benchmarks and "real world" interpretation...

[drifter106]

in looking at various websites and the data that is given, the benchmarks for amd 64 processors do not score as high as the top end pentiums. But by the same token, amd scores better in their respective areas. My question is this... except for 2 or 3 benchmarks the amd processors where normally within 10% of top intel performer (as far as numbers go) i.e. p4 = 1,000 and 64fx = 990. Many times the scores were + or - 5 points. Now how should we disagragate this data in real world time? Is performance 5... 10% lower... will downloading pics from camera take, i.e. 11 minutes rather than 10? Will multitasking take longer? IF it does how much longer.

You must understand where i am coming from...my world exists within the parameters of k2 and durons at school... oh yea and that thing they call a mac...am I being sarcastic...neverless i am soon to enter another world and find out what a "fast" computer can do... just want to make sure I go in the right direction and am comfortable with what i choose..btw...will probably go with a barton or low 64 or 2.8 or 3.0... gaming, pics, music, burning cd''s and dvd's,...

stumbled onto this site 3 weeks ago... VERY educational to say the least...have noticed that it does lean towards amd but thats the nature of the beast... keep up the fantastic dialogue..

later...

 

[Mik3y]

ummm....thanks...i guess....

 

[Megatomic]

In most tasks you would never notice the difference in real world performance between the two platforms. Now if you were talking about a large video encoding job you might save a few minutes with the Intel but you might make those minutes up with several smaller mp3 encoding jobs. And gaming is often a dead heat.

I myself went AMD to save money, but I'd love to play around with a P4HT system. I do plan to do alot of video processing jobs in the coming year, I could use the extra power in that area. But my Barton will do the job well for me.

 

[Sideswipe001]

I tend to agree. I have used primarily AMD, but currently I have dual Xeons (thanks HP!). It depends on what you do and what your budget is. I rarely encode, mostly just game. I never noticed a differece in multitasking between and Intel and a AMD processor, because most multitasking just doesn't stress the processor.

You should take into account 64 bit processing though. Assuming microsoft and software people throw out a nice solid 64 bit windows, with good drivers, and 64bit programs, you might well take a large lead over the current P4s just by chaning OSes.

 

[drifter106]

sideswipe...

does that mean if you do go with a 64 bit processor and utilize a 64 bit OS. it would have a positive effect on your performance...say strengthen your weak areas and improve where you are strong...

 

[Markfw]

I think it sounds like you should try out the 64 bit Athlon64. From what I have seen so far in 64 bit mode, it makes a big difference even in 32bit apps.

 

[drifter106]

thanks for the info... will look into that ...

 

[RussianSensation]

Generally the rule of thumb is that you will not notice the difference in performance if it does not exceed 10% for any application except the ones that use time which you can actually time and "feel" the application ending sooner.

Of course each platform has its advantages and disadvantages and in those areas the difference is substantially greater than 10% (ie amd = gaming, mathematical, p4 = encoding, multitasking etc.)

Overall right now i'd get Barton Mobile for the best bang for the buck or P4 2.8 and overclock it to 3.5 or if you do not want to overclock A64 3000

 

[Cerb]

Right now they are about even. Across reviews, you can get +/- 5% easy, and it takes around 10% to really feel the difference.
At the moment, the A64 has the gaming and general use edge--butnot enough that you should care. Price and features should influence your purchase more than anything. P4s OC, but if you're not so keen on that, then P4s encode faster (come on socket 939...), though the FX is making big strides there. For anything else, they are either about even or AMD has the edge. But as previously stated, the are ahead by a very small amount; enough that your drivers and background apps will make a bigger difference than CPU choice.

I'd stick w/ AMD right now just for CnQ.
And whatever you, don't even think of going w/ a Celeron.

 

[Sideswipe001]

Originally posted by: drifter106
sideswipe...

does that mean if you do go with a 64 bit processor and utilize a 64 bit OS. it would have a positive effect on your performance...say strengthen your weak areas and improve where you are strong...

Yes. It's pretty much exactly what I mean. At least, from what I've seen, encoding scores went through the roof for the A64s when they were running in 64 bit mode. And really, unless the drivers were just crappier, there is no reason to expect performance to get worse. Soon, UT2004 will be released in a 64 bit flavor as well. We'll then really get to see what it can do for games.

 

[dullard]

Benchmarks are fuzzy numbers. NEVER put too much emphasis on just one benchmark. Why? Since they are far too easy to manipulate. While the benchmarkers generally are telling the truth, they are never telling the whole truth. For example, processor X may easilly surpass processor Y on a game on level 1. But if you benchmark the same two processors on the same game on level 2, the story often is completely different. Processor Y may easilly surpass processor X on that same game on level 2. If a benchmarker only shows one of those two levels, you'll be misled.

Benchmarks also cannot be extrapolated to apply to different tasks. A benchmark that says the top AMD processor is 5% faster on a game is useless for the time it takes to download pics from a camera. Why? That task does not require a fast processor. The download time may be bottlenecked by the camera itself (ie it may be going as fast as it can and no processor could change that) or the download time may be bottlenecked by the type of interface with the computer (ie USB 2.0 is far faster than USB 1.1).

Things are even more cloudy when you multitask. The newer P4s have special capabilities that boost its multitasking capability - how much depends on exactly what you are multitasking. No other processor has these capability (note you can go dual processor but that is a whole other story). Thus you could take two programs where AMD wins by 10% over a P4 individually, but when run at the same time, the P4 MAY win on both. I highlighted MAY since the P4 can only multitask well if the two programs are doing different types of calculations (something that the user often won't know and certainly cannot control).

Basically, a benchmark applies ONLY to the item being benchmarked. If you make any changes, the benchmark is useless. I think this discussion on Apple G5 claims illustrates that point quite well. Apple used one benchmark and claimed it applied to everything.


As for the comments on the 64-bit OS, that is something we are left to wonder. Clearly some uses (such as encryption) will get a major boost. But how often are you sitting around waiting for your computer to encrypt data? As for other tasks, it is anybodies guess. Wild claims of 30% boosts are out there without any proof. Then there are the real tests we can make with BETA drivers - with 30% decrease in performance. Obviously the final drivers will be faster, but where in the -30% to 30% range it'll end up is anyone's guess.