When Tuesday does the G70 NDA expire?

Because they can earn more with the Expensive Editions?

 

No, it's not. It's an evolution of the Pentium III architecture, like the Pentium III is an evolution of the Pentium Pro architecture.

 

They sell about 2 Extreme Editions a month. It's just a willy waving part with no grounding in any kind of sales or performance reality.

It's more to do with customer expectation. Introduce a 4GHz part and customers want to see the rest (4.2, 4.4, 4.6 etc) of the range, since they think you'll finish what you started, since that's been the kind of clock progression PC customers have been used to for over a decade. Incremental steps from one speed to the next.

The P4's gone from 1.3GHz to 3.8GHz with nearly every step in between. Intel start with 4GHz and they're on a road to nowhere unless they can work miracles with 65nm.

The 1.13GHz P3 incident hurt them, with customers losing confidence in the P3 getting any faster as a result. Sales of the Athlon which went up past 1GHz without much issue (although it did get really hot ) went up pretty sharply iirc, because of that.

So while 4GHz P4s exist in their labs, marketing refuse to let them loose I reckon, since much higher than that is asking for it. Dual-core P4 will rule inside of Intel instead for a while, Intel happy to let AMD kick their ass in the single-core high-end since what the hell, Dell sell a bucketload to clueless customers anyway, so who really cares, right?

Edited to clarify what I mean about rule

 

Except that AMD also has dual-core processors, and once the load-balancing issues get ironed out, they'll also be better-performing under heavy load. AMD could also stand to do something similar to Hyperthreading to further improve performance.

 

Keyword "clueless".

 

which incident was that?

 

Intel released a 1.13GHz P3 not too long after the 1GHz. It was to be a low-volume product, but it turned out to be unstable. Intel pulled it, and I don't think it ever made it to market. Instead the P4 made its debut later that year.

 

yes, but nevertheless they continued with the tualatin which went considerably above 1.13

ed: googled for the 1.13 incident - appears it was with the coppermine.

 

Oh, right, I guess I'd forgotten that. I'm willing to bet that outperformed the first P4's by a decent margin, too. Intel really screwed the pooch with the P4 architecture.

 

The Tualatin cores were faster then the first higher clocked P4s. However, anyone that claims Intel "screwed the pooch" with the Northwood core is fooling himself.

At any rate, Intel will be launching dual core Pentium Ms next year with higher clocks, 667 MHz fsb and x86-64, so we'll see how they perform. The problem with the Pentium M architecture is that is performs exceptionally well in some applications (outperforming the A64 at lower clocks in games for instance) but relatively poorly in other applications like encoding. Intel will have to change the architecture before they adopt it as their primary processor.

As for AMD's future. Their latest roadmap shows up to the FX-61 at 3.2 GHz and the X2 5200+ at 2.8 GHz. 3.8 GHz for AMD is a long ways off, if they even make it that far.

 

They may have benefitted from better marketting from the higher clockspeeds, but the P4 architecture starts from just a bad idea: high clocks at the expense of IPC, and now they're reaping their rewards. Not that they haven't won by doing this in the past, but this time Intel's been hit by the big stick of basic physics, and the P4 architecture has come to a grinding halt.

The Pentium-M architecture, on the other hand, has two big advantages moving into the future:
1. It's designed to be a low-power core. Power consumption is going to become a larger and larger limiting factor moving into the future, so this will serve Intel well.
2. It's much higher IPC than the P4. Since clockspeed is limited by basic physics, the higher IPC of the Pentium-M will give it higher performance in the long run.

Yes, Intel may want to work to shore up the situations where the Pentium-M underperforms the Pentium 4 if they make the architecture switch, but I doubt it'll be that big of a deal.

 

^^^^ Sure sign we're already bored with the new release, less than a week later.

Quick, somebody find something juicy in it. New optimization, something.

Or. . .maybe 16 hidden pipes. 8) (tho I think Wavey is sincere, even if it looks to me like that's a thin branch he's standing on).

 

You're forgetting that despite what route Intel chose to go, they have had chips that performed quite well up until recently, and even then their performance isn't what hurts, it's the power consumption and thus heat. Yes Intel has hit the barrier and they know something has to change, this is why they renamed their processors using numbers rather then frequency. The Pentium M is obviously their future architecture unless they make some breakthrough to achieve higher frequency without the cost of huge power consumtion. I've heard they've managed 10 GHz but we don't know how feasible that is atm.

 

ANova, "up until recently," is the key phrase there. The way I see it, the Intel has been limping along with the Pentium 4, an architecture that has a whole lot of technologies designed just to keep it competitive. If they had just gone with a lower-clock architecture and implemented half of those technologies (SSE2, trace cache, etc.), AMD wouldn't have had a chance in the high-end space.

 

That was my point. Chalnoth.......along with the fact that it will be much more like a Athlon 64 then any Intel......

Guess sarchasm isn't in vogue ATM.......

 

what works for one architecture may not work well with another
For example, gains from HT on K7/K8 architecture, will be small if any
Consider:
P4 - small, very fast L1 cache. Instructions fetched from trace cache ( up to 1 op per tact)
K8 - big, not-so-fast L1 cache. Wide decoder (3 ops per tact) + using ecc bits in cache for storing decoding information
And so on and so on

 

I would tend to think that the gains would be larger. Hyperthreading is useful because it allows parallelism of the various pipelines within the CPU via software. The K8 architecture has more units, and thus Hyperthreading would allow it to make better use of those units.

The K8, after all, has three FP units, if I remember correctly. It'll be much easier to keep them all fed if they're allowed to work in parallel at least some of the time.

 

I agree with Chalnoth and also do not agree with him.

I agree that the Pentium M route is the way to go, for gaming at least. It seems to have about 3-7% better performance than 1Mb AMD such as FX-55 and be lower power, lower heat than P4 and lower cost than Fx-55.

I disagree that P4 was poor in entirety. The P4 Northwoods of 2.8-3.2GHz on 865/875 chipsets was the most powerful processor of it's time and caused little problems in regards to heat and power. It all went pear shaped with Prescott's long pipeline and leaky 90nm process though.

 

Well, I don't think it was poor entirely. I think it was a poor start. Intel brought forth a lot of good technologies for use with the P4 that barely kept the architecture afloat. But with the bad idea at the center of the architecture, well, they couldn't really push the architecture ahead.

 

Yes, there were all those rumours of the Intel engineers competing against the Intel marketing guys. The problem is of course that, to the man on the street, big numbers must mean fast machines. IPC is just not sexy. Buy a 2GHz machine or a 3GHz machine ? Well the faster of course ! You cannot blame Intel of course, it was so successful that AMD had to hide behind a numbering system ( which Intel likewise are now using of course for the very same reason ) that had big numbers in it so you could compare.

At present the fastest 3dmark2001 score using just one video card is in a system that uses a Dothan, that's quite a compliment to Dothan considering 2001 is mainly a cpu bench now.