XFX RX590 Fatboy OC+ Tips

[bignerdguy]

Hi all,

This thread was my way of passing some helpful tips on the XFX 590 Fatboy 8GB OC+ card i found when setting it up and trying to do some folding. I found that the GPU Overclocked setting it uses as its default clock speed does not work for folding at all. It will cause WU's to fail every time after a few minutes of folding. Also the card tends to run too hot with AMD's default settings for the fans which causes a system crash after a few minutes of folding. What i suggest is the following:

Either using the latest ATi drivers (v20.3.1 as of this writing) or via a utility like the MSI Afterburner, set the fan profile to a more aggressive setting which has the max speed occur at something like 77-78 degrees C. I found the card REALLY doesn't like temps much beyond 79 degrees and it will crash quite quickly after it hits 81 degrees. If you set the curve to be a lot steeper and have it ramp up to at least 50% at around 60 degrees, it will hover at no higher than 71-72 degrees even during heavy graphics or folding. I ran 3DMark at the highest settings for the better part of an hour and it barely hit 70 with fans running at around 75% or so.

Second issue is the clock speed. The card defaults to 1545 MHz but this is too fast for F@H. To fix this, use a utility to lower it to the chipset default of 1469 MHz and then bump it up 10-20 MHz at a time, letting it run for a while before continuing, until the WU's fail after a few minutes. Then bump it back down to the previous setting and it should remain stable. If not, bump it back down till it does or you could just run it at the default of 1469 if you dont feel comfortable trying any of this. My card is currently set for 1520 MHz and works normally at that setting with WU's processing without failing. Yours may work differently so dont take my setting as the best for you, test it.

I should mention i have a lot of room and good cooling in the case so the temps were not caused by the airflow being restricted. i found when running MSI Afterburner that AMD's default setting had the fan profile tool low as it was set to really rise at like 75-80 degrees with the peak at 90. the card cant handle this setting so setting it as i discussed is advised.

My system specs are:

AMD Ryzen 5 2400G w/ Vega 11 Graphics (Graphics part disabled with e 590 installed)
XFX RX 590 Fatboy OC+ w/ 8 GB of memory (default clock of 1545 reduced to 1520)
16 GB of Corsair Vengance memory set to DDR-2666
WD NVMe 250 GB SSD model: WDS500G3X0C
and other hardware not relevant to this program.

System runs fine in all other respects and the new card now screams through WU's, unlike the Ryzen vega built in card, which took the better part of a day to process 1 WU. now it can finish in less than 4 hours.

Hope this helps anyone with similar issues. Just reducing the clock speed and upping the fan control will help in many cases, though this will be a system by system fix and each will be different.

[muziqaz]

Thanks for your insight.
Yes, any (even factory) OCs can be unstable running fah, as this software uses the very latest of instructions available on our GPUs.

In your instance it would be better to use 590 for folding and your CPU set to use all the cores minus 1 for the GPU. Integrated vega GPU is not very fast for these types of situations. If you didn't have 590, then Vega might be able to add extra PPD for you
Thank you for your contribution

[Jan]

Very interesting. I am using a XFX RX 570 8GB (same chip, different clock speed) - there have been no problems with temps or stability. I stresstested with Furmark for about 30 minutes and reached about 82° max. Didnt bother the system. Maybe you have a very sensitive one or the clock speed is the big difference. Personally I believe the latter. While folding, temps dont go beyond 73° with an absolute minimum of fan use and I have not had any issues that I could connect to the hardware itself. Only the driver problems discussed in other threads...

In any case: This should underline your approach to lower the clocking. Try it, people.

*edit* P.S.: Around 3.5hrs for a WU with the 570 at around ~1286MHz. You dont really miss out on a lot of performance lowering the clock speed it seems.

[muziqaz]

Clock speeds are not vital for folding. Shader count (CU) is

[Jan]

Well, there you go. Even more the reason.

[bignerdguy]

Yeah the only drawback with the Vega graphics is it takes forever to churn through a work unit. about 22 hours vs about 3 for three for the 590. At the moment i have it set to the chipset default of 1460 as it was making F@H too unstable any higher. Also i don't know why but the factory overclock setting was making my whole system unstable even with the fan set higher. I think there may be another component in my PC that is causing this. Right now i am waiting on my stimulus check and am going to purchase new memory and a Ryzen 7 3800x CPU for it and see what happens. I had originally bought the Ryzen 5 2400G because i didn't want to buy a new video card too. Now i am regretting that decision. With the setup as is, and clocking the 590 at 1520 MHz it runs benchmarks and everything Except F@H just fine. Like i said, one of my other hardware devices is an issue, probably the memory as it wont run reliably past DDR-2666 and was supposed to run to DDR-3000, even the MB manufacturer said they tested it at that. Either way i have new memory and other stuff already picked out, just waiting on Trump to get the lead out and pay us...

[sam6861]

My vega 11 completes in around 8 to 11 hours, at a locked 1240 MHz Vega 11 graphics core.
My AMD RX 580 is about 3 hours. with 1250 MHz core at 1.0 volts, higher fan speed.
My NVidia GT 1030 is slow, takes more like 12 - 20 hours.

Watch the integrated Vega 11 core speed with some programs like Hwinfo64. Vega 11 might slow down too much from maxed CPU usage with CPU turbo core, from power or temperature limit.
I have my motherboard BIOS settings to have integrated graphics set to 1240 MHz so it stays locked at fast speed. Locked GPU frequency allows CPU+iGPU to use more then 65 watts of power. I have CPU turbo core off for less heat, less watts usage. Using 150 TDP be quiet heatsink so CPU stays nice and cool, about 80 C, with maxed 3.6 GHz CPU and maxed 1240 Mhz Vega 11 compute 1.

I have DDR4 3200 3.5 volts. At 3200, beeps, black screen, won't start. At 3000, random crash or freeze within a few hours. Drop it to 2933 MT/s, 3.5 volts and it works and is stable. Ryzen 2400g specs shows up to DDR4 2933, so I guess memory controller in my CPU can't handle faster speed.

Ryzen 2400g, 32 GB (8x4) DDR4 at 2933 MT/s, Asus B450-F, AMD RX 580 (top PCI-E slot), NVidia GT 1030 (bottom PCI-E slot)
All 3 GPU is usable at the same time in Folding, just make sure motherboard BIOS settings have iGPU multi monitor turned on and multiple GPU are shown in windows device manager, then maybe add a GPU folding slot if both not shown. I have iGPU UMA buffer set to a low 128 MB so integrated graphics don't steal away too much system memory.

[bignerdguy]

Ryzen 2400g, 32 GB (8x4) DDR4 at 2933 MT/s, Asus B450-F, AMD RX 580 (top PCI-E slot), NVidia GT 1030 (bottom PCI-E slot)
All 3 GPU is usable at the same time in Folding, just make sure motherboard BIOS settings have iGPU multi monitor turned on and multiple GPU are shown in windows device manager, then maybe add a GPU folding slot if both not shown. I have iGPU UMA buffer set to a low 128 MB so integrated graphics don't steal away too much system memory.

I tried to enable the multiGPU function but apparently there is something not quite right with the system as it caused the PC not to boot. i had to remove the 590, clear RTC/CMOS and boot on the Vega. then enter bios and set defaults, shut down and reinstall the 590 to get it back up. I will mess with it more this weekend but for now am going to leave it alone. It works quite well as it is right now so i am happy. Plus with the Vega running it does tend to slow the CPU processing times.

[BobWilliams757]

Do you have the CPU or the Vega 11 on board overclocked?

Just thinking about possible answers, as the 2400G onboard on my system will do a work unit much faster than you are stating it was taking for you. It's no powerhouse by any means, but can do 16-17,000 points per day.

Also curious what your CPU processing numbers are, as well as the settings. That might be an indication if something is wrong, as compared to other systems. I've only just today done CPU WU's, and had my cores (threads) set to just 4 since I wanted less power draw/noise. But if you want I can run a WU on settings the same as your system to give you a direct comparison.

[muziqaz]

16-17k from Integrated GPU?
That's exactly the same as 1c/2t 3000series Ryzen

[BobWilliams757]

16-17k from Integrated GPU?
That's exactly the same as 1c/2t 3000series Ryzen

1c/2t = 1 core/2 threads?

If so I didn't know such a Ryzen existed. I know of the 3s and the 5s, but only know of 2 onboard graphics integrated types, being Vega 8 and 11. If you are speaking of the supposed upcoming low power CPU's I have heard about them, but don't know much.

Clarify? I'm not sure if your statement is intended as "that's good" or "that sucks"!

If it's the later, yeah... it's onboard graphics so not much to be expected.

[muziqaz]

I got 3900x, and since 23/22threads are no go, while one thread is left for GPU, I fold with 21threads on one slot, and second slot is for 2 threads (which is 1 core). So that slot is producing 16-17k consistently. If I pause 21thread slot, that 1 core boosts to the max, and starts producing around 25-30k. But that is rare occasion.
My statement about PPD comparison should not be taken as critique I'm purely interested from scientific point of view. How different configs work out.
Now, what you guys can take away from this, is to see if its worth an effort to make integrated GPU fold or not.
Do you just leave one thread for dedicated GPU, and the rest of the threads for CPU slot, or do you spend your time, and figure out how to make integrated GPU fold and do one slot for dedicated GPU, one for integrated GPU and one slot for CPU.
Actually when you come to think of it, getting to work integrated graphics is worth the time.
Case 1: 1 thread for Integrated GPU, 1 thread for discreet GPU, 6 threads for CPU
Case 2: 1 thread for discreet GPU, 6 threads for CPU (since -nt7 automatically drops to -nt6 aka uses 6 threads).

So Case 1 will produce 16-17k PPD more than Case 2 even if you ignore the fact that 2400G is Zen1 based and much much slower than 3000series Zen2 based CPUs.

So my conclusion is: Vega11 is no powerhouse, but welcome addition to overall PPD output

[BobWilliams757]

Well that just shows us how fast the technology grows. The 3900x is giving fairly amazing CPU throughput with folding. Are your numbers for all available cores running in line with your example of just the two cores running on a slot?

As for core/thread resources, I have few hassles. I've only run a couple of CPU WU's, and haven't really played with the settings enough to see if it would be worth creating two slots for the CPU. But in my case with so little overall resources, it probably wouldn't help much. I don't have a dedicated GPU in my system, just the onboard, so that consideration is not there for me.

Using the standard configuration and allowing the client to deal with threads, FOH uses 1 for the GPU, and 7 for the CPU. My overall throughput jumps from the 16-17K PPD range to about 21K. But it's still nice and quiet, cool, and not drawing much for power.

[muziqaz]

Well that just shows us how fast the technology grows. The 3900x is giving fairly amazing CPU throughput with folding. Are your numbers for all available cores running in line with your example of just the two cores running on a slot?

As for core/thread resources, I have few hassles. I've only run a couple of CPU WU's, and haven't really played with the settings enough to see if it would be worth creating two slots for the CPU. But in my case with so little overall resources, it probably wouldn't help much. I don't have a dedicated GPU in my system, just the onboard, so that consideration is not there for me.

Using the standard configuration and allowing the client to deal with threads, FOH uses 1 for the GPU, and 7 for the CPU. My overall throughput jumps from the 16-17K PPD range to about 21K. But it's still nice and quiet, cool, and not drawing much for power.

If you have only fairly modern integrated GPU from AMD, it is always worth a try to set that going, providing you have good cooling. In most cases modern AMD integrated GPUs are theoretically a lot more powerful than CPUs next to them.

Are your numbers for all available cores running in line with your example of just the two cores running on a slot?

I'm not following you here
but
3900x@24threads produces 390k PPD
3900x@21Threads produces 300-350k PPD + 2thread slot of 16-17k + Radeon 7 with 1m-1.5m-2m PPD

When 21 Thread slot is running alongside 2 Thread one, boost is holding at 4075Mhz (which is probably upper limit for AVX256 workloads for the CPU), but AMD boost technology has been known to get the most out of CPUs, so if you pause 21 Thread slot and leave 2 Threads to run on itself, that 1c can boost to max ~4.4ghz.
The main 3000 series advantage over 2000 series Ryzens is full AVX256 support, instead of 2x128bit AVX on 2000 series. FAH still sees that as AVX256, but it kinda runs at half the speed of full AVX256.
So lets say my 3900x@21Threads is roughly twice as fast (in PPD produced) than my 2700x@15Threads which is pretty amazing.
With current influx of COVID projects which are quite small, my 3900x is not scaling to well, but still

[bignerdguy]

Well there are a lot of reasons why thew 3000 series CPU's are faster: memory pipes, Threads, etc but i only have the 2000 series in the Ryzen 5 class. I leave the setting on 4 (1 thread for GPU, and the rest for F@H) which means it uses all but one thread for the CPU and the last is for the GPU (the Ryzen 5 2400G has 4 CPU cores w/ 8 threads). Remember the setting in the config is for CPU's not threads as in my system setting it to 4 maxes out all 4 Cores/8 threads totally. I should also mention i was wrong about how long the GPU (Vega) was taking, it took about 8-9 hours, not 22. It was also maxed out as well the entire time. The CPU would also take about as long with the Vega turned on. Now with the Vega disabled the CPU and 590 GPU complete WU's in a little over 3 hours. I have also just ordered a Ryzen 7 3800x and am waiting on amazon to deliver it. Once i get that i will install it and see how fast it runs things. I will let you all know, but for now i am leaving the Vega disabled since it isn't nearly as fast as the 590.