dear all,

i heard that performance of processor is based on clock speed, then how an intel E7300 core2duo @2.66GHz is better than a Pentium 4 @3.00Ghz??
please let me know the terms...

The Core 2 Duo are dual core CPUs, whereas the P4 is a single core CPU, so the Core 2 Duo would offer better performance I would think.
http://en.wikipedia.org/wiki/Pentium_4
http://en.wikipedia.org/wiki/Core_2_duo#Core_2_Duo

2 members found this post helpful.

Clockspeed is only a part when it comes to determining the speed of a CPU. The Core2 architecture is more efficient than the Netburst-architecture (Pentium 4) when it comes to instructions per second. That means, if you compare a single core of a Core2Duo with a single core Pentium 4 at the same clockspeed, the Core2 CPU will be faster.
That only the clockspeed matters is a marketing myth from Intel from the Pentium 4 days, they needed an argument against AMD CPUs that where as fast as their Pentium 4 while running at lower clockspeed.

3 members found this post helpful.

+1 to TobiSGD (though the 'clockspeed is most important' goes back to before the P4 days).

Its not just about how efficient Core2 is vs P4, CPU cache and FSB also play a part.

Whcih P4 @ 3.0GHz? There has been a few-

2002-
'Original' P4 3.0GHz- 512k, 400MHz FSB.

2003-
P4 HT 3.0Ghz- 512k, 800MHz FSB.

2004-
P4 HT 3.0GHz 'E'- 1024k, 800MHz FSB.

2005-
P4 HT 630- 2048k, 800MHz FSB.

All quite a bit older (and slower) than the Core2Duo-

2008-
Core2DuoE7300- 3MB, 1066MHz FSB.

Intel would not let 3 years pass and be producing slower CPUs (in the same basic price range). Not unless AMD shuts down anyway.

Its not just the number of cores. Core 2 Solo (single core) is faster than Pentium 4, and Core 2 Duo is faster than Pentium D (dual core p4).

2 members found this post helpful.

I want to add here that the disadvantage of the old P4 is that due to the high frequency it produces much heat and therefore the machines are very loud due to the fan.

We have some Computers at work which look very similar (Fujitsu-Siemens) some have AMD-Sempron 3000+ and some have a P4, I don't have to look at the computer which processor is inside, I can hear the difference. And if I can't hear it, I can tell from the heat which machine it is.

Note that this is also valid for DualCore Cpus from Intel they are only a Pentium with two cores. One has to distinguish between DualCore and CoreDuo.

Markus

1 members found this post helpful.

As said above, nowadays CPU clock speed means very little. Much more important are FSB speed, bus/core ratio, L2 cache, etc.

2 members found this post helpful.

This is true. Indeed, this was part of the AMD's marketing for their CPUs. For example the AMD socket 754 and socket 939 CPUs were given names like Athlon64 2800+. The "2800+" was meant to imply that this particular CPU was as fast as an Intel CPU running at 2800 MHz, even though the official clock speed of the AMD Athlon64 2800+ was only 1800MHz:
http://en.wikipedia.org/wiki/List_of...icroprocessors

2 members found this post helpful.

FSB is dead, dude!

That actually started with the 'socket A' Athlon XP CPUs.

By the way, if its an athlon/athlon 64/etc, the XXXX+ is measured against a Pentium 4. If its a duron/sempron, the XXXX+ is measued against a celeron.

That is why you will get athlon XP 2400+ ('Thoroughbred A/B', 2000MHz, 266MHz FSB, 256k cache ) and the sempron 2800+ has almost exactly the same spec ('Thoroughbred-B', 2000MHz, 333Mhz FSB, 256k cache).

1 members found this post helpful.

Care to explain a bit ?

If it is dead, then why do they still list it in many processor specs ?

1 members found this post helpful.

I havent seen FSB listed in current CPU specs for ages. IIRC the last CPUs I saw with a FSB in the spec were the original atoms (non-DXXX models).

AMD doesnt list FSB for the phenom II or athlon II CPUs-

http://products.amd.com/en-us/Deskto...f10=&f11=&f12=
http://products.amd.com/en-us/Deskto...f10=&f11=&f12=

Intel doesnt list them for (current) atoms, i3, i5 or i7s-

http://ark.intel.com/products/37147

(too lazy to actually track down the i3, i5 and atom spec sheets right now LOL)

FSB is not used anymore due to much faster CPU-chipset interfaces (AMD Hypertransport, Intel QPI).

2 members found this post helpful.

AMD get rid of the FSB with the first Athlon64 CPUs, since them their connection to the chipset is called HyperTransport. I can't see a mention of FSB in the specs from AMD since then.
Intel get rid of the FSB with Core i-CPUs, I also can't see a mention of FSB in their specs.

1 members found this post helpful.

I know this is semantics, but I regard that as wrong - the P4D is a dual core P4. Now wikipedia does seem to classify the P4D as a separate device, but as it is a P4 with two cores, I've always regarded it as a two core P4.

In any case, with a Core 2 Duo and a P4, at the same clock speed, the C2D would be faster (and cooler).

Before we get to the explanation, a little analogy (which I hate, as an explanation, but it might get you into the right frame of mind). If someone came up to you and asked 'Which of those two cars is faster?' and you answered 'Well, the small Japanese 4 cylinder will do 8000 rpm, but the American 8 cylinder will only do 6000 rpm, so the Japanese one is obviously faster.' would that be an adequate answer? (To answer my own question, it wouldn't. Apart from the issues of whether we are talking about top speed acceleration or speed round corners, or some more general measure etc, etc, it would be a bit ridiculous to think of this as just an issue of engine revs, and engine revs are analogous to clockspeed here.)

I am sure that you have heard of cache and its importance in architectures - if a processor has a bigger cache, it can go faster, without needing more clockspeed. That is simple, but a minor part of the equation, but does illustrate that clock speed can't be the only thing involved.

• The C2D does some quite clever stuff internally that breaks the instructions up into 'effective RISC' instructions, that can be processed more rapidly than unmolested x86 instructions.
• without the effective RISC cleverness, the P4 relies on long pipelines, so that the instructions can be processed in a large number of (almost trivial) steps. This is fine when the code doesn't branch (each step being small, the whole pipeline can be clocked quickly, but this 'stalls', and all of those stages need to be refilled, when the code has to branch off somewhere new). So, the P4 arch needs to be 'protected' against pipeline stalls by large cache sizes (actually, this is only partial protection, but the 'cost', in time, of waiting for information from RAM is so relatively great, that it has a real performance impact) or it spends a large proportion of its time waiting for pipeline re-fills.

So the P4 needs high clock speeds, big caches (is aided by linear code), and needs bandwidth to ram to achieve any performance, where a C2D is lighter on those aspects. It was possible to get the P4 up to high clock speeds (Intel knew of these disadvantages when they architected the part, but also new that this long pipeline arch would allow them to push to higher clock speeds than short pipeline parts that would have to do more in each pipeline stage) and Intel thought that, by now, we'd all be running P4s pushing on for 10 GHz, and that would get over the problem. If nothing else, that would have presented a real cooling problem and we are all feeling the advantages of them seeing sense (eventually).

+1.

Heres some bencharks backing that up-

http://www.tomshardware.com/reviews/...mark,2974.html

No the best test IMO, far to much emphasis on artifical benchmarking, but hey, you take what you can get.

I'd sort of disagree with some of this, but thats getting way off subject.

Intel got the Core2Duos up to some pretty high speeds. I dont know if they ever shipped E8700s (3.5GHz) but they sure shipped E8600 (3.33Ghz, 6MB of cache)-

http://ark.intel.com/products/35605

Not that far short of the 3.8GHz that intel got the P4s up to.

IMO intel miscalculated with the P4 series, it was flawed from the start. Optomising the CPU for RAMBUS was a mistake for various reasons. Intel was planning on getting the P4s 10GHz but could never deliver, the engineering was just to hard. Even a 10GHz P4 (without major cache increases or architectural changes) would be slower than modern CPUs anyway, even for single core use.

True, but that's after a couple of die shrinks. I can only guess what the current P4 clock speed would be up to, under the same circumstances, but I'd have to guess it would be well over 6 GHz. After all, people were overclocking P4 to over 5GHz by the time of its death.

My opinion (and I have no hard evidence for this) is that marketing thought that they couldn't sell it. I imagine a conversation like this (but with powerpoint slides):

Engingeering: We could make a new part that uses a lower clock speed, but is faster in use than our existing product
Marketing: Whaaat! Why would that be a good idea?
E: Well, yields would be higher (eventually), and power consumption would be way lower. We'd be leveraging our Pentium M technology and...
M: But you are saying the clock speed would be lower..
E: Yeah, but
M: How could we sell that? Our customers would see a lower clock speed and assume that its a slower part. We'd have to sell it more cheaply...
E: You must be assuming that our customers are brain dead
M: Assuming? Have you seen some of the e-mails we get... there's the guy with the fire hose trying water cooling and who thinks we should give him a refund when his computer got wet...
E: But the reviews would say...
M: You are making the assumptions that they can read and will take some notice. They won't pay more for a slower part. Just how stupid are you?
E: It would not be slower, it would just have a slower clock rate
M: There you go; it would be slower. that's the number on the lid. Its a less big number. Less big number = pay less for the part.
E: No, why don't you listen...
M: Just like the customers won't.

* That mention of Pentium Ms: if the same conversation occurred earlier, which it probably did, it would say P3. They may even have recycled the same set of powerpoint slides.
(I'm not really sure how much humour marketing really have. that's a guess.)

Only with liquid nitrogen AFAIK (though its possible I'm wrong, and its also possible that somebody hit 5GHz with phase-change cooling, maybe even a huge pelt setup). With air cooling, low to mid 4GHz numbers were possible.

BTW, at least one person got windows to boot with a P4 overclocked to 8GHz-

http://hardware.slashdot.org/story/0...ocked-to-8-GHz

Though I doubt it was stable, I never saw any benchmarks, just a goofy CPU-z screen shot.

I wouldnt be suprised if marketing did play apart, they would have found it easier to sell a 2.2GHz P4 than a 1.4GHz P-M, even if the PM could well be faster all round.

But I also think that it was engineering who dropped the ball. Even if the '10GHz' figure was influenced by the marketing departement and slighly inflated, I think that its something that engineering were shooting for. They just couldnt deliver (or couldnt deliever without costs increasing to unacceptable levels). Not that unusual in engineering in general, and electronic engineering in particular.