raspberry-pi-pcie-devices
raspberry-pi-pcie-devices copied to clipboard
geerlingguy
Test AMD Radeon AI Pro R9700
The AI Pro R9700 has 32GB of VRAM, and was introduced in 2025.
Getting the card working on Pi OS 13 "Trixie"
See: Using AMD GPUs on Raspberry Pi without recompiling Linux.
It should. I may try to find a way to get one of these to test. The RX 9700 XT works fine (see #766), and the R9700 is similar architecture, more VRAM...
Edit: I should be getting one soon ;)
Test setup:
lspci output:
0001:03:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 48 [Radeon AI PRO R9700] (rev c0) (prog-if 00 [VGA controller])
Subsystem: XFX Limited Device 9801
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 188
Region 0: Memory at 1800000000 (64-bit, prefetchable) [size=256M]
Region 2: Memory at 1810000000 (64-bit, prefetchable) [size=2M]
Region 5: Memory at 1b80000000 (32-bit, non-prefetchable) [size=512K]
Expansion ROM at 1b80080000 [disabled] [size=128K]
Capabilities: [48] Vendor Specific Information: Len=08 <?>
Capabilities: [50] Power Management version 3
Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1+,D2+,D3hot+,D3cold+)
Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [64] Express (v2) Legacy Endpoint, IntMsgNum 0
DevCap: MaxPayload 256 bytes, PhantFunc 0, Latency L0s <4us, L1 unlimited
ExtTag+ AttnBtn- AttnInd- PwrInd- RBE+ FLReset+ TEE-IO-
DevCtl: CorrErr+ NonFatalErr+ FatalErr+ UnsupReq+
RlxdOrd+ ExtTag+ PhantFunc- AuxPwr- NoSnoop+ FLReset-
MaxPayload 256 bytes, MaxReadReq 512 bytes
DevSta: CorrErr+ NonFatalErr- FatalErr- UnsupReq+ AuxPwr- TransPend-
LnkCap: Port #0, Speed 32GT/s, Width x16, ASPM L1, Exit Latency L1 <1us
ClockPM+ Surprise- LLActRep- BwNot- ASPMOptComp+
LnkCtl: ASPM L1 Enabled; RCB 64 bytes, LnkDisable- CommClk+
ExtSynch- ClockPM+ AutWidDis- BWInt- AutBWInt-
LnkSta: Speed 32GT/s, Width x16
TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
DevCap2: Completion Timeout: Range ABCD, TimeoutDis+ NROPrPrP- LTR+
10BitTagComp+ 10BitTagReq+ OBFF Not Supported, ExtFmt+ EETLPPrefix+, MaxEETLPPrefixes 1
EmergencyPowerReduction Not Supported, EmergencyPowerReductionInit-
FRS-
AtomicOpsCap: 32bit+ 64bit+ 128bitCAS-
DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-
AtomicOpsCtl: ReqEn-
IDOReq- IDOCompl- LTR+ EmergencyPowerReductionReq-
10BitTagReq- OBFF Disabled, EETLPPrefixBlk-
LnkCap2: Supported Link Speeds: 2.5-32GT/s, Crosslink- Retimer+ 2Retimers+ DRS-
LnkCtl2: Target Link Speed: 32GT/s, EnterCompliance- SpeedDis-
Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-
Compliance Preset/De-emphasis: -6dB de-emphasis, 0dB preshoot
LnkSta2: Current De-emphasis Level: -3.5dB, EqualizationComplete- EqualizationPhase1-
EqualizationPhase2- EqualizationPhase3- LinkEqualizationRequest-
Retimer- 2Retimers- CrosslinkRes: unsupported
Capabilities: [a0] MSI: Enable+ Count=1/1 Maskable- 64bit+
Address: 000000fffffff000 Data: 0009
Capabilities: [100 v1] Vendor Specific Information: ID=0001 Rev=1 Len=010 <?>
Capabilities: [150 v2] Advanced Error Reporting
UESta: DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP-
ECRC- UnsupReq- ACSViol- UncorrIntErr- BlockedTLP- AtomicOpBlocked- TLPBlockedErr-
PoisonTLPBlocked- DMWrReqBlocked- IDECheck- MisIDETLP- PCRC_CHECK- TLPXlatBlocked-
UEMsk: DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP-
ECRC- UnsupReq- ACSViol- UncorrIntErr- BlockedTLP- AtomicOpBlocked- TLPBlockedErr-
PoisonTLPBlocked- DMWrReqBlocked- IDECheck- MisIDETLP- PCRC_CHECK- TLPXlatBlocked-
UESvrt: DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt- UnxCmplt- RxOF+ MalfTLP+
ECRC- UnsupReq- ACSViol- UncorrIntErr+ BlockedTLP- AtomicOpBlocked- TLPBlockedErr-
PoisonTLPBlocked- DMWrReqBlocked- IDECheck- MisIDETLP- PCRC_CHECK- TLPXlatBlocked-
CESta: RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr- CorrIntErr- HeaderOF-
CEMsk: RxErr- BadTLP- BadDLLP- Rollover- Timeout- AdvNonFatalErr+ CorrIntErr- HeaderOF-
AERCap: First Error Pointer: 00, ECRCGenCap+ ECRCGenEn- ECRCChkCap+ ECRCChkEn-
MultHdrRecCap- MultHdrRecEn- TLPPfxPres- HdrLogCap-
HeaderLog: 00000000 00000000 00000000 00000000
Capabilities: [200 v1] Physical Resizable BAR
BAR 0: current size: 256MB, supported: 256MB 512MB 1GB 2GB 4GB 8GB 16GB 32GB
BAR 2: current size: 2MB, supported: 2MB 4MB 8MB 16MB 32MB 64MB 128MB 256MB
Capabilities: [240 v1] Power Budgeting <?>
Capabilities: [270 v1] Secondary PCI Express
LnkCtl3: LnkEquIntrruptEn- PerformEqu-
LaneErrStat: 0
Capabilities: [2a0 v1] Access Control Services
ACSCap: SrcValid- TransBlk- ReqRedir- CmpltRedir- UpstreamFwd- EgressCtrl- DirectTrans-
ACSCtl: SrcValid- TransBlk- ReqRedir- CmpltRedir- UpstreamFwd- EgressCtrl- DirectTrans-
Capabilities: [2d0 v1] Process Address Space ID (PASID)
PASIDCap: Exec+ Priv+, Max PASID Width: 10
PASIDCtl: Enable- Exec- Priv-
Capabilities: [320 v1] Latency Tolerance Reporting
Max snoop latency: 0ns
Max no snoop latency: 0ns
Capabilities: [410 v1] Physical Layer 16.0 GT/s <?>
Capabilities: [450 v1] Lane Margining at the Receiver
PortCap: Uses Driver-
PortSta: MargReady+ MargSoftReady-
Capabilities: [500 v1] Physical Layer 32.0 GT/s <?>
Kernel driver in use: amdgpu
Kernel modules: amdgpu
And thanks to Micro Center for helping me on this project — I will be working on some special tests for next month ;)
But I'll continue testing in this issue...
A few benchmarks:
vkmark
WARNING: radv is not a conformant Vulkan implementation, testing use only.
WARNING: radv is not a conformant Vulkan implementation, testing use only.
=======================================================
vkmark 2025.01
=======================================================
Vendor ID: 0x1002
Device ID: 0x7551
Device Name: AMD Radeon Graphics (RADV GFX1201)
Driver Version: 104857607
Device UUID: c284624a6065d2a279721dfe5e3867cf
=======================================================
[vertex] device-local=true: FPS: 17396 FrameTime: 0.057 ms
[vertex] device-local=false: FPS: 757 FrameTime: 1.321 ms
[texture] anisotropy=0: FPS: 17329 FrameTime: 0.058 ms
[texture] anisotropy=16: FPS: 17482 FrameTime: 0.057 ms
[shading] shading=gouraud: FPS: 17510 FrameTime: 0.057 ms
[shading] shading=blinn-phong-inf: FPS: 17402 FrameTime: 0.057 ms
[shading] shading=phong: FPS: 17438 FrameTime: 0.057 ms
[shading] shading=cel: FPS: 17336 FrameTime: 0.058 ms
[effect2d] kernel=edge: FPS: 17602 FrameTime: 0.057 ms
[effect2d] kernel=blur: FPS: 17685 FrameTime: 0.057 ms
[desktop] <default>: FPS: 15897 FrameTime: 0.063 ms
[cube] <default>: FPS: 17477 FrameTime: 0.057 ms
[clear] <default>: FPS: 14055 FrameTime: 0.071 ms
=======================================================
vkmark Score: 15797
=======================================================
glmark2-es2-wayland
=======================================================
glmark2 2023.01
=======================================================
OpenGL Information
GL_VENDOR: AMD
GL_RENDERER: AMD Radeon Graphics (radeonsi, gfx1201, ACO, DRM 3.64, 6.17.5-v8-16k+)
GL_VERSION: OpenGL ES 3.2 Mesa 25.0.7-2+rpt3
Surface Config: buf=32 r=8 g=8 b=8 a=8 depth=24 stencil=0 samples=0
Surface Size: 800x600 windowed
=======================================================
[build] use-vbo=false: FPS: 941 FrameTime: 1.063 ms
[build] use-vbo=true: FPS: 9289 FrameTime: 0.108 ms
[texture] texture-filter=nearest: FPS: 10564 FrameTime: 0.095 ms
[texture] texture-filter=linear: FPS: 9515 FrameTime: 0.105 ms
[texture] texture-filter=mipmap: FPS: 10452 FrameTime: 0.096 ms
[shading] shading=gouraud: FPS: 10930 FrameTime: 0.091 ms
[shading] shading=blinn-phong-inf: FPS: 10482 FrameTime: 0.095 ms
[shading] shading=phong: FPS: 9220 FrameTime: 0.108 ms
[shading] shading=cel: FPS: 9452 FrameTime: 0.106 ms
[bump] bump-render=high-poly: FPS: 10325 FrameTime: 0.097 ms
[bump] bump-render=normals: FPS: 10312 FrameTime: 0.097 ms
[bump] bump-render=height: FPS: 10478 FrameTime: 0.095 ms
[effect2d] kernel=0,1,0;1,-4,1;0,1,0;: FPS: 10439 FrameTime: 0.096 ms
[effect2d] kernel=1,1,1,1,1;1,1,1,1,1;1,1,1,1,1;: FPS: 10355 FrameTime: 0.097 ms
[pulsar] light=false:quads=5:texture=false: FPS: 9750 FrameTime: 0.103 ms
[desktop] blur-radius=5:effect=blur:passes=1:separable=true:windows=4: FPS: 6129 FrameTime: 0.163 ms
[desktop] effect=shadow:windows=4: FPS: 6814 FrameTime: 0.147 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 380 FrameTime: 2.636 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=subdata: FPS: 593 FrameTime: 1.687 ms
[buffer] columns=200:interleave=true:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 405 FrameTime: 2.473 ms
[ideas] speed=duration: FPS: 4343 FrameTime: 0.230 ms
[jellyfish] <default>: FPS: 8772 FrameTime: 0.114 ms
[terrain] <default>: FPS: 4226 FrameTime: 0.237 ms
[shadow] <default>: FPS: 8124 FrameTime: 0.123 ms
[refract] <default>: FPS: 7164 FrameTime: 0.140 ms
[conditionals] fragment-steps=0:vertex-steps=0: FPS: 10835 FrameTime: 0.092 ms
[conditionals] fragment-steps=5:vertex-steps=0: FPS: 10703 FrameTime: 0.093 ms
[conditionals] fragment-steps=0:vertex-steps=5: FPS: 10293 FrameTime: 0.097 ms
[function] fragment-complexity=low:fragment-steps=5: FPS: 10461 FrameTime: 0.096 ms
[function] fragment-complexity=medium:fragment-steps=5: FPS: 10697 FrameTime: 0.093 ms
[loop] fragment-loop=false:fragment-steps=5:vertex-steps=5: FPS: 10472 FrameTime: 0.095 ms
[loop] fragment-steps=5:fragment-uniform=false:vertex-steps=5: FPS: 9871 FrameTime: 0.101 ms
[loop] fragment-steps=5:fragment-uniform=true:vertex-steps=5: FPS: 10506 FrameTime: 0.095 ms
=======================================================
glmark2 Score: 8280
=======================================================
GravityMark
Result: 58,587
And here's a video of it running, sucking down quite the power on this card! (Lots of variations of coil whine through the different scenes):
https://github.com/user-attachments/assets/fd1bfa12-3654-4650-a48f-47ad18c49797
I just noticed, the link status is reporting:
LnkSta: Speed 32GT/s, Width x16
And in nvtop, I see Gen 2 x1...
I have Gen 3 configured in /boot/firmware/config.txt, and the bridge itself is saying 8 GT/sec:
pi@cm5:~ $ sudo lspci -vv | grep -E 'PCI bridge|LnkCap'
0001:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries BCM2712 PCIe Bridge (rev 30) (prog-if 00 [Normal decode])
LnkCap: Port #0, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <2us
LnkCap2: Supported Link Speeds: 2.5-8GT/s, Crosslink- Retimer- 2Retimers- DRS+
0001:01:00.0 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Upstream Port of PCI Express Switch (rev 24) (prog-if 00 [Normal decode])
LnkCap: Port #0, Speed 32GT/s, Width x16, ASPM L1, Exit Latency L1 <64us
LnkCap2: Supported Link Speeds: 2.5-32GT/s, Crosslink- Retimer+ 2Retimers+ DRS-
Weird.
Running some AI Benchmarks here: https://github.com/geerlingguy/ai-benchmarks/issues/35
Power draw goes between 18.2W at idle (CM5 + card in eGPU dock, measured by wall outlet) to 289.5W at full tilt:
Testing the same card on an Ubuntu 25.10 install on kernel 6.17.x on an Intel Core Ultra 265K platform (see https://github.com/geerlingguy/sbc-reviews/issues/93):
- vkmark: 40285
- glmark2-es2-wayland: 21145
- GravityMark: 64,469
AI results for both cards: https://github.com/geerlingguy/ai-benchmarks/issues/35
The card has been quite stable. Interesting seeing the difference running it in a computer with very little limitation vs. on the Pi 5...
I want to test GPU-accelerated video transcoding (like I did with the 4070 Ti) with this card and h264_amf...
With the defaults using encoder-benchmark, I'm seeing:
Input #0, yuv4mpegpipe, from '720-60.y4m':
Duration: 00:00:30.00, start: 0.000000, bitrate: 663554 kb/s
Stream #0:0: Video: rawvideo (I420 / 0x30323449), yuv420p(tv, progressive), 1280x720, 60 fps, 60 tbr, 60 tbn
[vost#0:0 @ 0x5556232913c0] Unknown encoder 'h264_amf'
[vost#0:0 @ 0x5556232913c0] Error selecting an encoder
Error opening output file -.
Error opening output files: Encoder not found
So checking on which encoders are available, I see:
V....D libx264 libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (codec h264)
V....D libx264rgb libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 RGB (codec h264)
V....D h264_nvenc NVIDIA NVENC H.264 encoder (codec h264)
V..... h264_v4l2m2m V4L2 mem2mem H.264 encoder wrapper (codec h264)
V....D h264_vaapi H.264/AVC (VAAPI) (codec h264)
V....D h264_vulkan H.264/AVC (Vulkan) (codec h264)
Hmm... would maybe be best to test with vaapi as I think that's what the amdgpu driver supports?
I'll try running ffmpeg directly, instead, since VAAPI encoder support may be a future feature:
time ffmpeg -init_hw_device vaapi=foo:/dev/dri/renderD129 -i 720-60.y4m -filter_hw_device foo -vf 'format=nv12,hwupload' -c:v h264_vaapi -pix_fmt yuv420p -movflags +faststart 720-60.mp4 && \
time ffmpeg -init_hw_device vaapi=foo:/dev/dri/renderD129 -i 1080-60.y4m -filter_hw_device foo -vf 'format=nv12,hwupload' -c:v h264_vaapi -pix_fmt yuv420p -movflags +faststart 1080-60.mp4 && \
time ffmpeg -init_hw_device vaapi=foo:/dev/dri/renderD129 -i 4k-60.y4m -filter_hw_device foo -vf 'format=nv12,hwupload' -c:v h264_vaapi -pix_fmt yuv420p -movflags +faststart 4k-60.mp4
| Video file | File size | Time (sec) | Average fps |
|---|---|---|---|
| 720-60.y4m | 2.4G | 7 | 260 |
| 1080-60.y4m | 5.3G | 17 | 109 |
| 4k-60.y4m | 11G | 73 | 27 |
The card was using around 125W during transcoding:
I would expect the numbers to be substantially higher on the Intel PC as well, just like with the 4070 Ti...