Is mainstream desktop CPU development "completed"?

[frozentundra123456]

We seem to be coming back to these types of thread all the time. One thing that I have not really heard, is that the lack of progress on the desktop is also partially due to lack of a demanding killer application.

I dont agree that Intel is purposely holding back per core performance or performance per watt improvements, but if there was a popular consumer app that needed six cores, they could easily make a mainstream hexcore. But really, why should they? Most consumer and business apps still run more the fine on a pentium or i3.

 

[SOFTengCOMPelec]

We seem to be coming back to these types of thread all the time. One thing that I have not really heard, is that the lack of progress on the desktop is also partially due to lack of a demanding killer application.

I dont agree that Intel is purposely holding back per core performance or performance per watt improvements, but if there was a popular consumer app that needed six cores, they could easily make a mainstream hexcore. But really, why should they? Most consumer and business apps still run more the fine on a pentium or i3.

But that is a sort of chicken and egg situation.

If Intel made all mainstream (including entry level) processors, start at six core. Then in time, software would begin to (by default), use the extra processing abilities of the six cores.

A bit like some games console software does. Since they know it has 8 cores (or however many it has), so they can write software to exploit that fact.

As an example, consider the new instruction sets (such as AVX2, AVX3 etc), which come out, every few years (approximately). It is reckoned it takes a very long time (I think the figure was 10 years), BEFORE the new instructions are significantly used, by at least some of the software.

It is very hard to buy a brand new computer (with some exceptions at the bottom end, and/or VERY tiny computers), which is not at least dual core these days. Despite this fact, lots of software is mainly single core. Even when it would be very easy to use the extra cores, usefully.

E.g. Flash (not very long ago), would only ever use one core (if I remember, correctly). Much to my annoyance, because that would sometimes cause it to run very slowly.

Edit: tl;dr
Software producers don't want to waste time, writing software for many cores, when most computers, don't have many cores. (> Quad core, is still quite rare, on desktops).

 

[StrangerGuy]

We seem to be coming back to these types of thread all the time. One thing that I have not really heard, is that the lack of progress on the desktop is also partially due to lack of a demanding killer application.

I dont agree that Intel is purposely holding back per core performance or performance per watt improvements, but if there was a popular consumer app that needed six cores, they could easily make a mainstream hexcore. But really, why should they? Most consumer and business apps still run more the fine on a pentium or i3.

Sorry, they are holding back because more cores means unnecessary costs to them and also means pushing the "already way good enough" envelope further which makes future chips an even tougher sell when their current =<4 cores are still selling fine...However that has everything to do with the sheer inertia of the x86 Windows ecosystem rather than performance, IPC, process etc that this subforum likes to harp about.

 

[TheELF]

Sorry, they are holding back because more cores means unnecessary costs to them

It also would mean unnecessary costs to us as there is no need for 6 cores in mainstream, that's what haswell-e is for.

 

[mikeymikec]

I personally think that PCs will only need a maximum of 640KB RAM, and that there will be a need for maybe a maximum of five mainframes globally.

Innovation usually is a result of a desire or need for something. Does the average person need a desktop CPU that's say 25% faster/more efficient than ones available today? Is it sensible to assume that the average person won't ever need/want more from their PC than what they can get right now, is that what history teaches us?

IMO the emphasis on the cloud has nipped desktop/laptop development in the bud for the time being. Where I once expected a PC at home to be taking over and hooking up to various systems in the house, I think cloud services will be taking on that job.

 

[ninaholic37]

I personally think that PCs will only need a maximum of 640KB RAM, and that there will be a need for maybe a maximum of five mainframes globally.

:awe:

I think the secret to being a satisfied customer nowadays is to not visit CPU forums, and not pay attention to Intel/AMD news at all, and only buy a new desktop when your old one dies. Sort of like "hibernating for 12 years then waking up again only to fix it". Perhaps then, the improvements, if any, might be more noticeable. If not, you had to buy a new computer anyway, so no loss (hopefully). Same thing goes for buying a new car, I guess (if you like driving vehicles...).

 

[TheELF]

As I said, you can get haswell-e if you really need it,doesn't mean that intel should force you to spend more on 6 core CPUs by making its whole mainstream line hexacores.

 

[alcoholbob]

http://anandtech.com/bench/product/287?vs=1260

?

Let's see, you are comparing a 3.4GHz CPU from 2010 that can easily overclock to 4.5GHz on the stock cooler, versus a 2014 CPU that at stock turbos to 4.4GHz, and after overclocking you might hit 4.5GHz with an overclock.

So after 5 years you have 10% better performance. Thank you for pointing that out.

 

[Fjodor2001]

As I said, you can get haswell-e if you really need it,doesn't mean that intel should force you to spend more on 6 core CPUs by making its whole mainstream line hexacores.

With that kind of reasoning we'd be stuck with 1 core desktop CPUs. More cores need to go mainstream before they will be used by applications. It's a chicken and egg scenario, like others mentioned before too.

But this discussion is not only about number of cores. We're not seeing much frequency or IPC improvements either.

 

[Fjodor2001]

As I said, you can get haswell-e if you really need it,doesn't mean that intel should force you to spend more on 6 core CPUs by making its whole mainstream line hexacores.

In that case, do you think it's ok to pay for an oversized iGPU that will not be used to it's full potential by most people anyway?

Or should that be removed or shrunk too?

 

[mrmt]

I think some kind of Linux gaming console/desktop could be competitive too (Example: Valve offers a way for developers to launch games in Steam by community vote) , but we will need to see Intel and other companies make further advances in cheap relatively high performance desktop chips (like the G3258).

I don't see the problem of gaming PC as a CPU issue, but more of an ecosystem issue. Console games are more profitable to the companies that develop games, and Microsoft strategy for gaming doesn't pass through windows anymore, they are fully committed to the Xbox. Had Microsoft fielded DirectX 12 years ago two years ago we would be now seeing a lot of new games using the new APIs, games that wouldn't run as good in the consoles because of the low cost hardware that both console companies decided to use, and that would kill momentum for their console. If they launch DirectX 12 by the end of this year DX12 gaming will only gather steam by the last part of the console cycle, actually enhancing their case for a new generation. There's the competitive issue I talked earlier in this thread, Microsoft is too busy trying to compete with their business partners instead of enhancing its platform to the benefit of both. All these things don't make developers spend money in tools that would be used exclusively on the Windows PC, because the platform isn't really attractive anymore for established players, and even less attractive for new players. (This is why Windows 8 is such a disaster, because it destroyed the last Microsoft stronghold, Productivity, but that's for another thread...)

Vanilla Linux is simply unsuitable for gaming applications because of its business model. GPL is not everyone's cup of tea and things like their driver model won't work with the GPU companies, plus there is the matter of the API. Mantle was poised to die because of its business deficiencies, and Vulkan is about to get these same deficiencies, so the Linux gaming picture is not that bright. The effort Valve needs to put on their SteamOS is much more akin to what Apple did with FreeBSD or Google with Android, than what they are doing now, and I'm afraid the TAM for such venture is not big enough, especially because the steam boxes will have an inherent cost handicap against the consoles.

 

[mrmt]

If Intel made all mainstream (including entry level) processors, start at six core. Then in time, software would begin to (by default), use the extra processing abilities of the six cores.

Intel fields quad core/thread processors since 2007 but that doesn't prevent some software houses of not using them at all. But let me ask you, what is the mainstream application that *needs* six cores at 4Ghz?

 

[NTMBK]

In that case, do you think it's ok to pay for an oversized iGPU that will not be used to it's full potential by most people anyway?

Or should that be removed or shrunk too?

1. That is just one SKU, others are available with far smaller GPUs.
2. With high res displays increasingly common in laptops, they need that extra horsepower.

 

[cbn]

the steam boxes will have an inherent cost handicap against the consoles.

Yes, I think the current Steam boxes are priced too high.

What I think Valve needs to do is make some kind of nice looking slim enclosure with riser that folks can put their old DIY uATX/ATX hardware* (or budget new hardware like the G3258) in so the gear actually looks presentable in the living room.

That would be a lot cheaper than the expensive boxes OEMs are offering. And in many cases I'll bet even the old hardware will be faster than the processors I see some of the sub $500 steam boxes shipping with.

* the hardware that no longer uses the retail Windows license their new upgrade hardware is now using.

 

[mrmt]

That would be a lot cheaper than the expensive boxes OEMs are offering. And in many cases I'll bet the old hardware cpu will be faster than the processors I see some of the sub $500 steam boxes shipping with.

That would make harder for OEMs/ODMs to balance their supply chain, because basically they would be selling Intel hardware that Intel (or Nvidia, or AMD) wasn't currently selling anymore, that will make harder to place orders and manage RMAs.

Sony and Microsoft go to the market for a custom chip for cost reasons and for supply chain control reasons, by going with standard PC hardware they give up both, hence the high prices of the steam boxes and the need to use current hardware.

 

[cbn]

That would make harder for OEMs/ODMs to balance their supply chain, because basically they would be selling Intel hardware that Intel (or Nvidia, or AMD) wasn't currently selling anymore, that will make harder to place orders and manage RMAs.

I am actually thinking on a much smaller level than OEMs/ODMs selling used hardware.

Think individual user of tech with the desire to find a new use for that old motherboard they have. Or maybe the person who likes to buy those inexpensive G3258/motherboard bundles we see on sale at Newegg, etc.

Though certainly some kind of slim form factor case with riser (that can handle uATX or ATX board) could be used by any company if they wanted build a new system as well.

 

[ShintaiDK]

The problem is software. And thats not going to be solved anytime soon. Servers on the other hand already got 18 cores with 36 threads.

Its really hard to justify any core count above 4 on a desktop PC unless you do something silly all day and belong to a niche. For mobile I have trouble seeing the purpose at over 2+HT. Unless then better IGP and the usual 5% IPC plus power savings and more integration is all we get.

 

[cbn]

Its really hard to justify any core count above 4 on a desktop PC unless you do something silly all day and belong to a niche.

Why not go a step further and get rid of those quad cores. Bring mainstream desktop down to two cores?

Pentium = 2C/2T
Core i3 = 2C/4T
Core i5 = 2C/6T
Core i7 = 2C/8T

The cores would be wider, of course, but I think we are at the point in the frequency voltage curve where wider, but lower frequency cores would still be more efficient than going any faster with the current width cores. (especially when the additional threads per core are factored in)

 

[ShintaiDK]

You are not getting anything at all. Hence why we got turbo modes. A quadcore with 2 threads runs as fas as a dualcore could.

But perhaps adding 15-20MB of cache or something would be the best thing in terms of here and now increase. As seen with the bigger LGA2011 dies.

 

[mrmt]

I am actually thinking on a much smaller level than OEMs/ODMs selling used hardware.

Think individual user of tech with the desire to find a new use for that old motherboard they have. Or maybe the person who likes to buy those inexpensive G3258/motherboard bundles we see on sale at Newegg, etc.

This is exactly what they aim with SteamOS, but as I said, tweaking a linux distribution as they are doing today won't cut it, they have to fully rebuild the OS with a different set of business premises, as Google did with Android. As much as Mantle was AMD's Argo movie set to force Microsoft towards the direction they wanted (Hey, at least give some attention to the PC ecosystem even if your gaming strategy is still the Xbox), I think SteamOS is Valve's Argo movie set to force the Linux foundation towards the direction Valve wants (stable driver model, GPL improvements, etc).

 

[Fjodor2001]

1. That is just one SKU, others are available with far smaller GPUs.
2. With high res displays increasingly common in laptops, they need that extra horsepower.

1. Yes, but at least there are mainsteam SKU:s with a comparatively huge iGPU:s available on the market. But there are no corresponding maintream SKU:s where the CPU has been prioritized to a similar degree, e.g. with 6-8 CPU cores (and I don't consider the HEDT platform mainstream, but enthusiast).

2. What average non-gaming tasks would require an iGPU at Iris Pro 6100 class even when using a 4K display?

 

[Fjodor2001]

But let me ask you, what is the mainstream application that *needs* six cores at 4Ghz?

It's not invented or put on the market yet, since six cores are not mainstream yet, so there is no market for it yet...

Having said that, I actually don't think 6 CPU cores would be sufficient to result in any "revolution".

But let's say there was some real disruptive leap head, so we got e.g. 10-100 times higher mainstream CPU performance (regardless if via frequency, IPC or core increase). Then I'm pretty sure there would start popping up lots of interesting innovations making use of all that CPU power...

 

[ShintaiDK]

We had 6-8 threads mainstream/performance support since 2010.

And the hardware is not the reason why, its the software. And the very wast amount of software just cant be parallelized enough for it to make any difference vs the core count we got and some dream of.

So unless someone comes up with something nobody else thought about before in history that will revolutionize software multicore scaling coding to imaginary levels. Then its not going to change anytime soon. Even with only 25% serial code, more than 4 cores isnt worth it.

 

[Fjodor2001]

We had 6-8 threads mainstream/performance support since 2010.

I assume you mean CPUs with 3-4 cores and 6-8 threads? But 6-8 threads is not the same as 6-8 cores. You don't get as much benefit from multi threaded software on 6-8 thread CPUs as you do on 6-8 core CPUs. Hence less incentive to develop multi threaded software.

 

[SOFTengCOMPelec]

Intel fields quad core/thread processors since 2007 but that doesn't prevent some software houses of not using them at all. But let me ask you, what is the mainstream application that *needs* six cores at 4Ghz?

EDIT: Maybe, at the moment, it is NOT mainstream stuff, which benefits most, from having many cores.

Analogy:
It's a bit like a normal calculator versus a scientific calculator.

Most people are fine, most of the time with a normal 4 function, 8 digit, calculator.

But some people, some of the time, need a scientific calculator. Because they are at University, or doing engineering/science calculations, etc.

Most people, most of the time, would be fine, with a good clock speed, quad core computer.

The six core (or more), processors, would be useful to some people, some of the time, who want to perform, long/complicated video edits, play high end video games.
There are also rarer uses, such as compiling software.

(I DON'T know the actual figures) I would guess that something like 10%..20% of calculators, are scientific ones (as opposed to basic 4 function 8 digit, ones).

So maybe 10%..20% of desktop computers *need* the extra power of the 6 (or more) cores. At least some of the time. Such as when they are doing big video edits/conversions, or playing some of the higher end video games.
They may additionally have a rare hobby/work use, such as compiling stuff, or whatever interests that person, which needs heavy computer power.