OTHER than Intel's.
Translation: All your fabs belong to US (Intel, well, ok, an apple...ROFL). WAFERS to the highest bidders. DUH. Guarantee...LOL. Until person X writes a larger check than person Y. If AMD doesn't have a contract for years, the party is OVER. Intel and even NV can pay MUCH more per wafer start than AMD. Rumors are abound about the fighting over them already, and Intel supposedly already has an agreement with TSMC now. Huawei gone sept, so I'm guessing Intel outbids anyone but Apple for that stuff, and since apple is on 5nm, It's Intel vs. everyone else for those wafers. I'm guessing Intel's 23B NET INCOME per year matters here. I guess you missed the part where AMD SOLD their fabs (and everything else they owned but IP, due to stupid LOW PRICING for decades). IF you don't OWN the fab, anyone can ruin your day with a check (use your imagination, 100 ways to pay you without a check too, easily made legal). How do you stop someone from paying MORE than you can afford for a wafer?
What is AMD's option here, take or pay agreement again on wafers? How many years did it take to get out of that from GF? It cost them billions for that mistake. Intel put the fear of god in everyone (board makers couldn't put their name on mboards, Asus sold boards in whiteboxes etc, PC makers threatened, stores, etc etc), and AMD got stuck with years of wafers they couldn't use. Read The Art of War (or heck, Trump's Art of the Deal), and maybe you'll get how warfare works (business is ruthless, China reads trump's book...LOL, odd USA isn't using it in all schools).
The best product doesn't always win. Just ask Netscape, Beta, etc etc :) Intel can piss away another 4B a year just to stop you from getting wafers easily and still make nearly 20B NET. What happens with 5nm TSMC server chips from Intel? You are not listening to INTEL telling you what is coming. OTHER FABS making INTEL chips. Do the math.
rahvin - Monday, August 3, 2020 - link
Price to Earnings ratio has two equally valid calculations. Trailing P/E is based on the previous years earnings while forward P/E is based on the projected earnings. The trailing P/E is the one typically calculated because it's not based on projections. But in the case of AMD the trialing PE is distorted because of the cyrto-coin GPU market distortion and as a result the trialing P/E IMO is not a reliable indicator until that distortion is no longer in the calculation. Good investors investigate the basis of numbers before accepting them.You then spend a couple paragraphs scattershoting various topics that appear to be a stream of consciousness that makes little sense. If you'd like a reply please post something coherent.
AMD remains a very good investment, they have significant potential to not only hit their growth targets but exceed them, as they've done for the last several years, especially given that Intel continues to miss their goals and struggle with manufacturing.
guachi - Wednesday, July 29, 2020 - link
I can't believe I'll be able to retire much more comfortably in 2 years in large part to AMD's massive stock price increase on the last 5 years.ads295 - Wednesday, July 29, 2020 - link
Capitalism and human non-rationality FTW.beginner99 - Wednesday, July 29, 2020 - link
I'm still surprised how small their server cpu sales are compared to client. I guess consumers are more aware to performance/$ than corporations...WaltC - Wednesday, July 29, 2020 - link
Companies tend upgrade servers slowly...;)Spunjji - Wednesday, July 29, 2020 - link
There's a much, much longer lag between good hardware arriving and it being put into play on the server side of the equation.They have to make the CPUs, then the OEMs have to integrate them, then the resellers have to change their ideas about what products to sell, then the IT managers have to convince the people holding the purse strings to change architecture, then the solutions have to be designed... it's a slow process, end-to-end.
Compared to "shiny new product released, build self new PC" it's positively glacial.
rahvin - Monday, August 3, 2020 - link
It's not just those reasons, your typical enterprise that purchases large volumes likes to test new systems before deploying them. I'd argue your average enterprise server purchase is at least a year after commercial availability because that's as fast as they can test and deploy. But the cheaper cost of AMD is definitely having an impact.vladx - Wednesday, July 29, 2020 - link
Just got my new laptop with i7-10875H which beats every Ryzen mobile CPU except the 4900H after some undervolting. Maybe next time, AMD.aebiv - Wednesday, July 29, 2020 - link
Good for you?vladx - Wednesday, July 29, 2020 - link
Just reminding everyone that contrary to what many people say on the internet, Ryzen 4000 series doesn't beat Intel 10th gen in performance both single-core and multi-core.