Folding Forum

Juggy wrote:What settings do we use or simply leave it as-is and run for 5 minutes?

I am not sure, I'm at that same place right now. For now, I am going to run it with the defaults for 1 minute.

Foldy, do you know if there would be any benefit to running it for longer?

Juggy wrote:What settings do we use or simply leave it as-is and run for 5 minutes?

1min is enough for FAHbench to stabilize. (But if GPU gets hotter and temp limited then after more minutes GPU could get downclocked and slower.)

For GPUs be sure to use OpenCL, mixed precision and WU real.
Only real WU will also put enough load e.g. on pcie bus bandwidth. It has 65000 atoms count.

If you like you can also add a work unit with low atom count (25000) and a work unit with high atom count (150000).
AMD GPUs like Vega or RX 5x00 and NVidia GPUs like RTX 2080(ti) need work unit with high atom count to get fully used. Because they have so many shaders.

On Windows fast GPUs need pcie 3.0 x4 to not bottleneck, pcie 3.0 x8 is nearly optimum.
On Linux fast GPUs can also be run on pcie 3.0 x1 risers with only small performance loss.

Thanks foldy

Gigabyte RTX 2060 Super Gaming OC Windforce
PCIe 3.0 @ 16x
Windows 10 1909

Device - Geforce RTX 2060 Super
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 86.0815
Scaled Score - 259.764
Atoms - 64614

NoMoreQuarantine wrote:I assume that you meant the 5700 XT vs 2060 Super discussion. Even if I take the single precision performance listed by AMD and NVIDIA for those GPUs, the 5700 XT is still higher. 5700 XT @ 9.75 TFLOPs > 2060 Super @ 7.2 TFLOPS.

That's why I wrote that driver implementation of NV is better than AMD's.
And my experience is that NV can be better OC and UV.

NoMoreQuarantine wrote:... An alternative would be to run FAHBench at AMD/NVIDIA spec frequencies on cards with max VRAM, ideally on the same system or as close as possible.

VRAM doesn't matter for folding.

MSI GTX 1080ti gaming X on Windows 7 at pcie 2.0 x16 with 65% power limit

Device - Geforce GTX 1080ti
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 82.4945
Scaled Score - 248.939

Mobo: Asus Prime B350-Plus
Chipset: B350 Rev 51
BIOS: American Megatrends 4207 - AMD AGESA PinnaclePI-AM4 1.0.0.6
CPU: AMD Ryzen 5 2600 @ 3.4 GHz
RAM: 16 GB DDR4 @ 1067 MHz

Video Card: Asus 1002 67DF - 1043 056A
Bus: PCIe x16 3.0
GPU: AMD Radeon RX 590 @ 1545 MHz
VRAM: 8192 MB GDDR5 @ 2000 MHz
BIOS: 015.050.003.000
Driver: 19.50.29.20-200401a-353575E-RadeonSoftwareAdrenalin2020

OS: Windows 10 Home 64-bit 1909
FAHBench: 2.3.2 (listed as 2.3.1 in Help) OpenMM 7.4.1

Ambient temp: 23°C

Device: Ellesmere
Compute: OpenCL
Precision: Mixed
WU: Real
Accuracy Check: enabled
NaN Check: Disabled
Run length: 1 minute

Score: 41.8147
Scaled Score: 126.182
Atoms: 64614

Reran for 10 min and measured peak temp of 67°C with no thermal throttling. Foldy, it looks like the version shown in the Help > About window is still 2.3.1. Also, any recommendations for CPU benchmarking?

Device - Geforce GTX 1080ti
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled

AORUS GTX 1080ti on Win10 at pcie 3.0 x16, (AORUS stock clock)
COLD
Run length - 1 minute
test 1: Score - 92.5514 - Scaled Score - 279.288
test 2: Score - 93.7720 - Scaled Score - 282.971 - GPU Power Consumption 242W average according to GPU-Z

Run length - 15min
test 1: Score - 85.0094 - Scaled Score - 256-529
test 2: Score - 91.7769 - Scaled Score - 276.950 - GPU Power Consumption 243W average according to GPU-Z

___________________________
AORUS GTX 1080ti on Win10 at pcie 3.0 x16, (75% power limit)
COLD
Run length - 1min
test 1: Score - 84.5036 - Scaled Score - 255.002
test 2: Score - 86.9930 - Scaled Score - 262.514 - GPU Power Consumption 181W average according to GPU-Z

Run length - 15min
test 1: Score - 79.2501 - Scaled Score - 239.149
test 2: Score - 85.5493 - Scaled Score - 258.158 - GPU Power Consumption 181W average according to GPU-Z

EDIT: I added run lenght for information
EDIT2: added more test results
--> Scores are sometimes not consistent

REMARK: temp influence
I assume that initial temperature before test is important.
During 1min test, GPU temps rise from 25°C to 44°C
During 15min test, GPU temps rise from 25°C to 76°C. Temp saturation takes around 10 to 15min

REMARK: 'efficiency'
Comparing the stock clocks with 50% power limit scores for the 1min run:
Stock: 0,385 / Watts
50% power limited: 0,470 / Watts

@NoMoreQuarantine: Good catch Help>About string was not updated But it is latest source version.

You can run FahBench on CPU too if you switch "Compute" from OpenCL to CPU. (But the real FahCore_a7 runs with gromacs engine)

foldy wrote:@NoMoreQuarantine: Good catch Help>About string was not updated But it is latest source version.

You can run FahBench on CPU too if you switch "Compute" from OpenCL to CPU. (But the real FahCore_a7 runs with gromacs engine)

Thanks foldy. So we would need to add Gromacs to FAHBench for accurate benchmarking results. I am going to focus on researching and learning a database implementation for now, but eventually I will need to look at FAHBench to see if we can add Gromacs and maybe database integration (I might be getting ahead of myself).

Edit: For now though, it's Easter Sunday and I feel like I'm still at work; I'm going to give myself a break for a little while. Happy Easter everyone!

1070 on Windows 10

Score 62
Scaled Score 189.488
Atoms 64614

Actual ppd on Core22 is around 850Kppd

EVGA RTX 2080 XC Black
PCIe 3.0 @ 16x
Windows 10 PRO 10.0.18363 Build 18363
Video Driver - 445.75

Device - Geforce RTX 2080
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 103.568
Scaled Score - 312.533
Atoms - 64614

_________________________________

EVGA RTX 2080 XC Black Over Clocked
PCIe 3.0 @ 16x
Windows 10 PRO 10.0.18363 Build 18363
Video Driver - 445.75
Clock - 2025
Memory - 8000
Voltage - 1050 mv
Gpu Temp - 52 C

Device - Geforce RTX 2080
Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 113.15
Scaled Score - 341.446
Atoms - 64614

foldy wrote:

Juggy wrote:
On Windows fast GPUs need pcie 3.0 x4 to not bottleneck, pcie 3.0 x8 is nearly optimum.
On Linux fast GPUs can also be run on pcie 3.0 x1 risers with only small performance loss.

Core 22 and a 2080ti do saturate an x1 slot in Linux. Even an x4 slot is almost saturated (80%).
Windoze should run an x4 slot completely saturated.
At this point, the best hardware would be a dual core Celeron pushing 2 RTX GPUs, as no motherboard supports more than 2x pcie3.0 8x speed slots.

With pcie 4.0 on the horizon, one could run 3 or 4 GPUs on a $120 motherboard, though my experience with current hardware is, that it's very hard to find motherboards even supporting more than 2 Nvidia GPUs, (or 3 AMD GPUs) even if they have 3 full size slots and 3 or so x1 slots.
Something about manufacturers neglecting to program addresses for more than 2 or 3 GPUs in the bios.

Yep, I picked up two gtx 970 for $80 each and stuck them in an old Linux box. Not the greatest purchase of you're looking to fold and game, but just as a spare folding machine, it was the cheapest option I could find.

HaloJones wrote:I think you also have to take current prices more into consideration.

A 2080TI is $1200 give or take. And it produces >2m ppd.
A 1070 is $200 on ebay and produces 0.8mppd.

Then you have to think about power.

The 2080TI draws 260W, the 1070 draws 140W. If you're running a 2080TI in a hot country you may need to up your air conditioning using even more power.

It's awesome that people with 2080TI's are folding. But if you're looking to build a rig just for folding and you're not made of money, a 2060S or a 1070 running Linux would be the way to go. Get a used motherboard/cpu combo off ebay and don't bother using the cpu for folding.

Mobo: Gigabyte X570 Aorus Elite
Chipset: X570 rev.51
BIOS: AF11 - AMD AGESA Combo-AM4 1.0.0.4
CPU: AMD Ryzen 9 3900X @ 4.0 GHz
RAM: 32 GB DDR4(3600) @ 1800 MHz

ASUS ROG Strix RX5700XT OC 8G Gaming
PCIe 4.0 @ 16x
Windows 10 Pro Build 1909
Video Driver - 20.3.1
Clock - 2050
Memory - 3500
Gpu Temp - 58 C

Compute - OpenCL
Precision - Mixed
WU - Real
Accuracy check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 73.1214
Scaled Score - 220.655
Atoms - 64614

foldy wrote:@NoMoreQuarantine: Good catch Help>About string was not updated But it is latest source version.

You can run FahBench on CPU too if you switch "Compute" from OpenCL to CPU. (But the real FahCore_a7 runs with gromacs engine)

CPU benchmark

CPU - Intel Core i9 9900KF @ 4.9GHz
Motherboard - MSI Z390M Gaming AC mATX
Ram - 64GB @1600MHz (4 x 16GB G-Skill Ripjaw F43200C16-16GVK)

Compute - CPU
Precision - Mixed
WU - Real
Accuracy Check enabled
NaN Check - Disabled
Run length - 1 minute

Score - 1.39858
Scaled Score - 4.22041
Atoms - 64614