Why are desktop CPUs so slow at improving?

[TheELF]

Your opinion. That and a coupon for a free coke will get me a free coke. My opinion yes they would. Just as valid!


Again another free coke! No, and don't pretend to speak for me.


Scroll up you stating the obvious here captain! I've already made statements to this effect.


Would you keep your politics out of this! Because everything your saying is your opinion, and in mine, it's pure nonsense!


This is the first thing you said dealing with the topic! bravo! In future posts instead of making a bunch of opinions about me can you stick to the topic "Why are desktop CPUs so slow at improving?"


A slight error on my part HP was so dependant on Intel rebates they couldn't afford to take them, but Dell was complicit in Intel's anti-competitive activities as well. From 2001-2006 Intel paid Dell 6 billion not to ship any computers powered by AMD.

Perhaps CNET is bi partisan enough for everyone?

https://www.cnet.com/news/intel-and-amd-a-long-history-in-court/

TECH INDUSTRY
Intel and AMD: A long history in court

Monday's suit doesn't mark the first time AMD has accused Intel of antitrust violations. We look back at the companies' legal tangles.
BY MICHAEL SINGER
JULY 8, 2005 6:00 AM PDT

Intel and Advanced Micro Devices' long history of competing for microprocessor dominance has landed them in court before.

In the latest salvo, AMD this week filed an antitrust suit in U.S. District Court in Delaware. Here are some key moments in the companies' entwined histories:

1968--Intel is founded by Bob Noyce and Gordon Moore.

1969--AMD is founded by Jerry Sanders along with a team of former Fairchild Semiconductor employees.

Early 1980s--IBM chooses Intel's so-called x86 chip architecture and the DOS software operating system built by Microsoft. To avoid overdependence on Intel as its sole source of chips, IBM demands that Intel finds it a second supplier.

1982--Intel and AMD sign a technology exchange agreement making AMD a second supplier. The deal gives AMD access to Intel's so-called second-generation "286" chip technology.

1984--Intel seeks to go it alone with its third-generation "386" chips using tactics that AMD asserts were part of a "secret plan" to create a PC chip monopoly.

1987--AMD files legal papers to settle the 386 chip dispute.

1991--AMD files an antitrust complaint in Northern California claiming that Intel engaged in unlawful acts designed to secure and maintain a monopoly.

1992--A court rules against Intel and awards AMD $10 million plus a royalty-free license to any Intel patents used in AMD's own 386-style processor.

1995--AMD settles all outstanding legal disputes with Intel in a deal that gives AMD a shared interest in the x86 chip design, which remains to this day the basic architecture of chips used to make personal computers.

1999--Required by the 1995 agreement to develop its own way of implementing x86 designs, AMD creates its own version of the x86, the Athlon chip.

2000--AMD complains to the European Commission that Intel is violating European anti-competition laws through "abusive" marketing programs. AMD uses legal means to try to get access to documents produced in another Intel antitrust case, this one filed by Intergraph. The Intergraph case is eventually settled.

2003--AMD's big technology breakthrough comes when it introduces a 64-bit version of its x86 chips designed to run on Windows, beating Intel, which for the first time has to chase AMD to develop equivalent technology. AMD introduces its Opteron line of chips for powerful computer server machines and its Athlon line for desktops and mobile computers.

2004--Japan's Fair Trade Commission (JFTC) raids Intel offices in Japan searching for documents. Intel cooperates with the investigation but does not agree with the outcome. JFTC officials find that Intel's Japan unit stifled competition by offering rebates to five Japanese PC makers--Fujitsu, Hitachi, NEC, Sony and Toshiba--which agreed not to buy or to limit their purchases of chips made by AMD and Transmeta.

Actually what you are saying is that from 2001 onwards every Computer OEM could make a easy living with intel only products,while they would not be able to do the same with AMD only products...
If AMD had decent hardware then OEMs could completely switch over to AMD and let intel stand in the rain.
That's what aggressive competition is,AMD would match or increase rebates to the companies to make them switch to them.
So basically your complaint is not "let there be competition" your complaint is "give the useless company a chance"

 

[jpiniero]

Their gross profits are over 50% year over year!

It sounds high until you realize AMD has said that to be profitable, they need Gross Margin to be over 40%. And it's in the 60's actually.. and has been there for some time. Basically Intel's doggin it/higher prices/etc is their way of maintaining the margins in a time where PC sales are sliding.

 

[frozentundra123456]

People aren't quick to defend intel, more likely people are sick of your bitter, short sighted synopsis. With your opinionated, over the top posts and rhetoric (Like someone just slapped their daddy in the face), it seems that your "augmented view, devoid of reality" is likely augmented by some form of prescription drugs?
Well, no matter.. With 58 posts (and not many helpful), this is a good time to use the forums ignore feature. Good luck with that bitterness.. Let it go, or let it consume you..

Did it already. I think that is a record, for someone to earn the ignore button with 50 posts. Honestly, I dont see why anyone continues to respond.

 

[Verndewdimus]

for the history nut, IBM had the entire pc market before the securities and exchange commission demanded they distribute production, IBM chose Intel and AMD, both companies have had a close relationship with IBM since, in fact hyperthreading was "borrowed" from an IBM engineers work.

 

[Topweasel]

Intel released a true 64bit cpu in 2001!
Intel tried to innovate Well Whadya Know?!
Yeah it sucked because it was slow and needed completely new software but it was innovation and they got kicked in the face for trying to innovate.
On the other hand AMD just went with the cheapest and simplest solution of just adding extensions to fool the cpu into thinking that it is 64bit, like a year or two later...
https://en.wikipedia.org/wiki/Itanium#Itanium_(Merced):_2001

Do you know why Intel developed Itanium? It wasn't altruistic. They did see an opportunity to remove some of the shackles of the previous arch type by making something that could break from the x86 legacy.

But it was a two fold attack. 1.) Wanting to break into the high end server market and kill the competing Risc and other specialized server chip business. 2.) They specifically wanted something Sever first that as the major code converted to IA-64, they could then offer it as a professional chip, and then a consumer chip. The goal was that in 10 years they would have converted everyone over to IA-64 and away from x86 and therefore killing of AMD and all the forced cross licensing that went with it.

Itanium was about the eventual death of all CPU competition.

 

[Topweasel]

Actually what you are saying is that from 2001 onwards every Computer OEM could make a easy living with intel only products,while they would not be able to do the same with AMD only products...
If AMD had decent hardware then OEMs could completely switch over to AMD and let intel stand in the rain.
That's what aggressive competition is,AMD would match or increase rebates to the companies to make them switch to them.
So basically your complaint is not "let there be competition" your complaint is "give the useless company a chance"

It would be one thing if it was a discount for buying CPU's. It's another thing when the rebate (that was nearly a Billion a quarter just to Dell when they caved and started offering AMD) specifically limits them to not using a competitor.

The problem is AMD's infrastructure wasn't that great. They couldn't produce enough to be the only CPU supplier for any of these manufacturers let alone all of them and Intel used that to their advantage. But this just compounded the 2 previous decades of moves Intel did in an attempt to invalidate AMD specifically of any opportunity to compete. If Intel hadn't pulled what they did earlier with the 386 and 486 would AMD have been in the position they were in, that allowed Intel to further use the near monopolistic power to shut down AMD from entering the lucrative OEM market?

 

[FIVR]

It would be one thing if it was a discount for buying CPU's. It's another thing when the rebate (that was nearly a Billion a quarter just to Dell when they caved and started offering AMD) specifically limits them to not using a competitor.

The problem is AMD's infrastructure wasn't that great. They couldn't produce enough to be the only CPU supplier for any of these manufacturers let alone all of them and Intel used that to their advantage. But this just compounded the 2 previous decades of moves Intel did in an attempt to invalidate AMD specifically of any opportunity to compete. If Intel hadn't pulled what they did earlier with the 386 and 486 would AMD have been in the position they were in, that allowed Intel to further use the near monopolistic power to shut down AMD from entering the lucrative OEM market?

Intel's "Meet Comp" program actually paid the vendors according to how few AMD chips they sold. if they sold 0 AMD chips, they got the full rebate. If they sold 10% AMD chips, they got a smaller rebate. 30% AMD chips, even smaller still and so on.


The reason desktop CPUs have stagnated is because Intel has stagnated. They realize "gamers" and others in this thread will buy their CPUs no matter what, even if they only provide a .05% increase in performance every generation. Not only will they buy them, they'll also defend intel endlessly and claim that "the laws of physics" say that Intel can't do any better.

 

[Topweasel]

The reason desktop CPUs have stagnated is because Intel has stagnated. They realize "gamers" and others in this thread will buy their CPUs no matter what, even if they only provide a .05% increase in performance every generation. Not only will they buy them, they'll also defend intel endlessly and claim that "the laws of physics" say that Intel can't do any better.

More important and the direct responsible party is mobile computing. Specifically Intel realized that Laptops were going to replacement desktops as the primary computing device and had the team in Israel come up with a new mobile chip because there was nothing Intel could do with the P4 to make it competitive in notebooks. Conroe became the breaking point. Pretty much as fast as the P4 on desktop power, malleable to work well in laptops. Intel even though they have been large enough to do as many dies as needed has always tried to deal with as few as possible (up to x99 Intel only used 2 dies a 2 core and 4 for all consumer products). They figured out that if they spent enough time an effort into minimizing power usage they could have one CPU that worked well in lower power units and then from there they would see what feeding power would do for clocks. The Last 7 years hasn't been about making faster CPU's, its been basically working on taking a Sandy Bridge arch and working it down to the point that they can finally offer a 15w 4 core CPU.

 

[daveybrat]

Folks, let's keep this thread on track and on topic. Any further Intel/AMD bias or personal attacks will result in infractions handed out and/or closing the thread.

Thanks,

Daveybrat
AT Moderator

 

[VirtualLarry]

intel could have made a competitive move in releasing the 6c/12t CPU a week before amd started the pre-orders for ryzen,come on even you have to admit that they had that sku ready to go ever since amd first announced ryzen,but that would not cause any race toward innovation this would just have killed off amd right away.

Intel's OEM customers would have been highly displeased, though, because they rely on Intel's regular yearly model release cadence, to advertise and sell their systems. (Even though, due to the relatively tiny IPC increases that we've been getting from Intel, render their yearly upgrades nearly mute, and people are still able to manage, even gaming, with a 2500K. Whether that's an indication of how awesome the 2500K is, or an indication of how little Intel has "innovated" since then, I don't know. But surely, Coffee Lake's introduction of a mainstream six-core CPU is a form of innovation, even though it doesn't have any IPC increases.)

 

[goldstone77]

Intel's OEM customers would have been highly displeased, though, because they rely on Intel's regular yearly model release cadence, to advertise and sell their systems. (Even though, due to the relatively tiny IPC increases that we've been getting from Intel, render their yearly upgrades nearly mute, and people are still able to manage, even gaming, with a 2500K. Whether that's an indication of how awesome the 2500K is, or an indication of how little Intel has "innovated" since then, I don't know. But surely, Coffee Lake's introduction of a mainstream six-core CPU is a form of innovation, even though it doesn't have any IPC increases.)

Everything they released this year was planned from the leaked roadmaps with the exception of the i9's. But like you said the scheduled release date were definitely moved up! I mean it's not like they were going to just cancel Skylake-X or any so close to release anyway. Just release them all as fast as they can to try and capitalize on any money they could make from them, and drop CFL early to reduce the desire to buy Ryzen. We increase our core count in HEDT from 10 to 18, that's 80% this year, with a price reduction ~$1800-~$959 almost 50%. I think we can point at this as the most innovation this year based on competition for a single product line since the i9's were addon in response to 16 core ThreadRipper. Recap 80% more cores HEDT, and ~50% price reduction in the previous highest core count cpu.

 

[dullard]

they rely on Intel's regular yearly model release cadence, to advertise and sell their systems.

The regular yearly model release cadence isn't very regular (no matter how you measure it), nor very yearly (unless you count refreshes).

 

[whm1974]

OK so all of the low hanging fruit has been picked for improving desktop CPUs, however how much high hanging fruit is left on the tree? In other words I think there is some room left for increasing IPC, adding more cores, and lower the power needed by the CPUs. But it is getting harder to this and so will require more effort. And how much effort?

 

[Verndewdimus]

Its been said that 5nm and below will require exotic technology. I dont think we will see any major increases until we have light pipe tech, and photon transistors etcA photon transistor was developed by a team in france around 2006, but I dont know that we are any closer to really having an idea on how such things can be implemented. All this complaining about incremental advances is useless, as pc tech isnt even 100 years old yet, material sciences depend on theories that prove to be true or untrue, and chip technologies take exacerbating amounts of time and money to identify isotope qualities like those in the hafnium High K research days, were talking billions of dollars in research for a 10 - 30 percent performance gain , with a never ending latency challenge.

IF we were at a complete virtual reality online experience frame of technology, we wouldnt be griping about incremental gains, and were not that evolved as a society to even see what the world looks like in that setting , so what is a complaint about our current state ? you tell me how self realized and societally realized you arent to validate a complaint that cant take into account the myriad of factors that make this time what it is and isnt. No offense man, but heres your big reality pictture, 100 percent node shrink doesnt mean 100 percent perf gain, nor does one more core, one more thread or doubling the size of the cache, performance simply doesnt scale with our current tech , the way you want it to and no complaining will change that.

 

[PeterScott]

OK so all of the low hanging fruit has been picked for improving desktop CPUs, however how much high hanging fruit is left on the tree? In other words I think there is some room left for increasing IPC, adding more cores, and lower the power needed by the CPUs. But it is getting harder to this and so will require more effort. And how much effort?

More cores is obvious but is of limited utility for many tasks, and diminishing returns in many others. It's really only in "Embarrassingly Parallel" workloads that you can maintain gains to make it worth the cost of the extra cores. Most people don't run these enough to make much higher core count desktops a priority.

So for across the board improvement you are looking at improving IPC and/or clock speed. Though we are diminishing returns here as well. I doubt clock speed will see much enhancement on silicon. We would need some newer kind of tech.

So that leaves IPC. I suppose you might look to server only chips like Power 8 to see if there are some features there, that might eventually trickle down to the desktop, like desktop features trickle down to mobile.

AFAIK Power 8 is wider issue than Intel x86, but that doesn't seem to translate into a significant single thread gain, and even if it did, it's the kind of thing you are not going to keep iterating. It looks like every step on issue width costs much more, delivers much less, so again; Diminishing returns.

IMO, no matter where we look, we are at diminishing returns. It will probably take some kind of revolution in quantum computing, to deliver something completely different, to significantly move the bar.

For now, we have diminishing returns to look forward to. Not that this is a bad thing. We have crazy CPU power on the desktop these day, and you CPU is likely going to be good enough for as long as it keeps functioning.

 

[whm1974]

More cores is obvious but is of limited utility for many tasks, and diminishing returns in many others. It's really only in "Embarrassingly Parallel" workloads that you can maintain gains to make it worth the cost of the extra cores. Most people don't run these enough to make much higher core count desktops a priority.

So for across the board improvement you are looking at improving IPC and/or clock speed. Though we are diminishing returns here as well. I doubt clock speed will see much enhancement on silicon. We would need some newer kind of tech.

So that leaves IPC. I suppose you might look to server only chips like Power 8 to see if there are some features there, that might eventually trickle down to the desktop, like desktop features trickle down to mobile.

AFAIK Power 8 is wider issue than Intel x86, but that doesn't seem to translate into a significant single thread gain, and even if it did, it's the kind of thing you are not going to keep iterating. It looks like every step on issue width costs much more, delivers much less, so again; Diminishing returns.

IMO, no matter where we look, we are at diminishing returns. It will probably take some kind of revolution in quantum computing, to deliver something completely different, to significantly move the bar.

For now, we have diminishing returns to look forward to. Not that this is a bad thing. We have crazy CPU power on the desktop these day, and you CPU is likely going to be good enough for as long as it keeps functioning.

Thanks. I was thinking that was the answer or at least a good one for now.

 

[goldstone77]

Last paragraph of this blog seems interesting!

IEDM 2017: Intel’s 10nm Platform Process
Leave a reply
By Dick James

Back in March we were told that the 10-nm process shrinks beyond the usual 50% to 37% of the 14nm technology:


And that this actually brings them back on to a two-year cadence from the 45-nm node, assuming high-volume production as of the second half of this year.

It’s a bit close to the end of the year for that to happen, but if we see product in the New Year they won’t be too far off – we look forward to it!

I had hoped to fit in some commentary about the GLOBALFOUNDRIES 7nm paper given in the same session, but in the interest of brevity I will have to make a separate blog, maybe in the New Year.

And that this actually brings them back on to a two-year cadence from the 45-nm node, assuming high-volume production as of the second half of this year.

It’s a bit close to the end of the year for that to happen, but if we see product in the New Year they won’t be too far off – we look forward to it!

This would be an interesting change of event since Intel has been saying it's getting too hard, and that cadence will space out 4 years or more! That would be a surprising turn of events if all of a sudden things aren't so tough, and they magically return to a 2 year cadence! WOW!

 

[AMDisTheBEST]

Monopoly of windows and monopoly of intel on windows is the two deadly combos(wintel) that are fixing this market monopoly in place and creating stagnant competition and hence innovation.

The fact Apple uses exclusively intel cpu for its macs doesn’t help either.

For other platforms, like smartphones, we have not seen something like qualdroid yet.

 

[Verndewdimus]

we need cache that can inject quantum super positions so cache misses dont suffer latency from misses, with a superposition the cache is always right and wrong

 

[dullard]

Monopoly of windows and monopoly of intel on windows is the two deadly combos(wintel) that are fixing this market monopoly in place and creating stagnant competition and hence innovation.

The fact Apple uses exclusively intel cpu for its macs doesn’t help either.

For other platforms, like smartphones, we have not seen something like qualdroid yet.

Duopolies are almost negligibly better than monopolies (especially if they can differentiate their products enough in the consumers' minds to essentially have two monopolies such as AMD fans and Intel fans). For competition to really help, we need several more chip companies. But the cost of entry is just too high. Qualcom and ARM are going to spur on the industry far more than if we just stuck with AMD and Intel.

 

[TheELF]

Qualcom and ARM are going to do more to spur on the industry far more than if we just stuck with AMD and Intel.

Nope,the same thing will happen that happened to PC,qualcom samsung and apple are going to destroy every other company and only one of them will remain dominant.
When this happens ARM cores will still be way behind x86 in general performance.

(ARM is the one giving out licences they don't make anything)

 

[AMDisTheBEST]

Nope,the same thing will happen that happened to PC,qualcom samsung and apple are going to destroy every other company and only one of them will remain dominant.
When this happens ARM cores will still be way behind x86 in general performance.

(ARM is the one giving out licences they don't make anything)

There are tons of SoC makers. Arm isn’t just about the cpu, what these company really make are SoCs that use arm and they can either license the archeturture from Arm or create their own custom cores.

 

[kwalkingcraze]

i7-2600K was the last major redesign in CPUs, I think. Since then, new CPUs only have better integrated graphics and lower power consumption, and only slightly better CPU performance.

 

[Sonikku]

What sucks for WoW players is the game is single threaded so we need more IPC. But few are to be had, without paying out the damn nose for every scrap.

 

[PeterScott]

What sucks for WoW players is the game is single threaded so we need more IPC. But few are to be had, without paying out the damn nose for every scrap.

If all you need is single threaded IPC you can just get just get a fast dual core, and you don't have to pay through the nose for that.