Something doesn't add up...

J

JimNoble

Registered
http://www.apple.com/powermac/specs.html

That Photoshop performance chart suggests that a dual 1GHz PMG4 is only 13% faster than a Single CPU 933MHz PMG4 - ie, by no more than the difference in CPU speed (what's the 2nd processor doing?)...

Or does this just show that the speed of the powermac is fundamentally limited by the memory sub-system speed (both 933MHz and Dual 1GHz PMG4s have the DDR L3 Cache...)

Jim
 
I would say PS isn't using the dual CPUs effectively.

I was ripping CDs at home, using iMovie, browsing, receiving emails etc with my dual 450 and things just cruised along. Here at work I tried ripping a CD and doing email at the same time (733 Quicksilver) and the thing was a dog. Properly written apps simply smoke on the (old) duals.

Now, the results I see when i compare my dual 450s with my 733 at work MIGHT indicate that the higher clock rates are really not a benefit to most apps, i dunno. Is the instruction pipeline longer on the 733s than the older dual 450s? I've not done much investigation of the G4 chips after I bought my dual 450 machine, but suffice to say I am not impressed with the single 733 i have at work.
 
I don't know how Apple tested those specs, but I would tend to agree with testuser. Wait until Apple actually tells you how they measured the performance or until some actual reviews of the new PMG4s come out.

AlanCE: Photoshop uses dual processors fine. It's been optimized to use both the Velocity Engine of the G4 AND to use dual processors. So I don't think that's the problem. And by the way, every G4 has a pipeline that is seven stages long -- this won't be changed until the next generation chip (i.e.: the G5).

The other thing is that you need to be wary of what the percentages actually mean. I myself am confused on this point, too. Anybody care to explain what "50% faster" actually means? Let's say a certain function takes 50 seconds. Would 25 seconds be 50% faster? *** simX thinks.
 
Originally posted by simX
And by the way, every G4 has a pipeline that is seven stages long -- this won't be changed until the next generation chip (i.e.: the G5).
Click to expand...

Wrong.

Older G4's have a four-stage pipeline.
Apple (well, Motorola) changed with the 733 and up.
 
Originally posted by simX
The other thing is that you need to be wary of what the percentages actually mean. I myself am confused on this point, too. Anybody care to explain what "50% faster" actually means? Let's say a certain function takes 50 seconds. Would 25 seconds be 50% faster? *** simX thinks.
Click to expand...

No, 25 seconds would be 100% faster. 100% faster = twice as fast.

50% faster would be 1.5 times faster, so something that took 50 seconds would be 33.3333333 seconds.

Someone check my math on that second one, but I'm 99% sure that's correct.

Another way to look at it is like this: 200MHz is 100% faster than 100MHz. 150MHz is 50% faster than 100MHz. 300% faster is something multiplied by 4 -- NOT 3, as it may seem. Since 100% = 2x, 200% = 3x, and 300% = 4x faster.

I agree -- percentages can be a little difficult to understand at first... I tend to understand "four times as fast" or something like that quicker than "300% faster," just because the percentage requires a little math-computation in order to turn it into a number you can more easily understand.
 
Originally posted by starfleetX


Wrong.

Older G4's have a four-stage pipeline.
Apple (well, Motorola) changed with the 733 and up.
Click to expand...

Hmm, really? Didn't know that. I guess you learn something every day...

About those nasty percentages, I'm still iffy on how to understand them. I guess 100% faster is twice as fast, but it's still really weird.

Help me out here -- if 100% faster is twice as fast, which means effectively where the benchmark system is at 100% speed (meaning it's as fast as the benchmark, which is itself), then the faster system is at 200% speed, right? So if, let's say, the Pentium that Apple is using as a benchmark is set at 100%, and it performs a certain task in 100 seconds. The Dual GHz G4 is supposedly 72% faster, so it is 172% the speed of the benchmark Pentium. So how long would it take to perform the same task on the Dual GHz G4?

UPDATE: OK, I think I actually get it now. So if 100% faster is effectively 200% of the benchmark system, then this is 2x as fast. So you take the time it takes for the benchmark system to do a certain job and divide it by the factor, in this case 2. So something 100% faster than the base Pentium Apple is using would take 50 seconds to do a task that the Pentium took 100 seconds to do.

So, using that, if the Pentium took 100 seconds to do a task, then the 800 MHz G4 would take 79.37 seconds, the 933 would take 66.23 seconds, and the dual 1 GHz would take 58.14 seconds. And using the same reasoning, I can use the 933 as a benchmark for the dual 1 GHz, and use the reverse of the method described above to get that the dual 1 GHz machine is 13.91% faster than the 933 MHz machine.

Whew.

So now, if you say 933 MHz is the base clock speed, then 1 GHz is 107.3% of that clock speed, which means the clock speed is 7.3% faster, contrasted with the real-world estimate (by Apple, anyway) that the dual 1 GHz is actually 13.91% faster. So it seems maybe it does add up, eh?
 
ok, i am not a hardware techie, but everywhere else in the world of statistics 100% means the same as the original or the complete amount. you cannot be 100% more or less than anything. it is impossible to be 100% faster. 200% is twice 100% or twice as fast. 50% is half 100% or half as fast. apple's use of statistics are confusing the way they are worded. even the graphical representations are misleading to anyone used to using stats with percentages in any other discipline. :confused:

the relative values make sense, but the actual values don't
 
The difference is "faster" vs. "as fast"

If something is 100% AS FAST, then it is identical in speed.

Apple has purposely chosen "faster", because it's a better phrase than "as fast". But percentage-wise, saying 150% as fast comes out better than 50% faster.

I'm not sure which I like better, I guess I'm leaning towards Apple's "faster" approach.
 
testuser - when you talked about consumer products that use the 20% more approach instead of saying you get 120% of this product i realized why apple has taken this approach although it still doesn't make much sense to me. it is the old 2 is a bigger number than 1 so i'll go first approach. it would make more sense to me to say it gets the same thing done in 75% of the time or that it is 125% faster with bar graphs that are smaller either way. i understand why now, i just don't think it makes sense still.:rolleyes:
 
posted by my friend Ed
it would make more sense to me to say it gets the same thing done in 75% of the time or that it is 125% faster with bar graphs that are smaller either way. i understand why now, i just don't think it makes sense still.:rolleyes:
Click to expand...

If a process is done in 75% of the time, isn't it running 33% faster (that is at 133% of the original process)?

:confused:
 
If a G4 leave a station in LA at 50 mph, and a Pentium 4 leaves a station in NY at 60 mph...

which one runs my current apps better?









ANSWER: The G4 :D
 
If a process is done in 75% of the time, isn't it running 33% faster (that is at 133% of the original process)?
Click to expand...

how did we go from quarters to thirds?


this way of stating percentages is as screwed up as the rest of business math, which i remember now has no ties to logic but is purely convention.

i still want a bar graph that shows less if its faster or bigger if it is showing how much more it can do. but i am expecting logic in an entirely unlogical presemtation. my bad;)
 
The percentage thing is very simple physics or maths.

speed is speed, time is time. The link is distance. In a benchmark test distance is represented by the task being done. This is constant.

If we say that something is 50% faster then we are saying that its speed is 50% more than it was, where the original base was 100.

So, if the speed was 50 units, then the new speed would be (50 * ((100+50)/100) = 75 units.

How does this affect the time?

Well, distance = speed * time

Distance is constant so this means that (speed * time) is constant. If speed is 1.5 times what it was, then time must decrease to counter it, and this means that time is 33% less than it was.

1/(1.5) = 2/3

That is why processor makers like to use speed, because it produces larger %'s than time.

The reason why we have gone from halves to thirds is due to factors of numbers. A 50% increase in speed represents (150/100) which when simplified equals 3/2. We perceive this as 'half more' than it was before. As we have seen, in order to keep distance constant time must go to (100/150) or 2/3. We perceive this as two thirds what it was.

Hope that this helps a little.

R.
 
roger, thanks! that was by far the clearest explanation. i totally get it. i stilllll don't agree with representing it that way but i now understand all the numbers, what they mean and why tech heads would want to use them that way!! and since real tech heads and really naive people are probably the only ones that care about these fractional differences in most real time processes (not all, I know) then it makes sense to do it this way. again thanks for taking the time and effort to put it so straightforwardly even i can get it!!:p
 
Glad to see that my expensive education hasn't gone totally to waste.

Ed - I have edited the above post a bit to explain the thirds/quarters question that you had.

R.
 
If it's a photoshop benchmark, are they running it in classic? If so then it CAN'T really utilize the second processor. If it's in 9, then yeah, SDRAM seems to be an issue. As for the percentages ... I hate marketing people. There are plenty of clear and concise methodologies by which we can show relative differences. %difference is used not because it's clear, but because it's so unclear. Although technically correct, the nembers lead to emotional factors which are used by the marketmonkeys. I always state things in "as fast" terms instead of "faster" terms because I find that te be more clear. I'm not a salesperson. I'd beat myself in the head with a tack hammer while chewing on broken glass first. I like explaining things to people, I hate confusing them. And marketing literature seems to be geared toward misleading, not revealing.

$24.95 is roughly equal to $25, but no morketing phreak will EVER price something on the dollar. There are liars, there are damned liars, and there are statisticians. It's all a sham. I hate it all. I fear where capitalism may go now that it feels righteously dignified at the fall of the communist empire. I'd like to go back to not lying to people at same point.

I apologize for ranting here ... it just - came out. I couldn't stop. sorry.
 
This damn thread has given me a headache. Let's just leave it at this: the new macs are faster than the old macs. That's all I need to know. Percentages are usually fluffed in marketing anyway.
 
You must log in or register to reply here.
Back
Top