Dual 2 two RX 580 Final Cut Pro X

[h9826790]

I see you used the settings for the WX7100, but have you actually tested this card I am very curious.

This end up being my daily use setting. I use this setting for about 2 months now. No problem at all.

I pick this setting is because I don't need super power, if yes, I will go for a more powerful card.

And after memory timing mod, it gives me roughly the same performance as the original setting (on compute).

And the GPU is really running cool and quiet now. I just try another 10min Furmark test. This is what I get.

75C, 50% fan speed, little bit fan noise but still fairly quiet, and remain at the target temperature.

And of course it handle the OS well. Also, very decent performance in FCPX, Blender (including Radeon Pro Render), and some gaming (macOS) etc.

If there is some simple test you want to see. Please let me know. If I can access the required software, I should able to run the test for you.

 

[startergo]

This end up being my daily use setting. I use this setting for about 2 months now. No problem at all.

I pick this setting is because I don't need super power, if yes, I will go for a more powerful card.

And after memory timing mod, it gives me roughly the same performance as the original setting (on compute).

And the GPU is really running cool and quiet now. I just try another 10min Furmark test. This is what I get.
View attachment 804116
75C, 50% fan speed, little bit fan noise but still fairly quiet, and remain at the target temperature.

And of course it handle the OS well. Also, very decent performance in FCPX, Blender (including Radeon Pro Render), and some gaming (macOS) etc.

If there is some simple test you want to see. Please let me know. If I can access the required software, I should able to run the test for you.

So is this an actual WX7100 or just the settings of it applied to RX-580? I wanted to see the performance of the actual card. Server type card with 3 years of warranty and twice the price of RX-580 should have an adequate cooling for its design, right (or no )?

 

[h9826790]

So is this an actual WX7100 or just the settings of it applied to RX-580? I wanted to see the performance of the actual card. Server type card with 3 years of warranty and twice the price of RX-580 should have an adequate cooling for its design, right (or no )?

Oh, you mean the actual WX7100 card? NO, I don't have that, and no intention to buy one as well.

I just extract the setting and apply that to my RX580.

I am sure the cooling is adequate (but not necessary anything more than adequate). Also, server card never need to care about noise. This is why I prefer gaming card more.

With the mod, now my card has more than adequate cooling. The overkilled cooling ability makes the card can run very quiet even under high stress.

If the PULSE RX580 cost only half, and have 2 years warranty. Then that means for the same cost, I can have 4 years warranty. Which assume the RX580 still selling with the current price 2 years later. Obviously it won't happen.

 

[startergo]

Oh, you mean the actual WX7100 card? NO, I don't have that, and no intention to buy one as well.

I just extract the setting and apply that to my RX580.

I am sure the cooling is adequate (but not necessary anything more than adequate). Also, server card never need to care about noise. This is why I prefer gaming card more.

With the mod, now my card has more than adequate cooling. The overkilled cooling ability makes the card can run very quiet even under high stress.

If the PULSE RX580 cost only half, and have 2 years warranty. Then that means for the same cost, I can have 4 years warranty. Which assume the RX580 still selling with the current price 2 years later. Obviously it won't happen.

Mine is actually DELL OEM single fan 110W and I am curios do I have to tweak anything apart from maybe the target temperature? What would you suggest?

 

[h9826790]

Mine is actually DELL OEM single fan 110W and I am curios do I have to tweak anything apart from maybe the target temperature? What would you suggest?

From memory, that's with the RX480 setting, therefore, already "downclocked" to 1266MHz.

And AFAIK, that 110W is assuming you run the card in Windows with the default setting.

In Windows, AMD driver default "power play = 0", and this number will override the setting inside the ROM. Therefore, the card is not allowed to draw more than the "max power limit" defined in the power tune section. Which is 110W for your card.

However, in macOS, due to we use Apple driver, and the driver use the power play setting in the VBIOS, which should be 50% for your card (I cannot quite remember if I read the details about your card's ROM, but 50% is the default for RX480 reference card).

Which means, the card can actually draw 165W. In fact, this number explained why the RX480 reference card can draw ~160W in power draw test. And can draw more than 75W from the slot, end up killed quite a bit gaming PC's motherboard.

Since your card should has the same setting, which also means under high stress, most likely it will draw very close (or even more than) 150W. Which will slightly exceed the official limit that the slot and a 6pin can deliver. But shouldn't be a matter for cMP due to its high build quality.

So, if you really want your card stay at or below 110W, you should mod the "power control limit" in power play section to zero.

Or may be you can mod it to 30%, therefore, it should never draw above 150W.

However, in any case, once you lower the power control limit, your card's performance should drop in macOS. If you care about it, and your setup is working exactly what you want. You better not to change any setting.

For target temperature, your card should set that to 80C already. Not much you can do. You can lower it, but expect higher fan noise (thermal throttling occur at max temperature, which should be 90C for your card). And no point to raise it because 80 is quite high already.

 

[startergo]

this is my card:
https://www.techpowerup.com/vgabios/194686/194686
But I have dual mini 6 pin to 6 pin connected so the power should be plenty. I was thinking more about resetting the target temperature to 50C maybe instead of 75?
BTW look at the specs of the WX7100 https://www.amd.com/Documents/radeon-pro-wx7100-datasheet.pdf
display output and video acceleration

 

[h9826790]

this is my card:
https://www.techpowerup.com/vgabios/194686/194686
But I have dual mini 6 pin to 6 pin connected so the power should be plenty. I was thinking more about resetting the target temperature to 50C maybe instead of 75?
BTW look at the specs of the WX7100 https://www.amd.com/Documents/radeon-pro-wx7100-datasheet.pdf
display output and video acceleration

NO, that's just plenty "power available" for the 6pin. However, the card is drawing power according to how it programmed, but not how much power available to the 6pin.

Even you give it 100000W available on the 6pin. It will still draw ~50% of what it actually use, and the remaining 50% from the PCIe slot. And if the card draw more than 150W in total, there will be more than 75W draw from the slot regardless how much buffer you gave to the 6pin.

As per your link, your card's target temperature is 80C, not 75C. And if you set it to 50C, the fan will spin up much much earlier. However, won't change the behaviour that still allow to GPU warm up to 90C, and then get into thermal throttling. Unless the cooling system can handle all the heat.

For room temperature 22C, your card expected to stabilise at about 83C. Since it's above the default target temperature, which means even though you set the target temperature to 50C. It will still stabilised at 83C, just the fan spin up earlier, and slower to reach 83C.

 

[orph]

@startergo
as it's a dell card i think dell/amd will have the power under control to work in OEM systems that tend to have weaker PSU's & motherboards so i think it will be safe at default settings.

if you pull up something like hardware monitor in osx you can watch power use and see what it pulls under stress, i used luxmark GPU stress test and give it 10 mins or so and watch power use from both 6 pin and pci slot.

in windows i dont think you can see the full power use of the card ?

if you do want to make any changes to the bios first back up your original and know there is some risk of bricking the card. (and read all the posts with instructions!)
only change i think is good for you to do is to use the memory timeings patch as that will give a speed bost but wont use a lot more power (dont think it will use much more at all if any)

id not push core clock or anything relay all you will do is get the fan loud and push power use higher.

I set my card to run at 1300mhz your card is 1266mhz thats only 33mhz slower which relay almost no difference, im vary happy with the speed & in a lot of ways im tempted to say just use it without messing with it.

 

[Balbuzard2A]

Hello,
I found this topics few days ago.
I have a Mac Pro 2009-5.1 Dual Xeon 5690 - 32gb Ram - Ssd- Mojave with 2 Sapphire Pulse rx580 radeon, dedicated to video editing.
2xSapphire RX580 Pulse Radeon, feeded by an external power source.
1 in the 16x PCI slot the second into a 4X PCI slot (to large to use both 16x PCI slots)

Here are the benches I did today for you :
I have 2 UHD 28" monitors (one plugged to each radeon)
I did all test with all 3840/2060 display resolution.

Geekbench with one GPU selected at a time :
Geekbench OpenCL : gpu(16x) = 39643
Geekbench OpenCL : gpu(4x) = 39131
Geekbench Metal : does not display results... I do not know why but the test time is really short.

Unigine Valley (default settings) 24,1 fps 1226 score (only test on the 16x radeon out)

test with BruceX (FCP 10.4.6)

I unplug the second rx580 for test with only one GPU (16x slot) render to ssd (sata2 original motherboard)
1 GPU only Bruce X : 16 seconds
I plug the second GPU
2 GPU (16x and 4x PCI slots) Bruce X : 10,7 seconds

As expected : no changes in Cinebench score : 3367

amha, 2 RX580 is 15-40% faster than one depending the task FCPX runs

Hope this helps.

Regards

 

[Balbuzard2A]

I updated my config :

Big Sur with OpenCore
64 GbRAM

Pcie1 : RX580
Pcie2 (with Pcie Riser) : 5700XT
Pcie3 : Inatek usb3.1gen2 (20gb/s)
Pcie4 : decklink video monitor 4k

Internal 6 SATA devices :
1 ssd for System
2 HDs in Raid0 (12To total)
3 HDs in Raid0 (24To total)

an impressive "all boost" !