CPU performance chart

[Outsider]

I just came accross this CPU performance chart that Tom's Hardware has put together:
http://www23.tomshardware.com/cpu.html

What supprised me most, excluding the top 15 perfomers which have a very steep price/performance ratio, you're only looking at under 2x the performance.
I would have thought there would be more spread in the speed of the CPUs. between the low end Semprons 2500+ and the high end (yet not crazy priced) Athlon X2 4800+.

Am I missing something?

P.S. I only looked at the test that are purely CPU intensive. Playing a game is only intensive to a certain degree. Since the GPU does a lot of the processing, throwing a faster CPU beyond a certain point makes very little difference. I looked at things like MP3 encoding, Studio Max7, all the CPU benchmarks, etc.

[Cafe_333]

Yes you are missing something - gaming performance lol. ;)

Gaming performance is very cpu intensive - just play any fps game and you'll see your cpu temp climb. The environment in games are unpredictable, the cpu needs to react to user decisions. This is why AMD tends to do better than Intel in gaming performance because it has a shorter pipeline each time it needs to process information. With AV encoding the process is more predictable, so everyone does a pretty good job here - although Intel generally does this better benefitting from its longer pipeline.

Anyway, the gaming performance is what really seperates the cpu's. Other than that, low end cpu's are still good enough for today's average user if their needs don't go beyond email, surfing and light desktop applications. This is the reason why everytime someone starts a thread asking to recommend a computer, one of the first questions always posted is what they intend to use their computer for.

[Outsider]

I don't know if you actually checked out the site, because if you do look at the game performance (for the 3 or 4 games tested on that site) you'll notice that the increase in performance (between the low end semprons and X2 athlons) is quite minimal. Much more minimal then pure CPU number crunching applications.
Look at Doom III 1280: Sempron 2500+ 68.6fps vs Athlon X2 4800+ 83.2fps
How about UT2004: Sempron 2500+ 101.3fps vs AthlonX2 4800+ 164fps
FarCry: Sempron 2500+ 126.9fps vs AthlonX2 4800+ 194fps

Sure there is a difference, but nothing to write home about... To keep things in perspective, we're talking about a CND$62 CPU vs a CND$750 CPU... (priced at Canadacomputers) I would hope that for almost 10x the price one could get more then twice the number crunching performance... this is what I find strange...
Does this not show that games are not _really_ that CPU dependant?

[willy]

I don't know if you actually checked out the site, because if you do look at the game performance (for the 3 or 4 games tested on that site) you'll notice that the increase in performance (between the low end semprons and X2 athlons) is quite minimal. Much more minimal then pure CPU number crunching applications.
Look at Doom III 1280: Sempron 2500+ 68.6fps vs Athlon X2 4800+ 83.2fps
How about UT2004: Sempron 2500+ 101.3fps vs AthlonX2 4800+ 164fps
FarCry: Sempron 2500+ 126.9fps vs AthlonX2 4800+ 194fps

Sure there is a difference, but nothing to write home about... To keep things in perspective, we're talking about a CND$62 CPU vs a CND$750 CPU... (priced at Canadacomputers) I would hope that for almost 10x the price one could get more then twice the number crunching performance... this is what I find strange...
Does this not show that games are not _really_ that CPU dependant?
To test out the CPU performance under the gaming category, I think we should look at the scores at the *low(er)* resolution - ie. Doom III 1024. In fact, it would be nice if they have 800x600. Once the resolution reaches certain point, it becomes video card bound.

[supernerd]

I would have thought there would be more spread in the speed of the CPUs. between the low end Semprons 2500+ and the high end (yet not crazy priced) Athlon X2 4800+.

Am I missing something?


Processor improvements are always incremental, even between two different cores of the same family, which usually means marginal improvements in speed. The "major" improvements between many current generation CPU's have been increased cache size, faster on-die memory controllers, smaller fab process. There are no major ISA or pipeline design changes. Thus, performance will be still very dependent on clock speeds (when comparing the same family of chips i.e. athlon vs athlon or P4 vs P4).

So, a sempron 2500+ at 1.4GHz is expected to be half the speed of say a FX-57 at 2.8GHz (in terms of raw CPU power). Moving to dual core, you expect to see another speed improvement of 2x if the benchmark is written well and the application is actually suitable for parallel computation.

what im really trying to say is that the "megahertz myth" is only a myth between different cpu architectures. within cpus of the same architecture, the clock rate is going to make the biggest difference in raw CPU power (which is why I dont really like AMD's naming scheme. A64 3800+ plus double the cache turns into a A64 4000+? Sort of implies a 5% improvement just for increasing cache size... which is unlikely in many applications)

[supernerd]

To test out the CPU performance under the gaming category, I think we should look at the scores at the *low(er)* resolution - ie. Doom III 1024. In fact, it would be nice if they have 800x600. Once the resolution reaches certain point, it becomes video card bound.

Games will never be a truely good measure of raw CPU performance. For one, there are too many other factors that come into play, such as memory, video, and I/O performance in general. Secondly, until games utilize a second thread for useful parallel computation, dual cores will appear the underperform.

The OP was right to only look at tasks such as encoding and sandra for comparison of raw CPU power.

[jm20]

Yes you are missing something - gaming performance lol. ;)

Gaming performance is very cpu intensive - just play any fps game and you'll see your cpu temp climb. The environment in games are unpredictable, the cpu needs to react to user decisions. This is why AMD tends to do better than Intel in gaming performance because it has a shorter pipeline each time it needs to process information. With AV encoding the process is more predictable, so everyone does a pretty good job here - although Intel generally does this better benefitting from its longer pipeline.

Anyway, the gaming performance is what really seperates the cpu's. Other than that, low end cpu's are still good enough for today's average user if their needs don't go beyond email, surfing and light desktop applications. This is the reason why everytime someone starts a thread asking to recommend a computer, one of the first questions always posted is what they intend to use their computer for.

Thats a terrible analogy. Shorter pipelines to not react "better" to user input in any way. I suggest you read more ArsTechnica, and I'll start you off link (http://arstechnica.com/articles/paedia/cpu/amd-hammer-1.ars)

[JoKeRr]

Thats a terrible analogy. Shorter pipelines to not react "better" to user input in any way. I suggest you read more ArsTechnica, and I'll start you off link (http://arstechnica.com/articles/paedia/cpu/amd-hammer-1.ars)

Agreed. This guy really doesn't know much about computer architecture as a whole, no need to mention pipelining and IPL etc

[SmartBen]

I would have thought there would be more spread in the speed of the CPUs. between the low end Semprons 2500+ and the high end (yet not crazy priced) Athlon X2 4800+.


X2 CPUs are dual core CPU. Up to now, there isn't many program/game/application utilized to use more than 1 cpu to function. So your X2 4800+ (2 2400Mhz CPU) will mostly work as a Athlon 64 4000+ (1 2400Mhz CPU) [well, correct me if I am wrong on the numbers]. So you are comparing a lower end 2500+ with 4000+. The different is should not be too big..

And to be honest, it is difficult to write mutlipe-thread programs that uses more than 1 CPU. There are inter-process communication delay, data sync problem which slow the whole process down even you have twice as man power... When I was till in school, they told us that if you could archive 20% performance improvement by changing the program from single thread to mutliple thread, you are already doing a great job.

X2 4800+ gets a higher model number not because it has GREAT processing power, it means in general situation, it will give you better mutiple processing power and processES will be able to finish quicker in general.

Normal user will see GREAT performance improvement since there are 2 CPU handling all the requests. The small yet detectable delay between running all those hidden or active programs in your window will decrease even more with X2, and like my friend keep saying, it would be nice if you could play game and at the back encode/decode some movie/video with a X2 CPU. (And yes, you will be killing yourself if you only have 1 CPU)

[ShadowVlican]

xbitlabs has performance comparisons as well

i wouldn't go judging these cpus based on one hardware site along, especially if that single judgement was based on tom's hardware guide ;) :lol:

[jm20]

xbitlabs has performance comparisons as well

i wouldn't go judging these cpus based on one hardware site along, especially if that judgement is based of Intel's Marketing Department

FIXED

[ShadowVlican]

:lol:

[Cafe_333]

Thats a terrible analogy. Shorter pipelines do not react "better" to user input in any way. I suggest you read more ArsTechnica, and I'll start you off link (http://arstechnica.com/articles/paedia/cpu/amd-hammer-1.ars)Instead of flaunting your ego why don't you try asking me how I came up with that 'analogy' instead? Learn to have some respect for other posters because I don't appreciate the insinuation that I would need to re-educate myself on the validity of my knowledge. Even if I were wrong, do you really feel that conceited to make such a condescending remark? Next time either ask or post an educated rebuttal. I am sure you are capable of delivering a more mature mannered response as opposed to a cocky tone that only serves to belittle others.

Agreed. This guy really doesn't know much about computer architecture as a whole, no need to mention pipelining and IPL etcYou don't know me at all. Are you that shallow to judge someone on the basis of one piece of information and then write me off as someone who doesn't understand much about computer architecture?? Judging books by its cover is a cliche we're all supposed to know, but clearly some never seem to get the better of. Either explain why I am wrong or ask why I thought that way in my post. I am sure you are intelligent enough to comprehend that than resorting to insulting remarks without justification.

......................

With regards to the context of my initial post, I have only stated that the shorter pipeline in AMD is one of the biggest reasons why they tend to perform better in GAMES than Intel. I don't believe this to be incorrect. As this is a public forum, anyone is free to correct me if I am wrong. I can admit fault, but honestly, is there *really* a need for condescending insults?? Anyway, here's my explaination of shorter pipelines performing better in games found in the second response of this post:
http://www.redflagdeals.com/forums/showpost.php?p=2566489postcount=1