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.
External GPU (eGPU) Resources
- Thread starter F-Train
- Start date
- Sort by reaction score
I think that purchasing a GPU is like purchasing RAM. More is always better. The question is, how much is enough?
If I was making the kinds of photographs that Edward Burtynsky and Jeff Wall make, I'd have 128GB of RAM minimum and the fastest video card on the planet. For some of his recent photographs, in which he's started experimenting with Virtual Reality, I wouldn't be surprised if Burtynsky is using what filmmakers call a rendering farm.
On the other hand, I figure that Henri Cartier-Bresson would have happily got on with 8GB-16GB of RAM and an RX 480/580/590.
Edward Burtynsky, Salinas #2, Cádiz
Jeff Wall, A Sudden Gust of Wind (After Hokusai)
Hernri Cartier-Bresson, Hyères, Côte d'azur
Last edited:
I wouldn't put myself in that class of photographers. My talents and work are certainly amateur.
I shoot with an Olympus EM5 Mk2 that is 'only' 16MP. The raws are relatively small compared to my previous Sony cameras, and a lot less than some of the big shot cameras out there now.
As for the choice between a RX590 and Vega 56, who knows really. I think I'll just start shopping around and see what sort of a deal I can scrounge up.
I shoot with an Olympus EM5 Mk2 that is 'only' 16MP. The raws are relatively small compared to my previous Sony cameras, and a lot less than some of the big shot cameras out there now.
As for the choice between a RX590 and Vega 56, who knows really. I think I'll just start shopping around and see what sort of a deal I can scrounge up.
The RX 580 launched in April 2017 at US$230. If you're a bit patient, you can purchase it now at under $200.
The RX 590 launched last month at $280. That is the current price, although I saw one attempt on Amazon to sell it at $320. I doubt that it will go down from its launch price in the near future.
The RX Vega 56 launched in August 2017 at a nominal price of $400. I say nominal because this was in the midst of the cryptocurrency mining craze and it looks like the card was not actually available at that price. Also, the "partner" custom versions didn't launch until December 2017/January 2018.
Right now, Vega 56 and 64 prices are all over the map. We are starting to see Vega 56 prices sub-$400, and I have seen "sale" pricing on the Vega 64 as low as $340, which is lower than any price that I have seen for the Vega 56.
On the Vegas, reference cards (which use blower ventilation and come stock in terms of clocking) are less expensive than the "custom" cards. Among the partners, Sapphire and Asus appear to be very reluctant to significantly cut the price of their custom versions of the Vega 56 from MSRP. ASRock, which only started making AMD cards this past summer, appears to be the most aggressive on price cutting. It was the partner that offered the Vega 64 reference card for $340. It may be focused on acquiring market share.
Last edited:
The basic difference is between reference cards, which have one fan, and custom cards, which have two or three. The debate about which is "better", from the points of view of cooling and noise, is invariably based on the premise that the card is going into a computer case that has other components that may be affected by the heat that the card generates.
This premise goes out the window if we are talking about a card that is going to live by itself in an external GPU enclosure. Maybe others have a view, but it is unclear to me which of the two designs, if either, is "better" for what we are talking about here. I am inclined to think that it doesn't matter a whole lot, and reference cards, when they are available, are cheaper. That said, I purchased Asus's three-fan Vega 56, which is not the least expensive option on the market, just because there are things about it, including how it handles heat dissipation and cooling, that I like.
I believe that prices are all over the place, as you put it, because the vendors don't know at the moment what the market will bear. They are trying to find out by testing pricing. I would not assume, just because a card has an asking price, that people are actually buying at that price. Within days of the release last month of the RX 590 at US$280, there was an attempt to sell it on Amazon at $320. Did anybody actually pay that? Who knows. What I do know is that the asking price for the RX 590 seems to have settled, for the moment, at $280. That suggests that the attempt to sell at $320 was not a roaring success.
I think that the only sensible strategy is to keep original launch prices firmly in mind when considering asking prices. The caveat to this is that custom cards do attract something of a premium over reference card launch price. For example, in August 2017, the notional launch price of the Vega 56 reference card was $400. However, I know that in December 2017/January 2018, when Asus launched its three-fan Vega 56, it launched at a notional price of $450/$460. As in my last post, I'm using this word notional because until recently the Asus cards, in reality, commanded a great deal more than that. Right now, their asking price actually is $450/$460. And can you purchase a Vega 56, especially a reference version, for more like $400, maybe even $380? Yes.
If I wanted to be cynical, I'd say this... Graphics cards are fundamentally commodities. The "partners" are in the business of trying to make a commodity look like a differentiated, value added product. In other words, they are basically marketers. As a purchaser, it comes down to how much you buy into their attempt at differentiation. Gamers apparently buy into it hook, line and sinker. I may be halfway there
Last edited:
Apple recommends either the MSI card or the Sapphire Pulse (at least for the cheese grater Mac Pro, I'm sure the same can be said for eGPUs).
I'm waiting for bootcamp support.
If that happens, the free games that currently come with the purchase of certain AMD GPUs (see post #293) might actually be worth US$180, instead of their current value to a Mac user, which is $0
Last edited:
I don't know whats holding them back. It would actually make macs viable gaming rigsIf that happens, the free games that currently come with the purchase of certain AMD GPUs (see post #293) might actually be worth US$180, instead of their current value to a Mac user, which is $0
[doublepost=1544747241][/doublepost]
as far as i looked into it, it works if you do some tinkering...
(edit GTX 770 is Keppler so metal is supported, should work)
When I was building my PC over the summer, the message I got from my research with the RX580 was to buy the cheapest non-reference (I do not think many people buy reference cards of any GPU after launch) you could find from any of the major brands. If I remember right, it came down to the RX580 not overclocking very well and essentially all of them being very similar. I do know that this is not always the case, but with that card, that was the advice I had gleaned. The only other thing I found was people's hesitation towards certain brands because of bad experiences, so I would take all of those arguments with a grain of salt unless you see a trend form.
One last thing on reference cards, I believe most PC builders stay away from them due to cooling. This would be a much less significant issue if you are not planning on overclocking or have a eGPU box that has sufficient cooling. In other words, I think most looking for an eGPU on a Mac would likely be fine with a reference card.
[doublepost=1544748117][/doublepost]@F-Train your benchmarks with the Vega 56 vs RX590 are interesting. Can you share what it translates to in real world usage? For example, I think you have said you work with 4K video, how much of a time savings are we talking between those two cards?
I'm waiting for bootcamp support.
anyone has Razer Core X? seems cheaper than the Sonnet... much louder?
I only own Core X, and it's very quiet with AMD Vega 56 inside. I couldn't hear much noise out of it.
What I'd like to know is why AMD would offer US$180 in incentives to get people to purchase its GPUs, but only offer incentives to people who use Windows.
[doublepost=1544749094][/doublepost]
I've only had the Vega 56 since yesterday and I hope to do some tests over the weekend. See post #270 if you have anything specific in mind.
Last edited:
I plan to spend some time this weekend comparing how Final Cut Pro X performs using my Mac mini's internal GPU, the RX Vega 56 GPU and perhaps the RX 590 GPU.
This morning, I did a quick export test with interesting results. The computer is a 2018 Mac mini i7/8GB/512GB, and the test file was a 3840x2160 .mov video that is 06:47 in length and 1.99GB in size.
With the Vega 56, the export was almost twice as fast:
Share/export to Apple Devices 4K using the Vega 56: 08:04
Share/export to Apple Devices 4K using the internal GPU: 15:12
Post-export, both files were 1.03GB.
Interestingly, Final Cut appears to have used both the Vega and the internal GPU for the faster export. If the screen capture below accurately reflects what happened, it would appear that Final Cut may have distributed the task between the two. On the left of the capture, we see the GPUs post-export, with the Vega idling and the internal GPU not even registering on the graph.
I believe that the latest release of Compressor, which is now 64-bit, will be faster at this than Final Cut. At the moment, Compressor will not use an external GPU. Nevertheless, I'm going to include it in this weekend's tests.
This morning, I did a quick export test with interesting results. The computer is a 2018 Mac mini i7/8GB/512GB, and the test file was a 3840x2160 .mov video that is 06:47 in length and 1.99GB in size.
With the Vega 56, the export was almost twice as fast:
Share/export to Apple Devices 4K using the Vega 56: 08:04
Share/export to Apple Devices 4K using the internal GPU: 15:12
Post-export, both files were 1.03GB.
Interestingly, Final Cut appears to have used both the Vega and the internal GPU for the faster export. If the screen capture below accurately reflects what happened, it would appear that Final Cut may have distributed the task between the two. On the left of the capture, we see the GPUs post-export, with the Vega idling and the internal GPU not even registering on the graph.
I believe that the latest release of Compressor, which is now 64-bit, will be faster at this than Final Cut. At the moment, Compressor will not use an external GPU. Nevertheless, I'm going to include it in this weekend's tests.
Last edited:
GFX Bench has a new benchmark test, based on the Metal API, called Aztec Ruins. I think that this test is particularly interesting because Apple has deprecated OpenGL in favour of Metal. It would be great if more people would try this new test.
AnandTech has an article that explains Aztec Ruins at https://www.anandtech.com/show/13271/kishonti-releases-vulkan-gfxbench-5
Aztec Ruins, Intel 630 vs RX Vega 56 (Test 1 of 2)
Setting for Test 1: High Tier Offscreen, 2560x1440
Computer: 2018 Mac mini, i7/8GB/512GB
GPU enclosure: Asus XG Station Pro
Monitor: Asus ProArt 32" 3840x2160 (PA32UC)
First Screenshot: Intel UHD Graphics 630 (2018 Mac mini internal graphics card)
Second Screenshot: RX Vega 56 (Asus AREZ Strix, OC Edition). AREZ Strix is the new name for the ROG Strix series.
[doublepost=1544904820][/doublepost]Aztec Ruins, Intel 630 vs RX Vega 56 (Test 2 of 2)
Setting for Test 2: Normal Tier Offscreen, 1920x1080
Computer: 2018 Mac mini, i7/8GB/512GB
GPU enclosure: Asus XG Station Pro
Monitor: Asus ProArt 32" 3840x2160 (PA32UC)
First Screenshot: Intel UHD Graphics 630 (2018 Mac mini internal graphics card)
Second Screenshot: RX Vega 56 (Asus AREZ Strix, OC Edition). AREZ Strix is the new name for the ROG Strix series.
AnandTech has an article that explains Aztec Ruins at https://www.anandtech.com/show/13271/kishonti-releases-vulkan-gfxbench-5
Aztec Ruins, Intel 630 vs RX Vega 56 (Test 1 of 2)
Setting for Test 1: High Tier Offscreen, 2560x1440
Computer: 2018 Mac mini, i7/8GB/512GB
GPU enclosure: Asus XG Station Pro
Monitor: Asus ProArt 32" 3840x2160 (PA32UC)
First Screenshot: Intel UHD Graphics 630 (2018 Mac mini internal graphics card)
Second Screenshot: RX Vega 56 (Asus AREZ Strix, OC Edition). AREZ Strix is the new name for the ROG Strix series.
[doublepost=1544904820][/doublepost]Aztec Ruins, Intel 630 vs RX Vega 56 (Test 2 of 2)
Setting for Test 2: Normal Tier Offscreen, 1920x1080
Computer: 2018 Mac mini, i7/8GB/512GB
GPU enclosure: Asus XG Station Pro
Monitor: Asus ProArt 32" 3840x2160 (PA32UC)
First Screenshot: Intel UHD Graphics 630 (2018 Mac mini internal graphics card)
Second Screenshot: RX Vega 56 (Asus AREZ Strix, OC Edition). AREZ Strix is the new name for the ROG Strix series.
Last edited:
F-Train: I continue to be impressed with the benchmark results from your setup. To recap my setup is i7/8/512, with ASUS RX Vega 64 in a Razer Core X, and an LG 27UK850-W. I'm not sure if I am running the tests correctly and just selected the two tests without changing any other settings in the GFX Metal app or on my mini. I do wonder what a bump in my RAM from my current 8GB might achieve. Here are my results:
So, after Christmas next week, I think I'll go ahead and purchase the Asus Station Pro and a Sapphire RX590. Just shy of $600 at current prices. I looked into the Vega cards, but they're priced way up there... out of my budget.
I think that you are making the right decision. Based on how you intend to use the GPU, and having used both the RX 590 and Vega 56 with Photoshop and Lightroom, I don't believe that you would get any additional benefit from the Vega. Indeed, I think that you'd be fine with the RX 580, which would save you US$50-$80.
If the Amazon and Newegg prices are the same when you make your purchase, I would suggest that you buy from Amazon. It has a significantly better return policy on GPUs.
If you are interested in the free games, a caveat... The retailer supplies the codes necessary to claim the games. Newegg delivers the codes within 48 house of purchase. With Amazon, you may have to follow up with customer support. Of course, this may be quite academic. The games will not run on Mac OS, and at the moment Apple does not support the use of an eGPU with Bootcamp.
Last edited:
In case anyone is interested, here is how System Information reports the GPUs and eGPU enclosure for my Mac mini.
First screen capture: Intel UHD Graphics 630 (built-in GPU)
Second screen capture: RX Vega 56 (external GPU)
Third screen capture: Asus XG Station Pro (eGPU enclosure)
In the second capture, the display serial number is blank because I removed it in Photoshop.
First screen capture: Intel UHD Graphics 630 (built-in GPU)
Second screen capture: RX Vega 56 (external GPU)
Third screen capture: Asus XG Station Pro (eGPU enclosure)
In the second capture, the display serial number is blank because I removed it in Photoshop.
Last edited:
With one exception, I have had no problems using the Asus XG Station Pro to run the Vega 56.
GFXBench Metal has 19 benchmark tests. Of those, one, an older GFXBench test that has apparently been updated, causes the Asus power supply to shut down. I don't know the reason.
Apart from that, I have had no power issues running benchmarks, running X-Plane (which is GPU intensive) or running Final Cut Pro X tasks such as 4K rendering, exporting and transcoding.
I'm also having no issues with noise or heat. I suspect that the talk about those issues is about video cards in computer cases, and especially cards that gamers have overclocked and are running hard. Or perhaps I'm just not particularly sensitive to fan noise.
Last edited:
Good to know! I might grab on early next year. Prices keep going down...
I believe that you are in Europe. When I checked Vega 56 prices, I had a look at European pricing and found two things. First, there seem to be quite a few more vendors than in the U.S. Secondly, prices ex-VAT were lower than U.S. prices before tax, perhaps due to there being more vendors/competition.
One of the things that sold me on the Asus XG Station Pro is that it comes with a 1.5m/5' Thunderbolt cable. If you get an enclosure with a 50cm/20" cable, consider purchasing a longer active cable. I really like the latitude that the Asus cable gives me in placement of the enclosure.
If you purchase the Asus... There have been some complaints about the GFX Metal tests suggesting that they can be wonky. Regardless, I don't place any stock in the fact that I ran into a power supply issue with one out of 19 tests, particularly since I have had no other issues. That said, if the power supply ever shuts down (it has a light that will go off), it's a matter of unplugging it and plugging it back in.
Last edited: