Its a shame, during the Broadcom hostile takeover Jacobs and Steve were pressured by the Shareholders on how their current business model was better than that of the Tan's mountain farmer strategy which pays off well but only focuses on the short term benefits.
EEtimes rick mentioned how that failed strategy was and how their M&A scene is destroying the innovation. Apple was to be enjoying highest benefits if that deal went through, glad that it fell off I cannot imagine Snapdragon processors without CAF, which is what I guess would happen once Broadcom gets hold of their SD platform.
But the jury is out, Jacobs left the company, they are releasing the 845 Source so late excusing the leak of the 6xx and 7xx platform in the code. The Centriq was the showcase of the Qcomms strategy of long term benefits and they had great progress, guess that's the reason why 835 and 845 don't have full custom core uarch like Kryo from 82x..and the recent trade war and dependency on China's MOFCOM for the NXP deal and how Huawei grips them for the 5G game. Ironically US govt loves M&A game with strong WallSt backing for Tan. But couldn't proceed due the Penetration of PRC - clue is Danhua Capital. And the latest news of CFIUS extension of reach bill.
Qualcomm is cornered and feeling the heat, thus they had to deal with Samsung again for their fabrication and manufacturing. And not to forget the blaze from Apple's predatory practice of shutting the small frys slowly (GTAdvanced, Imagination Tech)..
>And not to forget the blaze from Apple's predatory practice of shutting the small frys slowly (GTAdvanced, Imagination Tech)..
Both results were their own making
GTAdvanced was basically a company that wanted Apple business not matter what, even though they knew, according to their knowledge technically impossible.
Imagination Tech : No one uses PowerVR, they felt asleep for far too long. Drivers on Linux, Android were barely adequate. Bought MIPS which was idiotic, their wireless department were not performing, and did not even bother trying to get MediaTek on board after Apple has made the move.
Jacobs's move were more to do with taking the company private. It was mentioned as better for National Securities and Interest. And he is doing it while starting his own company XCom.
I am still now sure, or see a clear way how this Apple and Qualcomm turns out. It seems silly to reduce its IP price just for Apple. But I also don't see Qualcomm lower its IP prices as it has market wide implications. Qualcomm CEO said it should be resolved within 6 - 9 months time. I am waiting to see this.
Huh? You're trying to blame the failure of Centriq on Broadcom? Cornered and feeling the heat and so had to deal with Samsung? What are you talking about? What's Imagination Technologies have to do with server chips? Even in mobile GPUs, how is Qualcomm hurt by the loss of a competitor? Or are we supposed to interpret Qualcomm as a "small fry"? Qualcomm has been a bigger bully than Apple.
Qualcomm's core competency is cellular technology. They are the best in the world at it. They muscled their way into the mobile SOC business through their cellular IP. But they had no way to similarly muscle their way into the data center. It's not Broadcom's fault there isn't great interest in their server chips. Breaking into the server market is hard.
On the include link it states "He was previously Distinguished Engineer at Microsoft,"
Well not sure if he was involved with Microsoft Windows on ARM at Qualcomm, but it could be that this is sign that Microsoft Windows on ARM is going to be huge failure.
One interesting a wild thought, this is so soon after Intel's CEO leaving - but I don't believe this guy has the qualifications of replacing the CEO at Intel.
Seems like they're giving up the future ARM server market before really trying to create it - yet spent most of the R&D for it. Is the product really this bad?
It's much cheaper to research a project than it is to actually try to bring it to market in volume. The product doesn't have to be really bad in order to stop short of pursuing it. It just has to be deemed to be financially undesirable. If they really want to sell the chip they will have to commit to years of support, that could place a drag on their margins for years to come.
The product could be great, it wouldn't matter. There simply isn't a good reason to replace x68 in the server space at this time. There are low price options coming from both Intel and AMD, the power efficiency of both puts them in the budget of a typical datacenter, and the software stack is mature and performant.
ARM in servers is having the same problem Intel has had getting x86 in phones.
It's been out for more than two years now and I'm not aware of it even hitting 1% of the server market, even if you try to choose a specific niche. I'm not certain what your definition of 'seeing traction' is, but I'd expect to see it show up in market share charts which to date it does not seem to have. If it has somewhere, please let me know as I'm honestly curious.
I've also noticed that most of the news around Cavium's server products has gone dark since the Marvell acquisition, I'm honestly not certain they are focused on that market or considered it a major part of the assets when they acquired it.
Sorry, I was looking at the first gen. For the second gen, I'm still not seeing traction in terms of marketshare. The links provided are PR announcements, not marketshare metrics and may or may not get actually built (lots of stuff like that gets announced then never gets funded). Also, the number of actual servers being built appears to be quite small per those links, a 4096 CPU cluster is not all that impressive when its a 48 core chip..
Again, not saying its dead or anything, but this is a long way from what I'd consider to be significant traction. It looks exploratory more than anything else.
Well it was just fully released in May. You won't see market share, yet. You also won't see much for a while. The articles I posted aren't just PR they show two things about ThunderX2 that can't be said for Centriq. Firstly, that it is being put into HPC machines. Of course, Centriq isn't really HPC capable, but the point is that being put into actual machines, especially the 2.3 petaflops machine, is significant. Secondly, server makers are actually developing products based on ThunderX2. I only posted that one workstation but I remember seeing a server announced. I don't recall seeing anything like that for Centriq.
I'm not sure what your definition of traction is. You seem to be looking for being on the road. Traction, to me, just means that it's not just spinning its wheels, which is what ARM server chips have been doing for the past 10 years. Note that in 2008 GPUs had very little real revenue for compute acceleration, yet it was clear that it was something being taken seriously.
As far as those HPC projects, they are already funded, they are being built. Sometimes they get canceled because of technology/component cost shortfalls and disagreements about how to deal with them. But it's not like weapons system development. You can be pretty sure they will be built. The 2.3 petaflops system will be delivered soon, in fact. According to the article it will be installed later this year. 2.3 petaflops would put it in the top 100 systems in the world in the latest (June) Top 500 list, btw.
I'm not sure how you are defining traction. Qualcomm actually had some reasonably large scale real world deployments (most prominently CloudFlare). I don't see anything like that, and today's article does not emphasize servers at all. If its too soon to say that my expectations are 'down the road' then its too soon to say that its gaining traction. At this point the deployments don't actually exist, its just announcements. Even if they did, they are such a minuscule part of the markets they are in that they can best be described as experiments.
I think i explained what I meant by traction. It's getting the tires moving instead of just spinning in place. It's not rolling down the highway.
Where has Qualcomm had large scale real world deplayoments? They had PR releases, exactly what you say these Cavium links are. The difference is that these are real, publicly verifiable systems. When Qualcomm and Microsoft or whoever else say that they are putting Centriq in the cloud that doesn't mean anything by itself. Even if they put them in publicly available instances. Take, for example, the AMD GPUs in.. was it AWS? I don't remember. Yes, that public cloud provider made public instances of AMD GPUs. But the number of them is almost meaningless compared to the number of NVIDIA GPU instances. Look at it from a business perspective in order to gain an insight on what it means. What these cloud providers want, most of all, is pricing power against Intel. That's why Microsoft and whoever else encouraged Qualcomm to pursue an ARM server chip. And making announcements about these chips is part of that pricing pressure. But real world HPC shops making real world systems has nothing to do with such a game. It definitely shows real world interest in the chips. Now, maybe there is some real world interest in the Centriq, but I personally haven't seen it yet.
As far as the deployments don't exist. You are wrong. They do exist. A 2.3 petaflops HPC machine isn't ordered and built over night. It was started years ago in some form or another. It will be delivered in the next few months.
CloudFlare deployed as did a few other edge providers. It wasn't enough and it was a lot more than what Cavium has done so far. Your definition of 'gaining traction' appears to be "announced a product and a single customer"....which I guess is some motion but seriously virtually every product gets that far.
"There simply isn't a good reason to replace x68 in the server space at this time."
Yes there is. Or rather ARE. System power/performance for one. Core density is another. Modern x86 is not x86 or even 486 at all. The instruction set is so overloaded with "sometimes useful for some applications" stuff, the decoder has to be huge, meaning everything else does not scale down either without price/performance hit. For example, you practically have to have speculative OoO. Meaning the cores have to be big and inefficient to hit decent price/performance levels. ARM (v8+) suffers from the same problem but not nearly as much, especially if you throw out all the 32-bit stuff as those server processors do.
The decoder hit for x86 is estimated to be around 1%. It's not a significant factor. ARM also uses speculative OoO (which is why Meltdown/Spectre hit ARM as well), it has nothing to do with how complex a core is and is about ensuring all cores are kept as busy as possible regardless of how poorly optimized the code is.
The dirty secret is that ARM is not magically violating physics. As the capability of ARM based products have improved, price/performance has scaled to get closer and closer to Intel. At the same performance level in the same tasks they roughly use the same amount of power, assuming equally optimized code. Both architectures are very optimized for the tasks they perform, and as they add capabilities, core complexity, die size and power consumption rise, with die shrinks and power gating utilized to keep that rise from being linear.
Why has Cavium’s ThunderX2 got traction, and not Qualcomm? Dual Socket Support?
Anyway, Qualcomm's initial projection, and the revised projection, and now the 3rd and latest projection of ARM Server market shares has all but failed. And Given any Centriq deals for 2018 is done, and for most part 2019 deal has nothing much left.
Cavium's current ThunderX2 is positioned in the compute segment, this Qualcomm chip is more in the storage segment. That could have something to do with it. Cavium also has the advantage of already being in the server components market. Qualcomm just had ties with Microsoft. Unless Microsoft was so impressed to buy lots of their chips they have to convince others they are willing and capable to serve the segment long term.
A report came out that Centriq was in trouble and Qualcomm denied it. Now this. Too many companies just lie, lie, lie to their investors. Next thing the head of Intel's 5G modem project is gonna be resigning.
Despite the layoffs, they're still retaining a large number of employees in that division, and presumably the 2nd gen server part (worked on by the full staff) has taped by now out so they will have a good follow up part to release soon.
The latest comments from Cristiano Amon which come after both Chandrasekher and Bhandarkar’s departure indicate that they're folding the server unit into their cellphone SoC division:
"Amon told Reuters that the server chip business will be rolled into Qualcom’s Qualcomm CDMA Technologies unit, which designs and sells mobile phone chips, to gain cost efficiencies."
This makes sense given the high projected performance of the A76: a core design suitable for a high end cellphone may do very well in an energy efficient server design so long as it has a good un-core architecture.
Notably, the company plans to pursue deployment of Centriq to cell towers as edge compute:
"The company sees a future for its server efforts as higher-powered computing is added to cellular base stations. Those crucial parts of cellular networks will get chips that allow them to offer processing power to phones over wireless links. That capability will enable virtual reality and augmented reality in smartphones and offer an experience that’s comparable to those now only provided by personal computers, Amon said."
"Despite the layoffs, they're still retaining a large number of employees in that division, and presumably the 2nd gen server part (worked on by the full staff) has taped by now out so they will have a good follow up part to release soon.
The latest comments from Cristiano Amon which come after both Chandrasekher and Bhandarkar’s departure indicate that they're folding the server unit into their cellphone SoC division:"
I think the conclusion we can make, then, is that they will be moving those server SOC engineers back to working on custom mobile SOCs.
"Notably, the company plans to pursue deployment of Centriq to cell towers as edge compute:"
They are having trouble competing with Cavium in a segment new to both of them but they will be successful competing against Cavium's bread and butter? We'll see.
Their goal is to commoditize compute and wean the industry off of the x86 monopoly over several generations. So long as the industry uses x86, they will be under the yoke of Intel's restrictive licensing policies as well as their non-engineering tactics for suppressing competition. Having only AMD as an alternative is not enough to avoid these. I think the conclusion we can make is closer to what was said in the article:
"Amon said the restructured server chip unit will focus on large cloud computing players. He would not name U.S. targets, but said the company is hoping to sell to Chinese internet giants such as Alibaba Group Holding Ltd, Tencent Holdings Ltd and Baidu Inc through a joint venture in China.
Large internet companies tend to write much of their own internal software and can customize it for Qualcomm’s chips. That means they will not struggle like smaller players to adapt off-the-shelf data center software that is overwhelmingly written to run on Intel’s so-called x86 chips.
'It’s very clear to us that the ARM opportunity is focused on a few players where you don’t have the software x86 barrier to entry,' Amon said."
Thunder X2 targets higher performance than the Centriq but doesn't offer any advantages to owners over having x86 chips with comparable or higher power consumption and the same need for recompilation. The main thing customers may be looking for is having some leverage over Intel when it comes to pricing in supporting an alternative ecosystem when the parts are finally viable for real use.
Centriq addresses the most major issues for TCO over the lifetime of a server: power and cooling and it is in a different league in terms of power efficiency. Also, having cheap and cheerful compute in towers hasn't been a pressing need prior to 5G after which the amount of data available to process will increase substantially: it could enable light weight AR glasses by shifting compute to edge servers or augment autonomous vehicle sensor arrays by processing information from environmental cameras which can see LIDAR opaque dead spots etc.
"Their goal is to commoditize compute and wean the industry off of the x86 monopoly over several generations. So long as the industry uses x86, they will be under the yoke of Intel's restrictive licensing policies as well as their non-engineering tactics for suppressing competition."
I'm pretty sure you're projecting here. There goal is to try to take advantage of market opportunities that provide better return on investment than other options.
"Large internet companies tend to write much of their own internal software and can customize it for Qualcomm’s chips"
Yes. This is a statement from 10 years ago. But the fact is that the only ARM player who has shown any traction in the server market since than has been Cavium very recently. So if these companies are willing to do it now why does Centriq seem to be nowhere to be found even though Qualcomm has such huge capital resources compared to Cavium?
"Thunder X2 targets higher performance than the Centriq but doesn't offer any advantages to owners over having x86 chips with comparable or higher power consumption and the same need for recompilation."
How much it will cause people to switch, we'll have to see.
"Also, having cheap and cheerful compute in towers hasn't been a pressing need prior to 5G after which the amount of data available to process will increase substantially:"
But Cavium already has a business providing compute for software defined networks in base towers. If there is even more compute required who do you think companies will go with to add more? Someone they already have a relationship with for the core part of their business who also happens to have more compute available, or a complete newcomer?
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
34 Comments
Back to Article
Quantumz0d - Friday, July 6, 2018 - link
Its a shame, during the Broadcom hostile takeover Jacobs and Steve were pressured by the Shareholders on how their current business model was better than that of the Tan's mountain farmer strategy which pays off well but only focuses on the short term benefits.EEtimes rick mentioned how that failed strategy was and how their M&A scene is destroying the innovation. Apple was to be enjoying highest benefits if that deal went through, glad that it fell off I cannot imagine Snapdragon processors without CAF, which is what I guess would happen once Broadcom gets hold of their SD platform.
But the jury is out, Jacobs left the company, they are releasing the 845 Source so late excusing the leak of the 6xx and 7xx platform in the code. The Centriq was the showcase of the Qcomms strategy of long term benefits and they had great progress, guess that's the reason why 835 and 845 don't have full custom core uarch like Kryo from 82x..and the recent trade war and dependency on China's MOFCOM for the NXP deal and how Huawei grips them for the 5G game. Ironically US govt loves M&A game with strong WallSt backing for Tan. But couldn't proceed due the Penetration of PRC - clue is Danhua Capital. And the latest news of CFIUS extension of reach bill.
Qualcomm is cornered and feeling the heat, thus they had to deal with Samsung again for their fabrication and manufacturing. And not to forget the blaze from Apple's predatory practice of shutting the small frys slowly (GTAdvanced, Imagination Tech)..
Let's see how it pans out..
Quantumz0d - Friday, July 6, 2018 - link
Edit - The reason of 835, 845 not having is because they moved a lot of uarch engineering to this Centriq division for temporary.iwod - Friday, July 6, 2018 - link
>And not to forget the blaze from Apple's predatory practice of shutting the small frys slowly (GTAdvanced, Imagination Tech)..Both results were their own making
GTAdvanced was basically a company that wanted Apple business not matter what, even though they knew, according to their knowledge technically impossible.
Imagination Tech : No one uses PowerVR, they felt asleep for far too long. Drivers on Linux, Android were barely adequate. Bought MIPS which was idiotic, their wireless department were not performing, and did not even bother trying to get MediaTek on board after Apple has made the move.
Jacobs's move were more to do with taking the company private. It was mentioned as better for National Securities and Interest. And he is doing it while starting his own company XCom.
I am still now sure, or see a clear way how this Apple and Qualcomm turns out. It seems silly to reduce its IP price just for Apple. But I also don't see Qualcomm lower its IP prices as it has market wide implications. Qualcomm CEO said it should be resolved within 6 - 9 months time. I am waiting to see this.
peevee - Friday, July 6, 2018 - link
What is "Tan's mountain farmer strategy "?Quantumz0d - Sunday, July 8, 2018 - link
https://www.eetimes.com/author.asp?section_id=36&a...Read the last paragraph of Page 1.
Yojimbo - Saturday, July 7, 2018 - link
Huh? You're trying to blame the failure of Centriq on Broadcom? Cornered and feeling the heat and so had to deal with Samsung? What are you talking about? What's Imagination Technologies have to do with server chips? Even in mobile GPUs, how is Qualcomm hurt by the loss of a competitor? Or are we supposed to interpret Qualcomm as a "small fry"? Qualcomm has been a bigger bully than Apple.Qualcomm's core competency is cellular technology. They are the best in the world at it. They muscled their way into the mobile SOC business through their cellular IP. But they had no way to similarly muscle their way into the data center. It's not Broadcom's fault there isn't great interest in their server chips. Breaking into the server market is hard.
Raqia - Friday, July 6, 2018 - link
This had been on his LinkedIn since March: https://www.linkedin.com/in/dileep-bhandarkar-a724...HStewart - Friday, July 6, 2018 - link
On the include link it states "He was previously Distinguished Engineer at Microsoft,"Well not sure if he was involved with Microsoft Windows on ARM at Qualcomm, but it could be that this is sign that Microsoft Windows on ARM is going to be huge failure.
Ian Cutress - Friday, July 6, 2018 - link
He was purely server at QCT.HStewart - Friday, July 6, 2018 - link
One interesting a wild thought, this is so soon after Intel's CEO leaving - but I don't believe this guy has the qualifications of replacing the CEO at Intel.MrSpadge - Friday, July 6, 2018 - link
Seems like they're giving up the future ARM server market before really trying to create it - yet spent most of the R&D for it. Is the product really this bad?Yojimbo - Saturday, July 7, 2018 - link
It's much cheaper to research a project than it is to actually try to bring it to market in volume. The product doesn't have to be really bad in order to stop short of pursuing it. It just has to be deemed to be financially undesirable. If they really want to sell the chip they will have to commit to years of support, that could place a drag on their margins for years to come.Reflex - Saturday, July 7, 2018 - link
The product could be great, it wouldn't matter. There simply isn't a good reason to replace x68 in the server space at this time. There are low price options coming from both Intel and AMD, the power efficiency of both puts them in the budget of a typical datacenter, and the software stack is mature and performant.ARM in servers is having the same problem Intel has had getting x86 in phones.
Yojimbo - Sunday, July 8, 2018 - link
"The product could be great, it wouldn't matter. There simply isn't a good reason to replace x68 in the server space at this time."Then why is ThunderX2 seeing traction?
Reflex - Monday, July 9, 2018 - link
It's been out for more than two years now and I'm not aware of it even hitting 1% of the server market, even if you try to choose a specific niche. I'm not certain what your definition of 'seeing traction' is, but I'd expect to see it show up in market share charts which to date it does not seem to have. If it has somewhere, please let me know as I'm honestly curious.I've also noticed that most of the news around Cavium's server products has gone dark since the Marvell acquisition, I'm honestly not certain they are focused on that market or considered it a major part of the assets when they acquired it.
Yojimbo - Monday, July 9, 2018 - link
"It's been out for more than two years now"No it hasn't. It's been out since May of this year: https://www.cavium.com/news/cavium-announces-thund...
Seeing traction:
https://www.top500.org/news/sandia-to-install-firs...
https://www.hpcwire.com/off-the-wire/bright-comput...
https://www.zdnet.com/article/caviums-thunderx2-pr...
https://www.servethehome.com/cavium-thunderx2-gain...
http://gw4.ac.uk/isambard/
Yojimbo - Monday, July 9, 2018 - link
Sorry, I put a period in front of the first link. https://www.cavium.com/news/cavium-announces-thund...Reflex - Monday, July 9, 2018 - link
Sorry, I was looking at the first gen. For the second gen, I'm still not seeing traction in terms of marketshare. The links provided are PR announcements, not marketshare metrics and may or may not get actually built (lots of stuff like that gets announced then never gets funded). Also, the number of actual servers being built appears to be quite small per those links, a 4096 CPU cluster is not all that impressive when its a 48 core chip..Again, not saying its dead or anything, but this is a long way from what I'd consider to be significant traction. It looks exploratory more than anything else.
Yojimbo - Monday, July 9, 2018 - link
Well it was just fully released in May. You won't see market share, yet. You also won't see much for a while. The articles I posted aren't just PR they show two things about ThunderX2 that can't be said for Centriq. Firstly, that it is being put into HPC machines. Of course, Centriq isn't really HPC capable, but the point is that being put into actual machines, especially the 2.3 petaflops machine, is significant. Secondly, server makers are actually developing products based on ThunderX2. I only posted that one workstation but I remember seeing a server announced. I don't recall seeing anything like that for Centriq.I'm not sure what your definition of traction is. You seem to be looking for being on the road. Traction, to me, just means that it's not just spinning its wheels, which is what ARM server chips have been doing for the past 10 years. Note that in 2008 GPUs had very little real revenue for compute acceleration, yet it was clear that it was something being taken seriously.
As far as those HPC projects, they are already funded, they are being built. Sometimes they get canceled because of technology/component cost shortfalls and disagreements about how to deal with them. But it's not like weapons system development. You can be pretty sure they will be built. The 2.3 petaflops system will be delivered soon, in fact. According to the article it will be installed later this year. 2.3 petaflops would put it in the top 100 systems in the world in the latest (June) Top 500 list, btw.
Reflex - Monday, July 9, 2018 - link
I'm not sure how you are defining traction. Qualcomm actually had some reasonably large scale real world deployments (most prominently CloudFlare). I don't see anything like that, and today's article does not emphasize servers at all. If its too soon to say that my expectations are 'down the road' then its too soon to say that its gaining traction. At this point the deployments don't actually exist, its just announcements. Even if they did, they are such a minuscule part of the markets they are in that they can best be described as experiments.It's pretty fair to say its too soon to call.
Yojimbo - Tuesday, July 10, 2018 - link
I think i explained what I meant by traction. It's getting the tires moving instead of just spinning in place. It's not rolling down the highway.Where has Qualcomm had large scale real world deplayoments? They had PR releases, exactly what you say these Cavium links are. The difference is that these are real, publicly verifiable systems. When Qualcomm and Microsoft or whoever else say that they are putting Centriq in the cloud that doesn't mean anything by itself. Even if they put them in publicly available instances. Take, for example, the AMD GPUs in.. was it AWS? I don't remember. Yes, that public cloud provider made public instances of AMD GPUs. But the number of them is almost meaningless compared to the number of NVIDIA GPU instances. Look at it from a business perspective in order to gain an insight on what it means. What these cloud providers want, most of all, is pricing power against Intel. That's why Microsoft and whoever else encouraged Qualcomm to pursue an ARM server chip. And making announcements about these chips is part of that pricing pressure. But real world HPC shops making real world systems has nothing to do with such a game. It definitely shows real world interest in the chips. Now, maybe there is some real world interest in the Centriq, but I personally haven't seen it yet.
As far as the deployments don't exist. You are wrong. They do exist. A 2.3 petaflops HPC machine isn't ordered and built over night. It was started years ago in some form or another. It will be delivered in the next few months.
Reflex - Tuesday, July 10, 2018 - link
CloudFlare deployed as did a few other edge providers. It wasn't enough and it was a lot more than what Cavium has done so far. Your definition of 'gaining traction' appears to be "announced a product and a single customer"....which I guess is some motion but seriously virtually every product gets that far.Yojimbo - Friday, July 13, 2018 - link
Uhh. I showed you 4 different HPC shops using the ThunderX2 in real machines. Almost no products get that far.peevee - Tuesday, July 10, 2018 - link
"There simply isn't a good reason to replace x68 in the server space at this time."Yes there is. Or rather ARE. System power/performance for one. Core density is another.
Modern x86 is not x86 or even 486 at all. The instruction set is so overloaded with "sometimes useful for some applications" stuff, the decoder has to be huge, meaning everything else does not scale down either without price/performance hit. For example, you practically have to have speculative OoO. Meaning the cores have to be big and inefficient to hit decent price/performance levels.
ARM (v8+) suffers from the same problem but not nearly as much, especially if you throw out all the 32-bit stuff as those server processors do.
Reflex - Tuesday, July 10, 2018 - link
The decoder hit for x86 is estimated to be around 1%. It's not a significant factor. ARM also uses speculative OoO (which is why Meltdown/Spectre hit ARM as well), it has nothing to do with how complex a core is and is about ensuring all cores are kept as busy as possible regardless of how poorly optimized the code is.The dirty secret is that ARM is not magically violating physics. As the capability of ARM based products have improved, price/performance has scaled to get closer and closer to Intel. At the same performance level in the same tasks they roughly use the same amount of power, assuming equally optimized code. Both architectures are very optimized for the tasks they perform, and as they add capabilities, core complexity, die size and power consumption rise, with die shrinks and power gating utilized to keep that rise from being linear.
iwod - Friday, July 6, 2018 - link
Why has Cavium’s ThunderX2 got traction, and not Qualcomm? Dual Socket Support?Anyway, Qualcomm's initial projection, and the revised projection, and now the 3rd and latest projection of ARM Server market shares has all but failed. And Given any Centriq deals for 2018 is done, and for most part 2019 deal has nothing much left.
Yojimbo - Saturday, July 7, 2018 - link
Cavium's current ThunderX2 is positioned in the compute segment, this Qualcomm chip is more in the storage segment. That could have something to do with it. Cavium also has the advantage of already being in the server components market. Qualcomm just had ties with Microsoft. Unless Microsoft was so impressed to buy lots of their chips they have to convince others they are willing and capable to serve the segment long term.peevee - Friday, July 6, 2018 - link
It is "Chandrasekher", not "Chandrasekhar".Ryan Smith - Friday, July 6, 2018 - link
Ack! There's an embarrassing error.You are correct. The article has been fixed.
Yojimbo - Saturday, July 7, 2018 - link
A report came out that Centriq was in trouble and Qualcomm denied it. Now this. Too many companies just lie, lie, lie to their investors. Next thing the head of Intel's 5G modem project is gonna be resigning.Raqia - Saturday, July 7, 2018 - link
Despite the layoffs, they're still retaining a large number of employees in that division, and presumably the 2nd gen server part (worked on by the full staff) has taped by now out so they will have a good follow up part to release soon.The latest comments from Cristiano Amon which come after both Chandrasekher and Bhandarkar’s departure indicate that they're folding the server unit into their cellphone SoC division:
"Amon told Reuters that the server chip business will be rolled into Qualcom’s Qualcomm CDMA Technologies unit, which designs and sells mobile phone chips, to gain cost efficiencies."
https://www.reuters.com/article/us-qualcomm-chips/...
This makes sense given the high projected performance of the A76: a core design suitable for a high end cellphone may do very well in an energy efficient server design so long as it has a good un-core architecture.
Notably, the company plans to pursue deployment of Centriq to cell towers as edge compute:
"The company sees a future for its server efforts as higher-powered computing is added to cellular base stations. Those crucial parts of cellular networks will get chips that allow them to offer processing power to phones over wireless links. That capability will enable virtual reality and augmented reality in smartphones and offer an experience that’s comparable to those now only provided by personal computers, Amon said."
https://www.bloomberg.com/news/articles/2018-06-14...
This makes sense as major telecoms are planning to replace more traditional hardware with whiteboxes utilizing software defined networks:
http://about.att.com/story/att_deploying_white_box...
Yojimbo - Sunday, July 8, 2018 - link
"Despite the layoffs, they're still retaining a large number of employees in that division, and presumably the 2nd gen server part (worked on by the full staff) has taped by now out so they will have a good follow up part to release soon.The latest comments from Cristiano Amon which come after both Chandrasekher and Bhandarkar’s departure indicate that they're folding the server unit into their cellphone SoC division:"
I think the conclusion we can make, then, is that they will be moving those server SOC engineers back to working on custom mobile SOCs.
"Notably, the company plans to pursue deployment of Centriq to cell towers as edge compute:"
They are having trouble competing with Cavium in a segment new to both of them but they will be successful competing against Cavium's bread and butter? We'll see.
Raqia - Sunday, July 8, 2018 - link
Their goal is to commoditize compute and wean the industry off of the x86 monopoly over several generations. So long as the industry uses x86, they will be under the yoke of Intel's restrictive licensing policies as well as their non-engineering tactics for suppressing competition. Having only AMD as an alternative is not enough to avoid these. I think the conclusion we can make is closer to what was said in the article:"Amon said the restructured server chip unit will focus on large cloud computing players. He would not name U.S. targets, but said the company is hoping to sell to Chinese internet giants such as Alibaba Group Holding Ltd, Tencent Holdings Ltd and Baidu Inc through a joint venture in China.
Large internet companies tend to write much of their own internal software and can customize it for Qualcomm’s chips. That means they will not struggle like smaller players to adapt off-the-shelf data center software that is overwhelmingly written to run on Intel’s so-called x86 chips.
'It’s very clear to us that the ARM opportunity is focused on a few players where you don’t have the software x86 barrier to entry,' Amon said."
Thunder X2 targets higher performance than the Centriq but doesn't offer any advantages to owners over having x86 chips with comparable or higher power consumption and the same need for recompilation. The main thing customers may be looking for is having some leverage over Intel when it comes to pricing in supporting an alternative ecosystem when the parts are finally viable for real use.
Centriq addresses the most major issues for TCO over the lifetime of a server: power and cooling and it is in a different league in terms of power efficiency. Also, having cheap and cheerful compute in towers hasn't been a pressing need prior to 5G after which the amount of data available to process will increase substantially: it could enable light weight AR glasses by shifting compute to edge servers or augment autonomous vehicle sensor arrays by processing information from environmental cameras which can see LIDAR opaque dead spots etc.
Yojimbo - Monday, July 9, 2018 - link
"Their goal is to commoditize compute and wean the industry off of the x86 monopoly over several generations. So long as the industry uses x86, they will be under the yoke of Intel's restrictive licensing policies as well as their non-engineering tactics for suppressing competition."I'm pretty sure you're projecting here. There goal is to try to take advantage of market opportunities that provide better return on investment than other options.
"Large internet companies tend to write much of their own internal software and can customize it for Qualcomm’s chips"
Yes. This is a statement from 10 years ago. But the fact is that the only ARM player who has shown any traction in the server market since than has been Cavium very recently. So if these companies are willing to do it now why does Centriq seem to be nowhere to be found even though Qualcomm has such huge capital resources compared to Cavium?
"Thunder X2 targets higher performance than the Centriq but doesn't offer any advantages to owners over having x86 chips with comparable or higher power consumption and the same need for recompilation."
Then why is there actual interest in it? It does offer advantages. https://www.nextplatform.com/2018/05/16/getting-lo...
How much it will cause people to switch, we'll have to see.
"Also, having cheap and cheerful compute in towers hasn't been a pressing need prior to 5G after which the amount of data available to process will increase substantially:"
But Cavium already has a business providing compute for software defined networks in base towers. If there is even more compute required who do you think companies will go with to add more? Someone they already have a relationship with for the core part of their business who also happens to have more compute available, or a complete newcomer?