Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Keep or return

T

The Mercurian

macrumors 68020
Original poster
Mar 17, 2012
2,160
2,443
So,

finally bought new 15" MBP after much deliberation. Judging by geekbench scores expected 40% improvement versus my old machine. Just run an exhaustive test running same statistical model on old machine (2011 15" MBP) versus new one. Very disappointed that the real life speed test was only about 12% quicker. Thinking to return. Main reason for expensive new machine was more speed and I didn't get the speed boost I expected.

Opinions ?
 
The Mercurian said:
So,

finally bought new 15" MBP after much deliberation. Judging by geekbench scores expected 40% improvement versus my old machine. Just run an exhaustive test running same statistical model on old machine (2011 15" MBP) versus new one. Very disappointed that the real life speed test was only about 12% quicker. Thinking to return. Main reason for expensive new machine was more speed and I didn't get the speed boost I expected.

Opinions ?
Click to expand...

The 2011 15" and 17" MBPs all have a manufacturing flaw that causes the GPU to fail.

So I'd keep the new 15" if I were you.

And you won't know when you'll need the extra power. It's always good to have extra power under your hood even if you don't need it now, because you might need it in the future.
 
yjchua95 said:
The 2011 15" and 17" MBPs all have a manufacturing flaw that causes the GPU to fail.

So I'd keep the new 15" if I were you.

And you won't know when you'll need the extra power. It's always good to have extra power under your hood even if you don't need it now, because you might need it in the future.
Click to expand...

You are the guy with 11 macs right ? I'm not sure I should listen to you :p

I'm aware of the radeongate stuff - the older one is still under applecare however (bought in 2012)
 
The Mercurian said:
You are the guy with 11 macs right ? I'm not sure I should listen to you :p

I'm aware of the radeongate stuff - the older one is still under applecare however (bought in 2012)
Click to expand...

Yes, I'm that guy. I'm also the guy from Queenstown in New Zealand :)

I used to have an early-2011 15" MBP that succumbed to Radeongate 3 days after AppleCare expired.

And when comparing with the mid-2012 15" cMBP with 1GB GT650M, the performance gap between it and the 6750M was pretty huge.

It's up to you whether you want to listen or not, I'm just stating my opinion :)
 
The Mercurian said:
So,

finally bought new 15" MBP after much deliberation. Judging by geekbench scores expected 40% improvement versus my old machine. Just run an exhaustive test running same statistical model on old machine (2011 15" MBP) versus new one. Very disappointed that the real life speed test was only about 12% quicker. Thinking to return. Main reason for expensive new machine was more speed and I didn't get the speed boost I expected.

Opinions ?
Click to expand...

A lot of the improvements in the last couple years have been focused around power-efficiency. What is the battery life improvement compared to the 2011 model? Also, if you got the 750M model, GPU performance improvements should be huge.
 
Freyqq said:
A lot of the improvements in the last couple years have been focused around power-efficiency. What is the battery life improvement compared to the 2011 model? Also, if you got the 750M model, GPU performance improvements should be huge.
Click to expand...

Battery life is alot better although that is not overly relevant since I more or less use this as a desktop replacement.
GPU irrelevant to me.

I guess summary is that I spent about €3000 for performance improvements that I won't achieve. That seems wasteful to me and not enough benefit over the old one - which I can continue to run - if radeongate gets if then they will have to fix it and even if that doesn't happen I could just return this one and keep the 3k aside for Broadwell model without really having lost anything.

Decision made. Return. Thanks folks for you opinions! I do listen yjchua95 - I'm merely cantankerous and prone to disagreement :D

Enjoy the upcoming ski season you lucky git :D
 
I'm not sure where the 40% were coming in, but I've never seen jumps in performance like that (I wish I did :) )

Anyways, its still a fine machine, I'd keep it, since the Broadwell chipset is being delayed. Enjoy what you have :)
 
maflynn said:
I'm not sure where the 40% were coming in, but I've never seen jumps in performance like that (I wish I did :) )

Anyways, its still a fine machine, I'd keep it, since the Broadwell chipset is being delayed. Enjoy what you have :)
Click to expand...

Not comparing sequential models - there are a couple of iterations in between remember. Geekbench score for 2011 model is about 10,000, for 2013 model about 14,000. Given that my application is basically a pure CPU task I would have expected about a 40% improvement. I'm getting about 12%. Don't think thats worth 3k.

Lets put it this way - take 4 hours (240mins...ish) to do the base model - so can do 2 in a working day. If it really was 40% quicker that would be about 144 mins to do a model (2hrs 20mins) and I could do 3 in a day. But as it happens it works out about 215 mins ( 3hr 35min) on the new machine - which frankly is not worth 3k to me.

Anyhow as it happens I feel kind of sick today and don't want to drive 2 hours to the Apple store so I will keep it to ruminate for another day or two until I feel a bit better.
 
I would keep it just to avoid the GPU failure issue.

Artificial benchmarks seldom have much to do with real world results. Most of the time, you'll never see an improvement as dramatic as the benchmarks make you believe.
 
556fmjoe said:
I would keep it just to avoid the GPU failure issue.

Artificial benchmarks seldom have much to do with real world results. Most of the time, you'll never see an improvement as dramatic as the benchmarks make you believe.
Click to expand...

That lesson has been learned! :D
 
The Mercurian said:
Not comparing sequential models - there are a couple of iterations in between remember. Geekbench score for 2011 model is about 10,000, for 2013 model about 14,000. Given that my application is basically a pure CPU task I would have expected about a 40% improvement. I'm getting about 12%. Don't think thats worth 3k.
Click to expand...

I think your mistake is that you're taking benchmark results and expecting real work results to the the same. While your app maybe CPU intensive, there are other aspects that impact performance. How multithreaded is your app, can it take advantage of all that newer processors have to offer. The benchmarks of course do, most apps don't.

I generally expect to see a 10 to 12% increase in performance when I upgrade, and I genially try to upgrade every few years.

As for worth, it certainly sounds like you made up your mind, and while you ruminate on it further, you seem to be set that given the performance you're seeing, its not worth the price.

I agree with others about the GPU and peace of mind, but 3,000 is a lot of money and if you are not getting value out of it, then return it.
 
The Mercurian said:
Decision made. Return. Thanks folks for you opinions! I do listen yjchua95 - I'm merely cantankerous and prone to disagreement :D

Enjoy the upcoming ski season you lucky git :D
Click to expand...

Mercurial: subject to sudden or unpredictable changes of mood or mind (e.g. his mercurial temperament)

I just had to grab this off the dictionary and couldn't resist :D

Thanks, I hope I end up in one piece at the bottom of the ski slopes :D

PS it's bloody cold :/ I'd happily trade winter for a tropical weather like in Malaysia (where I used to work before moving down to Melbourne and closer to my homeland). But the ski slopes and having snowball fights are just too tempting to stay there :D

Oh and for your 2011 MBP, I recommend that you reapply the thermal paste to reduce the chances of Radeongate happening :)
 
yjchua95 said:
Mercurial: subject to sudden or unpredictable changes of mood or mind (e.g. his mercurial temperament)

I just had to grab this off the dictionary and couldn't resist :D
Click to expand...

Haha - this i actually very accurate :)

Oh and for your 2011 MBP, I recommend that you reapply the thermal paste to reduce the chances of Radeongate happening :)
Click to expand...

Ah yeah but not while it is still under AppelCare. I was going to sell it, but since I will stick with it now I will ride out AppleCare (another 11 months or so)

I would definitely take Queenstown over Malaysia!!:D
 
The Mercurian said:
Not comparing sequential models - there are a couple of iterations in between remember. Geekbench score for 2011 model is about 10,000, for 2013 model about 14,000. Given that my application is basically a pure CPU task I would have expected about a 40% improvement. I'm getting about 12%. Don't think thats worth 3k.

Lets put it this way - take 4 hours (240mins...ish) to do the base model - so can do 2 in a working day. If it really was 40% quicker that would be about 144 mins to do a model (2hrs 20mins) and I could do 3 in a day. But as it happens it works out about 215 mins ( 3hr 35min) on the new machine - which frankly is not worth 3k to me.
Click to expand...

Is the statistical model fully parallelized? I'm also trying to find out what you compared exactly.

32 bit single core:
late 2013 2.6 GHz: 3324
late 2013 2.0 GHz: 2812
late 2011 2.5 GHz: 2697
late 2011 2.2 GHz: 2298

32 bit multi core:
late 2013 2.6 GHz: 12605
late 2013 2.0 GHz: 10732
late 2011 2.5 GHz: 9629
late 2011 2.2 GHz: 8328

It depends a lot which models of which generation you compare, and the performance gains seem to be larger in multi-core applications. In any case I only see 40%+ gains when you compare low end 2011 models to high end 2013 models. Still I'm surprised that you only got 12% speed increase... I would have expected at least 20%. Which machines exactly are you comparing?

Of course if you were really bottle-necked by CPU power in your work, then you should invest in a few Mac Pros (or a remote workstation, if mobility is necessary). But I guess that was just an example.
 
thundersteele said:
Is the statistical model fully parallelized? I'm also trying to find out what you compared exactly.

32 bit single core:
late 2013 2.6 GHz: 3324
late 2013 2.0 GHz: 2812
late 2011 2.5 GHz: 2697
late 2011 2.2 GHz: 2298

32 bit multi core:
late 2013 2.6 GHz: 12605
late 2013 2.0 GHz: 10732
late 2011 2.5 GHz: 9629
late 2011 2.2 GHz: 8328

It depends a lot which models of which generation you compare, and the performance gains seem to be larger in multi-core applications. In any case I only see 40%+ gains when you compare low end 2011 models to high end 2013 models. Still I'm surprised that you only got 12% speed increase... I would have expected at least 20%. Which machines exactly are you comparing?

Of course if you were really bottle-necked by CPU power in your work, then you should invest in a few Mac Pros (or a remote workstation, if mobility is necessary). But I guess that was just an example.
Click to expand...

Its partly parallelised - in other words I've written R code to invoke multiple instances of another program - so it is as many cores as I want to have. But regardless of that each core once invoked should simply churn through the numbers since it is all loaded into RAM at the start and there is plenty of RAM so swapping and disk access does not come into it* (in fact it crashes if it maxes the RAM). I compared the 64 bit geekbench scores (since the various software I'm using are all 64 bit) for the following.

Geekbench Late 2011 15": 2.5 Ghz 16GB RAM
64bit single core: 2957
64bit multiple core: 10630

Geekbench Late 2013 15": 2.6 Ghz 16GB RAM
64bit single core: 3714 - 26% faster
64bit multiple core: 14431 - 36% faster

Actual increase - 12%.

Ok not 40% but 26 - 36% is still alot more than 12% faster.


Re: Mac pros - could barely afford this and I need portability so servers etc not an option. But it is ok I can struggle on with my old one until future faster chips come out.

*note - if disk speed did come into it then the new one would be even faster again since the new SSD has transfer rates about 4 times higher than the old one.
 
The Mercurian said:
So,

finally bought new 15" MBP after much deliberation. Judging by geekbench scores expected 40% improvement versus my old machine. Just run an exhaustive test running same statistical model on old machine (2011 15" MBP) versus new one. Very disappointed that the real life speed test was only about 12% quicker. Thinking to return. Main reason for expensive new machine was more speed and I didn't get the speed boost I expected.

Opinions ?
Click to expand...

What specific model you bought? haswell series? dGPU? clock speed? Memory?
 
I think you keep everything too simple... I mean, the CPU is about 30% faster than your 2011, which isn't that bad IMHO.

However, there's much more: retina display, faster RAM, much faster SSD, much lighter, slimmer, better battery life.. seems that's of no value to you?

If you only value CPU, then you are going to have a tough time.. I'm not too confident that the next intel CPU generation is much faster than the current one. They seem to be focusing on other things: power consumption and graphics performance.

I don't know how this situation is in the windows world... have you checked Windows workstation class machines?
 
PatrickVogeli said:
I think you keep everything too simple... I mean, the CPU is about 30% faster than your 2011, which isn't that bad IMHO.

However, there's much more: retina display, faster RAM, much faster SSD, much lighter, slimmer, better battery life.. seems that's of no value to you?
Click to expand...
Good point. The new one does have a much bigger faster SSD. But thats not what I stumped up cash for - I wouldn't stump up cash just for that - as good as these things are.

If you only value CPU, then you are going to have a tough time.. I'm not too confident that the next intel CPU generation is much faster than the current one. They seem to be focusing on other things: power consumption and graphics performance.
Click to expand...
Yeah but even 5% faster than current is a bigger jump compared to my 2011 one. Spend 3k now or wait a few months then spend 3k for one that is a bit faster as opportunity cost.

I don't know how this situation is in the windows world... have you checked Windows workstation class machines?
Click to expand...
They all suck - not all for the same reasons. But you can't get a windows machine that will overall match the MBP. Now usually around now people start jumping in suggesting machine x or machine y that happen to ahve one advantage over a MBP - but none of them will match it when all things considered. And none can match the disk speed. Any that come close to matching are hideous huge tinder-block like behemoths that cost as much or more than top end MBP
 
The Mercurian said:
Re: Mac pros - could barely afford this and I need portability so servers etc not an option. But it is ok I can struggle on with my old one until future faster chips come out.
Click to expand...

Thanks for the details. Hard to understand why you didn't get at least 20%-30% speedup. You could of course run geekbench on both machines - just to see how they perform, it won't help with your code of course.

Btw, a crazy idea: How about you keep both and let them both do the work. 100% more work done!
 
thundersteele said:
Thanks for the details. Hard to understand why you didn't get at least 20%-30% speedup. You could of course run geekbench on both machines - just to see how they perform, it won't help with your code of course.

Btw, a crazy idea: How about you keep both and let them both do the work. 100% more work done!
Click to expand...

Have geekbenched them both and they are in keeping with above scores.

Haha - crazy indeed - I can't afford that!
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back