5k Mac buying options - graphics vs processor

[marclondon]

I'm finally going to buy an iMac after being a long time Mac user - I think the 5k machines are just too good to miss now, plus I get a big discount.

I''m not going to max one out but will probably opt either for the i7 upgrade on the mid-range model or get the top model with either i7 or the 395x graphics processor (or just stick with the base machine).

This will be a family machine and the kids will want to run Windows 10 and Steam games.

Question - is the i7 or better graphics the best long term option?

 

[Smashnet]

In your family machine context the I7 upgrade will probably not yield any recognizable advantage. You will surely be happier with the better GPU if you're up to playing games.

 

[mroffbeat]

Kids touching my iMac? I'd go for a locked door

Just ordered mine and went full metal jacket. Today I went to the store to gaze at my soon to be delivered 5k goodness. I think i7 is a great speed boost on long term. With your savings buy your kids a ps4

 

[Strider64]

I've not much of gamer now a days, but I've been gaming since 1980. I say a better cpu is preferable over a better gpu. Just my opinion. A computer is more than just a gaming machine and I think games are a great way of introducing children to the world of computers. Just my .02 cents.

 

[makrom]

The 5k iMacs aren't ideal for gaming, and whether you got an M395 with or without X doesn't do much to change that. Benchmarks show a 10% difference in GPU demanding situations. The difference grows a lot when memory becomes an issue, but this is usually the case when playing at resolutions way above what either GPU is capable to handle decently.
If gaming is the main aspect of optimisation, the GPU might still be the better choice, 10% is still better than 0%. But I would still go with the CPU. Once you start using your iMac for CPU demanding tasks like rendering in FCPX for example, it can easily give you a 50% advantage. And anyone who ever used FCPX knows that waiting for rendering to be done makes up a lot of the time of using it.

 

[Catholic13]

The 5k iMacs aren't ideal for gaming, and whether you got an M395 with or without X doesn't do much to change that. Benchmarks show a 10% difference in GPU demanding situations. The difference grows a lot when memory becomes an issue, but this is usually the case when playing at resolutions way above what either GPU is capable to handle decently.
If gaming is the main aspect of optimisation, the GPU might still be the better choice, 10% is still better than 0%. But I would still go with the CPU. Once you start using your iMac for CPU demanding tasks like rendering in FCPX for example, it can easily give you a 50% advantage. And anyone who ever used FCPX knows that waiting for rendering to be done makes up a lot of the time of using it.

10% can actually makes a HUGE difference in gaming.
It can turn an otherwise unplayable...playable.

As for the steam games...indie or "AAA?"
I doubt it'll really matter if it's just for indie games as those don't tend to be graphically intensive... and while you can *technically* upgrade your CPU- it's not something I would recommend.https://www.ifixit.com/Guide/iMac+Intel+27-Inch+Retina+5K+Display+CPU+Replacement/30515
I chose the m395x over i7 for mine...I can't say much about it though because I'm not going to receive it until the middle-end of next week......................I don't game much, but when I do, I hate having to set my graphics to low.......

 

[makrom]

10% can actually makes a HUGE difference in gaming.
It can turn an otherwise unplayable...playable.

No, it can't. It will turn 10FPS into 11FPS, 24FPS into 26FPS and so on, how is any of that a huge difference?

 

[T'hain Esh Kelch]

Since you mention gaming as a primary reason, I would definitely go for the GPU upgrade. The i5 is more than sufficient to keep the card saturated, and the GPU will be the bottleneck in that system.

 

[maflynn]

In this day and age, a powerful GPU is going to be better for gaming then a CPU upgrade, and in fact. I'm not sure you'll see a big enough jump in performance by going to the i7 - at least in terms of game playing.

 

[dyn]

CPUs have reached the limits of physics and won't be coming noticeably faster unless they start using different materials. All of the improvements are geared towards energy efficiency and the integrated GPU. There is very little need to upgrade a CPU. GPUs are a completely different world. There still are big improvements in performance and in energy efficiency as well.

 

[jerwin]

http://barefeats.com/imac5k13.html

but beware, the m395x he tested came with an i7; the m395 with an i5. It's not entirely clear which component is most responsible for the marginal improvements.

The i7 is clocked higher, has 4 additionalvirtual cores, and more cache. The m395x has 12.5 percent more cores, and twice the vram.

 

[throAU]

CPUs have reached the limits of physics and won't be coming noticeably faster unless they start using different materials. All of the improvements are geared towards energy efficiency and the integrated GPU. There is very little need to upgrade a CPU. GPUs are a completely different world. There still are big improvements in performance and in energy efficiency as well.

Not really. Clocks are materials limited, but new instructions or increased threading (i.e., getting an i7 instead of an i5 which has been crippled by intel) can speed things up significantly. It really depends on what you're doing. e.g., the hardware AES crypto instructions (introduced a few generations ago) speed up AES when used by up to 30x over previous CPUs. There hasn't been much new like that for a little while, but that doesn't mean there won't be in future.

Me? Unless you're doing a lot of video processing or other CPU demanding stuff i'd skip both upgrades and put the money aside to upgrade to a new one sooner. Spending hundreds of dollars to gain 10% on graphics today is just far less bang for buck than putting the money aside to trade up to a machine with newer generation GPU that's 50% faster sooner - you'll lose less money turning over the old machine in 2-3 years rather than 5+ too.

 

[maflynn]

CPUs have reached the limits of physics and won't be coming noticeably faster unless they start using different materials

People have been saying that for 20 year and at some point that will be the case but I think there's still plenty of innovation and improvements left in what they can do

 

[Santabean2000]

Not really. Clocks are materials limited, but new instructions or increased threading (i.e., getting an i7 instead of an i5 which has been crippled by intel) can speed things up significantly. It really depends on what you're doing. e.g., the hardware AES crypto instructions (introduced a few generations ago) speed up AES when used by up to 30x over previous CPUs. There hasn't been much new like that for a little while, but that doesn't mean there won't be in future.
Me? Unless you're doing a lot of video processing or other CPU demanding stuff i'd skip both upgrades and put the money aside to upgrade to a new one sooner. Spending hundreds of dollars to gain 10% on graphics today is just far less bang for buck than putting the money aside to trade up to a machine with newer generation GPU that's 50% faster sooner - you'll lose less money turning over the old machine in 2-3 years rather than 5+ too.

This. And in this day in age where we have a lot of subscription based apps and the app store, it really is no chore to migrate to a new machine.

I'd also skip the upgrades, but rather add the leftover money towards AppleCare; sell after 2 years for minimal depreciation, (easy sell with one year off full cover left). This way, you're always under warranty, getting the best every other year. Will see a lot more for your $$.

 

[DarthVader!]

I'd also skip the upgrades, but rather add the leftover money towards AppleCare;

That's not a bad a idea, I guess I backed into that way of doing things, by getting a stock iMac and including AppleCare. I do wonder though if I should have opted for the SSD, but that upgrade did not make sense. I would have to go all in on the 1TB SSD to fit my data and that upgrade was well out of reach.

 

[jerwin]

there's no question that a m395 is superior to a m390. And if you go with a fusion drive, the 2tb model has 128 gigs of flash, rather than the paltry 24 gigs of the 1tb fusion drive.
do note that bootcamp puts the windows 10 partion entirely onto the spinning disk. Allegedly, you can move the bootcamp partion onto an external ssd, but it's not a trivial operation. WinClone (non basic) can handle this, though.

as for the 4gb of the m395x, gaming at 4k and 5k can start to take advantage of that. The synthetic benchmark 3dfirestrike ultra requires 3gb, for instance. Of course, it's likely to run slowly on anything less than a desktop card, but such is life.
The directx12 game Ashes of the Singularity is also supposed to take advantage of veam beyond 2gb. A taste of games to come?

With recent games though, such as Witcher 3, Shadows of Mordor, Star Wars Battlefront , 4gb vram and especially 8gb vram don't provide a significant performance improvent, except on settings which willl overwhelm a single desktop card, and certainly a laptop card such as the m395x.

http://www.techspot.com/review/1114-vram-comparison-test/page5.html

 

[briloronmacrumo]

People have been saying that for 20 year and at some point that will be the case but I think there's still plenty of innovation and improvements left in what they can do

Not with silicon. Articles written ( mostly in the Wall Street Journal Tech section in 2015 - don't have the links handy but I could get them ) describe the inability of silicon to keep up with Moore's law. This is happening now. It is true that newer materials provide some options but they aren't production yet.

 

[satchmo]

This. And in this day in age where we have a lot of subscription based apps and the app store, it really is no chore to migrate to a new machine.

I'd also skip the upgrades, but rather add the leftover money towards AppleCare; sell after 2 years for minimal depreciation, (easy sell with one year off full cover left). This way, you're always under warranty, getting the best every other year. Will see a lot more for your $$.

Wow, lots of different opinions in this thread.

But do people really upgrade that often? I keep my Macs for 6-10 years. I buy it, enjoy it, and focus on making money with it in my business. For me, flipping it after only 2 years in seems rather quick, but I get it.

I suppose if you have the time or inclination to re-install apps and putting the Mac on classifieds and having strangers come into your home, it's an okay option.

 

[grayskyz]

Wow, lots of different opinions in this thread.

But do people really upgrade that often? I keep my Macs for 6-10 years. I buy it, enjoy it, and focus on making money with it in my business. For me, flipping it after only 2 years in seems rather quick, but I get it.

I suppose if you have the time or inclination to re-install apps and putting the Mac on classifieds and having strangers come into your home, it's an okay option.

I used to upgrade that often and I fell behind a bit on my last iMac which resulted in a loss of about 400-500 bucks. If you're using your rig for gaming and business, you should be wiping off your data and starting a fresh install and apps less than two years apart anyway!!! Keep everything running right and clears away old cache files and prefs you don't use anymore.

I'm okay with stranger danger...haha.

 

[throAU]

Not with silicon. Articles written ( mostly in the Wall Street Journal Tech section in 2015 - don't have the links handy but I could get them ) describe the inability of silicon to keep up with Moore's law. This is happening now. It is true that newer materials provide some options but they aren't production yet.

Process changes are becoming more difficult, sure. But progress will move on. Guaranteed. You'll see bigger CPU dies with more cores on them, and more specialised instruction sets. more intelligence built into power management and clock up/down based on thread count vs. clock rate.

Moore's law is not the be all end end all of processing power, it's just a statement based on transistor density and is in fact nothing to do with performance. Transistor density does generally correlate with improved performance but is not the only way to improve.

And there's no guarantee we'll continue on silicon anyway. Progress will move on.

 

[dyn]

Not really. Clocks are materials limited, but new instructions or increased threading (i.e., getting an i7 instead of an i5 which has been crippled by intel) can speed things up significantly.

I'd say you are over-exaggerating the definition of the word "significantly". The problem with current CPUs and their materials is about what happens with electrons, transistors and so on. You can only go so far. Other materials are better and can give somewhat of a significant difference but that doesn't necessarily mean a higher clocked CPU frequency. If you want to integrate other components on the CPU that requires room which means you need to shrink transistors again. Whatever you do, it is not easy to escape that reality.

Btw, increased threading is not going to make things faster as there are two kinds of workloads: sequential and parallel. Sequential is not going to benefit from increased threading. Increased threading also demands more from the developers. They need to change their apps in such a way it can use that increased threading. Sadly this is rather complex to do and as we've seen now from years of experience with multithreading there aren't that many apps using this properly and efficiently.

People have been saying that for 20 year and at some point that will be the case but I think there's still plenty of innovation and improvements left in what they can do

I think you misunderstood them. What they said is that this is inevitably going to happen but they didn't know when. Moore's law has been changed a couple of times because of that (they went from 12 months to 24 months some 30 years ago already!). It has been motivated by statistics as well as technology at that time. The fact that more and more factories have difficulty with die shrinks says enough (e.g. 10nm seems so difficult that Intel had to postpone it) and if you ask any researcher in nanotech they'll explain that silicon isn't very suited for that (mind you, CPUs are moving towards nanotech). As said above, if you want to do something else you quickly run into that issue. You are more looking into a complete redesign of the CPU.

I'd also skip the upgrades, but rather add the leftover money towards AppleCare

As an EU citizen I'd rather spend that money on the CPU upgrade as the AppleCare is completely useless here due to regulations. As a Dutch citizen I even have a longer warranty. On a machine with a 5K display the GPU is going to be its biggest bottleneck, hence why I'd rather upgrade the GPU than anything else.

But do people really upgrade that often? I keep my Macs for 6-10 years. I buy it, enjoy it, and focus on making money with it in my business. For me, flipping it after only 2 years in seems rather quick, but I get it.

There are some enthusiasts here that do this and it's about the only thing you can do with a Mac if you want to upgrade. Macs have never been upgradable, that's an exclusive feature of PCs. Some of us upgrade because we need to (the newer machine is more performant, the old one is hanging by a string) or because we want to switch from a desktop to mobile setup (or vice versa) and some of use upgrade because they want to have the newest product.

And there's no guarantee we'll continue on silicon anyway. Progress will move on.

If we want progress we'll have to move towards something else than silicon. Silicon isn't very well suited for nanotech. Nanotech is still very new to us.

 

[throAU]

Btw, increased threading is not going to make things faster as there are two kinds of workloads: sequential and parallel. Sequential is not going to benefit from increased threading. Increased threading also demands more from the developers. They need to change their apps in such a way it can use that increased threading. Sadly this is rather complex to do and as we've seen now from years of experience with multithreading there aren't that many apps using this properly and efficiently.

Well aware of that. However much of the hard core number crunching type stuff people are waiting on is what could be classed as "embarrassingly parallel". 3d rendering, video processing, transcoding, etc. are all in this category and scale almost linearly with thread count.

This is why Apple put so much work into grand central dispatch, and why this is important. Its why GPUs are being used so much for things lately - not because they have magical powers that a CPU does not (or rather, can not have in the future), but because they are themselves massively multi-threaded and good for this sort of workload. CPUs will become more this way, because unfortunately we're not going to be running much faster than 4 Ghz for the foreseeable future. The only way to speed up is special purpose hardware instructions and increasing thread count.