{"slug": "llm-adventures-with-intel-arc-pro-b60s-p2p-vllm", "title": "LLM Adventures with Intel Arc Pro B60s, P2P, VLLM", "summary": "Intel Arc Pro B60 graphics cards can run large language models effectively on a consumer platform, according to a builder who assembled an 8-card system for a fraction of the cost of a server setup. The builder used a PEX88096 PCIe switch and an 11-slot backplane from AliExpress, achieving performance that rivals a used RTX 3090 setup. The system runs Ubuntu 26.04 LTS with kernel 7.1.3 and compute-runtime 26.22.38646.4, and the builder recommends avoiding Intel's official BKC platform in favor of a more flexible configuration.", "body_md": "I’m not sure how to begin this so here’s my life story and ramblings as I try and justify my intel purchase.\n\nI feel I need to get this information out to a place that can be scraped by agents or found by future adventurers rather than lost in scattered thoughts and sporadic messages.\n\nI’ll hopefully keep updating this as I go\n\nThe build began in December 2025, an awful time to start, made do with what I had already laying around.\n\nA Maxun Z890, an intel 225F and 64gb of questionable latency DDR5.\n\nI bought two Intel B60s around Christmas just to see what it’s like, no way you can buy brand new intel cards cheaper than a used 3090 and run AI well, right? turns out you can, but you really have to work for it.\n\nI had four cards by January, and now in July a PEX88096 10-port switch, 11-slot pcie “backplane” with plans to bring my total up to 8 cards by August.\n\nPlatform\n\nNaively I began with the official intel endorsed config which was not a good time, Intel’s BKC/platform at that time was Ubuntu Server 25.04 with pinned dependencies, an offline installer and overall it was just a bad experience, I should’ve known better, they themselves would have known better, no one did better.\n\ncurrently my platform and recommendation is kernel 7.1+ and compute-runtime being 26.22.38646.4 or newer and realistically any OS where you can install the dependencies and a modern kernel.\n\n([GitHub - intel/compute-runtime: Intel® Graphics Compute Runtime for oneAPI Level Zero and OpenCL™ Driver · GitHub](https://github.com/intel/compute-runtime))\n\nif you’re lazy or starting out, with Ubuntu 26.04 LTS you can follow:\n\nWhich is relatively new documentation and was not what intel were preaching earlier this year or even what is probably still in the llm-scaler documentation, avoid that.\n\nCurrent Hardware\n\n```\nOS: Ubuntu 26.04 LTS (Resolute Raccoon) x86_64\nHost: X870 Pro-A WiFi\nKernel: Linux 7.1.3-070103-generic\nCPU: AMD Ryzen 9 9900X (24) @ 5.66 GHz\nGPU 1: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 2: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 3: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 4: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 5: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 6: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 7: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\nGPU 8: Intel Arc Pro B60 @ 2.00 GHz [Discrete]\n```\n\nI recently bought 4 more cards bringing my total to 8, it required swapping the 265k plus/Z890 for a 9900x/X870.\n\nhaving multiple cards on a consumer platform is a bad time, before buying the pcie switch it was running four cards, two at 5.0 x8 and two at 4.0 x4 across different root ports, the performance was awful and the latency was even worse.\n\nbut it ran and it was *cheap*.\n\nI eventually purchased a PEX88096 10-port pcie switch and an 11-slot 8654 to PCIE board off aliexpress, with the current price of ECC memory and server platforms it was about a 10th the price of doing things properly.\n\nexample listings of the parts I bought, I did not use these stores, this is an example:\n\n```\nhttps://www.aliexpress.com/item/1005011868399389.html  \nhttps://www.aliexpress.com/item/1005012272021905.html\n```\n\nIt’s worth noting that battlemage is pcie 5.0 and I halved my theoretical bandwidth using a 4.0 switch, however the results seem great and I can’t justify the price of 5.0 switches with the cards themselves being so cheap.\n\nAdded bonus of the B60 in particular is it’s wired x8, I only had to run a single 8654 8i cable per slot, per card, that’s at least 20 bucks of savings each.\n\nWith the particular 10 port switch I bought, it conveniently gives me the possibility of up to 10 x8 cards with dip switches on the pcie card itself for x16, x8/x8, 4*x4\n\n`xpu-smi topology -m`\n\n```\n              GPU 0/0     GPU 1/0     GPU 2/0     GPU 3/0     GPU 4/0     GPU 5/0     GPU 6/0     GPU 7/0     NIC0     CPU Affinity\n  GPU 0/0     S           PIX         PIX         PIX         PIX         PIX         PIX         PIX         PHB      0-23\n  GPU 1/0     PIX         S           PIX         PIX         PIX         PIX         PIX         PIX         PHB      0-23\n  GPU 2/0     PIX         PIX         S           PIX         PIX         PIX         PIX         PIX         PHB      0-23\n  GPU 3/0     PIX         PIX         PIX         S           PIX         PIX         PIX         PIX         PHB      0-23\n  GPU 4/0     PIX         PIX         PIX         PIX         S           PIX         PIX         PIX         PHB      0-23\n  GPU 5/0     PIX         PIX         PIX         PIX         PIX         S           PIX         PIX         PHB      0-23\n  GPU 6/0     PIX         PIX         PIX         PIX         PIX         PIX         S           PIX         PHB      0-23\n  GPU 7/0     PIX         PIX         PIX         PIX         PIX         PIX         PIX         S           PHB      0-23\n  NIC0        PHB         PHB         PHB         PHB         PHB         PHB         PHB         PHB         S        0-23\n```\n\n`lspci -tv`\n\n```\n-[0000:00]-+-00.0  Advanced Micro Devices, Inc. [AMD] Raphael/Granite Ridge Root Complex\n           +-00.2  Advanced Micro Devices, Inc. [AMD] Raphael/Granite Ridge IOMMU\n           +-01.0  Advanced Micro Devices, Inc. [AMD] Raphael/Granite Ridge Dummy Host Bridge\n           +-01.1-[01-2f]----00.0-[02-2f]--+-00.0-[03-0c]----00.0-[04-0c]--+-10.0-[05-08]----00.0-[06-08]--+-01.0-[07]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[08]----00.0  Intel Corporation Device e2f7\n           |                               |                               \\-18.0-[09-0c]----00.0-[0a-0c]--+-01.0-[0b]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                                                               \\-02.0-[0c]----00.0  Intel Corporation Device e2f7\n           |                               +-04.0-[0d-1e]----00.0-[0e-1e]--+-00.0-[0f-12]----00.0-[10-12]--+-01.0-[11]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[12]----00.0  Intel Corporation Device e2f7\n           |                               |                               +-08.0-[13-16]----00.0-[14-16]--+-01.0-[15]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[16]----00.0  Intel Corporation Device e2f7\n           |                               |                               +-10.0-[17-1a]----00.0-[18-1a]--+-01.0-[19]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[1a]----00.0  Intel Corporation Device e2f7\n           |                               |                               \\-18.0-[1b-1e]----00.0-[1c-1e]--+-01.0-[1d]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                                                               \\-02.0-[1e]----00.0  Intel Corporation Device e2f7\n           |                               +-08.0-[1f-2a]----00.0-[20-2a]--+-00.0-[21]--\n           |                               |                               +-08.0-[22-25]----00.0-[23-25]--+-01.0-[24]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[25]----00.0  Intel Corporation Device e2f7\n           |                               |                               +-10.0-[26-29]----00.0-[27-29]--+-01.0-[28]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n           |                               |                               |                               \\-02.0-[29]----00.0  Intel Corporation Device e2f7\n           |                               |                               \\-18.0-[2a]--\n           |                               +-0c.0-[2b-2e]----00.0-[2c-2e]--+-14.0-[2d]--\n           |                               |                               \\-15.0-[2e]--\n           |                               \\-1c.0-[2f]----00.0  Broadcom / LSI PEX880xx PCIe Gen 4 Switch\n```\n\n`nvtop`\n\n```\n Device 0 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 2000MHz MEM N/A MHz  TEMP  33°C FAN    0RPM POW   7 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 1 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 2000MHz MEM N/A MHz  TEMP  31°C FAN 1305RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 2 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 2000MHz MEM N/A MHz  TEMP  30°C FAN 1305RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 3 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 2000MHz MEM N/A MHz  TEMP  33°C FAN    0RPM POW   8 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 4 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  32°C FAN    0RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 5 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  32°C FAN    0RPM POW   8 / 440 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 6 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  32°C FAN    0RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 7 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  32°C FAN    0RPM POW   9 / 440 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n```\n\n##\nold intel topology:\n\nI hit the limitation of this intel consumer platform and only 6 cards will post, any more throws a D4 qcode and fails to post, MMIO space can’t be user adjusted on intel consumer platforms, at least not mine. 6/8 was the final total.\n\n`xpu-smi topology -m`\n\n```\n              GPU 0/0     GPU 1/0     GPU 2/0     GPU 3/0     GPU 4/0     GPU 5/0     NIC0     NIC1     CPU Affinity    \n  GPU 0/0     S           PIX         PIX         PIX         PIX         PIX         SYS      SYS      0-19            \n  GPU 1/0     PIX         S           PIX         PIX         PIX         PIX         SYS      SYS      0-19            \n  GPU 2/0     PIX         PIX         S           PIX         PIX         PIX         SYS      SYS      0-19            \n  GPU 3/0     PIX         PIX         PIX         S           PIX         PIX         SYS      SYS      0-19            \n  GPU 4/0     PIX         PIX         PIX         PIX         S           PIX         SYS      SYS      0-19            \n  GPU 5/0     PIX         PIX         PIX         PIX         PIX         S           SYS      SYS      0-19            \n  NIC0        SYS         SYS         SYS         SYS         SYS         SYS         S        PHB      0-19            \n  NIC1        SYS         SYS         SYS         SYS         SYS         SYS         PHB      S        0-19\n```\n\n`lspci -tv`\n\n```\n-+-[0000:00]-+-00.0  Intel Corporation Device 7d1b\n |           +-04.0  Intel Corporation Device ad03\n |           +-06.0-[01-29]----00.0-[02-29]--+-00.0-[03-0c]----00.0-[04-0c]--+-10.0-[05-08]----00.0-[06-08]--+-01.0-[07]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                               |                               \\-02.0-[08]----00.0  Intel Corporation Device e2f7\n |           |                               |                               \\-18.0-[09-0c]----00.0-[0a-0c]--+-01.0-[0b]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                                                               \\-02.0-[0c]----00.0  Intel Corporation Device e2f7\n |           |                               +-04.0-[0d-1b]----00.0-[0e-1b]--+-00.0-[0f-12]----00.0-[10-12]--+-01.0-[11]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                               |                               \\-02.0-[12]----00.0  Intel Corporation Device e2f7\n |           |                               |                               +-08.0-[13-16]----00.0-[14-16]--+-01.0-[15]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                               |                               \\-02.0-[16]----00.0  Intel Corporation Device e2f7\n |           |                               |                               +-10.0-[17-1a]----00.0-[18-1a]--+-01.0-[19]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                               |                               \\-02.0-[1a]----00.0  Intel Corporation Device e2f7\n |           |                               |                               \\-18.0-[1b]--\n |           |                               +-08.0-[1c-24]----00.0-[1d-24]--+-00.0-[1e]--\n |           |                               |                               +-08.0-[1f-22]----00.0-[20-22]--+-01.0-[21]----00.0  Intel Corporation Battlemage G21 [Arc Pro B60]\n |           |                               |                               |                               \\-02.0-[22]----00.0  Intel Corporation Device e2f7\n |           |                               |                               +-10.0-[23]--\n |           |                               |                               \\-18.0-[24]--\n |           |                               +-0c.0-[25-28]----00.0-[26-28]--+-14.0-[27]--\n |           |                               |                               \\-15.0-[28]--\n |           |                               \\-1c.0-[29]----00.0  Broadcom / LSI PEX880xx PCIe Gen 4 Switch\n |           +-06.1-[2a]----00.0  INNOGRIT Corporation NVMe SSD Controller IG5236 [RainierPC]\n |           +-07.0-[2b-54]--\n |           +-07.1-[55-7e]--\n```\n\n`nvtop`\n\n```\n Device 0 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  37°C FAN    0RPM POW   6 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 1 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  36°C FAN    0RPM POW   8 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 2 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  34°C FAN    0RPM POW  11 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 3 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  36°C FAN    0RPM POW   8 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 4 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  37°C FAN    0RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n\n Device 5 [Battlemage G21 (Arc Pro B60)] PCIe GEN 4@ 8x RX: N/A TX: N/A\n GPU 400MHz  MEM N/A MHz  TEMP  38°C FAN    0RPM POW   9 / 400 W\n GPU[                            0%(eff 0%)] MEM[                     0.039Gi/23.906Gi]\n```\n\nP2P isn’t working on this awful intel platform existence\n\nThis one was rough, I ended up having to do all the ACS in grub and escape them for some reason, confirm they’re all there after a reboot in /proc/cmdline\n\n```\nGRUB_CMDLINE_LINUX_DEFAULT=\"pci=disable_acs_redir=0000:07:01.0\\\\;0000:05:10.0\\\\;0000:03:00.0\\\\;0000:15:01.0\\\\;0000:0f:08.0\\\\;0000:03:04.0\\\\;0000:11:01.0\\\\;0000:0f:00.0\\\\;0000:0b:01.0\\\\;0000:05:18.0\"\n```\n\nthe addresses will be different, easiest way is to fire up VLLM or another backend that’ll attempt P2P and extract the system’s cries from dmesg.\n\nmay also not hurt to try `intel_iommu=off`\n\nwhile you’re at it.\n\nThere’s probably a better way to do this than manually adding all the addresses like this, but it worked.\n\nI have a feeling this just works on AMD but I had intel.\n\ntips, tricks, problems and solutions\n\nunderclocking and performance\n\nThese cards don’t draw a lot of power individually and the 220wish TDP is quite extreme for AI workloads.\n\nUnder load/decoding it’s about 100w, 150w doing batches quite aggresssively.\n\nI’ve found reducing maximum clock from 2400mhz to 2000mhz nets around a 15% (80-85w vs 100w) reduction in power for a negligible loss in single request LLM performance, haven’t tested further, should be explored.\n\nidle without working ASPM can be rough, around 35w per card.\n\nWith working ASPM and an active process such as nvtop in the backround they’re closer to 10w, without a process they end up closer to 35w again.\n\ncards don’t show up in xpu-smi discovery | dmesg complains about bar space/allocations\n\n```\nGRUB_CMDLINE_LINUX_DEFAULT=\"pci=realloc=off\"\n```\n\nturns out this is an XE quirk that I don’t care to investigate further, likely related to the XE VF passthrough mode, didn’t go away with sr-iov enabled either.\n\nthis relates to: a vbios update broke my cards from showing up in xpu-smi/xpumanager.\n\nThere’s likely performance implications to this but implications are better than a non-functioning system.\n\nP2P isn’t working AMD edition\n\nThis one was rough, I ended up having to do all the ACS in grub and escape them for some reason, confirm they’re all there after a reboot in /proc/cmdline\n\n```\nGRUB_CMDLINE_LINUX_DEFAULT=\"pci=disable_acs_redir=0000:07:01.0\\\\;0000:05:10.0\\\\;0000:03:00.0\\\\;0000:15:01.0\\\\;0000:0f:08.0\\\\;0000:03:04.0\\\\;0000:11:01.0\\\\;0000:0f:00.0\\\\;0000:0b:01.0\\\\;0000:05:18.0\"\n```\n\nthe addresses will be different, easiest way is to fire up VLLM or another backend that’ll attempt P2P and extract the system’s cries from dmesg.\n\nThere’s probably a better way to do this than manually adding all the addresses like this, but it worked.\n\nI had a feeling this was an intel quirk but it happens on AMD, I likely need to configure the switch itself but it works for now so good enough.\n\nfans don’t update when the cards are idle, temperature doesn’t update while idle\n\nFor the last 7 months this has remained a problem for me, I currently run nvtop in a screen as part of my startup jobs, however the most reliable way to update the cards is to simply probe them by running something such as `xpu-smi discovery`\n\n.\n\nyour user will require video/render groups for this to work sudoless.\n\nhigh power idle after a load, cards wedge, errors when waking cards\n\nthis one was relatively common sense but I forgot at some point during my platform migration and noticed they were idling at 31w each, fix was enabling\n\n`iommu=pt`\n\nto stop them getting wedged when P2P is enabled, I also noticed errors reported with this on/full.\n\nanother problem I had was any kind of load caused the cards to wedge and never return to a proper idle, “solution” was to disable d3cold\n\n``` bash\n#!/bin/bash\n# Find all BDFs for Intel Arc B60 (8086:e211)\nGPU_BDFS=$(lspci -d 8086:e211 -D | awk '{print $1}')\n\nif [ -z \"$GPU_BDFS\" ]; then\n  echo \"No Intel Arc GPUs (8086:e211) found.\"\n  exit 1\nfi\n\nfor bdf in $GPU_BDFS; do\n  echo \"Pinning $bdf out of D3cold...\"\n  echo 0  | sudo tee /sys/bus/pci/devices/$bdf/d3cold_allowed\n  echo on | sudo tee /sys/bus/pci/devices/$bdf/power/control\ndone\n```\n\n", "url": "https://wpnews.pro/news/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm", "canonical_source": "https://forum.level1techs.com/t/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm/252516#post_4", "published_at": "2026-08-22 05:08:09+00:00", "updated_at": "2026-08-22 05:12:56.896064+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-infrastructure"], "entities": ["Intel", "Arc Pro B60", "Maxun Z890", "AMD Ryzen 9 9900X", "PEX88096", "Ubuntu 26.04 LTS", "AliExpress"], "alternates": {"html": "https://wpnews.pro/news/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm", "markdown": "https://wpnews.pro/news/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm.md", "text": "https://wpnews.pro/news/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm.txt", "jsonld": "https://wpnews.pro/news/llm-adventures-with-intel-arc-pro-b60s-p2p-vllm.jsonld"}}