RX 580 - no Sapphire Pulse 8GB available - "pure cosmetics?"

[owbp]

For the price, I guess I'll keep it in and will swap it for the GTX980Ti when I need CUDA power for Resolve. Which I only need once every couple of months.

While being at it, would you mind testing it in Resolve too?
How does it compare with GTX980ti in all time OpenCL vs CUDA(Resolve or PP)?
I've just sold off my "Mac GPUs" and I'm really close to buying RX580, but with GTX980ti being just a little bit more expensive (second hand), I'm really close to finally abandoning FCPX and starting fresh with Resolve

P.S. I've found out that R9 280X, although being much faster in rendering, is a little less forgiving with preview of unrendered timeline. GTX980 (non TI) wasn't used at all in FCPX rendering, but i could preview with her forever after adding any effect i could think of. Wonder if it is same for RX580?

 

[h9826790]

Received it today. The box is super tiny - well compared to the giant GTX 10x0 boxes I've tried in the past couple of months. But the card is big. It's a 2 slots, but it needs some wiggling to get into slot 1. Out of curiosity, I put the Nvidia GT120 in slot 2 and it blocks one of the fans - not only it covers it, but it also touches it. I tried to move the USB3 card in slot 2 to see, and it's a very tight install. I don't know why they can't just make a card that fits the 2 slot specs.
Plugged it, booted up and it just works. Feels more "solid" than the GTX980Ti which has its glitches now and then.
I was surprised to see that it's just a RXX Radeon something in Sierra, where is HS it shows as RX 580 properly. Isn't it supposed to do that in Sierra too?
For the price, I guess I'll keep it in and will swap it for the GTX980Ti when I need CUDA power for Resolve. Which I only need once every couple of months.

No, the ident is only correct in High Sierra, but not Sierra. I think only HS officially support eGPU, and this RX580 is a part of the eGPU developer kit, therefore, Apple has no need to ident this GPU correctly in Sierra.
[doublepost=1512084701][/doublepost]

While being at it, would you mind testing it in Resolve too?
How does it compare with GTX980ti in all time OpenCL vs CUDA(Resolve or PP)?
I've just sold off my "Mac GPUs" and I'm really close to buying RX580, but with GTX980ti being just a little bit more expensive (second hand), I'm really close to finally abandoning FCPX and starting fresh with Resolve

P.S. I've found out that R9 280X, although being much faster in rendering, is a little less forgiving with preview of unrendered timeline. GTX980 (non TI) wasn't used at all in FCPX rendering, but i could preview with her forever after adding any effect i could think of. Wonder if it is same for RX580?

Can you preview properly with 280X by using proxy? I know the quality will drop a bit, but in general, there is no need to preview down to exact pixel accuracy, as long as the "big picture" looks correct, it's usually good enough to "know" if the editing is correctly done.

 

[owbp]

Can you preview properly with 280X by using proxy? I know the quality will drop a bit, but in general, there is no need to preview down to exact pixel accuracy, as long as the "big picture" looks correct, it's usually good enough to "know" if the editing is correctly done.

Yes, that is not problem, proxy or optimized video, i could preview both if i go easy on transitions and visual effects.
The thing that amazed me is that GTX980 could preview with no stutter or dropped frames much more heavily edited footage without rendering.

To be perfectly clear, in terms of rendering speeds R9 280X is king of the cards i've tried with FCPX.
But amount of effects you could add before frames being dropped even GTX680 had better results than R9 280X, of course GTX980 could handle a lot more IIRC.
And you know how much your GTX1080TI is under-optimized in FCPX and how much it isn't used at all... Same was for 980 in my case.

So I'm just interested where RX580 sits in all this (knowing that rendering and export speeds are great).

 

[h9826790]

Yes, that is not problem, proxy or optimized video, i could preview both if i go easy on transitions and visual effects.
The thing that amazed me is that GTX980 could preview with no stutter or dropped frames much more heavily edited footage without rendering.

To be perfectly clear, in terms of rendering speeds R9 280X is king of the cards i've tried with FCPX.
But amount of effects you could add before frames being dropped even GTX680 had better results than R9 280X, of course GTX980 could handle a lot more IIRC.
And you know how much your GTX1080TI is under-optimized in FCPX and how much it isn't used at all... Same was for 980 in my case.

So I'm just interested where RX580 sits in all this (knowing that rendering and export speeds are great).

Yes, I know the 1080Ti is pretty much idle in FCPX. A single 7950 can render faster than the 1080Ti in FCPX. I can see that the card is doing something with very light loading, it only draw ~17W most of the time (13W is the real idle draw in my case), and max at just ~60W.

 

[pierrox]

While being at it, would you mind testing it in Resolve too?
How does it compare with GTX980ti in all time OpenCL vs CUDA(Resolve or PP)?
I've just sold off my "Mac GPUs" and I'm really close to buying RX580, but with GTX980ti being just a little bit more expensive (second hand), I'm really close to finally abandoning FCPX and starting fresh with Resolve

P.S. I've found out that R9 280X, although being much faster in rendering, is a little less forgiving with preview of unrendered timeline. GTX980 (non TI) wasn't used at all in FCPX rendering, but i could preview with her forever after adding any effect i could think of. Wonder if it is same for RX580?

I actually tested the RX580 on page 2 of this thread. I first got a Nitro+ as it was available and I wanted to test one - thanks Amazon for their return policy! In Resolve, it didn't do much better than the R9 280X (I'm a big fan of this card too, well integrated and flashable). But like the R9 280X, it feels solid, probably because the drivers are better integrated in the OS - if not better written, but that's another debate!

But from what I read here and there, RX580 does pretty good in FCPX. Which shouldn't be surprising as it's newer tech than the R9 280X... and come on, it has nearly 3 times more RAM!

Having said all that, if you want to go for Premiere or Resolve, then definitely go for a Nvidia. CUDA is being used for real and there's no comparison. I get twice the frame rate when I transcode rushes in Resolve with my 980Ti. The CPU becomes the bottleneck, so no need to go extra big. I think currently the sweet spot on a cMP would be the 1080 - similar perf than the 980Ti, less power draw, and less pricey than a 1080Ti.

 

[h9826790]

I actually tested the RX580 on page 2 of this thread. I first got a Nitro+ as it was available and I wanted to test one - thanks Amazon for their return policy! In Resolve, it didn't do much better than the R9 280X (I'm a big fan of this card too, well integrated and flashable). But like the R9 280X, it feels solid, probably because the drivers are better integrated in the OS - if not better written, but that's another debate!

But from what I read here and there, RX580 does pretty good in FCPX. Which shouldn't be surprising as it's newer tech than the R9 280X... and come on, it has nearly 3 times more RAM!

Having said all that, if you want to go for Premiere or Resolve, then definitely go for a Nvidia. CUDA is being used for real and there's no comparison. I get twice the frame rate when I transcode rushes in Resolve with my 980Ti. The CPU becomes the bottleneck, so no need to go extra big. I think currently the sweet spot on a cMP would be the 1080 - similar perf than the 980Ti, less power draw, and less pricey than a 1080Ti.

Not sure if that Pascal’s issue. In my own test (Resolve), CUDA won’t perform better than leave the acceleration at “Auto” (more or less the same speed). Also, if manaually select CUDA. The output video may be seriously interlaced. This “bug” only happen when I manually select CUDA as the hardware acceleration method. All OpenCL, METAL, Auto won’t show this problem.

 

[Squuiid]

But the card is big. It's a 2 slots, but it needs some wiggling to get into slot 1. Out of curiosity, I put the Nvidia GT120 in slot 2 and it blocks one of the fans - not only it covers it, but it also touches it. I tried to move the USB3 card in slot 2 to see, and it's a very tight install. I don't know why they can't just make a card that fits the 2 slot specs.

Yes, this was an issue for me too. This card is BIG despite supposedly being a 2 slot card. I had to remove a plastic gromit from my NIC which protruded out the back and touched a fan on the RX 580.
I tried the Allegro Pro USB card instead and that flat out just blocked the entire fan from functioning.
Be aware of this before buying.
Other than that, I think it's the perfect card for a cMP right now.

 

[h9826790]

Yes, this was an issue for me too. This card is BIG despite supposedly being a 2 slot card. I had to remove a plastic gromit from my NIC which protruded out the back and touched a fan on the RX 580.
I tried the Allegro Pro USB card instead and that flat out just blocked the entire fan from functioning.
Be aware of this before buying.
Other than that, I think it's the perfect card for a cMP right now.

I think you have the PULSE, and he has the Nitro+. Your card is a 2 slot card, and his card is a 2.2 slot card.

I went to the computer shop yesterday to look at the real thing, and I realise even with the same design, the Nitro+ clearly looks bigger than the PULSE. I was fooled by the photos on the web, because of the same design, they look identical on the monitor, but actually they are not.

 

[Squuiid]

I think you have the PULSE, and he has the Nitro+. Your card is a 2 slot card, and his card is a 2.2 slot card.

I went to the computer shop yesterday to look at the real thing, and I realise even with the same design, the Nitro+ clearly looks bigger than the PULSE. I was fooled by the photos on the web, because of the same design, they look identical on the monitor, but actually they are not.

I agree, the Nitro+ is far bigger however, I'm pretty sure Pierrox also ordered a Pulse which he ended up keeping (he returned the Nitro+).
If he was talking about the Nitro+ then there is no way you could fit another card in Slot 2 I don't think.
I could be wrong though

 

[h9826790]

I agree, the Nitro+ is far bigger however, I'm pretty sure Pierrox also ordered a Pulse which he ended up keeping (he returned the Nitro+).
If he was talking about the Nitro+ then there is no way you could fit another card in Slot 2.
I could be wrong though

Oh yeah, he just got the PULSE yeaterday.

In fact, I replied to his post about Sierra not showing the ident RX580 properly.

 

[pierrox]

Oh year, he just got the PULSE yeaterday.
In fact, I replied to his post about Sierra not showing the ident RX580 properly.

Yep that's correct!

I first got a Nitro+ which is very bulky - and that coming from someone who also tried a triple fan 1080Ti! As soon as I realized it covers slot 2, the return label was printed!
And I got the Pulse yesterday and it's still pretty bulky. The backplate is thick and forces you to wiggle it into the slot. Then the top box is very close to the slot above. As I said, keeping the GT120 in there at the same time is totally impossible, only a thin card without any circuit underneath can fit in there (and not too wide as it would block the air flow).

All in all rather happy with it for the price, but I hope it benefits from more driver improvement in the next HS updates because it's showing glitches with 3 monitors (when they wake up from black screen) which I don't get with Sierra. Anyway, HS has long to go as we discover everyday... #rootpasswordfiasco

 

[tabarus]

Anyone meanwhile tested all ports of a Sappire RX580 Nitro+ ?,
are there any other issues beside that it is shown a generic card ?

 

[owbp]

Anyone meanwhile tested all ports of a Sappire RX580 Nitro+ ?,

There shouldn't be any problem with ports on unflashed cards, aka cards that are using generic Radeon Framebuffer.
Dead ports appear if you assign specific framebuffer (usually via Clover/Chameleon) or you flash the card with EFI that has different port mapping than the card EFI is taken from.

 

[tabarus]

There shouldn't be any problem with ports on unflashed cards, aka cards that are using generic Radeon Framebuffer.
Dead ports appear if you assign specific framebuffer (usually via Clover/Chameleon) or you flash the card with EFI that has different port mapping than the card EFI is taken from.

So it shoud be fine to safe some dollars and get a nitro instead of the pulse ? (on amazon.de is the nitro cheaper)

 

[h9826790]

So it shoud be fine to safe some dollars and get a nitro instead of the pulse ? (on amazon.de is the nitro cheaper)

Yes, if you don't mind it's a 2.2 slot card (effectively occupy 3 slots). And you know how to power it properly.

 

[tabarus]

Yes, if you don't mind it's a 2.2 slot card (effectively occupy 3 slots). And you know how to power it properly.

thank you !

 

[owbp]

thank you !

Since you're mentioning amazon.de, take a look at idealo.de also. Pulse is cheaper than Nitro there.

 

[tabarus]

Since you're mentioning amazon.de, take a look at idealo.de also. Pulse is cheaper than Nitro there.

yeah i already had a look there, i also read some reviews about the sellers there, especially when it comes to return things. Therer were some negative things. On Amazon I never had any issues when returning things.

 

[owbp]

yeah i already had a look there, i also read some reviews about the sellers there, especially when it comes to return things. Therer were some negative things. On Amazon I never had any issues when returning things.

Ok, got it. I'm not from Germany, but have few friends there and they've told me to look at idealo for best prices. As i have understand, idealo just search through .de stores and display prices, the rest of up to a seller or the store you're contacting.
But i get it, Amazon is one of the safest options out there.

 

[Oneballs]

Anyone have any problems with a Pulse not being fully recognized? Bought one off Newegg when they got back in stock, but not getting the card fully recognized in “About this Mac.” Not super concerned but one of the reasons I purchased the card was it’s 100% OOTB-ness. Running 10.13.1.

 

[Squuiid]

Anyone have any problems with a Pulse not being fully recognized? Bought one off Newegg when they got back in stock, but not getting the card fully recognized in “About this Mac.” Not super concerned but one of the reasons I purchased the card was it’s 100% OOTB-ness. Running 10.13.1.

Odd. Is it definitely 8GB and not 4? Perhaps they have started shipping a third part number? macOS only looks out for two specific part numbers.

https://www.techpowerup.com/vgabios/191713/sapphire-rx580-8192-170324

https://www.techpowerup.com/vgabios/191403/sapphire-rx580-8192-170310

 

[Oneballs]

Odd. Is it definitely 8GB and not 4? Perhaps they have started shipping a third part number? macOS only looks out for two specific part numbers.

https://www.techpowerup.com/vgabios/191713/sapphire-rx580-8192-170324

https://www.techpowerup.com/vgabios/191403/sapphire-rx580-8192-170310

ah, I had not even thought about that. Yes, definitely a 8GB.

Part number for my pulse is 113-1E3870U-O45.

 

[h9826790]

ah, I had not even thought about that. Yes, definitely a 8GB. View attachment 739927

Part number for my pulse is 113-1E3870U-O45.

This is a new part number. It should be very easy to “fix” the ROM. But I bet you only want to use it OOTB, and don’t want to Hex edit / flash the ROM.

 

[Squuiid]

Part number for my pulse is 113-1E3870U-O45.

That’s a bummer! Very unlucky, although it’s just cosmetic and won’t cause you any problems.

 

[Synchro3]

Same here. Pulse RX 580 from NewEgg. The card is not fully recognized in “About this Mac.”

However, it works just fine.