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

4K = bad performance. what should I upgrade?

D

damian0815

macrumors newbie
Original poster
Oct 22, 2014
8
0
I recently upgraded my GPU to an Nvidia GTX660 (edit: was 640) to support 4k@60Hz, and it looks and feels absolutely amazing - but I'm taking a severe performance hit on text rendering, which is a problem when running Xcode full screen at 4k, which is what I do most of my day.

My CPU is a Core i7-870 running at 2.93GHz, which although it's nearly 5 years old, is still a solid CPU as far as I understand it. I'm running off an SSD, I have maxed out the RAM with the fastest DIMMs that the system will support (1600MHz), but it's still not fast enough.

So I'm looking at buying a new system, but on clock speed alone, the new iMac Core i5 with Retina display doesn't actually look that much faster than what I have (3.5GHz vs 2.93Ghz).

Has anyone gone a similar route? I mean, 2010-era Mac upgrade to 2014 era with similar clock speed - do you really notice a performance boost, even if the clock speed of the CPUs stays roughly the same?


Edit: Derp! I had the card in a PCIex4 slot. Put in a PCIex16 slot, everything is fine.
 
Last edited:
CPU clock speed alone is a terrible performance indicator. Each CPU generation gets faster at similar clock speeds, you'll notice that the speed itself hasn't really gone up all that much but there's probably at least a 2x overall performance increase in the last 5 years. I believe AnandTech or some other website did a comprehensive study of CPU performance, you might want to go take a look at that. Right now, I'm running a Core i7 4770K at 3.5GHz, and noticed a massive improvement over my old 2010 Mac Pro with a 3.2GHz Nehalem processor.

Having said all of that, your GPU will be the limiting factor at 4K resolution, especially if you're only using a GTX 640 (the low-end GK107). You have the luxury of being able to drop in a PCIe graphics card, and should look at investing in a modern mid-range to high-end GPU to drive such a large display.
 
damian0815 said:
So I'm looking at buying a new system, but on clock speed alone, the new iMac Core i5 with Retina display doesn't actually look that much faster than what I have (3.5GHz vs 2.93Ghz).
Click to expand...

You cannot compare the processors from different architectures using clock speed. It won't tell you anything.

I hate myself for doing this, but Geekbench is the easiest to way to show you the difference

GB3 (32bit free test): (single core - multi core)

i7 870 ~2200 - ~8000

i5 4690k ~4100 ~11500

The i5 iMac will smoke your CPU and if your Xcode projects are complex, it will speed up compilation times a lot.
 
Asgorath said:
Right now, I'm running a Core i7 4770K at 3.5GHz, and noticed a massive improvement over my old 2010 Mac Pro with a 3.2GHz Nehalem processor.
Click to expand...

Oh, that's great to hear - thanks!

Asgorath said:
Having said all of that, your GPU will be the limiting factor at 4K resolution, especially if you're only using a GTX 640 (the low-end GK107).
Click to expand...

Actually I mis-typed - it's a GTX660 that I have, and I just bought and am just out of the 2-week right-of-return period. TBH I struggled to find a video card that would support 4K@60Hz on OSX - couldn't find any information about if any of the 700-series were supported.

Anyway, When I profile Xcode itself using Instruments, the performance hit is dominated by the time to assemble and draw the NSTextView driving the code editing window. For comparison I can make Firefox or Chrome full-screen and I don't have any of the same problems I do when editing code, which leads me to believe I'm CPU bound...
 
damian0815 said:
Actually I mis-typed - it's a GTX660 that I have, and I just bought and am just out of the 2-week right-of-return period. TBH I struggled to find a video card that would support 4K@60Hz on OSX - couldn't find any information about if any of the 700-series were supported.

Anyway, When I profile Xcode itself using Instruments, the performance hit is dominated by the time to assemble and draw the NSTextView driving the code editing window. For comparison I can make Firefox or Chrome full-screen and I don't have any of the same problems I do when editing code, which leads me to believe I'm CPU bound...
Click to expand...

https://forums.macrumors.com/threads/1440150/

I have a hard time believing you're CPU bound at 4K, but YMMV.
 
Hmmm...

The 990x these days is going for 400ish and delivers:

Geekbench3 (64-Bit test): (single core - multi core)
2905 - ~16640

Just how many Pro apps use a single core these days?

On the cMP, one can upgrade to a PCIe x4 ssd. Apple's latest chips are the fastest on the market, at 1.4GB/s. Slower versions pulling 1.1GBs and can go for $600ish. Internal PCIe SSD's on the riMac are limited to an x2 interface and 800MB/s. Unless you pay for the high-priced Thunderbolt options...

The $300 970GTX already squishes the 290X. The $450 980X takes the crown. The 290x can also be picked up for $300ish.

A fast video card and PCIe ssd will take most workflows to the stratosphere for a much cheaper investment.


theSeb said:
You cannot compare the processors from different architectures using clock speed. It won't tell you anything.

I hate myself for doing this, but Geekbench is the easiest to way to show you the difference

GB3 (32bit free test): (single core - multi core)

i7 870 ~2200 - ~8000

i5 4690k ~4100 ~11500

The i5 iMac will smoke your CPU and if your Xcode projects are complex, it will speed up compilation times a lot.
Click to expand...
 
Last edited:
handheldgames said:
Hmmm...

The 990x these days is going for 400ish and delivers:

Geekbench3 (64-Bit test): (single core - multi core)
2905 - ~16640

Just how many Pro apps use a single core these days?

On the cMP, one can upgrade to a PCIe x4 ssd. Apple's latest chips are the fastest on the market, at 1.4GB/s. Slower versions pulling 1.1GBs and can go for $600ish. Internal PCIe SSD's on the riMac are limited to an x2 interface and 800MB/s. Unless you pay for the high-priced Thunderbolt options...

The $300 970GTX already squishes the 290X. The $450 980X takes the crown. The 290x can also be picked up for $300ish.

A fast video card and PCIe ssd will take most workflows to the stratosphere for a much cheaper investment.
Click to expand...

Agreed. This would be a better path to follow
 
Hello,

are all Maxwell equipped nVidia cards without any further tweaking 4K enabled in 10.10.?
I am looking into a GTX 970 myself and would love to know.
Thank you so much!
 
More confusion!

I just tried out my friend's Macbook Pro 2014 with discrete GPU (GTX 750M). After we completely disabled the internal Iris Pro, he was able to drive both his internal 3880x2160 (Retina) screen and my 3840x2160 simultaneously from the GTX 750M, running Xcode fullscreen without any lag at all.

Since the GTX660 is supposed to be more performant than the GTX750M (http://www.tomshardware.com/reviews/gaming-graphics-card-review,3107-7.html), this would point to some other problem with my system compared to his. The two obvious ones are the PCIe connector (mine is PCIe x4, his is PCIe x8) and the CPU (mine is Nehalem, his is Haswell).

----------
ThoreEliasson said:
Hello,

are all Maxwell equipped nVidia cards without any further tweaking 4K enabled in 10.10.?
I am looking into a GTX 970 myself and would love to know.
Thank you so much!
Click to expand...

It would seem so. Check the TonyMacx86 forums, the Hackintosh folks are some of the best ways to find out about GPU compatibility on Mac Pros -- what works on a Hackintosh ought to work on a Mac Pro, running the same OS version at least.
 
Ok, so if you put a PCIex16 GPU in a PCIex4 slot you get ... a quarter of the performance.

Derp.

Turns out a GTX660 is plenty usable, if it's in the right PCIe slot.
 
Last edited:
Final word on this: My friend let me try my GTX660 in his Core i5-4670 based Hackintosh, with my 4K monitor.

Smooth as butter.

So I'm definitely CPU bound.
 
It's probably more of a driver issue. a GTX 660 is no slouch and can't be compared to a 750M. The 660 is more powerful by a factor of 3.

An 870 and a 4670 is around 35% of each other in performance. I doubt that is causing an issue.

Having a 660 in a x4 slot won't affect 4k performance unless you're playing a game. No discussion there as I run a 780ti at 1080p over thunderbolt 1 at 80-90% of the performance.

Who knows what else it can be lol maybe that x4 slot was only running x1 mode? lol
 
damian0815 said:
Final word on this: My friend let me try my GTX660 in his Core i5-4670 based Hackintosh, with my 4K monitor.

Smooth as butter.

So I'm definitely CPU bound.
Click to expand...

No..

You upgraded graphics and the graphics is smoother the processor has nothing to do with it in this instance.
 
No,

GTX660 on my i7 870 (in the PCIe-x16 slot) = acceptable (but not great) performance
GTX660 on my friend's i5 4670 = amazing performance

changing variable = CPU :. (probably) CPU caused the difference.
 
damian0815 said:
No,

GTX660 on my i7 870 (in the PCIe-x16 slot) = acceptable (but not great) performance
GTX660 on my friend's i5 4670 = amazing performance

changing variable = CPU :. (probably) CPU caused the difference.
Click to expand...

FWIW, Xcode's on screen layout happens on the CPU, not the GPU.

In OS X only entire windows hit the GPU. For the most part, the actual drawing of everything within a window is done on CPU. There are a few exceptions for apps that use things like OpenGL, but Xcode does not.
 
Oh, that's very useful to know.

So, right - if I'm getting drawing bottlenecks it's most likely CPU related. That pretty much confirms what the evidence says.
 
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