Intel Core generations?

If I'm not mistaken Arrandale is the first generation Intel Core processors and today's Haswell is the 4th generation. But why do they calculating from Arrandale, when earlier we had Core Duo and later Core 2 Duo too? Are these two generations not calculated? Should not be called today's Haswell as 6th generation?

 

Core was based on Pentium M. Core 2 used the Core uarch. Core i have been based on Nehalem, Sandy Bridge, and Haswell. All very different although Intel kept the Core brand name the same.

 

It is quite complicated. Gone are the days of simplicity. Now we just have as much obfuscation as possible. It isn't "Core" generations, it is "Core i" generations.

The first Core processor was Yonah, based on the Pentium M architecture.

Then came Core 2, based on the Core architecture.

Then came the first generation Core i, based on the Nehalem architecture, of which the first processor was Bloomfield, not Arrandale, by a long shot.

Then came the 2nd generation Core i, based on Sandy Bridge.

The 3rd generation Core i is also based on Sandy Bridge, but the 22nm Ivy Bridge die shrink.

The current 4th generation Core i is based on Haswell.

In total we have 6 different Core generations based on 5 different microarchitectures, of which there are 4 different Core i generations based on 3 different microarchitectures, none of which are actually the Core microarchitecture.

Another way to think of it is that Core is 32-bit, Core 2 re-introduced 64-bit, and Core i re-introduced hyperthreading.

 

Yeah; Core Duo/Core 2 Duo does not equal Core i...

Haswell is 4th gen and it is at least two orders of magnitude above Core (Pentium M).

 

Uh-huh.....

 

Gordon Moore be rolling over in his grave right now...

 

So, we see how fast we forget how slow things used to be huh?

When Core products ruled the earth (right after the Pentium M...) the passmark scores were somewhere between 200 to 400 'points'.

Try using one of those to do anything with today (right 'up' there with the AMD E350 apu's, score-wise, but single threaded/single processor based - it's much worse in real world and 'current O/S' usage).


If we take specific tasks (such as transcoding video, ), the gap is much, much more than just two orders of magnitude, for example.


I never said they had 100x the transistors (sigh... it's only about an order of magnitude there, not two) - but the perceived performance advantage in modern workflows. Don't forget to include the new hardware instructions that the years have brought us.

Yeah, I know what you're thinking (I've seen the thread): nothing has changed since SNB... how wrong that stance is.

 

Good lord I haven't. I recently fired up a vintage Dell Pentium D system and holy cow, dual Prescott was so slow, even in the HDD era. Stuff just takes forever to happen and you get these random slowdowns. And the best part? the "stock" cooling system Dell provided was the size of your average modern day enthusiast air cooler.

I think the trend now as time goes on is people won't be upgrading processors so much for raw perfomance like ye olde days but more for the features of the chipset + CPU each generation brings. X58 era brought the IMC and turboboost, Arrandale introduced a functional on chip GPU, Z68 ushered in chipset SATAIII, Z87 finally introduced semi-universal USB3.0, Z97 introduced chipset M.2 support.

 

The pentium m was massively potent for its day and while restricted to 32 bit I would still take it over some stuff that ships today, it would make the e 350 look like a joke if it was made on a modern process.

 

I still have a Toshiba laptop with a 2Ghz Yonah Core Duo in it running Windows 7. Browsing the web, doing office tasks, and watching movies with it are fine. I can barely tell the difference between it and a more powerful computer for doing those tasks. It normally doesn't hit 100% CPU load.

It is definitely more powerful than the netbook I have that uses a single core 1.7Ghz Athlon K125. This computer does suffer from performance issues related to the processor, but the processor is a bit less than half as powerful as the Core Duo.

Where do you get this nonsense from, really? Not even a 12-core Xeon is anywhere close to 100 times faster than a Core Duo......

If we say that Haswell is 2x as powerful, clock for clock, than Yonah, when running at full turbo on four cores, an i7-4940MX is still only 6.5 times as powerful as a 2.33Ghz Core Duo T2700.

 

Thanks for the explanation. So what is Arrandale than? Or can we say Arrandale is a 2nd gen based on Nehalem, Ivy Bridge a 2nd gen based on Sandy Bridge, Broadwell is a 2nd gen based on Haswell?

 

Pretty much. Nehalem microarchitecture is 45nm and 32nm. The 32nm die shrink of Nehalem is called Westmere. Arrandale is a Westmere processor. Bloomfield is 45nm Nehalem. Up to this point in time, Intel was giving random names to everything, which makes it very confusing. Starting with Sandy Bridge, they reduced the random name clutter a lot.

 

Arrandale is the code name for a family of laptop CPU's; Westmere is the actual uarch which itself is a die shrink of Nehalem.

 

Ok, but still not sure than what is the first generation mobile Core I processor than? For example I thought Core i5-560M is a first generation, but it is not bloomfiled, instead the newer Westmere (Arrandale).

 

Clarksfield 45 nm i7 quads were the first ones. Clarksfield is the mobile version of Lynnfield. There was never a mobile version of the older Bloomfield.

 

I think it's called 4th gen because it's the 4th gen of the "true" quad cores (ignoring the C2Q/C2QX chips). We have Nehalem -- > Sandy Bridge --> Ivy Bridge --> Haswell

 

It makes no sense because Ivy Bridge is a die shrink of Sandy Bridge. If we're gonna count die shrinks, we might as well insert Westmere between Nehalem and Sandy Bridge (the Gulftown chips were still called Core i7). Which would make Haswell the 5th generation Core i series.

 

Arrandale is the first mobile processor with IGP in one package, top of it with DDR3 Support and Hyper Threading. That is where the generations begins.

 

Except that Clarkdale has no IGP and is also first generation Core i. Bloomfield is also first generation Core i, has no IGP, and came out well over a year before the mobile Core i processors. DDR3 isn't part of it either, as DDR3 support first came out on Core 2.

The only things that really differentiate the Core i processors from Core 2 processors is the integrated memory controller and hyperthreading.

But really, it is pointless to argue about this stuff. The names we are discussing is 100% marketing department branding and doesn't really have anything to do with the inherent properties of the hardware.

 

True but Westmere only had hex cores for desktop i7 chips, the quad cores were limited to Xeons only. Even mobile i7 Westmere (Arrandale) chips were dual core. Hence the focus really was on quad core.

 

Qing; no nonsense. And you're confused with Core Duo (not talking about the second generation...).

When Intel cpu's transcode video at up to 10x faster than AMD and even some gpu's can (sure; possibly at a different quality) you think that a 10 year old cpu stands a chance? Lol...


See:
AMD Richland vs. Intel Haswell: Integrated Graphics Performance Review. Page 7 - X-bit labs



No doubt there were some great mobile platforms from yesteryear - but they couldn't power a watch today (yeah; exaggerating ) and the reason that they're not made on current process nodes is that they would still be inferior as a platform choice in a modern notebook.

Time marches on and so does the way tech is measured - it is not a single metric that counts (Qing... comparing Hex Core Xeons to Core Duo - lol...) - but the platform as a whole that gets WORK done.

I'm comparing platforms from about a decade apart - I don't think I would be just '6.5' times less productive with an oldie but goldie Core - I think I would literally starve to death trying to get one job done.

 

I'm not confusing anything. A Core Duo is a first generation Core processor.

Can you even comprehend those benchmarks? They are testing the multimedia hardware acceleration of several IGP's, including using software that wasn't updated yet to take advantage of hardware acceleration on the AMD IGP. And if you look closely, you see that the Haswell i5 gets destroyed by the Ivy Bridge i3.

So your conclusion is that Haswell processors are more than 100 times more powerful than Yonah processors, just because you feel that they should be?

Yeah, Haswell processors are better than Yonah processors at everything. But the most powerful Haswell mobile processor is only about 6.5 times more powerful than the most powerful Yonah processor. That really is a huge difference, but it only seems small when compared to your imaginary claims. Of course, the difference can be more or less, depending on what you are doing, but based on some research comparing different cores, it sounds about right that on average Haswell gets 2x the work done per core per clock cycle compared to Yonah.

 

Qing, sorry I'm having a very hard day and you're just too argumentative (all the time) to deal with.

You think you're right in your mind (you're wrong). Try reading for comprehension and a possible new way of looking at things.

And please; don't try to make a point by making things up and/or making conclusions for others. Makes you look stupid.

Especially when you do it repeatedly.

And I don't think you are.

Thanks.

 

Well who are you to say that you're right and others are wrong?

You do it repeatedly, or, should I say, in every post you make, and it makes you look stupid and bigoted.

 

No, I'm trying to have a conversation. You? Not so much...

 

So talking down on people is your definition of a conversation? Over here we call that preaching.

 

That is the problem with the written word (and a poor writer: me).

Again; No, not my intent at all. Try to read the intent as well as the content: I am trying to have a conversation (really) that ultimately results in sharing what we each know.

Thank you for at least trying (by asking questions) to participate.

 

I find it ironic (and somewhat hypocritical) that for someone who makes statements like these:

that you should end up using these "scores" that you oh so disapprove of to try to make a point.

(yeah yeah apples and oranges scores whatever, it's the principle that counts)

 

Yeah; it is ironic. But at least I'm trying to show in a way most of you understand, right?

The 'principle' to me is that we exchange something meaningful. I don't care looking 'ironic' while doing so.

 

Tiller, I'm not the one going into threads spreading nonsense I dreamed up as hard facts. This stuff you are saying about Haswell being at least two orders of magnitude more powerful than Yonah ranks pretty high up there on the bologna meter. I mean come on. Think about it this way: Haswell is the sixth processor generation (Clarksfield and Arrandale are counted as one) after Yonah. In order for us to reach over a 100x performance difference after six generations, each successive generation needs to be greater than 2x as powerful as the last. This has never happened; although it was close, not even going from a dual-core Merom Core 2 Extreme X7900 to a quad-core Penryn Core 2 Extreme QX9300 reached a doubling of performance.

However, I don't think anything you will ever say on this forum can ever top you claiming that you could feel differences in time down to the microsecond range.

 

**BOOKMARKED**

If I need a lesson on how to talk down to people in a hypocritical manner I'd refer to this thread!