OSNews

I still think they are. I will believe this when they deliver these processors. A roadmap is on paper. I think Intel has about as much a chance of delivering on this promise and Microsoft did "Cairo." Product problems will likely be the word of the day.

You do realize Yonah is a notebook chip and the other chips in that review were not. That it was able to perform reasonable close to an equally clocked desktop dualcore processor from AMD does nothing to support an opinion of it being an overpriced piece of junk.

>So far latest yohan benchmark show that intels processors are still overpriced pieces of junk:

Go you, so for twice the price, you get a slight advantage...And one is a laptop proc while the other is the desktop one.
So if "intel piece of junk" are overpriced, what are the dual core amd ones?

competes with PR. Promising it will make the next and greatest thing , while currently delivering nothing. When I will see any intel processor showing comparbale performance to amd offers I will believe it . The article tries to stress advantage of intel in manufacturing proccess which is irrelevant for the bottom line of the user (which is performance and price/performance ratios) .

So far latest yohan benchmark show that intels processors are still overpriced pieces of junk: http://anandtech.com/cpuchipsets/showdoc.aspx?i=2627

Piece of junk?

You sir, are a "I don't have a f*cking clue, try-hard techie" idiot on the highest level.

Without competition, everyone has to pay more. Did you even reach puberty when PIII was the only CPU that took the ladder? Do you remember what happened when AMD's K7 came along? Prices dropped.

Mobile CPUs have restricted requirements that are dictated by the application. They have to operate in tighter spaces, lower voltages, etc.

Desktop CPUs have less stricter thermal requirements, they don't need to save power, and they ain't designed/made to handle temps of 100deg/C. So efficiency is not a primary goal.

If you bother actually READING the article and carefully looking at the benchmarks, you'll see both the Pentium-M and K8 solution are pretty much even in overall performance.

The Athlon64 may win here, and the Pentium-M may win there, but overall, its not gonna be clear as night and day. It'll be highly dependent on the application being used.

And this is to be expected...

A64 (or K8) used the K7 as its basis. Obviously modified slightly, tweaked for more performance and added AMD64 capability.

Pentium-M used PIII as its basis (but heavily modified with some P4 features and power saving features).

If you look in the past, PIII vs Athlon, you'll see the Athlon (K7) will win in FPU intense apps.

Today, Pentium-M vs A64 is no different.
We're just repeating history all over again.

AMD will just edge out, but overall, most people won't be able to tell the difference. (Not like the P4 vs A64 difference).

This isn't just a PR exercise, its an admission from Intel that they are getting their asses handed to them by a competitor with a pretty good design, but is much smaller than them. (both in manufacturing and marketing resources).

Did you know the original Intel P7 project (not later Itanium) involved having a 32bit CPU with 64bit capability? Ironically, this is what the AMD's K8 is!

If you think about it, it felt like Intel made a major mistake. If they followed on with the original P7, we'd have a very different story today. (There would be no inefficient Pentium 4 or the ridiculously expensive Itanium).

Oh well, at least they're throwing the P4 out at the end of 2006!

Itanium would still be be here, Intel was working on Merced ever since they came out with the P6 (Pentium Pro) core.

While you are more or less correct about most of what you have said, I have to be nitpicky about three things:

1. Efficiency with desktop CPUs is just as important as mobile CPUs -- the Pentium 4 is a result of Intel ignoring this, and now look at what's happening.

2. The Pentium M might be more or less clock-for-clock with the Athlon 64, but ...
2a. The Athlon 64 has an on-die memory controller, and much, much lower memory latencies than the Pentium M (and Yonah). This is a major benefit when it comes to certain applications.
2b. The Pentium M (and Yonah) have absolutely no 64-bit capabilities. This is an automatic win for the Athlon 64 in terms of 64-bit code. Also, A64 + 64-bit code automatically leaves the A64 winning most benchmarks against the Pentium M running 32-bit code.

3. The K8 is much, much more than a "tweaked slightly" K7 core. Come on, you're a smart guy ... do some research please.

2b) You're very right about this one. 64-bit code gives the K8 another 10-15% usually. That margin would make AMD win almost all of the benchmarks, and lead in the others. The difference may not be huge, but a 20% margin at the same clockspeed* is nothing to scoff at.

* It should be noted that Yonah is 2.16 GHz at 65nm, while the X2 goes up to 2.4 GHz at 90nm. If AMD doesn't muck up the 65nm transition (that's not a trivial 'if'), the X2 is going to have a significant clock-speed edge over the Intel chips as well, in addition to performance per-clock. Either way, Conroe should be an interesting matchup.

I am seriously looking forward to Conroe and Merom ... they'll be the first Intel processors to be excited about since 5 years ago.

The derision of the P4 is undeserved. Prescott might have been silly, but the original P4 wasn't exactly a bad design. It was simply one that was optimized for scenarios that the market moved away from. Raw multimedia performance became less important, as GPUs offloaded geometry-processing and video decoding. Integer performance remained important, because game engines moved to more complex physics, and more complex culling algorithms to keep the GPU fed. Most importantly, the bottom fell out of the power situation.

The P4 managed to keep Intel on top for quite awhile during the K7 era. If the leakage current situation hadn't become so bad, the P4 would be at 5GHz as they expected, and still be a performance leader. If AMD64 hadn't taken off so well, Intel wouldn't have had to invoke the sub-par 64-bit support in the Prescott core. Really, the only thing bad about the P4 that its designers could have anticipated at the time was that it never really made sense as a server processor.

What exactly do we need with a 128bit processor? As it stands 64bit is widely available but not widely needed.

Shrinking the die size also means less power consumption and room for more things on the chip like more on-die cache, and even the processor that has 2 dual-core dies on it, which is something you suggested.

Also, did you read the article at all? They have plans for 8-core dies.

"Think about it, look at all the room in a computer, a laptop even, there's plenty of space in there to put dozens of cool chips, but no, they want to hog all the space with a giant heat sink water cooling apparatus."

That's because most software still isn't designed to handle multi-CPU well. One 3 GHz CPU will perform better than four 2 GHz CPUs (when 3 of those CPUs are idle and the CPU that is doing something has the extra overhead of SMP synchronization).

On second thought, though :
There are plenty of OSes and applications that are multi-CPU aware. Word and Outlook may not be examples of such, but then again, thats not where the extra horsepower is needed (e.g. scientific computing, cancer research, manufacturing etc)

One thing though is that Windows (don't know if Linux does or not, I'm sure it does) will distribute different apps onto the different processors, and you can even say that you want X processor to run on processor 2.

Should be "X process to run on processor N"

If you look at Moores law you'll realize that in the last several years nothing really has changed drastically in the processor world.

The processor is undergoing incredible changes right now. The goal of maximum single threaded performance - which they've been chasing for 34 years has all but gone.

We should be at processors 10x as fast as todays offerings, in fact the Cell processor is supposed to be the wave of the future, too bad it's hot as hell. huh?

At 3.2GHz it's not far above laptop CPUs.

Im not impressed at all with the way Intel is going, they keep banging their heads trying to make smaller and smaller dies so they can extract the most profit per process, when in fact they should be thinking about keeping the dies the same and go with more cores. Even 128 bit processing even.

All processors want smaller dies, it costs less to make them.
Intel's next Itanium will be one of if not the the biggest die in the industry.
128 processing would be useless in all but the most esoteric apps.

Think about it, look at all the room in a computer, a laptop even, there's plenty of space in there to put dozens of cool chips, but no, they want to hog all the space with a giant heat sink water cooling apparatus.

You haven't been following Intel very closely, they're almost completely focusing on low power now.

If chip makers would produce volume at a high enough NM and low power requireements, we could have three dozen processors in a laptop and it would be just or even cooler than todays machines.

That is planned but it'll take some radical re-design to achieve it. Software will need to be completely redesigned to make use of such a chip.

Edited 2005-12-05 18:27

"That is planned but it'll take some radical re-design to achieve it. Software will need to be completely redesigned to make use of such a chip."

Indeed. Why does anyone think AMD64 killed the Itanic? Poor design or performance? Now, it was the simple fact that the Itanium sucks att running legacy code. AMD64s don't.

In the end, it's all about software.

"Even 128 bit processing even."

LOL. Intel can't even get 64 bit right (do some reading into the details of EM64T and you'll see that it really is nothing more than a cheap knock-off of AMD64, and not a proper implementation).

I doubt that Intel will get EM64T on par with AMD64 before Merom, and based on their track record so far, I'm unsure I'd trust them to do it right even then.

If you want a good consumer processor, buy AMD. They aren't the ones who are going to be playing catch up for the next third of a decade.

LOL. Intel can't even get 64 bit right (do some reading into the details of EM64T and you'll see that it really is nothing more than a cheap knock-off of AMD64, and not a proper implementation).

I doubt that Intel will get EM64T on par with AMD64 before Merom, and based on their track record so far, I'm unsure I'd trust them to do it right even then.

EM64T in the Prescott was a quick&dirty hack, as was the dual core. With Merom these things have been designed in from the start.

If you want a good consumer processor, buy AMD.

Fair enough. But in a year's time the situation will look quite different, with Intel at least back on par.

They aren't the ones who are going to be playing catch up for the next third of a decade.

Hopefully AMD themselves aren't that complacent. They did well while Intel went down the Netburst cul-de-sac, but they should be very worried about Merom, especially now that Intel has more or less given up on Itanium and is properly committed to x86-64.

"Hopefully AMD themselves aren't that complacent. They did well while Intel went down the Netburst cul-de-sac, but they should be very worried about Merom, especially now that Intel has more or less given up on Itanium and is properly committed to x86-64."

Here's hoping.

I second that. Were at a point where we can advance software and hardware to a point where it is insanely fast and efficient, but so hard to maintain and work with that it just leads to more problems rather then solving new problems.

Why not rip out the x86 translation layer and implement it in software instead?

Because these days it's pretty insignificant compared to other things on a processor, particularly caches. And in spite of its rather baroque design x86 has one advantage over 32-bit RISC encodings: it's significantly more compact, thus saving valuable cache space.

SSE/SSE2/3 anyone?

Vector processing is overrated. On an out-of-order processor it doesn't make all that much difference whether e.g. the multiplication of two four-element vectors is expressed in one instruction or four, you still need to schedule and execute four multiplications. Vector instructions can reduce code size though, and the dedicated SSE register are nice to have as well.

I can easily use a 2 Ghz CPU. I can easily use 2. I could easily max out 4 or 8 cores, too, but I might have to do more than one thing with all those extra cycles. Usually I max out a 2 Ghz CPU for 2 to 4 hours at a time encoding video or compiling software. If I had more CPUs I'm sure I could find a use for them too.

I'm sorry you just don't see the need for such a processor. I have friends who do all their work on a Pentium 133 on Linux in a text console. They won't see a need for such a processor either. There's no need for graphical email and web. If you really get down to it, from a client perspective, nobody really needs more than a couple hundred Mhz of CPU and about 25 lines of text at 9600+ bps. So why waste our time with your post?

I'm sorry you just don't see the need for such a processor. I have friends who do all their work on a Pentium 133 on Linux in a text console. They won't see a need for such a processor either. There's no need for graphical email and web. If you really get down to it, from a client perspective, nobody really needs more than a couple hundred Mhz of CPU and about 25 lines of text at 9600+ bps. So why waste our time with your post?

You don't represent the average Dell PC buying crowd. You are probably in the top 1% of power users that just buy the latest and greatest hardware becasue you think it makes you look cool. The best way to mesure the PC power market is to look a the latest and greatest game out today and see what the minimum system requirements are. I will give you a hint...it hasn't changed much for more then a year so. People like you suck up the hype and believe people need 4 core cpu's on their desktop ultimately it just amounts to intel selling processors.

Probably you are right: "Intel is almost as over as Microsoft is". In fact I don't see either of them being dead anytime soon. And this comes from a long time Mac user who has had his good share of reasons to hate Microsoft.

Even in browsers, a market where Microsoft is losing ground I doubt anyone believes Firefox will win the battle if by winning you mean being the most used browser in the world. I think even Firefox programmers know that: they will hit their target, have their success, if they manage to get, say 15-20% of the market. That alone will push everyone to consider Firefox as a reality, for example pushing web-developers to use true W3C standards.

Similarly Apple will "win" if they manage to get 10-15% of the marketshare, which will push many more developers to program for it.

Looks like they will go there: http://tinyurl.com/bt94w

That will certainly be an interesting time, when quantum effects start to affect processor performance and design. If anyone cares to potificate on that whole scenario, please do.