not sure if this is the right place but - intels future?

[tommo123]

sidestepping the intel vs arm thing on the mobile stage for a moment and just focusing on the desktop/laptop side.

someone on here (maybe IDC) said that intel isn't really a cpu company. they're a monstrosity of a foundry that just happens to design cpus to be able to keep the foundries going.

with intels biggest (IMO) competitor being themselves a year earlier, how are they going to get people to buy new PCs since for most people a core2duo from 2006 will still do them quite nicely. anyone think intel will expand into the software side of things in a meaningful way? i mean make software that can take advantage of their hardware to get people to upgrade to something new a year later?

if they rely on hardware alone then we're already well past the point of "good enough" for the vast majority of people right? 😱

 

[NTMBK]

anyone think intel will expand into the software side of things in a meaningful way? i mean make software that can take advantage of their hardware to get people to upgrade to something new a year later?

Well, they already produce an autovectorizing compiler which makes it easy to produce code that will run super fast on newer processors.

 

[tommo123]

sorry, i meant something for the end user? hell, maybe even an OS to compete with windows

 

[dma0991]

I think what lies with Intel's future is taking that "good enough" performance and constantly reduce its power consumption till it gives increased battery life. People may not need that extra performance but the extra hours they could gain is always a benefit. It is the best course of action that Intel is undertaking if they want to bring the battle to ARM based devices like tablets, that is if Microsoft don't screw it up. We all know that Windows 8 is going to be a major flop the best OS for portable devices with a touchscreen.

 

[pelov]

Well, they already produce an autovectorizing compiler which makes it easy to produce code that will run super fast on newer processors.

I think the OP's point was that people are quite happy with the performance on older processors so the introduction of AVX2 isn't going to do anything in that regard.

They're well aware of the problem you're describing. One way to address is it to stimulate software development such that newer ISAs and increased CPU throughput is required for must-have applications. If you look at the microarchitectural side of things, though, that isn't what's happening. Ivy was a 10% bump and Haswell appears to be another 10% bump while the GPUs are growing far faster and getting much more attention. Clearly increasing CPU performance isn't their top priority.

Intel has been sexifying the laptops in order to generate sales, OP. Along with that, they're focusing heavily on increasing GPU performance of their chips as well, chewing up the sales of low end discrete GPUs. Ivy, Haswell and Sky Lake all appear to be GPU-focused architectures. Power consumption has dropped drastically and perf-per-watt appears to be king as far as goals go. Intel is also adding on a lot of goodies to their platform including charging your smartphone wirelessly via an Ultrabook, WiDi, aka wireless display and Thunderbolt.

While all of this sounds crazy to enthusiasts, they only make up a very small portion of the market. An overwhelming number of PCs sold today are laptops and not desktops and the even more mobile lower end is seeing bigger numbers. Small, slim and efficient mobile products with added features seems to be Intel's current focus with the desktop getting the leftovers 😛

 

[tommo123]

I think what lies with Intel's future is taking that "good enough" performance and constantly reduce its power consumption till it gives increased battery life.

the problem there is that ignores desktop users. once you have a laptop that lasts you a few hours on battery doing whatever you normally do - if that's good enough then why would you upgrade? longer battery life wouldn't then be a benefit.

I think the OP's point was that people are quite happy with the performance on older processors so the introduction of AVX2 isn't going to do anything in that regard.

They're well aware of the problem you're describing. One way to address is it to stimulate software development such that newer ISAs and increased CPU throughput is required for must-have applications. If you look at the microarchitectural side of things, though, that isn't what's happening. Ivy was a 10% bump and Haswell appears to be another 10% bump while the GPUs are growing far faster and getting much more attention. Clearly increasing CPU performance isn't their top priority.

Intel has been sexifying the laptops in order to generate sales, OP. Along with that, they're focusing heavily on increasing GPU performance of their chips as well, chewing up the sales of low end discrete GPUs. Ivy, Haswell and Sky Lake all appear to be GPU-focused architectures. Power consumption has dropped drastically and perf-per-watt appears to be king as far as goals go. Intel is also adding on a lot of goodies to their platform including charging your smartphone wirelessly via an Ultrabook, WiDi, aka wireless display and Thunderbolt.

While all of this sounds crazy to enthusiasts, they only make up a very small portion of the market. An overwhelming number of PCs sold today are laptops and not desktops and the even more mobile lower end is seeing bigger numbers. Small, slim and efficient mobile products with added features seems to be Intel's current focus with the desktop getting the leftovers 😛

i don't buy the 'laptop will save them' idea tbh. reason is, everyone i know (only a handful tbf) that bought a laptop either use it alongside a PC or brought it as a PC replacement. but, then they found they never actually take it anywhere and realise for the same money they could have had a PC and a nice big monitor.

i am right in thinking that intel got involved with a gaming studio a while back right? during them larabee attempt?

 

[pelov]

i don't buy the 'laptop will save them' idea tbh. reason is, everyone i know (only a handful tbf) that bought a laptop either use it alongside a PC or brought it as a PC replacement. but, then they found they never actually take it anywhere and realise for the same money they could have had a PC and a nice big monitor.

While that may be the case for your friends, laptops are still outselling PCs at a 3-to-1 or 4-to-1 ratio. Laptop sales weren't exactly spectacular last quarter but then again neither were tablets. Over the long run, yes, both Microsoft and Intel are going to see a diminishing market share (if you believe the study I read this morning). Soon PC sales should actually be decreasing as tablet sales and laptop sales increase.

Intel isn't gunning for laptops and PCs only, mind you. They have another Larrabbee in the works (the first one went down in Hindenburg-like fashion) and their stranglehold on the server hasn't yet been seriously challenged. They're aiming for perf-per-watt and lowering power consumption because they're also going to enter the tablet and smartphone market in full force.

 

[dma0991]

the problem there is that ignores desktop users. once you have a laptop that lasts you a few hours on battery doing whatever you normally do - if that's good enough then why would you upgrade? longer battery life wouldn't then be a benefit.

I don't think there will ever be a point where a user might think that battery life is enough. Would I want a laptop that runs forever? Definitely, there are no downsides to having infinite power. Is it achievable with current technology? Nope. One might think that 8 hours is sufficient for majority of use cases, I certainly don't. Why not make it last a week? I don't have to bother plugging it at the end of every day.

Users are given a choice, compute performance that is sufficient and on the go versus compute performance that is way better but it is heavy and not practical. The sales numbers don't lie, desktop numbers are dwindling while laptops are increasing. It is what the bigger majority of the consumers want, doesn't make sense if you're not going after a market that has an up trend and gives you the most revenue.

 

[BenchPress]

if they rely on hardware alone then we're already well past the point of "good enough" for the vast majority of people right? 😱

Only for today's applications.

The real problem here is that application's haven't been pushing the envelope recently. A big part of this is due to the great difficulty to make use of all the extra theoretical processing power. Multi-core and SIMD have been very promising on paper but for very few companies they're worth the effort they demand. I expect this to change quite considerably with Haswell. Its TSX technology makes it feasible to create tools and programming language extensions that help developers take advantage of more cores much more easily. Likewise AVX2 presents a major leap in SIMD technology, by enabling the successful SPMD programming model used on GPUs, but in a form that's easier to program than heterogeneous computing.

Another reason why things have slowed down is because we're still waiting on the next generation of consoles. Games are a big catalyst of hardware sales but the games that run on an XBox 360 or PlayStation 3 also run fine on a old mediocre PC. This will change when the new consoles are launched. Mark my words, a Core 2 Duo will no longer suffice.

 

[sefsefsefsef]

BenchPress, will AVX2 benefit Intel or AMD more, or both equally, once they both support it?

 

[Lepton87]

Benchpress you sound more like an AVX2/Haswell advertisement than a real person. Do you have a one post where you did not mention AVX2/Haswell?

 

[Puppies04]

with intels biggest (IMO) competitor being themselves a year earlier, how are they going to get people to buy new PCs since for most people a core2duo from 2006 will still do them quite nicely.

I hear this a lot on this forum, all the time infact and from my experience it doesn't quite work like that in the real world. I have replaced 6 or 7 pc's this year that were more than powerful to do what the owner needed. Unfortunatly they had that much spyware and rubbish on them that they ran like crap. Now I always give the owner a choice, I will get them running like new again or I can put them together a nice little I3 rig that is more power efficient and faster than the old PC ever could be, most choose option number 2 and that is where I would guess a large percentage of intels customers come from.

Also consider someone whos mobo/ram/hdd/PSU has just died, most people don't even bother trying to get it fixed, it is just "broken" and needs replacing so they trot off down to pc world or wherever and buy a brand new one.

I also hear the "why would someone with SB upgrade to IB" opinion a lot and the answer is they wouldn't. Intel can't stand still, they need to release faster and more efficient CPUs each year or else AMD will catch them up. Now they aren't trying to sell this years new CPU to someone who bought last years CPU. There are more than enough potential customers out there that as long as people are upgrading there machine on a 4-5 years basis then intel can carry on making money.

Now couple all that with the fact that intel make a big chunk of their profit from server spec CPUs where 2006s hardware just won't cut it today and you can see why intel won't be going out of business any time soon.

 

[biostud]

If they needed something else to produce in their fabs, they could always expand to dedicated GPU business. But as with so many things it's the mobile sector/server farms that hunger for the best perf/watt and that can always get better with new processes.

 

[BenchPress]

BenchPress, will AVX2 benefit Intel or AMD more, or both equally, once they both support it?

That depends. AMD could opt to implement AVX2 the same way they implemented AVX, by combining two 128-bit SIMD units shared between two integer cores. But that means an 8-core AMD chip would have only half the vector throughput of a 4-core Intel chip. My hope is they'll widen the SIMD units to 256-bit each, in which case they'd benefit roughly equally from AVX2 than Intel.

 

[NostaSeronx]

That depends. AMD could opt to implement AVX2 the same way they implemented AVX, by combining two 128-bit SIMD units shared between two integer cores.

The current implementation of AVX 256-bit width ops is that they are split in 128-bit width ops. The FPU doesn't need to execute both ops at the same time. The combining units is just marketing talk it isn't completely true. If executing 256-bit ops most likely than not you are going to execute only half the op per cycle rather than the full op per cycle.

The floating point for Bulldozer is also overly simplified in marketing:
Port 0a -> FMA
Port 0b -> IVEC FMA
Port 1a -> FMA
Port 1b -> XBAR
Port 2 -> Integer AVX, SSE, MMX
Port 3a -> Integer AVX, SSE, MMX
Port 3b -> Load-Store Unit

Each port should be 128-bits long and can be executed simultaneously. So, if you have 1 128-bit FMA and 1 128-bit Integer FMA both can be executed under port 0 at the same time.

If Bulldozer is the above, then Steamroller is:
Port 0a -> FMA
Port 0b -> IVEC FMA
Port 1a -> FMA
Port 1b -> IVEC FMA
Port 2a -> Load-Store Unit
Port 2b -> XBAR

 

[yottabit]

They'll just need to cross there fingers and hope for a new killer app or crippling OS

3D 4k Video Editing??

 

[jpiniero]

I was thinking about this when I read about the warning. It's pretty clear that PC sales are slipping, and there may not be much Intel can do about it.

Basically, we are at the point where most Corporate users could last much longer on their machines. Imagine what would happen to sales if, say companies that refreshed went from 3 years to 5; 4 to 6; 5 to 7-8 etc. It'd be a big hit on business. With the economy the way it is, and not expected to get better any time soon, it makes sense to cut back big time on corporate pc sales. On top of this is the diaster that is Windows 8.

As for Home users, they sure seem to love their cheap ARM Post PC devices. It was only a matter of time before they started to eat into Intel's sales beyond things like the netbook. Even if Intel manages to get a foothold in there (good luck getting Apple or Samsung to switch), the margins will not be what they would like.

Now couple all that with the fact that intel make a big chunk of their profit from server spec CPUs where 2006s hardware just won't cut it today and you can see why intel won't be going out of business any time soon.

Last time I checked, the Server group was only like 20% of revenue. This percentage will likely increase, but that's not something Intel would be too proud of.

 

[Idontcare]

with intels biggest (IMO) competitor being themselves a year earlier, how are they going to get people to buy new PCs since for most people a core2duo from 2006 will still do them quite nicely. anyone think intel will expand into the software side of things in a meaningful way? i mean make software that can take advantage of their hardware to get people to upgrade to something new a year later?

Within my sphere of influence of family and friends, by far the number one reason I see people asking me to upgrade/replace their computers has solely to do with the creep in bloatware that seems to accumulate on their rigs over time.

Everything auto-updates now, you add in some anti-virus software and suddenly a rig that was snappy 2 yrs ago runs like a dog with disk thrashing everytime you try and open a pdf attachment or visit shutterfly to upload some recent vacation photos (stuff regular people do with their multi-core desktops 😉).

Even if the hardware is good enough, the good enough factor decays over time because the system itself is not a static environment. Bloatware and auto-update everything sneaks up and robs performance.

Sometimes I redo their system installs, but mostly they just want new hardware and they aren't really interested in me talking them out of getting a new computer.

In short, Intel's future is secure so long as the software industry continues to evolve towards the business model which holds the consumer to be the beta-tester, the bug finder, and whose harddrives are to be a growing repository of auto-updated bloatware.

 

[jpiniero]

That's where SSDs/memmistors/etc come in the picture. I'm sure Symantec is hard at work at finding ways to make SSDs slow though.

 

[Cerb]

with intels biggest (IMO) competitor being themselves a year earlier, how are they going to get people to buy new PCs since for most people a core2duo from 2006 will still do them quite nicely.

They won't. That's one of the reasons the desktop is, "dead," not because nobody uses them, but because fewer users need to upgrade every couple years, now (fewer users need to than do, as well, as IDC already mentioned).

Thus, the push for ultrabooks, tablets, and phones, as those markets still have some growth room.

anyone think intel will expand into the software side of things in a meaningful way? i mean make software that can take advantage of their hardware to get people to upgrade to something new a year later?

They probably are doing that with McAfee, and they have been doing that with their compiler for years. Their good support of Linux is also related, but not a direct cash-grab. In reality, it takes a few years for software to catch up, so you see a much smoother transition than what you're thinking. I don't think there's any doubt that all the hype and early support for AVX2, however, is an attempt to do just that: make new CPUs desired by users earlier than usual, by letting software be ready when the hardware is released.

if they rely on hardware alone then we're already well past the point of "good enough" for the vast majority of people right? 😱

Yes, and that's where really where the Intel v. ARM thing comes in to play. ARM's whole ecosystem is built around fairly tight margins, while Intel's is built around very large margins. But, people need to be believe a new product is worth the cost. That could pose a long-term problem. It's not Intel v. ARM because ARM is the greatest ISA and company ever, but because they have managed to create a healthy development community and large markets by skimming a little off the top.

I'm sure Symantec is hard at work at finding ways to make SSDs slow though.

😀

Last time I checked, the Server group was only like 20% of revenue. This percentage will likely increase, but that's not something Intel would be too proud of.

Also keep in mind that server reliability features tend to creep down to consumer parts over time (fault checking and fault prevention become more necessary as everything shrinks and runs faster), so much of the server R&D today can be worth it in the future for all products, on top of the high-margin revenue it brings in (not to mention how it marginalizes any other entries to the market).

 

[Anarchist420]

I'm looking very forward to avx2 and TSX because it should make it so emulators can become replicators, although I'm disappointed that they're adding whole GPUs rather than more avx2 units, texture units, and display device logic/tmds transmitters and whatnot. I think it would be a good idea if intel moved the USB, SATA, audio codec, and lan controllers onto the CPU die while adding the vrms onto the CPU die also. I hope they use solder like they did with sb though and never again do what they did with IB.

 

[BTRY B 529th FA BN]

Everything auto-updates now, you add in some anti-virus software and suddenly a rig that was snappy 2 yrs ago runs like a dog with disk thrashing everytime you try and open a pdf attachment or visit shutterfly to upload some recent vacation photos (stuff regular people do with their multi-core desktops 😉).

Even if the hardware is good enough, the good enough factor decays over time because the system itself is not a static environment. Bloatware and auto-update everything sneaks up and robs performance.


In short, Intel's future is secure so long as the software industry continues to evolve towards the business model which holds the consumer to be the beta-tester, the bug finder, and whose harddrives are to be a growing repository of auto-updated bloatware.

That's why smart phones will always be fast when new. All those crap apps and other things installed crash out phones, make them stall, etc etc

 

[BenchPress]

The floating point for Bulldozer is also overly simplified in marketing:
Port 0a -> FMA
Port 0b -> IVEC FMA
Port 1a -> FMA
Port 1b -> XBAR
Port 2 -> Integer AVX, SSE, MMX
Port 3a -> Integer AVX, SSE, MMX
Port 3b -> Load-Store Unit

Each port should be 128-bits long and can be executed simultaneously. So, if you have 1 128-bit FMA and 1 128-bit Integer FMA both can be executed under port 0 at the same time.

No, each port can only execute a single operation each cycle. It's physically impossible to feed seven execution units simultaneously at such speeds/latency. That's why there are ports in the first place.

 

[OVerLoRDI]

Within my sphere of influence of family and friends, by far the number one reason I see people asking me to upgrade/replace their computers has solely to do with the creep in bloatware that seems to accumulate on their rigs over time.

Everything auto-updates now, you add in some anti-virus software and suddenly a rig that was snappy 2 yrs ago runs like a dog with disk thrashing everytime you try and open a pdf attachment or visit shutterfly to upload some recent vacation photos (stuff regular people do with their multi-core desktops 😉).

Even if the hardware is good enough, the good enough factor decays over time because the system itself is not a static environment. Bloatware and auto-update everything sneaks up and robs performance.

Sometimes I redo their system installs, but mostly they just want new hardware and they aren't really interested in me talking them out of getting a new computer.

In short, Intel's future is secure so long as the software industry continues to evolve towards the business model which holds the consumer to be the beta-tester, the bug finder, and whose harddrives are to be a growing repository of auto-updated bloatware.

This.

In my sphere of influence I get people asking me all the time about hardware configurations for their basic computer needs, when really 95% of what is on the shelf is overkill for what they want to do.

I always have them bring me the computer and clean it out. It is stunning how terribly slow these brand new machines are with all their bloatware. The PC industry is really shooting themselves in the foot with all this crap.

 

[Idontcare]

The PC industry is really shooting themselves in the foot with all this crap.

Unfortunately the truth of the matter is that they are laughing all the way to the bank and back again with all this crap.

The computer folks get paid to add the bloat to the prebuilt machines. I recall when I bought my DELL laptop they had an option where I could pay some ridiculous fee - $50 IIRC - and they would not install the bloatware.

It really felt like extortion the more I thought about it.

The Kindle deal is like that too. Pay extra money to not be bombarded with advertisements 😕 Mind you this is on top of you already paying for the broadband connection fees to Amazon and buying the Kindle to boot.

I'm waiting for yahoo.com to get hold of this idea and they will start extorting offering a service in which you have to pay $25/yr if you want your anti-spam features to work. Otherwise they intentionally spam-bomb your inbox since you didn't opt-out of that new feature.