I have recently seen discussions on weather or not the new M-2 Sempron CPU was actually slower than the old 754 Sempron it replaced.

There was an article in the Inquirer sayin it was slower, and today I found a review that showed it to be quicker, especially when Overclocked!

I am not trying to single out this site or this writer, I am just using it for an example, YOU MUST learn to question results and read the fine print, and sometimes, read between the lines...

One major review site that needs not to be named has a "Intel" fetish, and in there reviews I have caught many things that would skew things over the years that tend to put Intel in a more favorable light, even if every other review site shows just the opposite!

Now that is not realy the the point of this note, on the surface by quickly reading the article below, it would be easy to assume that the new Sempron realy is much faster, I realy don't care about that for the purpose of this thread, but after looking a bit closer something caught my eye that pretty much skewed the results of his testing.

http://www.legionhardware.com/document.php?id=560&p=3

He didn't do a real "Apple to Apples" test it appears to be on the surface, and did nothing to explain it, other than noting DDR300 speeds in the charts.

The socket 754 system must have used a big memory divider to only be running at DDR300 memory speeds! Sandra clearly shows it having less bandwidth at 2.4GHZ (300x8) then at stock 1.6GHZ! (200x8)

Now the synthetic tests clearly show that it is down, so how can you compare things fairly in an overclocked situation like this? Real world gaming tests don't show it hurting performance as much as the synthetic test would lead you to beleive.

I have said many times AMD 64 is not realy hurting for Memory Bandwidth, meaning that going to from DDR 400 to DDR500 at the same clock speed will not realy increase real world apps like gaming significantly, a few percent at best.
Going down to DDR333 from DDR 400 will not kill real world performance much either...but DDR-2 667vs DDR-1 300 will certainly show some differences, especialy Sandra.

Also of note is the slower the clock speed, the less hungry for Memory bandwith AMD 64 is... 2GHZ going from DDR400 to DDR 500 will not nearly show as much of an improvement as if the CPU was tested at 3GHZ.

Gaming with AMD-64 is more responsive to CPU speed increases than memory speed, wich can clearly be seen on this page.
http://www.legionhardware.com/document.php?id=560&p=5

Overclocked at 2.4GHZ Quake 4 in the old 754 rig that is crippled by running DDR300 was 95.2FPS, the AM-2 was only a tick faster at 95.6FPS with 1 to1 DDR2 667!

My guess is if he would have been running the 754 rig at DDR 400, the gap would have closed completley. If he had ran DDR1 without a divider and true DDR600 speeds, I bet it would have been slightly faster than AM-2!

In my opinion, the only "Valid" comparison would be the stock 1.6GHZ numbers, the rest is unfair, and his conclusions about Overclocked performance are skewed by it.

Again, I am not pointing fingers here, just trying to make people aware of how some fine print (Or even worse, things that may not tell you about the configuration) can effect the numbers and conclusions people may draw from reading just one review, or listening to heresay.

I do not beleive the writer of the article was trying to mislead anyone for there own purposes, just a simple oversight, most reviews I read are very accurate and unbiased...just food for thought!

wow great article as usual necro
well i know i learned a thing or two thanks again


ps..hey whats that sound do i hear something STICKY ?

Some interesting points Solid. I agree that the stock results are the only valid ones there. IMO AM2 Sempron's are bound to preform better than there S754 counterparts. Dual Channel RAM would be a big help.

Its pretty weird that he used different hardware for the different systems. It would have been good if he'd used DDR 400 2-2-2-5-1T and DDR2 667 3-3-3-5-1T, I reckon that would give similar preformance. Also he used different harddrives for each system.

Zaltan, Great catch on the timings and the HD's!

(Though HD's rarely will reflect in gaming performance other than load times, there are other tests that will)

I was so blinded by the glaring of the DDR300 speeds, I did not bother too look any farther!

You are doing exactly what I hope all people will do, dig deep and look at ALL the details!

+Rep

Quote: Originally Posted by $SOLID$ Necro Zaltan, Great catch on the timings and the HD's! (Though HD's rarely will reflect in gaming performance other than load times, there are other tests that will) I was so blinded by the glaring of the DDR300 speeds, I did not bother too look any farther! You are doing exactly what I hope all people will do, dig deep and look at ALL the details! +Rep

Thanks, I try my best .

This is aproblem anyoen who writes or reads reviews has to deal with. It is very, very difficult to do "apples to apples" comaprisons of hardware, especially hardware from different "generations". I know the last mobo review I wrote I had no other A64 system to compare it to. So I compared it to a 2yr old P4 3.0ghz machine. Why? To give some point of reference, otherwise the numbers are meanigless.

Benchmarks, both real world and synthetic, have their place in the buying decision process. However, they should never be the deal maker unless you happen to have 5-10 reviews that all say the same thing. You have to read the review and pay attention to the details. They are very rrarely apples to apples.

Even when sites like Toms Hardware where they can run 10 motherboards through their paces with identical hardware you have to make sure you only compare the results in that review. You can never compare results that one reviewer got to what another did, it will never be a valid comparison.

I guess what I am rambling about is that Necro is 120% correct.