Intel or AMD

[ALstonLoong]

I was just wonder, in terms of calculation ....which processor is better ? P4A or AMD xp ?

 

[fkloster]

In terms of # crunching power per clock cycle? AthlonXP hands down is better...

 

[pecel]

Another vote for AMD AthlonXP

 

[smp]

Do you want to OC?

Do you have mucho expendable income?

Are you on a budget?


Choosing a processor is a very subjective process, tell us what you want out of a CPU and we'll tell you which is better.

 

[CableDude]

AthlonXP

 

[Don66]

I have to vote for XP just unlocked mine.

 

[smp]

I was just wonder, in terms of calculation ....which processor is better ? P4A or AMD xp ?

Okay .. just reread your question .. in which case yeah, Fkloster is right...

In terms of # crunching power per clock cycle? AthlonXP hands down is better...

 

[ALstonLoong]

errr... i mean engineering maths calculation.

 

[microAmp]

Clock for Clock, AMD.

 

[Mem]

I think the answer is AMD .

 

[Adul]

at lest flokster is can admit that

just joking. Athlon hands down.

 

[Wolfsraider]

athlon without a doubt

 

[InFecTed]

The authentic bad boy is named Cyrix.

 

[PC166]

Duron running at 100FSB is more then enough to pit against P4 clock for clock. Very strong processor only 5 percent slower then an Athlon XP. Mhz isn't everything thought if a 1gig Athlon can run like 1.5gig PV and 1.53gig like 2.4gig hmm then you know you be paying alot lesser for same performance. My friend is an intel freak he think Intel like Microsoft is better, more stable and faster too like recently he just bought a 2.2 P4 computer and was bragging about how much faster it is then Athlon and then for budget computer he chose Celeron over duron cost a little more at comparable clock for clock thinking that its faster. Because a duron turn a SDRAM 100mhz to 200mhz effectively it is like 200mhz FSB or 133 to 266 for that matter and with DDR it double again! amazing.

 

[dullard]

<< I was just wonder, in terms of calculation ....which processor is better ? P4A or AMD xp ? errr... i mean engineering maths calculation. >>



Please define "better" for us.

1) The people on this board are correct that generally the Athlon XP will perform more operations per clock. This comes from the Athlon XP's shorter pipe length. Computers try to predict the answer to a question before they have enough information to answer the questions. The P4 has a 20 stage pipe, and if it predicts incorectly it just lost up to 20 stages worth of work (wasting time). The XP's pipeline is about half that size, so with an incorrect guess it wastes a lot less time. Now the question comes to this: does your engineering math calculation have lots of branches, and if so is it difficult to predict the correct path? This varies a lot from program to program. I do more engineering math calculations than probably anyone on this forum (it is my job and hobby). My most used program has very few branches so a 1.4 GHz Athlon XP performs about the same as a 1.4 GHz P4. I have seen other math calcuations that have many difficult branches and the 1.4 GHz Athlon XP will dominate the 1.4 GHz P4.

Sorry about the long paragraph above, but I've seen many people make mistakes and just blurt out the Athlon XP's math calculations are better no matter what (since they probably have only seen one scientific benchmark and it happened to be one that has many difficult branches).

2) However the P4A performs more clocks per second (unless you compare the very fastest XP to the very slowest P4A). This may or may not overcome the branch prediction problems listed above.

3) If you know data from both (1) and (2) then you can know which processor is faster (multiply the average calculations per cycle by the cycles per second). I always think of it as a Porche with a flat tire racing against a station wagon with 4 perfect tires. It is possible for either to win the race. Is this what you mean by better?

I know the SPEC benchmarks are hated benchmarks but they are based on widely available scientific calculations (higher is faster):
XP 2100+ Cint2000 (base): 720
XP 2100+ Cint2000 (peak): 749
2.4A P4 Cint2000 (base): 819
2.4A P4 Cint2000 (peak): 833

XP 2100+ Cfp2000 (base): 613
XP 2100+ Cfp2000 (peak): 660
2.4A P4 Cfp2000 (base): 806
2.4A P4 Cfp2000 (peak): 812

Thus on average the 2.4A is faster than the 2100+ in engineering calculations for both integer and floating point calculations. However, if you pick just a single program, the 2100+ XP may dominate the 2.4A.

Note: the 2100+XP is much cheaper than the 2.4A.

 

[Underclocked]

Athlon XP for maximum performance generally. Intel P4A for maximum performance with the least hearing damage.

 

[hypeMarked]

So, for performance as in games and application and no overclocking and for the least amount of money, is it XP or P4.

Underclocked, what do you mean by "hearing problem"?

 

[lax724]

AthlonXP

 

[ALstonLoong]

dullard
thanks for your explaination . so is there any intel processor that can perform better in maths calculation then athlon xp ? xeon ? itanium ? etc....... ?

 

[dullard]

<< so is there any intel processor that can perform better in maths calculation then athlon xp ? xeon ? itanium ? etc....... ? >>



That is a very loaded question. The answer will vary significantly with each program you use. So instead of answering it directly I'll post some things related to your question. (I won't mention the slower processors to keep this to a reasonable length).
1) At the moment, AMD sells the Athlon XP and Athlon MP. The XP is meant for single processor machines and the MP is meant for dual processor machines. They have the same core and operate at at the same speed when used in single processor machines.
2) Intel sells the P4 and the Intel Xeon. The Intel Xeon is meant for dual processor machines and the P4 is meant for single processor machines. The Intel Xeon has the same core as the P4 so one Intel Xeon performs the same as one P4.
3) If you have a math calculation program that can only use one processor, then your choice is significantly narrowed. At the moment you have the Athlon XP and the P4 to choose from. See my post above for average benchmarks for these.
4) If you have a math calculation program that can use multiple processors then you have lots to choose from.
5) Each program is different, but on average I'd use this rule: a multiprocessor machine runs X times faster than a single processor machine. Where X=SUM(i=1 to N) [0.85^(i-1)] and where N is the number of processors in the machine. I hope that formula makes sense to you. The constant I used, 0.85, will vary widely from program to program, but 0.8 is a good starting point.
6) If you know the speed you need to obtain (X), then you can solve for the number of processors you need (N).
7) If N=1 then you get the Athlon XP or the P4.
8) If N=2 then you get the Athlon MP or the Intel Xeon.
9) If N>2 then you have a tough choice. At the moment there are no AMD processors that can be put in a N>3 machine. Intel has three options. A) the Pentium 3 Xeon can handle 8 processors. B) the Intel Xeon MP can handle up to 8 processors - but 4 is the current limit since it is so new, C) the Itanium can handle as many processors as you want in a single super-computer.
10) Note: the P3 Xeon, and Intel Xeon MP do not run as fast as the best Athlon XP or the best P4 since their frequencies are much lower. The P3 Xeon maximum is 1GHz if N>2, the Intel Xeon MP maximum is 1.6 GHz and the Itanium maximum is 800 MHz. The Itanium is so different that at 800 MHz its math speed isn't too bad (although it isn't spectacular either). So if you calculated that you needed three 2.4 GHz P4's then you will need to buy about five Xeon MPs or about eight P3 Xeons.
11) If you need to run 64-bit programs, then the Itanium will shine. This is since the others will perform miserably at 64-bit programs. Go the Itanium route if N>8 or if you need to use 64-bit programs. However it has suffered from a lack of industry acceptance (these people would rather have Sun, SGI, etc). Also rumors of the Itanium 2's abilities have made the industry avoid Itanium. So Itanium is a flop so far.

Future processors (in order of release date):
1) In about 1-2 months a 1.0 GHz Itanium 2 will be released. It is supposed to be double the performance of the current Itanium (more with recompiled programs). This might finally gain industry acceptance. Or it might be a flop like the first. I fully expect Itanium 2 to shatter math calculation speed records at its release. But with industry heavilly invested in other proprietary solutions, they might still be unwilling to switch.
2) AMD will eventually have a new line of processors. These are codenamed Clawhammer for 1-2 processor computers (available sometime around the end of this year, no specific date is known for certain). The Clawhammer can run 64-bit programs. So if N<3 and you need 64-bit programs, then Itanium is a waste of money.
3) Sometime in the first half of 2003, AMD will release its Sledgehammer. The Sledgehammer will be AMD's first chip that allows N>2. Its maxium N will be eight. Sledgehammer is also 64-bit, so you can use 64-bit programs.
4) At about the same time as the Sledgehammer is released the Itanium 3 (it might have a different name) will be released. It should again be almost double the speed of Itanium 2. Thus if Intel stays on track, it will be 4 times as fast as the current Itanium for the same price. This begs the question, will industry buy Itanium 2 if Itanium 3 is just a year away? Intel's speed doubling every year is great if you need speed, but Intel shoots itself in the foot in the process...

Final complications:
1) If N>2 you MIGHT be able tol use single or dual processor machines. If you have a high speed connection between computers you can form a cluster of cheap machines. This cluster may outperform N>3 machines at a significantly lower price.
2) There are always Sun, SGI, and other proprietary processors also availble. I think their cost is just too high for the performance you get.

 

[ALstonLoong]

emm ... just wonder why intel design its processor with large bandwidth but interms of maths calculation still lag behind AMD. so wat are the benefit in the new design ? Just wanna get higher clock speed ?

 

[dullard]

<< emm ... just wonder why intel design its processor with large bandwidth but interms of maths calculation still lag behind AMD. so wat are the benefit in the new design ? Just wanna get higher clock speed ? >>



I posted an average of many common math calculations. That average shows Intel is faster. Then you say Intel "still lag behind AMD". I don't know how to respond.

Intel chose a large bandwidth so the memory won't be the bottleneck. Why did AMD switch from SDRAM to DDR SDRAM? So memory won't be a bottleneck. Bandwidth and clock speed really aren't very closely related.

The new design had one controversial part (unrelated to bandwidth) - this is the number of stages in its pipeline. The more stages, the higher the theoretical clock speed. However, in some programs a higher number of stages makes each clock less productive. So more stages makes the processor perform faster and more stages makes the processor perform slower. Anytime you have a situation like that, there must be an optimum. Intel decided at one time, that the P4 design was at the optimum. They may have estimated incorrectly (I'll let you make the decision on that), but they tried their best for the optimum performance. That is the reason for the new design.

A side effect of the new design is that the clock speed is high and the IPC is low. Thus marketing focuses on clock speed. If the optimum went the other way, with low clock speed and high IPC, then marketing would proclaim a great IPC. Either way you would have people complain about Intel's marketing.

 

[hopeless879]

<< So, for performance as in games and application and no overclocking and for the least amount of money, is it XP or P4. >>



For the best performance at the cheapest price without overclocking would be the AthlonXP.




<< Underclocked, what do you mean by "hearing problem"? >>



XP's generate more heat than a P4 therefore you need better, usually louder fans on them.

 

[ALstonLoong]

thanks again dullard for your explaination.

 

[Wolfsraider]

dullard i want to also thank you.

i look forward to reading more of your responses thanks