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

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,254
7,899
Lincolnshire, UK
That's the question I've asked myself again and again when I've suspected I could just get that little bit more performance out of my machine.....
So, spying a NVIDIA Quadro FX4500 on ebay for a reasonable(ish) price, I decided to find out. My existing card was a NVIDIA Geforce 6600 which is neither bottom of the barrel nor exceptional - just a stock no frills card for a G5 Quad I guess but the Quadro is one of the best cards available for the late G5...so there should be some improvement maybe?
These are my findings with some standard tests and not so standard tests I've cobbled together myself:

GioFX OpenMark
6600 - 5484
4500 - 15769

Let1kWindowsBloom
(http://www.vgg.com/rob/WindowsBloom.html)
6600 - Total time to create and dispose 1000 windows: 5 seconds (5925683 microseconds)
4500 - Total time to create and dispose 1000 windows: 5 seconds (5703893 microseconds)

Photoshop Filter Benchmark
(http://ksimonian.com/Blog/2010/02/2...for-both-mac-pc-free-radial-blur-filter-test/)
6600 - 74.7 seconds
4500 - 74.6 seconds

Playing a 4K(3840 X 2160) h264.mp4 in full screen
(http://hwcdn.net/j9t9v3v5/cds/Coastguard_H264.mp4)
6600 - Plays smoothly in VLC, stutters with MPlayer
4500 - Plays smoothly in VLC, stutters with MPlayer

Playing a 2K trailer in Flash on Youtube
(
)
6600 - stutters too much to be considered playable
4500 - stutters too much to be considered playable

Video encoding benchmark
(https://forums.macrumors.com/threads/diy-benchmark.1862039/)
6600 - 01:59
4500 - 03:26

Running multiple 360P h264 videos smoothly with Quicktime, all on screen under Expose
6600 - 5
4500 - 7

All in all, very disappointing. HD video playback is exactly the same and if the benchmark is to be believed (and I ran it 4 times to check), video encoding now takes twice as long!
I'd imagine the Quadro will have benefits for gaming judging by the huge leap in the GioFX benchmark and be great for dual monitors - two areas not applicable to myself.

Has anyone had noticeable improvements after upgrading their graphics cards - not including moving from non-Core Image cards etc?
 
That's the question I've asked myself again and again when I've suspected I could just get that little bit more performance out of my machine.....
So, spying a NVIDIA Quadro FX4500 on ebay for a reasonable(ish) price, I decided to find out. My existing card was a NVIDIA Geforce 6600 which is neither bottom of the barrel nor exceptional - just a stock no frills card for a G5 Quad I guess but the Quadro is one of the best cards available for the late G5...so there should be some improvement maybe?
These are my findings with some standard tests and not so standard tests I've cobbled together myself:

GioFX OpenMark
6600 - 5484
4500 - 15769

Let1kWindowsBloom
(http://www.vgg.com/rob/WindowsBloom.html)
6600 - Total time to create and dispose 1000 windows: 5 seconds (5925683 microseconds)
4500 - Total time to create and dispose 1000 windows: 5 seconds (5703893 microseconds)

Photoshop Filter Benchmark
(http://ksimonian.com/Blog/2010/02/2...for-both-mac-pc-free-radial-blur-filter-test/)
6600 - 74.7 seconds
4500 - 74.6 seconds

Playing a 4K(3840 X 2160) h264.mp4 in full screen
(http://hwcdn.net/j9t9v3v5/cds/Coastguard_H264.mp4)
6600 - Plays smoothly in VLC, stutters with MPlayer
4500 - Plays smoothly in VLC, stutters with MPlayer

Playing a 2K trailer in Flash on Youtube
(
)
6600 - stutters too much to be considered playable
4500 - stutters too much to be considered playable

Video encoding benchmark
(https://forums.macrumors.com/threads/diy-benchmark.1862039/)
6600 - 01:59
4500 - 03:26

Running multiple 360P h264 videos smoothly with Quicktime, all on screen under Expose
6600 - 5
4500 - 7

All in all, very disappointing. HD video playback is exactly the same and if the benchmark is to be believed (and I ran it 4 times to check), video encoding now takes twice as long!
I'd imagine the Quadro will have benefits for gaming judging by the huge leap in the GioFX benchmark and be great for dual monitors - two areas not applicable to myself.

Has anyone had noticeable improvements after upgrading their graphics cards - not including moving from non-Core Image cards etc?
I played Halo last night with all the settings maxed out.

It's something that's entirely doable with the Radeon 9800 Pro I had, but nothing I pushed the card on for any length of time (no heatsink fan).

With the new card I got last night (NVIDIA GeForce 6800 GT) the hottest the card got was 190º. My understanding is that 276º for that card is shutdown temperature so I was well within spec. After leaving Halo it dropped back to 167º where it seems to idle at.

But graphically wise I really had no issue playing Halo last night. The graphics seemed to better than what the 9800 Pro served up, but I'm not an expert. I have noticed that my Cinema Display seems MUCH brighter now. I wonder if the signal from the 9800 was poor.

In any case, my new card seems to be performing much better than the old.

Anecdotal evidence with no hard numbers I know - sorry! Just my surface observations from the small amount of time I had with the new card last night.
 
  • Like
Reactions: lepidotós
Anecdotal evidence with no hard numbers I know - sorry! Just my surface observations from the small amount of time I had with the new card last night.
Yeah, these high end cards shine with games - I'd just hoped for some tangible improvement in the UI - I certainly didn't expect a drop in encoding performance! I need to check that out - can't believe the card can be responsible.
 
Changing the card will make very little difference in general use apart from a few specific cases that heavily leverage 3D performance such as games, 3D rendering.

Core Image cards perform similarly in the OS X GUI as 2D and basic OpenGL performance is very similar between different cards, any marginal performance increase most likely comes from more VRAM.

I've said this before and I'll say it again - you will get no improvement with video playback on PowerPC Macs. OS X didn't get hardware acceleration offload support for H.264 until OS X 10.6.3.
 
I've said this before and I'll say it again - you will get no improvement with video playback on PowerPC Macs. OS X didn't get hardware acceleration offload support for H.264 until OS X 10.6.3.
Yes, I remember you saying that and I fully understand the reasoning, however, I was anticipating the extra VRAM maybe helping the frame buffering a little - I have seen a difference in playback between a 64Mb and a 128Mb card before.
If anything, I wanted to satisfy my curiosity as to why people endlessly recommend upgrading the GPU.
 
Sorry, wasn't being arsey. Honest!

More VRAM might help, but I'd be amazed if it's more than a couple of %, same as for a system RAM upgrade. Video decode is CPU bound primarily.
 
Sorry, wasn't being arsey. Honest!

More VRAM might help, but I'd be amazed if it's more than a couple of %, same as for a system RAM upgrade. Video decode is CPU bound primarily.
No worries :)
At the risk of sounding incredibly naive, is it not possible for specifically designed software to access the GPU and use it as a virtual processor?
As a side note, when I tried XBMC, the card fan spun up as it did running the GioFX benchmark, as if XBMC was passing processing over to the GPU - however, as XBMC crashes on my Quad regardless of what card I have installed I didn't investigate further.
 
I've said this before and I'll say it again - you will get no improvement with video playback on PowerPC Macs. OS X didn't get hardware acceleration offload support for H.264 until OS X 10.6.3.

Sad but true :(.

RE: Video playback, VLC's performance on my dual 1.8 G5 is not great but it's a lot better than QuickTime. Wish there was a really good video player for the PPC.
 
RE: Video playback, VLC's performance on my dual 1.8 G5 is not great but it's a lot better than QuickTime. Wish there was a really good video player for the PPC.
I've no complaints with VLC - you have to tickle the preferences a little to get maximum efficiency though and on the Quad it does use all 4 processors. Apart from my Quad where it crashes, I've found XBMC to offer best playback in most circumstances - it does seem to be very fussy on hardware though.
 
No worries :)
At the risk of sounding incredibly naive, is it not possible for specifically designed software to access the GPU and use it as a virtual processor?
As a side note, when I tried XBMC, the card fan spun up as it did running the GioFX benchmark, as if XBMC was passing processing over to the GPU - however, as XBMC crashes on my Quad regardless of what card I have installed I didn't investigate further.

To dumb it down slightly something in XBMC (probably the UI) is triggering 3D mode on the GPU via OpenGL so it'll spin up the fans automatically.

Not tried XBMC on my G5, will have to give it a whirl later. It seems fine with Quicktime MP4 in HD res, but H.264 is a bit too much of a stretch.
 
I've no complaints with VLC - you have to tickle the preferences a little to get maximum efficiency though and on the Quad it does use all 4 processors. Apart from my Quad where it crashes, I've found XBMC to offer best playback in most circumstances - it does seem to be very fussy on hardware though.

Since no current Kodi plugins work w/the PPC version of XBMC I hadn't loaded it. But I just gave it a spin and it seems to stream videos from my local server smoother than VLC. Thanks for the idea.
 
Since no current Kodi plugins work w/the PPC version of XBMC I hadn't loaded it. But I just gave it a spin and it seems to stream videos from my local server smoother than VLC. Thanks for the idea.
You might want to go into settings and change the default theme to Quartz - much cleaner and less CPU intensive.
 
You might want to go into settings and change the default theme to Quartz - much cleaner and less CPU intensive.

Thanks for the tip. Quartz seems much lighter and works great with a keyboard. However on my G5 it caused videos to randomly restart, had odd colour lines on the screen and always crashed XBMC. Maybe I'll jump into the Danger Zone and test other skins after I finish watching Top Gun :D
 
Thanks for the tip. Quartz seems much lighter and works great with a keyboard. However on my G5 it caused videos to randomly restart, had odd colour lines on the screen and always crashed XBMC. Maybe I'll jump into the Danger Zone and test other skins after I finish watching Top Gun :D
Ah, thanks for that info - that's why it keeps crashing on my Quad, installing Quartz is the first thing I do on every system - works ok without it in this case.
 
Yes, I remember you saying that and I fully understand the reasoning, however, I was anticipating the extra VRAM maybe helping the frame buffering a little - I have seen a difference in playback between a 64Mb and a 128Mb card before.
If anything, I wanted to satisfy my curiosity as to why people endlessly recommend upgrading the GPU.
The gfx card is bypassed entirely for video playback on VLC/XBMC/QT etc. It's all decoded using cpu & altivec ONLY.
 
Are there any programs that take advantage of the gpu for video playback?

Yes for applications making use of the Quicktime frameworks for MPEG/MPEG2. As previously stated - Apple added the same for H.264 derived codecs in 2010 with OS X 10.6.3 (hence Intel only).

Most GPUs of the PowerPC era only support MPEG/MPEG2 assist. From memory there's limited VC-1 or H.264 assist on the GeForce 6600 only, so it was bleeding edge stuff for an architecture that was in it's death throes. It didn't hit the mainstream until the later GeForce 8000 series cards, and the Radeon HD 2000 series.
 
Last edited:
val1984 a light bulb went off while reading your response to Dronecatcher. The CPU speed was set at SLOW, after bumping it to FAST videos play great and CPU usage is less. Thanks for the reminder....

MagicBoy your answer is not what I hoped for, but it is what I expected. Thanks.
 
To me, it looks like it's either one of these possibilities:
- you disabled half the cores with CHUD
- performance mode was set to reduced in one case and maximum in the other
- some other process was using half the cores

Negative on all three counts.
I have MenuMeters running so I can monitor the CPUs - they were all running and no other process was active apart from the mentioned MenuMeters - and that uses less than 0.4%
I'm trying to pretend it doesn't bother me...but I know at some point there'll be a fresh OS install to satisfy my OCD :)
 
Try playing Prey on the 6600 and then on the 4500, you'll definitely notice a difference.
 
  • Like
Reactions: tevion5
Negative on all three counts.

Are you sure the video card was the only difference? It should have nothing to do with the encode, so that's a bit of a mystery. When I was trying that bench, I could get vastly different results from forgetting to change all the settings the same way each time.
 
Are you sure the video card was the only difference? It should have nothing to do with the encode, so that's a bit of a mystery. When I was trying that bench, I could get vastly different results from forgetting to change all the settings the same way each time.
No, the settings are always consistent. I doubt it's the card - either somethings gone amiss in the OS or maybe the codecs used have been changed/updated by another app.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.