# I don't get this....

#### jrosenst

##### Distinguished
OK, I added up all the test results from the column, and this is what I got. I added 10% for the one test that the AMD FX-53 failed on...

Test AMD INTEL Less %
1 46.58 39.54 17.80%
2 90.59 87.51 3.52%
3 17.15 16.88 1.60%
4 16.52 15.87 4.10%
5 25.92 25.51 1.61%
6 170.1 166.3 2.29%
7 115.3 109.6 5.20%
8 79.2 77.1 2.72%
9 295 277 6.50%
10 97.9 84.4 16.00%
11 799 780 2.44%
12 6618 6627 -0.14%
13 166 118.1 40.56%
14 117.4 115.6 1.56%
15 164.8 155.9 y -5.40%
16 149.7 139 y -7.15%
17 92 89 y -3.26%
18 399 352 y -11.78%
19 162 150 y -7.41%
20 127 101 y -20.47%
21 85 57 y -32.94%
22 89 85 y -4.49%
23 167.2 98.4 y -41.15%
24 227 231 -1.76%
25 183 204 -11.48%
26 204 217 -6.37%
27 224 253 y 11.46%
28 111.4 119 y 6.39%
29 202 177 y -12.38%
30 95 89 y -6.32%
31 5.3 6.9 y 23.19%
32 52.9 65.9 y 19.73%
33 102 108 y 5.56%
34 4609 5290 -14.78% A Guess
35 5569 5568 0.02%
36A 4962 7514 -51.43%
36B 9955 10502 -5.49%
37A 23701 36090 -52.27%
37B 17948 26068 -45.24%
38A 5805 4967 16.87%
38B 5867 4987 17.65%

With the guess (Sum) -134.97%
Without the guess (Sum) -120.19%
With the guess (Avg) -0.032135714
Without the guess (Avg) -0.029314634

The y in the less column means less is better. This to me (If you weighted each test the same) say's the AMD FX-53 is around 3% slower then the Intel 3.4EE. Where is the justification for the line that reads:

"The new Athlon64 FX sets the new performance standard to beat, leaving the Pentium 4 3.4 GHz Extreme Edition far behind."

If your talking about price/performance neither of these chips comes close to the leader so I have to assume the column is talking about purely performance. Can someone enlighten me please?

Thanks,
Jeremy

#### Unolocogringo

##### Illustrious
You know I kind of had the same thoughts while reading the article.
But did not take the time to break the math down as far as you did.
Now everyone will say the tests were skewed and too many apps that favored Intel were used.

And welcome back frenchfry.

I aint signing nothing!!!

#### endyen

##### Splendid
If you take into account that these are GAMERS chips, and Intel won in exactly 0 games, you begin to understand. And yes the benches were exclusively from the Intel list.

#### jrosenst

##### Distinguished
Ok, I can see that. With just the game benchmarks the AMD is 10.69% faster (5.71% if you remove the X2 demo). So I would agree if the line read:

"For the gamer, the new Athlon64 FX sets the new performance standard to beat, leaving the Pentium 4 3.4 GHz Extreme Edition only faster for business purposes."

I am not here as a gamer, and if you remove the games the Intel is 6% faster. It would have been nice to of heard that somewhere in the article.

#### jrosenst

##### Distinguished
Interesting, I was looking at another forum and it showed a link to a different review site. It to had the AMD winning all the game benchmarks except one. The game it lost in was Quake III. So just out of curiosity I looked at the last CPU review here at Tom’s (P4 Prescott) and in that review was a Quake III benchmark. I wonder if it was just an oversight that it was left out of the AMD FX-53 review or if the Intel beat it so they chose not to use it?

#### Unolocogringo

##### Illustrious
So what game performance is all that matters??

Some people do more with there comps than game. To suggest that a cpu that is best in games is the best no matter what is rediculous.

So yes it did win the games section.But if you take out the overclocked results and look at only the standard results the conclusions that it wins outright is a fabrication,

Actually it wins about half of the benchmarks and looses the other half. If you add up the total percentage results the P4 system is overall faster.A non perceiveable amount faster but still overall faster.

I aint signing nothing!!!

#### darko21

##### Distinguished
Re: It would have been nice to of heard that somewhere in the article.

You will get over it. THG has a bit of a history of being somewhat intel biased over the last couple of years (in favor of intel). You could always email omid and complian how this upset you. No review is perfect and won't please all but this seems to be more in line with all the other review sites, not like in the past. even with the emphasis on the test that did not complete I'd say this is one of thg's better reviews.

If I glanced at a spilt box of tooth picks on the floor, could I tell you how many are in the pile. Not a chance, But then again I don't have to buy my underware at Kmart.

#### P4Man

##### Distinguished
That is a pretty useless math exercise really, as it tells you nothing other than that what the benchmark selection looks like. if all you do is encode MP3s all day, go get your P4EE. Does anyone ever convert a 600 Mb Wav file to MP3 ? Even if you do, it takes about one minute only, you do that like 20x a day ??

When I rip a CD, I rip straight from CD to MP3, and my CD drive's ripping speed is what holds me back mostly, not the MP3 encoding, and that is on a lowly 2500+. About the same applies to the this new CoolEdit benchmark. it might be interesting for one or two users heavily involved in audio processing, but really, how many people normalize 260 min (!) WAV files all day long ? I have never even seen a 260 minute long WAV file ! that is about a full seasons worth of the Soprano's, all into one huge audio file. If even that processes in a minute, then see if I care about that benchmark.

Video encoding seems much more usefull to me, as I presume quite a few people like making "backups" of DVD's, or converting downloaded Divx files to DVD, but processing gigabyte audio files which only takes a minute or two anyway seems like total nonsense to me. Whats next, measuring how long it takes to open 2000 emails at once ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =

#### P4Man

##### Distinguished
Oh, not to mention your statistics include utterly useless "benchmarks" like PC Mark or Sandra memory benchmark. Those synthetic benchmark can help in making an analysis, but if you are going to purchase a cpu because of its PC mark score, well,.. be my guest. I'd rather buy one that is faster encoding my Divx video or running my games, but hey, that is just me.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

#### pauldh

##### Illustrious
Yeah, I can see points made on both sides being valid. Overall, I call them equal too. Yet at those prices, I could care less. The fact that the FX does so well in Games clearly makes it worthy ot praise. But the fact that in other areas it compares so poorly to the P4's, should strip it of any overall speed king title. I mean in the MP3 creation test, a P4 2.8 Precott is 12 seconds faster! So to me it is a comparison of each company showing off what they are capable of right now. I think the A64/P4/AXP benchies are far more valuable to most readers than those overpiced exotics.

I thought the conclusion to the review was weak also. But they know that the readers are smart enough to look over all their review charts and use them wisely to pick what chips are best for them. The FX53 is only the Gamers speed king. Unless it is purely a rich man's gaming box why go there? And the P4EE, why go there at all?

For me games are important, but not worth sacrificing other areas. I can't imagine playing any games that the FX53's price is justifiable. Like one of the nice chips costing 1/3rd as much is going to hinder your gamerplay? What are the current or even known future games that an A64 3000+ or P4 3.0 will just so hinder the gameplay making us say, man if only had sold my car and bought the FX53 instead! If one ever comes out, upgrade then.

ABIT IS7, P4 2.6C, 512MB Corsair TwinX PC3200LL, Radeon 9800 Pro, Santa Cruz, TruePower 430watt

#### P4Man

##### Distinguished
>But the fact that in other areas it compares so poorly to
>the P4's, should strip it of any overall speed king title.
>I mean in the MP3 creation test, a P4 2.8 Precott is 12
>seconds faster!

Like I said, I don't really think that test is really significant to 99.9% of the potential customers, but then again, I agree you could use the same argument about 2% better gaming performance, especially if you are comparing 200 FPS to 205 or so.

What I would like to learn more about, is potential increase for media encoding and 3D rendering using 64 bit software. The doubled number of SSE2 registers in long mode should have some good potential here. I have no idea if it will enable K8 to close the gap with the P4 or not, but it shouldnt be too hard to install linux, and compile some opensource codecs/apps and see what it gives ? After all, its not like it will take 5 years for the OS and apps to arrive, and if AMD64 gives some huge increase (some apps already do) 6 or even 12 months from here, that is something to consider when making a purchase today. Its not like I would expect to (have to) replace a newly bought P4EE/A64/FX class machine again within 12 months.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

#### endyen

##### Splendid
For rendering and encoding, I found the change between the 3.2ee and 3.4ee was rather poor when compared to the FXs. I also liked Hardocp's DVD2avi benchmark. This is a real time consuming prog, and the fx shines. There are other encoding and rendering progs that favor the Amd chips, but you know the story.
For me the big test of Widows for A64 will be multitasking. I see it making HT look like dog sit.

#### Piccoro

##### Distinguished
I wonder if buying a SMP Athlon MP or Opteron 24* system instead of a P4 HT system would be better for those apps that are time sensitive and take advantage of SMP.
I would think that if SMP is your thing, that you would consider a dual system instead of a single cpu (even over a P4 with HT).

I would think a Opteron 240/242 system would be THE working enthusiast's machnine.
Low cost for a SMP system, killer SMP performance and a good gaming machine on top of that (game performance of about a P4 3ghz CPU w/800FSB).

I just don't see buying a P4 3ghz CPU with HT beneficial for these types of things. A P4 2.4C-2.8c w/HT maybe, because of the cost savings but a 3+ghz P4 when you could get a SMP system for not much difference in price?

Am I missing something?

#### P4Man

##### Distinguished
I disagree. A dual 242 would be considerably slower than a P4C HT in a lot of things, and likely be more expensive to boot. A "typical enthousiast" requires fast single threaded performance as well, something a dual 242 hardly delivers.

However, HT really does seem to help as much as it sometimes does because of windows XP's rather poor scheduling. I would like to see some HT tests using windows server 2003 (which has an improved task scheduler). I would not be surprised it would show substantially less improvement in the typical HT friendly multi tasking benchmarks.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

#### Spitfire_x86

##### Splendid
I mean in the MP3 creation test, a P4 2.8 Precott is 12 seconds faster!
Magix MP3 Maker Platinum benchmark result doesn't reflect real world performance just like Sandra, PCMark, Sysmark. Because nobody uses this software. For MP3, LAME is the best encoder (also most popular). Among non-MP3 formats, Ogg Vorbis the the best and popular.

Performance difference in LAME MP3 is small and in Ogg Vorbis, both are equal.

------------
<A HREF="http://geocities.com/spitfire_x86" target="_new">My Website</A>

<A HREF="http://geocities.com/spitfire_x86/myrig.html" target="_new">My Rig</A> & <A HREF="http://geocities.com/spitfire_x86/benchmark.html" target="_new">3DMark score</A>

#### pitsi

##### Distinguished
Raw numbers rarely represent the whote truth. Keep that in mind! An example is synthetic benchmarks, whose benchmark results I don't even bother reading. Just to be a little bit more accurate, you have there the Sandra Multimedia Test where the P4EE is ahead by 48% on average (no.37). Do those number have ANY affect on everyday performance? Of course not! Do those numbers affect your results! Of course yes! Btw, there are many more similar results in your list like the one mentioned above.

I would be really interested though if you did your calculations one more time (although they don't actually mean something), without including any synthetic benchmark tests this time. I suspect (!) that results will be a little bit different this time.

#### TheRod

##### Distinguished

You are not a mathematician or a statistician. You can't compare results this way.

I will explain this with a little exemple :

Look at these results example :

CPUA CPUB
4000 1200 (popular gameX)
1000 5000 (fully optimized appY)
3000 2200 (popular appZ)

TOTAL :
CPUA CPUB
8000 8400

With these results CPUB is the winner, but obviously, it clearly trails in non-optimized, general apps.

This little example, illustrate the fact, that you can't only sum numbers with benchmarks. You must do statistics and you must regroup results for tha same kind of CPU works. And you must consider the importance of each type of tasks.

Because, obviously, if the reviewer use 10 audio/video benchmark against 2 gaming benchmark, it's clear that Intel will win. On the other hand, if he choose 10 games and 2 audio/video apps, AMD will rocks!

And as other stated, synthetic benchmarks, should not be considered in "buying" decision. They are great to compare scalability and performance between processors of the same family, but they can't be used to choose a processor over another when comes time to buy.

To get a better picture, you should regroup your numbers, by type of benchmarks and don't count SYNTHETIC benchmarks like memory bandwidth, these numbers means nothing in real-world applications. They only show that if an applications or game is memory intensive the CPU with more memory bandwidth should have an advantage.

--
Would you buy a GPS enabled soap bar?

#### Carnivore

##### Distinguished
I could be wrong on this, but isn't the high-end enthusiast chip market driven by games? It's teh gamers who buy the majority of them, so they are speaking to the market. SO if a highend chip outperforms another in games, it is faster in the market it was built for.

I am in no way shape or form an AMD fan boy. I have laways used intel chips up until the 64 bit chip came out. I didnt buy an FX, cause 700 bucks for a chip is just wacky to me unless you have lots of cash lying around.

If you read some of the reviews, they actually state that the EE chip was designed for teh gamer, and the FX chip outperforms it in teh market it was intended for.

And let's face it. Only us gamers are dumb enogh to spend the stupid amount of cash it takes to get the top of the line chip.

#### pauldh

##### Illustrious
Maybe some of the other enthusiasts here are only gamers, but I find in the close circle of geeks I know, over 50% of us are video editors and use their DVD burners on a regular bases. Maybe that's because I/we are considered old timers to most gamers. But we also greatly still enjoy gaming when time allows. I see system builders and power ap users moreso as ethusiasts than gamers. Yet how many enthusiasts don't enjoy games? But to say Gamers are the enthusiasts, so FX53 is the King of all enthusiast chips, isn't acceptable to me. Nor would the video editors chip be able to claim that clear top spot. If a chip comes out that is equal in all areas but kicks butt in games, I'll claim it's the King right along with you. But to win one and lose the other?

Just think, the large number if not the majority of gamers (my guess here, not a statistic I read) who play on a daily basis, aren't even shaving yet. Are they likely to be the enthusiasts who will shell out \$750 for a few more fps? A64 does interest me greatly as I have 1 system that is almost strictly for gaming and a seperate for video editing. But I must say, I'd rather spend a little extra dough to cut my video encoding times down and ensure no dropped frames, than go from 230fps to 260fps in a game. Don't get me wrong, Both A64 and P4 (and FX53/P4EE) have areas they shine in. And both offer acceptable performance in all areas. I am not against either of them. I Just can't see one has the right to where a crown right now.

ABIT IS7, P4 2.6C, 512MB Corsair TwinX PC3200LL, Radeon 9800 Pro, Santa Cruz, TruePower 430watt

#### pitsi

##### Distinguished
Valid point there! Everyone should keep in mind that the reason the Pentium 4 Extreme Edition was introduced, was to offer the ultimate gaming performance (and please be my guest to correct me here if I am wrong at this). Apparently Intel has failed (and I don't think anyone can disagree), despite asking 20-30% more for their top-of-the-line CPU.

In my opinion AMD is doing very well lately. I hope they can keep it this way because this can only be good for us. Especially if they manage a smooth transition to 90nm (only time will tell), and have the San Diego out before Q4 2004, then Intel will have some serious reasons to worry.

#### lisabob2

##### Distinguished
There is a "Simpson's " episode that has Homer saying something like,
"80% of statisics can be manipilated to get any answer you want. 15% of people know that"
not exact quote, but I think it gets the idea across.
Basically you can look at whatever data supports your view and ignore data that opposes your view.
What does it prove??
Bob

#### pauldh

##### Illustrious
LOL, Are you saying that any of us here ignore data that doesn't go our way? :lol: I agree. We are all entitled to our opinions and can spend our money as we wish. But claiming the opposite opinion/preference is foolish, that's a hard thing to prove. Not to mention a waste of time really. But it must be fun since it happens so easily.

ABIT IS7, P4 2.6C, 512MB Corsair TwinX PC3200LL, Radeon 9800 Pro, Santa Cruz, TruePower 430watt

#### jrosenst

##### Distinguished
TheRod Wrote:
---

You are not a mathematician or a statistician. You can't compare results this way.
---

First off, you’re correct (kind of), I am not a statistician, what's required to add any real validity to my analysis. I am however, a mathematician. My analysis was a little better then the way you portrayed it. I didn't add the raw numbers; I added % better/worse in each test. I also assumed to the man doing the review each graph was equally important, thus weighted each one the same. If my primes it true, then the analysis is accurate.

But I think my point here has been a little misunderstood. I didn’t write the thread to say the Intel was better then the AMD or vise versa.

I wrote this thread because I think the editorializing part of that review didn’t match the facts presented. I agree a lot of those tests don’t really mean a whole lot, but the reviewer must have thought otherwise or he shouldn’t have used them. So, if he feels everyone of those tests are required to accurately convey a chips worth, he has an obligation to conclude what the chips worth is based on his results.

Let me put it a better way. This is the summary:
“The new Athlon64 FX sets the new performance standard to beat, leaving the Pentium 4 3.4 GHz Extreme Edition far behind….”

Would you, after seeing the results, make the same claim?

#### TheRod

##### Distinguished
Let me put it a better way. This is the summary:
“The new Athlon64 FX sets the new performance standard to beat, leaving the Pentium 4 3.4 GHz Extreme Edition far behind….”

Would you, after seeing the results, make the same claim?

First, I would never say this. Because today's new CPU's are so close in performance that it's would be a lie to say "far behind...".

But the FX-53 and P4EE are marketed has TOP IN GAMING CPU's. And when we chack games benchmark and price comparison, it's clear that AMD is the winner. Better gaming performance for much less money. Even if the P4EE shines in other areas, these processor are meant to be used in gaming rig.

So, if you enter a computer store and ask for the best gaming PC they can offer you, they should sell you an FX-53 machine. But if you ask for the ultimate Video Processing workstation, they should sell you a P4EE.

--
Would you buy a GPS enabled soap bar?