[jmelrose]

Here is Macs Only's post. Again, I'd love to hear people explain what might be going on here. Is it true? They equalized the OS and the RAM amount between the two machines.

"867 MHz PowerBook G4 Speed Tests Redux: Last Thursday we posted our speed tests of the new 867 MHz PowerBook G4. Astoundingly, the new PowerBook came out slower than our 800 MHz model. We thought it might be a bug in the special build (6E62) of 10.2.1 that the 867 MHz models were running. We were permitted to update them to 10.2.2 and re-run our tests. We also updated our 800 MHz model to 10.2.2 and pulled out a 512 MB RAM chip to make both the same at 256 MB of RAM.

Both PowerBooks ran a little slower on 10.2.2 than 10.2.1 but the results were the same. The new 867 tested significantly slower that the 800 and we have revised our speed test accordingly. As improbable as it seems, we can only conclude that both 867 MHz PowerBooks that we used in the tests are defective in some way. This will remain a mystery until we can get our hands on another 867 MHz PowerBook G4."

Macs Only!

[riverfreak]

I'll leave the technical explanation to others that are better informed. But I will say this, a serious, informed evaluation would avoid the use of hyperbole like:

( boldface mine in all)

Astoundingly, the new PowerBook came out slower than our 800 MHz model. The new 867 tested significantly slower that the 800 and we have revised our speed test accordingly.

Was it a significant difference? Can I see the stats?

As improbable as it seems, we can only conclude that both 867 MHz PowerBooks that we used in the tests are defective in some way.

Really? An N of two hardly seems adequate to conclude that the 867 models tested were defective. How about trying to actually explain the difference?

This will remain a mystery until we can get our hands on another 867 MHz PowerBook G4.

Ooh. A mystery. Nice work Macs Only. Really thorough.

river

[jmelrose]

I do want to say, befrore we start jabbing the folks at MacsOnly with spears, that they do a nice job on their website, and are delightfully anti-Windows, which makes it fun to read. If they are doing something wrong, I'd like to hear about it, but let's not slam others in the community for doing their best. I am 100% certain that the folks there are just trying to bring to light a problem they've found.


That being said, I'm hoping my 1GHz will be faster than the 800 when it arrives today... :-D

Jeff Melrose

[Eug]

We have benches here already showing that the 867 is slightly faster than the 800 (not surprisingly).

In other words, they're probably doing something wrong.

[riverfreak]

Originally posted by jmelrose:
...
let's not slam others in the community for doing their best. I am 100% certain that the folks there are just trying to bring to light a problem they've found.
...

Point taken. Three cheers for all Mac Brethren! Clarus! Clarus! Clarus!

[Nebrie]

Are they that lazy that they can't remove the extra 512 ram module in the 800 then run the test?

[jmelrose]

Originally posted by Nebrie:
Are they that lazy that they can't remove the extra 512 ram module in the 800 then run the test?

Um, re-read again. They did.
How's the crow taste today? Burn your mouth a little, did it?

[Eug]

Originally posted by jmelrose:



Um, re-read again. They did.
How's the crow taste today? Burn your mouth a little, did it?

You should note that we already have several comparisons that peg the 867 at slightly faster than the 800, so clearly there is something amiss with their assessment.

At best there's some sort of software glitch. At worst they don't know what they're doing.

Interestingly, the 800 scores 80. The 867 scores 67. The 1 GHz G4 TiBook apparently processor cycles to 666 MHz. One also wonders if the 867 cycles to 666 MHz. ie. software or settings glitch?

[ivi]

Interestingly, the 800 scores 80. The 867 scores 67. ... One also wonders if the 867 cycles to 666 MHz. ie. software or settings glitch?

Good point there. It does.

[crouchingtiger]

sheesh, what's the big deal? They didn't conclude that the 867 is actually slower than the 800 -- they concluded that their tests must be faulty!

Nitpicking the words "astoundingly" and "significantly" is also pretty weak. I think anyone comparing the hardware specs of the two computers would conclude that there is no reason that the 800 should be faster, hence the "astoundingly". There weren't enough repetitions to do a t-test analysis but anyone reasonable would look at those numbers even with just 2 repetitions and say that they are "significant".

True, the benchmarking could have been rigorous, but I think that they have done a good job of reporting and trying to troubleshoot their strange results without trying to sensationalize it in a misleading way.

[Podolsky]

I seem to recall that PB 550's outperforms the 667 "combo" PB. Though I don't recall the details, (something to do with the L2 cache), I do recall that it was real. So, with CPUs so close as 800 v. 867 there are a whole bunch of other factors that could come into play to place the 800 ahead of the 867. The big question is what are they and do they impact the 1 ghz???

[Eug]

Originally posted by Podolsky:
I seem to recall that PB 550's outperforms the 667 "combo" PB. Though I don't recall the details, (something to do with the L2 cache), I do recall that it was real. So, with CPUs so close as 800 v. 867 there are a whole bunch of other factors that could come into play to place the 800 ahead of the 867. The big question is what are they and do they impact the 1 ghz???

Like I said before, the 867 has already been shown (on other benches) to perform slightly faster than the 800, but obviously the 1 GHz is by far the fastest.

[Paul Huang]

I can guarantee you that if you put in a 5,400 RPM drive that has either 8 MB or 16 MB of level-2 cahce, the benchmark will easily be enhanced by at least an average of 20%.

[azark]

I seem to recall that PB 550's outperforms the 667 "combo" PB. Though I don't recall the details, (something to do with the L2 cache), I do recall that it was real. So, with CPUs so close as 800 v. 867 there are a whole bunch of other factors that could come into play to place the 800 ahead of the 867. The big question is what are they and do they impact the 1 ghz???

I think you're talking about the PB 500 not 550.

First Titaniums (400/500) have 1MB Level 2 cache.The 550/667 have only 256k of level 2 cache
(Since the 667/800 they have 256K L2 + 1MB L3...)

[gg1234]

Originally posted by Paul Huang:
I can guarantee you that if you put in a 5,400 RPM drive that has either 8 MB or 16 MB of level-2 cahce, the benchmark will easily be enhanced by at least an average of 20%.

Have you read this?

[Eug]

Originally posted by Paul Huang:
I can guarantee you that if you put in a 5,400 RPM drive that has either 8 MB or 16 MB of level-2 cahce, the benchmark will easily be enhanced by at least an average of 20%.

Well, in the case of the Macsonly.com benchmarks, that is not true. They are basically CPU benches.

As for Xbench, that may not be the case either since jhunt tried a 5400 rpm 8 MB cache IBM GNX drive in his TiBook GHz, and only got a minor boost compared to the Toshiba 4200 rpm drive.

Also, the 5400 rpm drives that the 800 MHz TiBooks come with are 2 MB cache drives.