When will we have CPUs that are 10x as fast as the current ones?

[happy medium]

yeah I think it is going to take at least 15 years for any company to figure out how to do that without making it cost $3k or use 400W.

hmm great minds think alike.

I'd say about 15 years or so. 30 core desktops @ 5ghz , with 3x the IPC will be normal by then.

 

[TheELF]

Intel make more profit selling smaller dies than large ones.

Of course they do,I believe their biggest money makers to be celerons and pentiums,still they produce the I5s and I7s not to mention the hex and octa ones.
So what?

Intel won't give us mainstream over 4 cores because the benefits for a mainstream user are close to zero,game performance even drops on the I7 for almost every game.

If there where a big difference, for the common user, they would bring out such a cpu for the mainstream faster then you could blink,because they would make a lot of money.

 

[Dresdenboy]

Let's start from the beginning. IIRC an high inherent ILP of x86 code is about 5 to 8. If we already get about 2 to 3 out of the code, there isn't much left to do.

 

[sm625]

Hi,

I just wonder how long you think it'll take until we will have CPUs that are 10x as fast as the current ones?

10x as fast at what? Just all around general performance? Lets take some common metrics for a current near top end PC:

Cold boot time: 30 seconds
Time to launch web browser and restore 10 open tabs: 4 seconds
Sunspider score: 150mS
Random video encode time: 40 seconds

Within 5 years I expect CPUs to have not only HBM but also HBF (high bandwidth flash). This will allow the OS to be stored on the SoC. It will then be possible to maybe deliver a 10x faster cold boot time of 3 seconds. Likewise the web browser launch time could easily hit 0.4 seconds with HBF. It is also likely that video encoding time could be reduced by 10 fold. But this would requiring some sort of "cheating" ie quicksync. The encode time would be faster, but you would lose some control over the process.

But for the Sunspider score? I'm not sure we'll see that heading to 15mS in 5 years. I think the absolute best score is 78mS today, but most typical highish-end PCs are almost double that. It will be close.

 

[Fjodor2001]

10x as fast at what?

Answered here.

 

[Cerb]

Within 5 years I expect CPUs to have not only HBM but also HBF (high bandwidth flash). This will allow the OS to be stored on the SoC. It will then be possible to maybe deliver a 10x faster cold boot time of 3 seconds.

No, it won't. Today, 30s is a pretty long boot time. It can be that long on high end PCs because they are high-end PCs, and have a ton of features that need to be initialized. A good $80 mobo can boot in half that time or less (even booting MBR); and a big OEM notebook, with less peripheral support to be concerned with, can boot in just over half that time again, with Windows 8+ on EUFI.

Likewise the web browser launch time could easily hit 0.4 seconds with HBF.

How are the CPUs going to improve enough to do that in 5 years? I don't see disk busy time being much of an issue, opening a browser. It's mostly CPU, once you're on an SSD.

 

[Enigmoid]

BIOS bootup on an OEM notebook is ridiculously quick. Like 1-2 seconds - fast enough that you have to look up what key to press to enter setup/BIOS/OEM selections (which may be non standard) because the screen flashes by so quickly you can't read anything.

 

[ShintaiDK]

My boot time is around 6 seconds with quickmode in BIOS.

To enter BIOS you blindly need to spam the key.

 

[Dresdenboy]

SSDs might do a lot to a system's performance. I have a gaming HTPC made of older components plus a new, cheap mobo. Athlon X4 635, 8 GB, Radeon 7950, 256 GB SSD, Win 7. Boot time is about 8 s. Desktop appears before the 4 color dots splash on the screen to build the logo.

 

[StinkyPinky]

Is boot time even an issue today? My PC boots up in maybe 10 seconds, which isn't much longer than it takes for my TV to turn on.

 

[tweakboy]

I would say 2040 ...

 

[Ken g6]

If you wanted it TODAY, then yes, I think a 10x processor could be developed with a targeted ASP of $3k and 400W power consumption envelope...but the pricepoint itself depends on there being enough people who actually will buy the thing.

I wonder what Intel could do with current high-core Xeons and a 400W TDP? 16 cores at 4GHz? :awe: Probably at least $5k, though.

Yes, I'm aware that the 5820k hex core dies could be smaller, since they contain more cores that are disabled (from the xeon range), though Intel obviously feel that this is the most profitable way for them to do it.

I hear they actually saw through the chips to cut back on cores. Maybe they can produce 2 smaller chips this way? I'm not sure.

 

[JimmiG]

Problem is CPU performance don't increase as much per year these days as it used to, so history does not tell us much.

This is unfortunately true. I miss the days when you'd buy a 500 MHz CPU, only to have a 1.2 GHz chip come out a year later, followed by a 2.4 GHz the following year and then a 2.4 GHz dual-core chip etc.

That just isn't happening anymore. We're just getting slightly tweaked versions of the same 3.5 - 4 GHz, 4 to 6 core CPUs every year.

 

[Dresdenboy]

I wonder what Intel could do with current high-core Xeons and a 400W TDP? 16 cores at 4GHz? :awe: Probably at least $5k, though.


I hear they actually saw through the chips to cut back on cores. Maybe they can produce 2 smaller chips this way? I'm not sure.

The dual core/single core Prescotts offered this "ability" by cheaply creating a dual core by connecting 2 adjacent dies to a on chip SMP system.

Those Xeons have their core surrounded and connected in a way preventing such measures. Instead they create different products while saving production costs (mask sets, different test systems, etc.).