CPU releases seem to be stagnated

[Hulk]

Intel released the P4 3.06 November 12, 2002. Now, 16 months later the fastest Intel chip is the 3.4, which is available in limited quantity. Things have really slowed down quite a bit. Intel has introduced the 865/875 chipsets and the "EE" chips in that time frame, but overall things haven't changed much.

AMD released the XP 2800+ Thoroughbred October 1, 2002 at 2250MHz. Today the fastest AMD chip runs at 2400MHz and again is available in limited quantities. To be fair, it must be pointed out that AMD had gone through the Barton core, and released the Opteron and A64 cores in this time period, which is a great achievement.

16 months ago the Athlon XP 2800+ competed against the P4 3.06.

Now the FX53 competes against the P4 3.4EE.

Which do you think shows the bigger performance increase?

P4 3.06 to P4 3.4EE

or

XP 2800+ to FX53

It's just strange, I've been running a P4 3.06 for over a year and there aren't substantially faster chips available. In the past there was a huge clockspeed/performance delta in 16 months.

 

[ForceCalibur]

No worries, Pentium M is coming...

 

[Hulk]

Actually, I forgot to mention the Pentium M. Although the Pentium M will be competitive with the faster P4's when it comes to integer dominated applications (business), from the benchmarks I've seen it will not be able to compete with the P4 clockspeeds when it comes to floating point operations, which include audio, video3D applications, and games, which I think are a major reason many people need a fast system.

 

[ForceCalibur]

have you heard? Everything in a few years is supposedly going to be based on Pentium M architecture, including 64 bit support and dual core (i believe).

Thats from intel anyway.

 

[Alkaline5]

Originally posted by: Hulk
Which do you think shows the bigger performance increase?

P4 3.06 to P4 3.4EE

or

XP 2800+ to FX53

It's just strange, I've been running a P4 3.06 for over a year and there aren't substantially faster chips available. In the past there was a huge clockspeed/performance delta in 16 months.

I've never seen benches comparing the 2800+ to the FX-53, but I'm positive the comparative performance gains would be greater for AMD. As you point out, AMD has at least changed architectures while Intel has just been releasing the occasional speed- or FSB-bump.

It's unusual that the 3.06 can still be considered a performance part over 1.5 years after it's release. You're definitely getting your money's worth out of that purchase. It almost makes me wish I'd plopped down the cash for a 3.06 and a 9700Pro back when they were new.

 

[Cerb]

No doubt the 2800+ to FX-53 is a huge step, even if they have about the same raw speed.

Intel had plans to get the P4 to 5GHz and beyond. They might make it that high, but it's breaking down. The Pentium M has what they need. It has good performance, good scaling (we don't know about overclocking and the like, but it didn't take long to get from 1.5GGHz to a shiny new and improved 2GHz) Give the Pentium M a 800MHz FSB, DC DDR for RAM, and watch it fly. It'll take work, of course, but it's an excellent chip on its own, and the performance per watt is very good. A desktop version will consume more power, but probably not as much as the P4s.
Remember, the 89W for Hammers is absolute theoretical max for the core.
The P4 TDB is expected use (IIRC, like 85% CPU use continuously).
It may lack a little here and there, but note (correct me if I'm wrong, of course):
Right now, it's 400MHz FSB. For the Intels right now, that means a single channel of PC3200 can fill the FSB. Not bad for a laptop, but a gaming desktop begs for DC DDR so the CPU can use a lot of bandwidth and have some left for the video card.
Given the A64 and FX performance in CC, it could make a world of difference.

 

[nick1985]

MHZ dont mean everything

 

[Titan]

i noticed this not too long ago as well.

What ever happened to Moore's Law?

 

[mrwxyz]

Originally posted by: tkotitan2
i noticed this not too long ago as well.

What ever happened to Moore's Law?

enter prescot

 

[SkipE]

Light only travels about 3 cm on a board at these frequencies.
That says nothing about the noise these signals generate at these frequencies.
I think it's amazing we're at this blazing speed.
Physics is beginning to dictate that we start looking for architectural changes for better performance....

SkipE

 

[Pariah]

Originally posted by: tkotitan2
i noticed this not too long ago as well.

What ever happened to Moore's Law?

Moore's Law is related to transistor density, not clock speed or anything performance related.

Though clock speeds have not gone up much, bus speeds have gone up as well as cache being added, so a 3.4EE will perform much better than a 3.4 Northwood would have. Both Intel and AMD have stated that the clock speed wars are over and that they will be focusing on adding more features to their CPU's rather than continuing to rapidly drive clock speeds up.

 

[jagec]

Originally posted by: Hulk
Today the fastest AMD chip runs at 2400MHz and again is available in limited quantities.

correction.

the fastest AMD chip runs at well above 3 GHz

naw, I know you mean stock though. Still, today is a GREAT day to be an overclocker: since new chips aren't coming out as fast, overclocking your old ones can give you a computer which is much faster than the fastest thing on the retail market. And will continue to be so for a good while.

 

[SkipE]

Good point Pariah.
Nonetheless, faster transistors mean better performance. And faster transistors have generally been obtained via smaller footprint, which means higher density. My point is we're reaching the physical limits of speed, which is pushing for architectural changes...

SkipE

 

[jose]

Right now I have 2 - 3.06 (3.4) systems & 1 - 3.0(3.3) system. I want something that has twice the performance. It's been quite a while since an upgrade 1yr+. But from the way things look, I'll have to wait another year to get double the performance of my current systems. Doesn't Intel want my money ?

What's the major diff between the P4 & the Pentium M chip ? is the "M" more similar to a P3 ?

Regards,
Jose

 

[nitromullet]

3.4EE will perform much better than a 3.4 Northwood would have
(italics mine)

The 3.4 Northwood exists...

Also, while it isn't quite as fast as the 3.4EE, I think that 'much' is a bit of an exaggeration, especially when one considers the price difference.

http://www6.tomshardware.com/cpu/20040322/prescott-06.html

In encoding benchmarks the scores between the EE and Northwood 3.4 are very close, in gaming the EE usually has about a 10-15% lead. In price, the EE "leads" by 150%, and the 3.4C isn't exactly cheap to begin with....

link also includes benchmarks for XP 2800+ and FX-53 for anyone interested.

 

[Pariah]

Originally posted by: SkipE
Good point Pariah.
Nonetheless, faster transistors mean better performance. And faster transistors have generally been obtained via smaller footprint, which means higher density. My point is we're reaching the physical limits of speed, which is pushing for architectural changes...

SkipE

True, but more densely pack transistors don't make CPU's faster. A 3Ghz .2µ CPU isn't going to perform any better than an otherwise identical 3GHz .1µ CPU. The smaller size and thus higher density will lower power requirements and reduce heat and potentially (though not guaranteed) allow for higher clock speeds.

When I said 3.4GHz Northwood, I meant a 533 bus version.

 

[SkipE]

I disagree. Why do you think they're more densely packed? It's because the gate length is shrinking. The speed (saturation current) pretty much scales as one over the gate length. This is true until we reach around 40-50 nanometer gate lengths, although other issues are coming into play also. The saturation current is directly related to the speed. It's kind of a chicken and egg thing. Whether you speak of more density or higher speed, they essentially come down to the same thing in terms of Moore's law. But that is coming to an end as we are reaching physical limits...

SkipE

 

[SkipE]

One other thing. THe smaller gate lengths inevitably lead to larger leakage -> higher off-state power. And in general for the high-performance transistors nobody cares, which is why these things are burning up. The M will be a lower off-state leakage by decreasing the short-channel effects, which will inevitably lead to lower speeds...

 

[biostud]

To me AMD has improved more than Intel. they've changed to a whole new structure and we still need software to support 64-bit to see it's full capcity.

 

[Hulk]

Originally posted by: biostud666
To me AMD has improved more than Intel. they've changed to a whole new structure and we still need software to support 64-bit to see it's full capcity.

I would have to agree there, but it's not like the both of them aren't trying. I think that Intel may have overestimated the P4's legs and underestimated the Athlon 64. It's fun to watch them scramble. We're lucky to have the competition.

I still don't see how the Pentium M is going to compete with a 4GHz P4 (if it tops out there) in audio, video, and 3D rendering benchmarks. There are some benches, actually quite a few, that clockspeed does seem to dominate.

I used to upgrade when I could get twice the performance out of my new rig (or more). 486-33 to P5 90to Cel 300A (at 450 of course) to PIII 850 to P4 2.4 and then a processor upgrade. Looks like those days are over. Which is good and bad, good because computers are quite fast and usable now and it's cheaper to hold on to one for a few years, bad because it was fun to boot up the new computer and watch all the old apps act like they're on speed. Until you get used to it of course and the cycle starts again...

The war at 90um should be a good one.

 

[imported_Aelius]

Personally I don't plan to wait that long.

If you look at my sig you can see where I'm at and where I'm heading.

In fact I'm giving my current PC away within a few weeks. XBox will hold me over until my new computer gets fully purchased within the next 1 to 2 months.

 

[Alkaline5]

Originally posted by: Aelius
Personally I don't plan to wait that long.

If you look at my sig you can see where I'm at and where I'm heading.

In fact I'm giving my current PC away within a few weeks. XBox will hold me over until my new computer gets fully purchased within the next 1 to 2 months.

If your just giving things away, then I'll take those Klipsch off your hands.

 

[Cerb]

Originally posted by: tkotitan2
i noticed this not too long ago as well.

What ever happened to Moore's Law?

Not much.
Density has increased, complexity has increased, number of transistors has increased...just that MHz are slowing down.