AMD fans rejoice, Intel has their own Hector Ruiz

[raghu78]

The only thing holding back ARM from dominating all the segments beyond smartphones is software. Microsoft is the key to ARM moving beyond smartphones into the enterprise and desktops. Microsoft is throwing its weight behind ARM and we will see the results over the next 5-10 years.

I would like to see AMD eventually release K12 or an improved version from what they originally started designing. Jim Keller hinted that K12 had a bigger engine than Zen. We already see Zen keeping up quite well with Broadwell (only slightly behind overall ). K12 could easily be the fastest custom ARMv8 core available even ahead of Apple's custom cores. K12 is likely to sport SMT as AMD has experience now with SMT designs. K12 could make for some very impressive ARMv8 server SoC . If AMD sees enough traction for ARM it will eventually be forced to launch K12.

 

[hansolo69]

The problem is financial statements aren't forward looking. They can tell you how dominant Intel was in the past (which no one is arguing against) but they can't tell you what Intel faces going forward.

I agree with that. Everyone's crystall ball is as good as Gartner's prediction. In future Intel can go bankrupt or be the same as today, nobody will know for sure.
In mobile (read phone) area there is a good planned obsolescence strategy to combat market saturation in every 2 years. Samsung and Apple did it very well, because there is are no real evolutionary applications worth to upgrade (yeah, games suck on phones).
Intel didn't have this strategy, nor Microsoft for home desktop users. Only games move home users to upgrade their desktops, so that's why we have flat/lower sales in this segment.
In mobile area Intel and Apple did good with ultrabooks/2-in-one, both for home users and professional users.
And in server market are any commercially available ARM servers better at ROI than s86 servers? None, or 0.000001%. And I doubt that Microsoft will improve that percent anytime soon.

In my opinion Mcafee was a bad move, Alterra the same, Mobileye is much better than those, the real competition in this market is only Nvidia.
And again, why nobody bashes the more nonsense aqusitions from Qualcomm or Softbank?

 

[formulav8]

Mobileye is much better than those

If I still had my Intel stock I would actually be pretty ok with that self-driving car company buy.

But the MacAfee buy I never understood and still don't. Anyone even remember their reasoning for that decision?

 

[Hi-Fi Man]

Another angle that hasn't been mentioned yet is IBM and OpenPOWER. Power is poised to make a better showing in servers than ever before (much like AMD, probably less so though) with talk about China being a potential large Power customer and the alliance finally coming out with better and cheaper commodity servers. Especially now that Intel is losing it's process lead, Zen and Power9 have a chance to be competitive.

 

[majord]

The thing i'm struggling with is how Mobileye was worth $15b? The seem quite a small outfit , with all there current products in the driver assistance (ADAS) space?. They appear to have some very interesting technologies for L3-4 automomous driving, with some good partnerships but it seems Nv has a big head start here?

and on the compute side of things, what is Intel using as the basis for their competing platform? (i.e against volta based NV platforms)

 

[hansolo69]

If I still had my Intel stock I would actually be pretty ok with that self-driving car company buy.

But the MacAfee buy I never understood and still don't. Anyone even remember their reasoning for that decision?

Some good explanation about McAfee here: https://arstechnica.com/business/2010/08/why-intel-bought-mcafee/
They dreamed about putting AV security in chips... and failed https://www.bloomberg.com/news/arti...ntel-s-mcafee-unit-in-4-2-billion-transaction

Problem with Mobileye aquisition isn't opportunity but it's inflated cost: http://www.recode.net/2017/3/13/14907184/why-intel-15-billion-mobileye-deal
This still looks like it could be an expensive purchase. Mobileye made $121 million in operating profit last year, making Intel’s $15 billion purchase price 124 times its profit.

 

[Valantar]

The only thing holding back ARM from dominating all the segments beyond smartphones is software and a big core with competitive IPC.

There, I fixed it for you. All ARM server products shipped so far have been based on anemic mobile cores, which is as much of a reason they failed as the lack of software - noone will develop software for a core they know is inferior to what they're currently using, even if it performs a few percent more work/watt.

We don't know how K12 performs, nor any of the recently announced server ARM chips, though they might fix this issue. I'll believe that when I see it, though.

 

[itsmydamnation]

We don't know how K12 performs, nor any of the recently announced server ARM chips, though they might fix this issue. I'll believe that when I see it, though.

AMD is quoted as saying K12 OOO Engine is 10% bigger then Zen's so why wouldn't it perform on par? Im sure if apple wanted to make a server with its anemic cores it would look pretty good next to the down right belimic skylake .

Even a sea of A72/73 (somewhere around C2D/nelehem perf per clock) would look very good for the right workload at the right price assuming someone get the hard part right (there is a reason AMD using using 64pci-e 3 lanes as interconnect between two sockets)

 

[Jason Ives]

Well, Intel may or may not be "doomed", but I feel that their golden years are behind them, the writing is on the wall. Without an Intel-leading / exclusive breakthrough in device physics R&D, to invent an entirely new form of transistor / semi-conductor (metal hydrogen, if superconducting?), then their most pressing competitors seem to be catching up to their process manufacturing lead, or at the very least, making it far less relevant as a decisive factor in the semiconductor market, and the competing architecture(s) seem to have a licensing advantage, compared to Intel. Maybe if they weren't so stingy about giving out x86/x64 licenses, then maybe the overall x86 market would be healthier, and in more devices. ARM is at the gate, and they're slowly beating it down. It seems, by unit volume, x86 is outnumbered by ARM, if not sales / revenue volume.

I recall this exact sentiment when Intel was in an infinitely worse position during Pentium D generation and right before Conroe. Remember? Its not everyday that a generation code name becomes a verb.
Intels golden years will never be behind them. They may dip into the silver now and a decade, but all they ever need to do is summon the grunt and they are on top again.

 

[NTMBK]

I recall this exact sentiment when Intel was in an infinitely worse position during Pentium D generation and right before Conroe. Remember? Its not everyday that a generation code name becomes a verb.
Intels golden years will never be behind them. They may dip into the silver now and a decade, but all they ever need to do is summon the grunt and they are on top again.

In 2006 (right before Conroe launched) Intel's key market, PCs, was still growing- it would grow another 50% before hitting its peak in 2011. But it's now been on a constant decline for 6 straight years, with no recovery in sight.

 

[VirtualLarry]

Yep, Intel's cash-cow market is shrinking, and unlikely to reverse course, and every other thing they've tried to stick their finger into, seems not to have worked out very well (mobile, etc.) Of course, Datacenter is doing well, and on track for growth, so they've got that at least. But AMD is pretty certain to make inroads with Ryzen's Server version. Intel's not going to have that market to themselves.

 

[Sheep221]

Sorry guys but this thread is clearly the result of the decades of brainwashing. The years of higher frequency, more cores(and more threads) and smaller nodes left you thinking that it's all there is. Well it's not, approaching physically smallest nodes is not the end for semiconductor industry, nor the X86, nor the big chip corporations. They are going to shift their focus to other areas of development where they can make computers better, some of these would be increasing energy efficiency, bringing down the cost and optimizing software, maybe bringing more components to SoC chips, increasing reliability etc. After certain point you don't get faster computers just by raising the physical stakes. Just like you won't create a semi truck from car just by fitting an engine with overly high displacement in it. There are limits and electronic/semiconductor conglomerates are well aware of them.
In the hardware terms I don't think computers are going to get much faster. And as such they don't have to be, most high end PCs in homes or offices hardly ever utilize full CPU power, if we want better graphics, better sound, better results we must focus on side things like output hardware and so on.
However if you need faster than that, you clearly have to order computing time on mainframe/supercomputer/quantum equipment to get the results beyond complexity of desktops, which is limited and those limits were not really extended over x86 existence.

 

[Valantar]

AMD is quoted as saying K12 OOO Engine is 10% bigger then Zen's so why wouldn't it perform on par? Im sure if apple wanted to make a server with its anemic cores it would look pretty good next to the down right belimic skylake .

Even a sea of A72/73 (somewhere around C2D/nelehem perf per clock) would look very good for the right workload at the right price assuming someone get the hard part right (there is a reason AMD using using 64pci-e 3 lanes as interconnect between two sockets)

The K12 might perform on par with Zen - who knows? We'll see when/if it actually hits the market in some tangible way. Until then, I remain skeptical, simply because we've yet to see a single ARM core that's remotely competitive with X86 outside of seriously power-limited scenarios.

And a sea of A72/A73 chips would quickly become an expensive proposition - even if you fit 50-60 of them per die (more would probably either make the die too big to be viable or too difficult to cool), you'd still need 8 or more sockets per machine to make this worthwhile. Which means skyrocketing costs with complex motherboards and interconnects.

As for the Apple analogy, if you're referring to the "The iPad Pro is faster than an Ultrabook" thing, that's been shown to be false in quite a few ways, no? GeekBench (which that story sprung from) is hardly a platform-neutral benchmark, no matter how much it tries to be. Besides, we have no idea about the power or frequency scaling of Apple's designs.

(and lastly, keep your metaphors straight. Anemic = lacking red blood cells, which (due to reduced oxygen carrying capacity in the blood) leads to low energy and weakness ("performance" in the case of this metaphor). It's a commonly used metaphor for something "weak". Bulimia is an eating disorder, which ... causes the CPU to throw up after eating? What do CPUs eat? Electricity? I don't quite see what you're getting at. Also, using serious mental illnesses as some sort of performance-related metaphor strikes me as more than a little tasteless.)

 

[hansolo69]

Yep, Intel's cash-cow market is shrinking, and unlikely to reverse course, and every other thing they've tried to stick their finger into, seems not to have worked out very well (mobile, etc.) Of course, Datacenter is doing well, and on track for growth, so they've got that at least. But AMD is pretty certain to make inroads with Ryzen's Server version. Intel's not going to have that market to themselves.

Market decline is important for investors, less for customers. Is Intel lossing money? Maybe then we will talk about bad decisions.
It will be good for all customers if AMD will have success and prices will go down reasonably.
It will be bad if we will see more competition from cheap chinese junks at the bottom of barrel (like in mobile or mediaplayer area). Intel contrarevenue was bad for competition and themselves, cheap junks are bad for everyone.

 

[sm625]

The Altera purchase I can see some synergy there. I've been predicting programmable execution units ie mini DSP blocks with access to L1 for massively increased IPC on a per application basis. For that they need the IP of a company like Xilinx or Altera. But the Mobileye purchase just seems like dumb money desperately grasping for yield, the kind of thing you see at the waning end of the business cycle. The unfortunate thing is that we could have to endure several more years of this madness before it pops. The damage companies are doing to their balance sheets is just laughably reminiscent of 1999-2000. Companies with lots of cash need to just sit on it for a few years. Wait for market caps to be cut in half, and then pounce. Funny how you didnt see very much M&A activity in 2009.

 

[Jason Ives]

In 2006 (right before Conroe launched) Intel's key market, PCs, was still growing- it would grow another 50% before hitting its peak in 2011. But it's now been on a constant decline for 6 straight years, with no recovery in sight.

But you are talking about the entire PC market trend, under which all pc hardware producers fall. Not sure what parallels you were drawing, but I was responding to what I quoted. Not talking about the pc market, but the actual caliber of products (in this case processors) a company could or would produce.

 

[NTMBK]

But you are talking about the entire PC market trend, under which all pc hardware producers fall. Not sure what parallels you were drawing, but I was responding to what I quoted. Not talking about the pc market, but the actual caliber of products (in this case processors) a company could or would produce.

The caliber of their products is tied to their financial health. Intel needs money to develop more and more advanced transistors, and more and more advanced designs using those transistors.

 

[Arachnotronic]

The caliber of their products is tied to their financial health. Intel needs money to develop more and more advanced transistors, and more and more advanced designs using those transistors.

Intel is not even close to the point where having enough money to do R&D for the future is even close to a concern.

The problem is that they're not getting the returns that they should on its R&D, which means that somebody is spending all that money in a sub-optimal fashion.

 

[TemjinGold]

Intel is not even close to the point where having enough money to do R&D for the future is even close to a concern.

The problem is that they're not getting the returns that they should on its R&D, which means that somebody is spending all that money in a sub-optimal fashion.

I don't think the issue is Intel not being able to fund R&D. The issue is more that Intel had been dominant because its rival could not. If we shift to an environment where that rival is able to fund R&D well, Intel may stop being so dominant.

 

[frozentundra123456]

I don't think the issue is Intel not being able to fund R&D. The issue is more that Intel had been dominant because its rival could not. If we shift to an environment where that rival is able to fund R&D well, Intel may stop being so dominant.

Recently not having money for r &d has been a problem for AMD, but the bigger problem was that even with relatively more money for R & D they somehow came up with Bulldozer. Ironically, when they were in a much worse financial position they came up with a very good product in Zen.

 

[Crumpet]

And yet, PC sales continue to fall:

Guess what? Ultrabooks and x86 tablets/2-in-1s have completely failed to reverse that trend. (2016, missing from that chart, was yet another decline.)

While the foundries have now caught up with Intel in process technology. Why do you think Intel has announced that they will go "server first" with future nodes? Because their competition has caught up. ARM server chips are finally appearing on nodes competitive with Intel (10nm Qualcomm server parts), with high performance custom cores capable of competing instead of just repurposed 32-bit phone CPUs.

Intel's financials have continued to look good because, as I said earlier, they haven't had any serious competition in the datacentre for years and years. They blew AMD out of the market with Nehalem and then AMD ran over their own foot with a Bulldozer, the RISC players (SPARC, POWER) were relegated to high end niches which were getting slowly eaten up by cheaper x86 machines... They had a free reign to crank out healthy profits, and slowly push up the prices. Well the fun times are over.

Does this include components?

I imagine with more people having access to youtube and forums etc, that perhaps more people are building their own.

 

[itsmydamnation]

T
And a sea of A72/A73 chips would quickly become an expensive proposition - even if you fit 50-60 of them per die (more would probably either make the die too big to be viable or too difficult to cool), you'd still need 8 or more sockets per machine to make this worthwhile. Which means skyrocketing costs with complex motherboards and interconnects.

Why would you want to do that and have death by NUMA? Just have 4 1P nodes per RU, 2.5inch ~6 drives a node. have 2x 10/25gbe phy a node, use 40/100gbe QSFP break out cables so 2 ports ( one from each TOR switch) services a single RU. copy paste to as many racks as you want.

Why would you want to kill your efficiency with horrible NUMA accesses? ( which is why AMD have 64 pci-e lanes interconnecting just 2 sockets, im guessing we will see a full mesh between memory controllers)

(and lastly, keep your metaphors straight. Anemic = lacking red blood cells, which (due to reduced oxygen carrying capacity in the blood) leads to low energy and weakness ("performance" in the case of this metaphor). It's a commonly used metaphor for something "weak". Bulimia is an eating disorder, which ... causes the CPU to throw up after eating? What do CPUs eat? Electricity? I don't quite see what you're getting at. Also, using serious mental illnesses as some sort of performance-related metaphor strikes me as more than a little tasteless.)

Nope i mean about intels quest for more SIMD width to grow into new markets without actually caring about what its current markets actual want. Trying to differentiate the harshness of anemic/ buliemic is pretty funny, how do the people with autoimmune diseases feel...............

 

[TemjinGold]

Recently not having money for r &d has been a problem for AMD, but the bigger problem was that even with relatively more money for R & D they somehow came up with Bulldozer. Ironically, when they were in a much worse financial position they came up with a very good product in Zen.

They took the wrong gamble on Bulldozer and it didn't pay off. It happens. But that doesn't mean AMD becomes less of a threat when it has more money if that's the argument you are somehow trying to make.

 

[scannall]

Recently not having money for r &d has been a problem for AMD, but the bigger problem was that even with relatively more money for R & D they somehow came up with Bulldozer. Ironically, when they were in a much worse financial position they came up with a very good product in Zen.

Brand new architectures are always a really big gamble. More flops than winners historically. Bulldozer, Itanium, Netburst and Larrabee are just a few.

 

[turtile]

I also don't think the latest CEO is doing a great job. It looks like he is buckling into the board's pressure to diversify the business but doing so as an after thought. Self driving cars is still a long ways away. Probably 8-10 years until we see some in cities. If they were going to invest, they should have done it a while ago. It's going to be a lot harder to compete with Google and others that already have a lot going on.