Discussion AWS Graviton2 64 vCPU Arm CPU Heightens War of Intel Betrayal

[exquisitechar]

https://www.servethehome.com/aws-graviton2-64-core-arm-cpu-heightens-war-of-intel-betrayal/

Pretty big deal for ARM in servers. Interested in seeing a comparison between this and Rome.

 

[Thala]

They took ARM's work on their design, licensed it and changed it.

That statement is fundamentally wrong. They did not take any design nor licensed it. With an architecture license you essentially gain the rights to develop your own design based on a reference specification delivered by ARM. There is NO design delivery by ARM that could potentially be changed.

This is in contrast to an IP/design license, where you actually get ARMs design and then doing modification if necessary.

 

[lobz]

X86 has control over servers like Italy has control over corona-virus.
Actual numbers looks good but they don't see what a disaster is coming.
Just imagine AMD wouldn't cancel K12 - Today they would have the best ARM core on market at almost perfect time. Shame on them.

You really do like to invent new ways to put your nonsense in more and more ridiculous contexts every single day, don't you?

 

[beginner99]

Today they would have the best ARM core on market

...that no OEM wants to sale and no costumer has an interest in. AMD has troubles getting their EPYC x86 into OEMs. How you think there chance of success with an ARM CPU would be? Let me help, you. ZERO.

 

[Richie Rich]

No matter how much I would have liked to get a K12, AMD did the right thing back then, they were too weak to track two targets, and x86 still is the preferred solution for servers and end user machines.

Edit: note AMD also failed terribly with their Cortex-A57 based chip which reinforces their decision to cancel K12.

As Steve Jobs said: "Go where the puck is going and not where it is right now (Wayne Gretzky)"
And this is exactly what Amazon is doing - They think in long term.
Now AMD has a bunch of money and still betting at dying horse. Not sure how long this gonna work. With new cores based on ARMv9, SVE2 the situation will become much harder for x86 than now with relatively small and weak A76.

 

[Richie Rich]

...that no OEM wants to sale and no costumer has an interest in. AMD has troubles getting their EPYC x86 into OEMs. How you think there chance of success with an ARM CPU would be? Let me help, you. ZERO.

I agree partly. But instead beating Intel by 10% better efficiency in x86 yard AMD would be able to smash Intel with 2x as good efficiency with K12. Such a huge advantage could bring much more customers than just x86 ISA legacy (and this will become even stronger in time). But who knows, maybe Zen3 will bring double IPC with double efficiency and x86 will enter smart phone market 🙂

 

[amrnuke]

As Steve Jobs said: "Go where the puck is going and not where it is right now (Wayne Gretzky)"
And this is exactly what Amazon is doing - They think in long term.
Now AMD has a bunch of money and still betting at dying horse. Not sure how long this gonna work. With new cores based on ARMv9, SVE2 the situation will become much harder for x86 than now with relatively small and weak A76.

I agree with Steve Jobs on that point, and appreciate Amazon pursuing ARM. However, AMD compared to Apple/Amazon is apples to oranges. They're in completely different places. Apple and Amazon created new markets and grew on the back of that.

AMD don't have the free cash to run a full court press on both x86 (which is necessary today) and ARM (which may be necessary tomorrow). If AMD backed out of x86 to devote resources to ARM, they'd lose a significant amount of revenue in an area where they have a technology lead. I don't think that's prudent.

Regarding Amazon, sure, they seem to be thinking long-term with servers, which is good, because their online marketplace for selling stuff is turning into a thrift store.

X86 has control over servers like Italy has control over corona-virus.

Wow.

 

[DrMrLordX]

Amazon has boatloads of money. They can afford to develop ARM CPUs/SoCs internally to avoid vendor lock-in from other companies, and they can develop their entire toolchain internally (and exploit existing Linux software). They're really that big. I'm still not 100% sold on Graviton/Graviton2 being the best use of their money, but I think I understand some of why they went that route anyway.

AMD isn't like that.

Down the road, the existence of Graviton2 instances on AWS will definitely change the server software landscape, and that may be the single best thing for ARM in the server that ever happened. Better than ThunderX2, better than Taishan, better than anything from Ampere.

 

[amrnuke]

Amazon has boatloads of money. They can afford to develop ARM CPUs/SoCs internally to avoid vendor lock-in from other companies, and they can develop their entire toolchain internally (and exploit existing Linux software). They're really that big. I'm still not 100% sold on Graviton/Graviton2 being the best use of their money, but I think I understand some of why they went that route anyway.

AMD isn't like that.

Down the road, the existence of Graviton2 instances on AWS will definitely change the server software landscape, and that may be the single best thing for ARM in the server that ever happened. Better than ThunderX2, better than Taishan, better than anything from Ampere.

If one ever wonders how big Amazon are...

AMD had a nice Q42019, with a net income of $170 million. If they repeated that, with 10% year over year growth in earnings, for 21 years straight, and they didn't spend any of that profit, they STILL wouldn't have as much money as Amazon has on hand right now.

Amazon's cash on hand is more than the GDPs of Bahrain, Iceland, Bolivia, etc.

 

[beginner99]

I agree partly. But instead beating Intel by 10% better efficiency in x86 yard AMD would be able to smash Intel with 2x as good efficiency with K12. Such a huge advantage could bring much more customers than just x86 ISA legacy (and this will become even stronger in time). But who knows, maybe Zen3 will bring double IPC with double efficiency and x86 will enter smart phone market

Yeah now AMD will make some cash and can afford to look into a k12 successor without risking the company. What AMD has shown in the last 1-2 decades is that it is too small of a company to influence the market. Intel and NV betting on "current" tech always won. First real quad-core? Didn't help AMD that much. The whole bulldozer fiasco. The uArch and idea behind it sounded reasonable on paper but it doesn't work unless you can provide the whole ecosystem to go with it. GCN with hardware scheduler and compute features? Didn't help. AMD was too early with that. Actually AMD was also first to tesselation. But then no one used it and later they gimped on it.

AMD is too financially constraint and too small to be a leader into a new "era".

And then performance, performance/watt and performance/$ isn't everything. Else intel should be going to 0% server market share in the next couple months. Because the will loose out on all of these already to Rome let alone Milan. You know what matters to corporation? Policy. And if policy is to buy server type X from OEM Y, then that is what you get. Because you can get mass rebate and have to train your people on 1 server. In fact since you can train them you can use "monkeys" to operate them instead of more expensive knowledgeable people that cost a lot more than power use.
(I'm currently fighting to get an AMD based server for my needs with corporate IT, I think I will get there but means dealying the whole thing for months...)

And then there also is the software. You know also what is policy in many corporate envs? Yes. Windows Server. Same reason. Monekys can work better with a GUI they already know than a cli. Yeah Amazon and co. buy a huge amount of server but so do all other big corps around the world combined. Much more. It's all the obscure intranet stuff no one outside actually knows about. Often this stuff is very old, internal legacy things. (when we moved from winxp to win7, everyone was moved to win7 32-bit due to 16-bit legacy apps). Only 1 year ago was I able to move to 64-bit win10. the big corporate world in non-tech is weird in that way and upgrading all your server stuff to work on ARM? Not going to happen. These people don't take risks. Why should they? Working with high power use is better than not working or delaying for months.

They can afford to develop ARM CPUs/SoCs internally to avoid vendor lock-in from other companies,

Exactly. K12 wouldn't have changed anything about this. It would still be vendor lock in. Amazon didn't create an ARM cpu because ARM is so cool. They did because they can't make an x86 one due to licensing. It's that simple.

 

[Richie Rich]

Down the road, the existence of Graviton2 instances on AWS will definitely change the server software landscape, and that may be the single best thing for ARM in the server that ever happened. Better than ThunderX2, better than Taishan, better than anything from Ampere.

This is an excellent point. No need to convince server providers to buy ARM bases HW. It's ready to deploy SW. This and Apple's ARM MacBook will start SW transition to ARM during 2020 and 2021. This gonna be huge market game change IMHO.

Wow.

It was joke about the corona-virus. However I will stop joking about that because the corona-virus death rate in Italy is actually horrible 26%. Assuming death after 7 days after detection, just compare WHO Situation Report no. 53 and 46, 1016 deaths per 3858 infected = 26.33%. USA has 24.32% and Spain 32.68%, China only 3.9%. Assuming death after 4 days after detection it's better Italy 14%, USA 17%, Spain 14%. However still pretty scary numbers. Sorry for the off-topic and I hope I'm wrong in my calculation.

 

[Andrei.]

It was joke about the corona-virus. However I will stop joking about that because the corona-virus death rate in Italy is actually horrible 26%. Assuming death after 7 days after detection, just compare WHO Situation Report no. 53 and 46, 1016 deaths per 3858 infected = 26.33%. USA has 24.32% and Spain 32.68%, China only 3.9%. Assuming death after 4 days after detection it's better Italy 14%, USA 17%, Spain 14%. However still pretty scary numbers. Sorry for the off-topic and I hope I'm wrong in my calculation.

This is not the place here in the CPU forum but wow talk about pulling figures out of your ass on such a topic? The confirmed cases are at 17660 with 1266 deaths, resulting in a 7% death rate, and you know there's tons more unconfirmed infected, bringing down the real percentage. Making up figures doesn't help your stance here.

 

[Richie Rich]

- DELETED POST and MOVED here if somebody interested in corona-virus numbers: https://forums.anandtech.com/threads/corona-virus-covid-19-predictions-is-it-real-disaster.2578219/
Sorry about that.






Three posts about corona virus by you.
Stop this off topic posting. Now.


esquared
Anandtech Forum Director

 

[Nothingness]

The whole bulldozer fiasco. The uArch and idea behind it sounded reasonable on paper but it doesn't work unless you can provide the whole ecosystem to go with it.

It never sounded reasonable. Sharing the front-end was a dumb idea IMHO.

And then there also is the software. You know also what is policy in many corporate envs? Yes. Windows Server. Same reason. Monekys can work better with a GUI they already know than a cli. Yeah Amazon and co. buy a huge amount of server but so do all other big corps around the world combined. Much more. It's all the obscure intranet stuff no one outside actually knows about. Often this stuff is very old, internal legacy things. (when we moved from winxp to win7, everyone was moved to win7 32-bit due to 16-bit legacy apps). Only 1 year ago was I able to move to 64-bit win10. the big corporate world in non-tech is weird in that way and upgrading all your server stuff to work on ARM? Not going to happen. These people don't take risks. Why should they? Working with high power use is better than not working or delaying for months.

I guess this varies a lot depending on what kind of job you have. I have always used UNIX then Linux for the last 25 years, in both small (<100) and large companies (>10k). All the servers I have interacted with run UNIX/Linux.

But yes I definitely agree with you that changing how things work in companies is a long and painful process.

Exactly. K12 wouldn't have changed anything about this. It would still be vendor lock in. Amazon didn't create an ARM cpu because ARM is so cool. They did because they can't make an x86 one due to licensing. It's that simple.

And because the ARM ecosystem is strong enough to make this viable even from a software point of view contrary to what many believe. It's obviously not reached the x86 maturity, but it's good enough for many tasks, both for server and end-user usages.

 

[lobz]

It was joke about the corona-virus. However I will stop joking about that(...)

Way to go being sensible 🙂 I give credit where it's due!

 

[Richie Rich]

It never sounded reasonable. Sharing the front-end was a dumb idea IMHO.

IMHO Bulldozer sounded reasonable when they kept everybody thinking there are 4xALUs (as an evolution from 3xALU K10). Like in Anandtech article from Nov30 2009: https://www.anandtech.com/show/2881/2

Before BD launch AMD revealed more detailed picture showing only 2xALU and since then everybody knew what disaster it's gonna be. Take a look at Anandtech article two years later on October 12 2011: https://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/2

Did somebody noticed that after Ampere Altra (A76/N1 based) is coming Mystique (much stronger A77) and Sirin (A78/Hercules)?

• 64-core A76 (Graviton2) is beating Zen1 in everything and put into trouble in some ways also Zen2 Rome
• 80-core A77 Mystique with +25% IPC jump will be clearly destroying Zen2 Rome systems and matching Zen3 64-core probably (while being cheaper and less power hungry). Assuming N7P TSMC process.
• 128-core A77 if Mystique will be manufactured on 5nm. It will smash Zen3 Milan systems. And all this one year before 5nm 96-core Zen4 enters market. Looks like X86 servers will be in real trouble before Nuvia enters market around 2024.

 

[DrMrLordX]

Haven't even seen Altra benched yet. Graviton2 looks great, but we haven't seen it head-to-head against last year's Rome so not sure what the breathless praise is for.

 

[coercitiv]

Graviton2 looks great, but we haven't seen it head-to-head against last year's Rome so not sure what the breathless praise is for.

Or the iPhone, why aren't we comparing it with Apple A13 like we do in every serious ARM vs. x86 server discussion? /s

 

[Thunder 57]

IMHO Bulldozer sounded reasonable when they kept everybody thinking there are 4xALUs (as an evolution from 3xALU K10). Like in Anandtech article from Nov30 2009: https://www.anandtech.com/show/2881/2

Before BD launch AMD revealed more detailed picture showing only 2xALU and since then everybody knew what disaster it's gonna be. Take a look at Anandtech article two years later on October 12 2011: https://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/2

Did somebody noticed that after Ampere Altra (A76/N1 based) is coming Mystique (much stronger A77) and Sirin (A78/Hercules)?

• 64-core A76 (Graviton2) is beating Zen1 in everything and put into trouble in some ways also Zen2 Rome
• 80-core A77 Mystique with +25% IPC jump will be clearly destroying Zen2 Rome systems and matching Zen3 64-core probably (while being cheaper and less power hungry). Assuming N7P TSMC process.
• 128-core A77 if Mystique will be manufactured on 5nm. It will smash Zen3 Milan systems. And all this one year before 5nm 96-core Zen4 enters market. Looks like X86 servers will be in real trouble before Nuvia enters market around 2024.

We get it. You love ARM, ALU's, and SMT4. x86 killed your loved ones, and you know better than some very smart people at all of these companies. Clearly they're being run by buffoons who know nothing. /s

 

[Nothingness]

Clearly they're being run by buffoons who know nothing. /s

That's not the place to discuss it but clearly some Intel management people were/are buffoons or at least they made some utterly stupid claims (and just plain lied with their process issues starting back to 14nm) and took wrong decisions.

 

[Richie Rich]

Or the iPhone, why aren't we comparing it with Apple A13 like we do in every serious ARM vs. x86 server discussion? /s

Because A13 Lightning core has about 4.5mm2 while Zen2 core has 3.8?mm2. It's IPC is huge but in the cost of die size.

• Cortex A76/N1 core used in Graviton2/Altra is about 1.4mm2 ...... compare to 3.8mm2 using Zen2
• Cortex A77 core next year server products has +17% transistors (1.6mm2) while delivering +25% IPC (beating Zen2 ST IPC and almost double MT IPC when Rome uses SMT2).

Less then half area size delivers IPC higher than Zen2 while using less than half power. What? Tell me how AMD and Intel could fight such a specs in large core count server CPUs. I see no way due to TDP limitation. x86 can hold HPC territory due to higher clocks. But then again only until Nuvia arrives with Apple's IPC.

In some ways this cheap ARM Neoverse uarch is even more dangerous for x86 than Apple's monster uarch.

 

[DrMrLordX]

Because A13 Lightning core has about 4.5mm2 while Zen2 core has 3.8?mm2.

I hope you realize you were deadpan responding to a sarcastic remark, right? Nobody in their right mind thinks that A13 belongs in any discussion about Altra or Graviton2 (except maybe in the context of "what if Apple made a server CPU" which by this point is overplayed).

 

[Richie Rich]

I hope you realize you were deadpan responding to a sarcastic remark, right? Nobody in their right mind thinks that A13 belongs in any discussion about Altra or Graviton2 (except maybe in the context of "what if Apple made a server CPU" which by this point is overplayed).

Yeah I know, I try to provoke them to get some numbers out of them. I hope somebody will prove me wrong but nobody delivers single digit. That's good sign. They deliver only whining, crying, sobbing, sarcasm and some old good hatred. This helps me to make a good decision for buying some stocks, especially now when everything is falling rapidly down, so you can double your money in few weeks.

 

[DrMrLordX]

Yeah I know, I try to provoke them to get some numbers out of them. I hope somebody will prove me wrong but nobody delivers single digit. That's good sign. They deliver only whining, crying, sobbing, sarcasm and some old good hatred. This helps me to make a good decision for buying some stocks, especially now when everything is falling rapidly down, so you can double your money in few weeks.

Nobody takes you seriously because A13 is not a server CPU. The ARM world is moving into the server business and they are not bringing Apple with them. They (the posters) are making fun of you for continuously hyping a very good mobile CPU. Not crying, sobbing, or hating.

The "numbers" are already available: Apple has yet to license any of their core designs to anyone who can or will put together a 64-or-more core design CPU. Huawei has done it, Ampere is doing it, Amazon has done it, and I'm sure I'm leaving out someone. Has Broadcom made it to 64c yet? Fujitsu is coming but their core is so niche that we may not hear much about it outside of some academic circles.

If you came to me two or more years ago and said, "A1x is a threat, Intel needs to watch out!" I might agree, assuming Apple was serious about rolling out big server CPUs based on their core designs with the proper I/O and interconnects. Apple has shown no interest. You can clearly see who are the players in the ARM server world moving forward. Sadly, we haven't yet seen any of these CPUs benched against Rome, and the benches we have for Graviton2 against Intel CPUs are kinda limited (though Cooper Lake isn't going to help Intel gain any ground).

Try to pay more attention to what is out on the market now or what is going to come out in the near future. Graviton2 is showing some promise. We may get more high-clockspeed ARM parts with many cores and capable interconnects that will do nicely in many server functions. Until then, people will continue poking fun at you over things like your sig.

 

[amrnuke]

Until then, people will continue poking fun at you over things like your sig.

And him adding coronavirus spread data to his sig...
He's literally trying to goad everyone into off-topic discussions. On topics he is clearly poorly-versed in (as a physician, his comments thus far are so factually incorrect as to be laugh-inducing, were it not for the severity of the condition).

So here's the real deal. Apple don't care about server market. Period.

I would agree with him, buy Apple stock. At P/E 19.6 it's a steal. But not because ARM is some magical unicorn that's going to dominate the world. It's because Apple have so many brand-loyal customers who are locked into their ecosystem.

 

[JasonLD]

Apple definitely isn't moving to their own ARM based processor unless they are ready to make a full transition within a calendar year. Mike Filippo hire was just last year, so at least I can see Apple is definitely trying to scale their processor beyond mobile, but I can't see that happening until 2022 at least.