Upcoming AMD gpus and big navi

[muziqaz]

Well, on 17320,17321 and 17322, my R VIIs under win 10 have higher PPD than my Titan XPs under Linux (2.8 Mppd versus 2.5).

Titan Xp released 2017 - 16nm
Radeon VII - released 2019 - 7nm

[Skajaquada]

Now my RX6800xt runs a 17321 with 4.5 Million PPD. That is almost the performance of a 2080Ti on the same project.
I think with more atoms the RDNA2 perform better then Nvidia with CUDA.

Now now, let's not run away to fantasy land
2080ti is 3rd fastest nVidia GPU

Nvidia got their boost by using Cuda cores some month ago. The 2080Ti got 20% better performance because of this, AMD does not get this extra power.
And without Cuda Support, the RX6800xt would be faster than a RTX2080Ti on high atom projects.

On 17322 the 2080Ti is 200k in front, that is only 5%. Without Cuda Support it would be 10-15% behind.
But the RDNA2 only shines on really big projects, the normal ones are really bad

[muziqaz]

There is no sugarcoating of AMD situation. This has been the case for years and years. AMD always was great on huge projects, yet fall behind massively in normal and smaller projects.
RDNA and RDNA2 at least don't fall off the cliff with small atom counts like GCN cards

[bruce]

What has been said about AMD vs. NV (above) is true, but there's more too it than that.

The performance of both AMD and nV GPUs both vary with the size of the protein. When there are relatively few atoms, the more powerful GPUs cannot achieve the high degree of parallelism that makes them shine. When there are lots of atoms, the additional parallelism achieved by having a large number of shaders is good for FAH performance.

Small proteins are ideal for small/narrow GPUs. Big proteins are ideal for wide GPUs. FAH is working on better ways to match the protein to the GPU but we haven't gotten all the kinks out of that process yet. Also, there's not a lot we can do when the choices of foldable WUs is limited.

There's always a difference between theoretical performance and actual performance ... but that difference is not linear.

[Skajaquada]

sorry but is there something wrong with the Projects on the Radeon RX6000 GPUs? Was there any change?
Today i installed my Ryzen 5800x and activated SAM, now my RX6800xt is a lot faster then a RTX3090 on many projects.
It is from 30% to 85% faster then average PPD, the highest score was about 8.3 million PPD...
Is that the Smart Access Memory or just a bug?

[muziqaz]

sorry but is there something wrong with the Projects on the Radeon RX6000 GPUs? Was there any change?
Today i installed my Ryzen 5800x and activated SAM, now my RX6800xt is a lot faster then a RTX3090 on many projects.
It is from 30% to 85% faster then average PPD, the highest score was about 8.3 million PPD...
Is that the Smart Access Memory or just a bug?

which project
And log file would help as well

[Skajaquada]

Sorry i cleared the log by restarting the pc. This is my current WU: project 17316; TPF 59sec; estimated PPD 7.1 million

Code: Select all

*********************** Log Started 2021-01-05T20:14:10Z ***********************
20:14:10:******************************* libFAH ********************************
20:14:10:           Date: Oct 20 2020
20:14:10:           Time: 13:36:55
20:14:10:       Revision: 5ca109d295a6245e2a2f590b3d0085ad5e567aeb
20:14:10:         Branch: master
20:14:10:       Compiler: Visual C++ 2015
20:14:10:        Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Zc:throwingNew /MT
20:14:10:       Platform: win32 10
20:14:10:           Bits: 32
20:14:10:           Mode: Release
20:14:10:****************************** FAHClient ******************************
20:14:10:        Version: 7.6.21
20:14:10:         Author: Joseph Coffland <joseph@cauldrondevelopment.com>
20:14:10:      Copyright: 2020 foldingathome.org
20:14:10:       Homepage: https://foldingathome.org/
20:14:10:           Date: Oct 20 2020
20:14:10:           Time: 13:41:04
20:14:10:       Revision: 6efbf0e138e22d3963e6a291f78dcb9c6422a278
20:14:10:         Branch: master
20:14:10:       Compiler: Visual C++ 2015
20:14:10:        Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Zc:throwingNew /MT
20:14:10:       Platform: win32 10
20:14:10:           Bits: 32
20:14:10:           Mode: Release
20:14:10:           Args: --open-web-control
20:14:10:         Config: C:\ProgramData\FAHClient\config.xml
20:14:10:******************************** CBang ********************************
20:14:10:           Date: Oct 20 2020
20:14:10:           Time: 11:36:18
20:14:10:       Revision: 7e4ce85225d7eaeb775e87c31740181ca603de60
20:14:10:         Branch: master
20:14:10:       Compiler: Visual C++ 2015
20:14:10:        Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Zc:throwingNew /MT
20:14:10:       Platform: win32 10
20:14:10:           Bits: 32
20:14:10:           Mode: Release
20:14:10:******************************* System ********************************
20:14:10:            CPU: AMD Ryzen 7 5800X 8-Core Processor
20:14:10:         CPU ID: AuthenticAMD Family 25 Model 33 Stepping 0
20:14:10:           CPUs: 16
20:14:10:         Memory: 31.94GiB
20:14:10:    Free Memory: 26.81GiB
20:14:10:        Threads: WINDOWS_THREADS
20:14:10:     OS Version: 6.2
20:14:10:    Has Battery: false
20:14:10:     On Battery: false
20:14:10:     UTC Offset: 1
20:14:10:            PID: 3688
20:14:10:            CWD: C:\ProgramData\FAHClient
20:14:10:  Win32 Service: false
20:14:10:             OS: Windows 10 Enterprise
20:14:10:        OS Arch: AMD64
20:14:10:           GPUs: 1
20:14:10:          GPU 0: Bus:10 Slot:0 Func:0 AMD:6 Navi 21 [Radeon RX 6800/6800 XT /
20:14:10:                 6900 XT]
20:14:10:           CUDA: Not detected: Failed to open dynamic library 'nvcuda.dll': Das
20:14:10:                 angegebene Modul wurde nicht gefunden.
20:14:10:
20:14:10:OpenCL Device 0: Platform:0 Device:0 Bus:10 Slot:0 Compute:1.2 Driver:3188.4
20:14:10:***********************************************************************
20:14:10:<config>
20:14:10:  <!-- Folding Core -->
20:14:10:  <core-priority v='low'/>
20:14:10:
20:14:10:  <!-- Folding Slot Configuration -->
20:14:10:  <cause v='COVID_19'/>
20:14:10:
20:14:10:  <!-- Network -->
20:14:10:  <proxy v=':8080'/>
20:14:10:
20:14:10:  <!-- Slot Control -->
20:14:10:  <pause-on-start v='true'/>
20:14:10:  <power v='full'/>
20:14:10:
20:14:10:  <!-- User Information -->
20:14:10:  <passkey v='*****'/>
20:14:10:  <team v='70335'/>
20:14:10:  <user v='Skajaquada'/>
20:14:10:
20:14:10:  <!-- Folding Slots -->
20:14:10:  <slot id='1' type='GPU'>
20:14:10:    <paused v='true'/>
20:14:10:    <pci-bus v='10'/>
20:14:10:    <pci-slot v='0'/>
20:14:10:  </slot>
20:14:10:</config>
20:14:10:Trying to access database...
20:14:10:Successfully acquired database lock
20:14:10:FS01:Initialized folding slot 01: gpu:10:0 Navi 21 [Radeon RX 6800/6800 XT / 6900 XT]
20:14:17:WU00:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\ProgramData\FAHClient\cores/cores.foldingathome.org/win/64bit/22-0.0.13/Core_22.fah/FahCore_22.exe -dir 00 -suffix 01 -version 706 -lifeline 3688 -checkpoint 15 -opencl-platform 0 -opencl-device 0 -gpu-vendor amd -gpu 0 -gpu-usage 100
20:14:17:WU00:FS01:Started FahCore on PID 5444
20:14:17:WU00:FS01:Core PID:8084
20:14:17:WU00:FS01:FahCore 0x22 started
20:14:17:WU00:FS01:0x22:*********************** Log Started 2021-01-05T20:14:17Z ***********************
20:14:17:WU00:FS01:0x22:*************************** Core22 Folding@home Core ***************************
20:14:17:WU00:FS01:0x22:       Core: Core22
20:14:17:WU00:FS01:0x22:       Type: 0x22
20:14:17:WU00:FS01:0x22:    Version: 0.0.13
20:14:17:WU00:FS01:0x22:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
20:14:17:WU00:FS01:0x22:  Copyright: 2020 foldingathome.org
20:14:17:WU00:FS01:0x22:   Homepage: https://foldingathome.org/
20:14:17:WU00:FS01:0x22:       Date: Sep 19 2020
20:14:17:WU00:FS01:0x22:       Time: 02:35:58
20:14:17:WU00:FS01:0x22:   Revision: 571cf95de6de2c592c7c3ed48fcfb2e33e9ea7d3
20:14:17:WU00:FS01:0x22:     Branch: core22-0.0.13
20:14:17:WU00:FS01:0x22:   Compiler: Visual C++ 2015
20:14:17:WU00:FS01:0x22:    Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Ob3 /Zc:throwingNew /MT
20:14:17:WU00:FS01:0x22:             -DOPENMM_GIT_HASH="\"189320d0\""
20:14:17:WU00:FS01:0x22:   Platform: win32 10
20:14:17:WU00:FS01:0x22:       Bits: 64
20:14:17:WU00:FS01:0x22:       Mode: Release
20:14:17:WU00:FS01:0x22:Maintainers: John Chodera <john.chodera@choderalab.org> and Peter Eastman
20:14:17:WU00:FS01:0x22:             <peastman@stanford.edu>
20:14:17:WU00:FS01:0x22:       Args: -dir 00 -suffix 01 -version 706 -lifeline 5444 -checkpoint 15
20:14:17:WU00:FS01:0x22:             -opencl-platform 0 -opencl-device 0 -gpu-vendor amd -gpu 0
20:14:17:WU00:FS01:0x22:             -gpu-usage 100
20:14:17:WU00:FS01:0x22:************************************ libFAH ************************************
20:14:17:WU00:FS01:0x22:       Date: Sep 7 2020
20:14:17:WU00:FS01:0x22:       Time: 19:09:56
20:14:17:WU00:FS01:0x22:   Revision: 44301ed97b996b63fe736bb8073f22209cb2b603
20:14:17:WU00:FS01:0x22:     Branch: HEAD
20:14:17:WU00:FS01:0x22:   Compiler: Visual C++ 2015
20:14:17:WU00:FS01:0x22:    Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Ob3 /Zc:throwingNew /MT
20:14:17:WU00:FS01:0x22:   Platform: win32 10
20:14:17:WU00:FS01:0x22:       Bits: 64
20:14:17:WU00:FS01:0x22:       Mode: Release
20:14:17:WU00:FS01:0x22:************************************ CBang *************************************
20:14:17:WU00:FS01:0x22:       Date: Sep 7 2020
20:14:17:WU00:FS01:0x22:       Time: 19:08:30
20:14:17:WU00:FS01:0x22:   Revision: 33fcfc2b3ed2195a423606a264718e31e6b3903f
20:14:17:WU00:FS01:0x22:     Branch: HEAD
20:14:17:WU00:FS01:0x22:   Compiler: Visual C++ 2015
20:14:17:WU00:FS01:0x22:    Options: /TP /nologo /EHa /wd4297 /wd4103 /O2 /Ob3 /Zc:throwingNew /MT
20:14:17:WU00:FS01:0x22:   Platform: win32 10
20:14:17:WU00:FS01:0x22:       Bits: 64
20:14:17:WU00:FS01:0x22:       Mode: Release
20:14:17:WU00:FS01:0x22:************************************ System ************************************
20:14:17:WU00:FS01:0x22:        CPU: AMD Ryzen 7 5800X 8-Core Processor
20:14:17:WU00:FS01:0x22:     CPU ID: AuthenticAMD Family 25 Model 33 Stepping 0
20:14:17:WU00:FS01:0x22:       CPUs: 16
20:14:17:WU00:FS01:0x22:     Memory: 31.94GiB
20:14:17:WU00:FS01:0x22:Free Memory: 26.73GiB
20:14:17:WU00:FS01:0x22:    Threads: WINDOWS_THREADS
20:14:17:WU00:FS01:0x22: OS Version: 6.2
20:14:17:WU00:FS01:0x22:Has Battery: false
20:14:17:WU00:FS01:0x22: On Battery: false
20:14:17:WU00:FS01:0x22: UTC Offset: 1
20:14:17:WU00:FS01:0x22:        PID: 8084
20:14:17:WU00:FS01:0x22:        CWD: C:\ProgramData\FAHClient\work
20:14:17:WU00:FS01:0x22:************************************ OpenMM ************************************
20:14:17:WU00:FS01:0x22:   Revision: 189320d0
20:14:17:WU00:FS01:0x22:********************************************************************************
20:14:17:WU00:FS01:0x22:Project: 17316 (Run 0, Clone 2783, Gen 74)
20:14:17:WU00:FS01:0x22:Unit: 0x00000000000000000000000000000000
20:14:17:WU00:FS01:0x22:Digital signatures verified
20:14:17:WU00:FS01:0x22:Folding@home GPU Core22 Folding@home Core
20:14:17:WU00:FS01:0x22:Version 0.0.13
20:14:17:WU00:FS01:0x22:  Checkpoint write interval: 25000 steps (2%) [50 total]
20:14:17:WU00:FS01:0x22:  JSON viewer frame write interval: 12500 steps (1%) [100 total]
20:14:17:WU00:FS01:0x22:  XTC frame write interval: 250000 steps (20%) [5 total]
20:14:17:WU00:FS01:0x22:  Global context and integrator variables write interval: disabled
20:14:17:WU00:FS01:0x22:There are 3 platforms available.
20:14:17:WU00:FS01:0x22:Platform 0: Reference
20:14:17:WU00:FS01:0x22:Platform 1: CPU
20:14:17:WU00:FS01:0x22:Platform 2: OpenCL
20:14:17:WU00:FS01:0x22:  opencl-device 0 specified
20:14:25:WU00:FS01:0x22:Attempting to create OpenCL context:
20:14:25:WU00:FS01:0x22:  Configuring platform OpenCL
20:14:36:WU00:FS01:0x22:  Using OpenCL on platformId 0 and gpu 0
20:14:36:WU00:FS01:0x22:Completed 50000 out of 1250000 steps (4%)
20:15:11:Saving configuration to config.xml
20:15:11:<config>
20:15:11:  <!-- Folding Core -->
20:15:11:  <core-priority v='low'/>
20:15:11:
20:15:11:  <!-- Folding Slot Configuration -->
20:15:11:  <cause v='COVID_19'/>
20:15:11:
20:15:11:  <!-- Network -->
20:15:11:  <proxy v=':8080'/>
20:15:11:
20:15:11:  <!-- Slot Control -->
20:15:11:  <pause-on-start v='true'/>
20:15:11:  <power v='full'/>
20:15:11:
20:15:11:  <!-- User Information -->
20:15:11:  <passkey v='*****'/>
20:15:11:  <team v='70335'/>
20:15:11:  <user v='Skajaquada'/>
20:15:11:
20:15:11:  <!-- Folding Slots -->
20:15:11:  <slot id='1' type='GPU'>
20:15:11:    <pci-bus v='10'/>
20:15:11:    <pci-slot v='0'/>
20:15:11:  </slot>
20:15:11:</config>
20:15:34:WU00:FS01:0x22:Completed 62500 out of 1250000 steps (5%)
20:16:32:WU00:FS01:0x22:Completed 75000 out of 1250000 steps (6%)
20:16:32:WU00:FS01:0x22:Checkpoint completed at step 75000
20:17:30:WU00:FS01:0x22:Completed 87500 out of 1250000 steps (7%)
20:18:27:WU00:FS01:0x22:Completed 100000 out of 1250000 steps (8%)
20:18:28:WU00:FS01:0x22:Checkpoint completed at step 100000

[Neil-B]

Past logs can be found in a sub directory of the FaH software folder ... might be worth editing your post, high lighting the log then clicking on the code button above the edit box then saving

[muziqaz]

59s on p17316 should give you ~4609904.91 PPD
Check your system clock. Recently I had similar issue with windows clock, and fahclient showing some unbelievable PPD figures for CPU

[Skajaquada]

Thx a lot guys, the issue is solved.
I think it was the system clock or anything similar. The client is showing more normal PPD for the last two projects now.
And if it were real, i had to see it on the stats. But there are normal points for all projects yesterday, so it must have been a problem with the client.

[muziqaz]

As I said somewhere else: welcome to 21st century Microsoft OS

[Folgore101]

Thx a lot guys, the issue is solved.
I think it was the system clock or anything similar. The client is showing more normal PPD for the last two projects now.
And if it were real, i had to see it on the stats. But there are normal points for all projects yesterday, so it must have been a problem with the client.

Hi, how much can you do with the RX6800XT? Not being a constant value if you can point to the scaffold you see most often and the minimum and maximum.

Thanks for your patience, it's hard to find this data for AMD cards.

[MeeLee]

I wished projects would add the average percentage of double precision calculations are added.
If some GPU projects use a lot more PCIE bandwidth, the RX 5000/6000 may have a minor advantage with their access to PCIE 4.0 x16 PCIE speed access (that is, if the motherboard and CPU support that speed).

[muziqaz]

Thx a lot guys, the issue is solved.
I think it was the system clock or anything similar. The client is showing more normal PPD for the last two projects now.
And if it were real, i had to see it on the stats. But there are normal points for all projects yesterday, so it must have been a problem with the client.

Hi, how much can you do with the RX6800XT? Not being a constant value if you can point to the scaffold you see most often and the minimum and maximum.

Thanks for your patience, it's hard to find this data for AMD cards.

6000 series is very inconsistent. PPD ranges from 2.2m up to 5m

[Folgore101]

Thx a lot guys, the issue is solved.
I think it was the system clock or anything similar. The client is showing more normal PPD for the last two projects now.
And if it were real, i had to see it on the stats. But there are normal points for all projects yesterday, so it must have been a problem with the client.

Hi, how much can you do with the RX6800XT? Not being a constant value if you can point to the scaffold you see most often and the minimum and maximum.

Thanks for your patience, it's hard to find this data for AMD cards.

6000 series is very inconsistent. PPD ranges from 2.2m up to 5m

Thank you.