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.
580X Vs Vega 48 in the real world?
- Thread starter Loa
- Start date
- Sort by reaction score
Jerwin (and anyone else interested), see the solution here (and please confine replies to that thread).
Wow, performance gain seems to depend drastically on the software used. Thanks for posting this, very interesting! (Especially since I have ordered the Vega 48.)
What I find strange is that the gain is only 9% using Metal, while Vega cards have been around in the iMac Pro line for a while. I’d have expected a higher gain, better optimization of Metal for the newer architecture.
Of course, for my intended use, it’d be more interesting to see comparisons of how much gain you get e.g. in export times in photo editing software. I hope C1 makes use of the additional power.
And it would be interesting to see if tasks that heavily rely on CPU+GPU really benefit from the Vega allegedly running cooler than the 580X.
What I find strange is that the gain is only 9% using Metal, while Vega cards have been around in the iMac Pro line for a while. I’d have expected a higher gain, better optimization of Metal for the newer architecture.
Of course, for my intended use, it’d be more interesting to see comparisons of how much gain you get e.g. in export times in photo editing software. I hope C1 makes use of the additional power.
And it would be interesting to see if tasks that heavily rely on CPU+GPU really benefit from the Vega allegedly running cooler than the 580X.
Of course, for my intended use, it’d be more interesting to see comparisons of how much gain you get e.g. in export times in photo editing software. I hope C1 makes use of the additional power.
And it would be interesting to see if tasks that heavily rely on CPU+GPU really benefit from the Vega allegedly running cooler than the 580X.
Same here. Here's hoping that Barefeats will provide such tests.
I should get my iMac this week, once I've set it up I'll try this test that was referenced here some time ago. Hopefully someone with the i9/580X config could do the same, to get some insight. Maybe we could even do the test with different editing software.
I edited a photo job last week, the export of 200 24MP RAWs for customer preview (with Auto Adjust and resizing to 1600px) took my 2011 MBP 15 minutes... really looking forward to that i9/Vega now.
I did the CaptureOne test linked above on CO Pro 12.0.3 - 45 seconds flat for the export test with or without GPU acceleration enabled.
sRGB was the Color Space profile, didn't list sRGB IEC61966-2.1 as an option.
i9 w/ 580X
sRGB was the Color Space profile, didn't list sRGB IEC61966-2.1 as an option.
i9 w/ 580X
Cool, thx for doing this-now we can compare when I get my iMac. My 2011 MBP (i7@2.2
) needs nearly 5 mins for this. Strangely, GPU enabling also had nearly no effect. When I export pictures that have been edited, switching off GPU acceleration makes the export much faster. Must be the old Radeon 6750 in that machine...BTW, the sRGB profile you can choose from the menu acutally is the IEC profile.
I have the C1 for Sony v 11.3.2 and get this:
Code:
iMac 2019 (10.14.4 (18E2034) Mojave)
C1 Pro (for Sony) version - 11.3.2
RAM 8GB
CPU - Intel Core i9 9900K (8-core, 16-threads, 3.60 GHz)
GPU - Radeon Pro Vega 48 8 GB
JPEG
CPU+GPU - 0:58,45 (58s)
CPU only - 0:58,90 (59s)
The GPU speed up factor is about 1.02x.
Last edited:
OK, from PC land... gaming performance (as this was mentioned as a use case above).
I have an RX480 (Which is essentially a slightly lower clocked 580) and have Vega 64 (2 of them).
Vega 64 is easily 2x faster in real world rendering performance than the 480 and can handle higher resolutions. In 4k gaming it is about 3-4x the frame rate of the RX480.
Given that the Vega 56 and 64 are pretty close (for gaming) at the same clocks, Vega 48 should be at least 3/4 the performance of Vega 64 at the same/similar clocks. i.e., somewhere near 1.5-2x RX480/580?
Vega also clocks higher than the RX480/580 and has higher bandwidth memory (HBM2).
So i'd suggest that a reasonable expectation would be 1.5-2x RX580 performance from Vega 48 in games that are GPU limited.
Whether apple's drivers are there yet is another thing entirely - but in raw performance Vega 48 will smoke the RX580 and that gap will get wider the higher resolution you go and as newer drivers/software take advantage of Vega only features like rapid packed math (which can double or quadruple throughput on smaller-than-32 bit integers - which Polaris chips can't do) and high bandwidth memory.
I have an RX480 (Which is essentially a slightly lower clocked 580) and have Vega 64 (2 of them).
Vega 64 is easily 2x faster in real world rendering performance than the 480 and can handle higher resolutions. In 4k gaming it is about 3-4x the frame rate of the RX480.
Given that the Vega 56 and 64 are pretty close (for gaming) at the same clocks, Vega 48 should be at least 3/4 the performance of Vega 64 at the same/similar clocks. i.e., somewhere near 1.5-2x RX480/580?
Vega also clocks higher than the RX480/580 and has higher bandwidth memory (HBM2).
So i'd suggest that a reasonable expectation would be 1.5-2x RX580 performance from Vega 48 in games that are GPU limited.
Whether apple's drivers are there yet is another thing entirely - but in raw performance Vega 48 will smoke the RX580 and that gap will get wider the higher resolution you go and as newer drivers/software take advantage of Vega only features like rapid packed math (which can double or quadruple throughput on smaller-than-32 bit integers - which Polaris chips can't do) and high bandwidth memory.
It could. My 32GB is still on its way from amzn.
Though, I figured, my disabling of Turbo Boost might be the real cause.
I will run another test tonight.
I have a iMac 2019 with i9, 40GB ram, 512GB SSD and 580X.
Just tell me what tests i need to do to make it comparable to yours with vega 48
Just tell me what tests i need to do to make it comparable to yours with vega 48
I have a iMac 2019 with i9, 40GB ram, 512GB SSD and 580X.
Just tell me what tests i need to do to make it comparable to yours with vega 48
The one I linked in this post above, if you use Capture One.
FWIW I'm new to the Apple universe and just ordered (an hour ago) my first machine, 27" i9/Vega/512SSD. I went back and forth on the Vega and 580X, and had read this (among many other threads) about the differences. The $$ isn't a small amount between them.
However I pushed my budget a little to get the Vega for a few reasons: I'm hoping to use this machine for Logic, general computing, and light gaming for 7 years; and half-floats MIGHT become important in Logic virtual instruments during those 7 years of which the Vega is MUCH better for. I already have a gaming PC with a 980 TI I'll continue to update for heavy gaming (and no longer general computing).
I know I was very conflicted on which to buy, hopefully my decision/reasons helps someone else.
However I pushed my budget a little to get the Vega for a few reasons: I'm hoping to use this machine for Logic, general computing, and light gaming for 7 years; and half-floats MIGHT become important in Logic virtual instruments during those 7 years of which the Vega is MUCH better for. I already have a gaming PC with a 980 TI I'll continue to update for heavy gaming (and no longer general computing).
I know I was very conflicted on which to buy, hopefully my decision/reasons helps someone else.
So here are my results for the C1 export test:
C1 12.0.3
i9/Vega 48/40GB RAM/1 TB SSD
CPU only: 1:15 minutes
GPU enabled: 55 seconds
I tried to press the "process" button and the "start" button on my stopwatch at the same time, might be +/- 1 second error. So I notice a difference with or without GPU.
C1 12.0.3
i9/Vega 48/40GB RAM/1 TB SSD
CPU only: 1:15 minutes
GPU enabled: 55 seconds
I tried to press the "process" button and the "start" button on my stopwatch at the same time, might be +/- 1 second error. So I notice a difference with or without GPU.
Last edited:
So:
first run with GPU: 45sec
second one with GPU: 45sec
first run without GPU: 1m 15sec
second one without GPU:1m 16sec
So.... Capture One is faster on a 580X? @andersburn scored 45sec and @Bohemien 55sec with GPU enabled and all other factors equal?
I don't think you can do realworld testing like this -
I'm in Denmark and it properly work a lot slower in 40 degrees than the 20 I have. And I counted from I pressed start till the counter disappeared, he might have done it differently.
But you can say that there's properly no difference between the to cards in Capture One.
It’s 14 degrees in Germany
I pressed “start” on the stopwatch simultaneously with “process” on the Mac and stopped when the orange bar disappeared. The 580X could be faster than the Vega for RAW export, there was a post sometime ago with a link comparing the detailed Geekbench compute results of the two GPUs, the Vega was faster in most categories shown there, but not in RAW processing (see edit below). Maybe this shows here? That point made me wonder if the 580X might be better for my intended use, but then there are so many other computing tasks where the Vega should have an advantage.
EDIT:
I checked out the Geekbench scores-what I meant was a comparison like on this page, where when you scroll down you can compare the results of the two GPUs in different categories (Vega is on the right, the page doesn't say that). I chose two Geekbench results that were close to the mean values (on this page), but here, it looks like the Vega 48 actually is faster in RAW processing, too. However, if you look at the list of posted results, you see there is a lot of variation going on-I got around 138000 in one pass, 143000 in the next.
So bottom line-I have no idea why @andersburn's 580X renders the C1 test 10 seconds faster than my Vega 48. Maybe it's really down to the weather...
I pressed “start” on the stopwatch simultaneously with “process” on the Mac and stopped when the orange bar disappeared. The 580X could be faster than the Vega for RAW export, there was a post sometime ago with a link comparing the detailed Geekbench compute results of the two GPUs, the Vega was faster in most categories shown there, but not in RAW processing (see edit below). Maybe this shows here? That point made me wonder if the 580X might be better for my intended use, but then there are so many other computing tasks where the Vega should have an advantage.
EDIT:
I checked out the Geekbench scores-what I meant was a comparison like on this page, where when you scroll down you can compare the results of the two GPUs in different categories (Vega is on the right, the page doesn't say that). I chose two Geekbench results that were close to the mean values (on this page), but here, it looks like the Vega 48 actually is faster in RAW processing, too. However, if you look at the list of posted results, you see there is a lot of variation going on-I got around 138000 in one pass, 143000 in the next.
So bottom line-I have no idea why @andersburn's 580X renders the C1 test 10 seconds faster than my Vega 48. Maybe it's really down to the weather...
Last edited:
Max just posted this - as always, a thorough analysis and a well-formulated assessment based on his findings:
His analysis is focused on video editing, but I think from his results one could deduce that while the Vega should undoubtedly (?) be the stronger GPU judging from the specs (more raw computing power, faster memory), a lot depends on how the software exploits the features of the newer hardware design.
One interesting observation on the side is that in his perception the fans seem to ramp up quicker with the 580X, as this was a point people speculated about from the day the 2019 iMacs were announced (the Vega 48 allegedly running cooler thanks to the HBM2 memory).
His analysis is focused on video editing, but I think from his results one could deduce that while the Vega should undoubtedly (?) be the stronger GPU judging from the specs (more raw computing power, faster memory), a lot depends on how the software exploits the features of the newer hardware design.
One interesting observation on the side is that in his perception the fans seem to ramp up quicker with the 580X, as this was a point people speculated about from the day the 2019 iMacs were announced (the Vega 48 allegedly running cooler thanks to the HBM2 memory).
Last edited: