Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
PC vs. Mac Pro?
- Thread starter Android who?
- Start date
- Sort by reaction score
last try. continuing on the last example..
1 core finishes calculation in 2 minutes.
.....
1 core = 2 minute
2 core = 1 minute (add 1 -- 100% increase)
4 core = 30 sec (add 1 -- 50% increase)
8 core = 15 sec (add 1 -- 25% increase)
the more cores you add, the less you gain.. the value of each core diminishes
(hint- stop trying to steer the math to fit your opinion -- this isn't about opinion.. let the math happen first then reach conclusion)
----------
haha. too funny.
for fun, let's see your calculation showing increase in core count being linear.
as well as an example of a diminishing equation.
Your original example is a machine I'd love to have. It actually has accelerating returns, not diminishing.
Going from 3 to 4 cores, a 33% horsepower increase, doubles performance!
Amazing!
Weird though that going from 1 to 2 cores, a doubling of horsepower, also only doubles the performance.
That fourth core must be magic to be worth 3x as much performance as the 2nd one.
----------
Stunning.
Although your math looks correct, your interpretation of how it works is WRONG.
the 'amazing' machine is what you guys are saying is happening.. you're saying the equation doesn't diminish.. the example was meant to point out how you guys are seeing it(linear) followed by an example of what really happens(diminishes)
now, you're actually taking the example of what you're saying happens-- and making fun yourself without even realizing it.
please tell me you're doing this on purpose and just pulling my leg.
----------
how does this statement make any sense ?
math is math.. you don't interpret it.
Lets look at your example.
1Core = 2min
8 core = 15sec
How many times faster is 15 seconds than 2 minutes? 800%
BOOM!!!!
Just stop it man.
Break down? Only if you don't know what you are doing or you bought cheap pieces of junk initially.
2007 c2d laptop works fine (CPU/RAM is limiting)
2009 c2q desktop works fine
2009 i7-920 desktop works fine
2000 pentium 3 dell desktop works (lol)
I can even power up the 1985 desktop with the 'turbo' button and get it to work.
I've never had a PC beakdown as badly as you seem to be describing. If you take care of your stuff it lasts.
Actually, not really. Since you can run Windows on it if need be (Boot camp or Virtual). So in effect you currently really can't beat the nMP hardware wise for either OS you wish to run.
what are you, some kind of genius or something?
you're right, that's what my example shows. as in, the example is correct and proves what I'm saying.
you're completely failing to miss the part about each additional core not having the same weight in the equation. if each CPU weighed the same, the other example, the linear one, would be true..
you're now doing the same thing bleaker is doing.. making fun of yourself without even realizing it.
Last edited:
you're confusing yourself with the way you're wording it (or-- the way you're wording it is showing that you're confused)
you're seeing 1 core is worth 100% or 8x100 = 800
that's good and all but answer this following the same logic--
how many time faster is 1 minute than 2 minutes? 200%
---
follow your same logic/wording and you see the additional core in the 2min->1min example is worth 200% whereas in your BOOM example, each additional core is worth 100%
Last edited:
The law of diminishing returns means a flattening of the output curve, despite fixed-rate inputs. It does *not* mean a diminishing *percentage of the total* output increase per unit input.
Of course the *percentage of total performance improvement* is less when going from 3 to 4 CPUs than from 1 to 2 CPUs. That has nothing to do with the law of diminishing returns. It's because each additional CPU is adding proportionately less of the total. If Amdahl's Law did not exist this would still happen. It has nothing to do with the topic under discussion.
In a basketball game, going from 10 to 20 points doubles the score. Going from 100 to 110 points (same 10 point increase) only increases the score 10%. Whether applied to CPU cores or game points, this is not the law of diminishing returns.
I used to have this dog that would chase his tail. Around and around he'd go. Always convinced he could get it. Never did.
I'm going to stop here, flat, as there is nothing I can do to help you.
I do sincerely encourage you to take some remedial math courses though, particularly to understand how compounding and percentages work. Failure to understand these concepts could have severe consequences in your personal finances.
If you are a student then perhaps you haven't covered this yet.
I'm going to stop here, flat, as there is nothing I can do to help you.
I do sincerely encourage you to take some remedial math courses though, particularly to understand how compounding and percentages work. Failure to understand these concepts could have severe consequences in your personal finances.
If you are a student then perhaps you haven't covered this yet.
and once you finally wrap your head around this, try to start bringing other factors into the equation.
one of which is that as core count increases, each individual core slows-- so, you're not really doubling speed when doubling the number of cores.. it really works out to around 60-70% (example.. 12core nmp is not twice as fast as 6core nmp)
after you figure that one out, bring in another factor- co$t.. as core count increases, the cost per performance that each additional core offers increases -- drastically and rapidly..
(an example - guessing on the cost-- 6core = $500 and renders in 1 minute... 12core = $2000 and renders in 40seconds.)
add all this up and you'll see how the more cores you buy, the more ripped off you get.. not only do you pay more for each core you add, you get less performance increase out of it-- then on top of that, you get another round of less because GHz is decreasing as well.
afterwards, watch bleaker brag about his 24core machine in a new light.
one of which is that as core count increases, each individual core slows-- so, you're not really doubling speed when doubling the number of cores.. it really works out to around 60-70% (example.. 12core nmp is not twice as fast as 6core nmp)
after you figure that one out, bring in another factor- co$t.. as core count increases, the cost per performance that each additional core offers increases -- drastically and rapidly..
(an example - guessing on the cost-- 6core = $500 and renders in 1 minute... 12core = $2000 and renders in 40seconds.)
add all this up and you'll see how the more cores you buy, the more ripped off you get.. not only do you pay more for each core you add, you get less performance increase out of it-- then on top of that, you get another round of less because GHz is decreasing as well.
afterwards, watch bleaker brag about his 24core machine in a new light.
sure, stop there.. that's fine.
I do notice however that you've failed to explain how all cores are created equal.
answer it this way--
again-
I have 1 core and it calculates an equation in 2 minutes
how much money do I have to spend to double my performance?
how much money do I have to spend for quadruple performance?
how much money do I have to spend for 8x performance?
what about 16x?
32?
then 64x?
please, pretty please-- answer those questions
and I'll even make it easier for you-- let's say 1cpu core costs $100 and you can put as many cores as you'd like into one machine.
(edit)
let's see it.
(edit-- changed the numbers a bit to make it even easier for you to figure out)
Last edited:
meh, that's nothing. at one point on this forum, there were 50 people trying to tell me how wrong I was about something.. ends up, I was right all along..
three people is nothing
(in the meantime, go figure out that scenario I posted above)
Sorry five, but you're wrong on this one. Let's take your example to a different scenario (using the same math):
Driving my car from point A to point B...
10 mph gets me to my destination in 10 minutes.
20 mph gets me to my destination in 5 minutes.
40 mph gets me to my destination in 2.5 minutes. And so on.
Under your "non diminishing" scenario we should expect:
10 mph gets me there in 10 minutes.
20 mph gets me there in 5 minutes.
30 mph gets me there in 2.5.
Granted, that would be one hell of a car, but that's not how the math works.
What you're really thinking about is price to performance ratio, which is a totally different thing. It's true that the more cores you add, the price increases at a different rate. But when we're talking about the workstation/server market it's not a very good argument. Once you're in the 12 core range, the price stepping halts. So every additional 12 cores added from there cost the same price. Apple made a decision to go with a single CPU model only, which might make sense for their business plan. But don't buy into the nonsense excuse that it's because everything's going to the GPUs. CPU necessary tasks aren't going anywhere and there are reasons why many wanted a dual CPU option.
right, the second paragraph is what i'm saying.. and applying it in this way:
if i have a 4core computer and would like to double my performance, i will need to spend $400 to do so (assuming 1 cpu costs $100)
if i have a 8core computer and would like to double my performance, i will need to spend $800
if i have a 16core computer and would like to double my performance, i will need to spend $1600
etc..
or, say i started computing with a core2duo.. i went to a quad and experienced 2x speed and it cost $200 for the experience
next time i want to do that, it will cost me $400
then the next, $800
i can't double my performance for $200 continually...
but i see what you're saying that this isn't flattening the curve if looking at it from "i have a 2core-- should i double or quadruple my performance when i upgrade" ... that is a linear decision..
----
however, when you consider the actual costs of cpus in a computer.. say, nmp quad being the base then considering pricing on 6core, 8core,12 core.. the resulting graph will not be a straight line.. the best value is the 6core then the curve flattens from there..
the curve flattens faster again once you consider you're losing GHz..
multi socket machines will help alleviate the secondary flattening since you don't have to take as big as a ghz drop (or, in certain configs, the ghz wouldn't drop at all)
i do think if you chart this out, you'll see that the law of diminishing returns does apply to computers..
----------
yep, i see what you're saying (as well what joema said)
threw me for a loop.. the price to performance ratio isn't my main thing though and i was saying that aspect would simply compound the problem which was already there..
but another way to say what i was saying is:
if i have a 12core machine and i turn one off the 'penalty' is much less severe than if i have a 2 core machine and i turn one off..
the less cores in a machine, the more valuable each one is..
likewise, the more cores in a machine, the less valuable each one is..
do you see what i'm thinking? (joema summed it up in the second paragraph of his i quoted)..
but i'm going to have a rethink.. it still seems that should chart as a curve instead of a line but now i'm not so sure.
right-- your example is saying the same thing i was saying when i originally posted it.. that scenario 2 makes no sense and that scenario 1 is true..
the key difference between your car scenario and what i'm saying about core count is that i'm talking about a cpu core working at 100% and you're talking about a single car working at a percentage of its total..
or- your car scenario would apply to what i'm saying if i were throttling a single cpu..
car engine = single-core machine
discussion = multi-core machine.
but i'm not-- i'm adding cpu cores in order to get faster performance and each time i add one, it's less valuable than the last..
your mph comparison has time on both sides.
my cpu comparison has time on the left and hardware requirements on the right..
do you see the difference?
[EDIT] though i still see the glitch in what i'm saying.. i'm seeing a curve because the time is being listed non linearly (1, 2, 4, 8, 16 --instead of-- 1, 2, 3, 4, 5) which will make a curve instead of a straight line..
i believe it's supposed to do that but i'm a bit hesitant to say it outright right now..
----------
nah.. just thrown for a loop is all.
(but don't worry-- if and when you guys prove me wrong.. i'll acknowledge etc..)
Last edited:
Well, it was my reality. And seriously, if y'all are such fans of Wintel systems, why are you here in this forum? Perhaps you should be in a forum dedicated to Wintel systems?
I would be seriously disappointed if this forum was not biased toward the promotion of Apple products.
Am I an Apple fanboy? Hell, yes. Am I biased? Meaning is my opinion non-objective? Hell, yes.
This particular thread's opinions are highly subjective, and I have given my two cents.
Do I think that Windows is going in the right direction with Win 8. Hell yes. It runs better than XP Pro on VMs which we use at work. In fact, I like Windows systems, but since the question was PC vs. MacPro, I gave highly biased opinions in favor of MacPro systems.
Buy a MacPro. Best investment you will ever make if you need the power.
Why are you getting so defensive and angry? You said
I'm saying there is absolutely no grounds for this statement. PC's (can if taken care of) last a long time.
I like apple products but I can see two sides of the story. Its perfectly fine to be biased but a qualifier should be mentioned, rather than stating it as fact.
The thread title is "PC vs. Mac pro"; I'm perfectly on topic.
The OP even stated
Meaning he would prefer an objective view. He does not want a biased answer.
Can One build a better PC than a Mac Pro for less money?
Many already said it: Depends on your needs.
Matching the nMP in terms of hardware and noise level as we'll reliability is hard and will always be costlier. Always.
But why is that?
Nobody here actually thought about the most expensive in a project like this. Time and labour.
You will have to do some research and a lot of planning is required in order to achieve this feat.
In addition some said that a Hackintosh will do the trick in terms of the OS.
No it won't. It will also take time. Doing an update can be tricky. It will also require some skills that a Dumb User In Front Of Screen (DUIFOS) in 99% of all cases doesn't have.
Ok then let's get Windows but add more power to it than needed compared to a nMP for the same task because you always need a Antivirus and other security apps.
How does Windows ask me to restart the machine on weekly basis in order to install the latest patches. That's the time that I cannot use for productivity. Same applies to Hackintosh. I need to do some research. Takes time, too.
My time is valuable. And while you all are drooling about your PC / Hackintosh that might be cheaper whenever you finished building it - how shall I say it gently?
Well I already earned some money with my workstation while you are still ordering the parts.
I while be working while you are still configuring your Hackintosh OS or trying to get rid of your unneeded apps on your Dell or HP.
The Mac works out of the box. Take that time into consideration. It's the most expensive piece of all.
Many already said it: Depends on your needs.
Matching the nMP in terms of hardware and noise level as we'll reliability is hard and will always be costlier. Always.
But why is that?
Nobody here actually thought about the most expensive in a project like this. Time and labour.
You will have to do some research and a lot of planning is required in order to achieve this feat.
In addition some said that a Hackintosh will do the trick in terms of the OS.
No it won't. It will also take time. Doing an update can be tricky. It will also require some skills that a Dumb User In Front Of Screen (DUIFOS) in 99% of all cases doesn't have.
Ok then let's get Windows but add more power to it than needed compared to a nMP for the same task because you always need a Antivirus and other security apps.
How does Windows ask me to restart the machine on weekly basis in order to install the latest patches. That's the time that I cannot use for productivity. Same applies to Hackintosh. I need to do some research. Takes time, too.
My time is valuable. And while you all are drooling about your PC / Hackintosh that might be cheaper whenever you finished building it - how shall I say it gently?
Well I already earned some money with my workstation while you are still ordering the parts.
I while be working while you are still configuring your Hackintosh OS or trying to get rid of your unneeded apps on your Dell or HP.
The Mac works out of the box. Take that time into consideration. It's the most expensive piece of all.