<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Thomas Website</title><description>Thomas&apos;s personal website</description><link>https://workfor.live/</link><language>en</language><item><title>Aoostar WTR Max Increase VM GPU VRAM</title><link>https://workfor.live/posts/wtrmax-increase-gpu-vram/</link><guid isPermaLink="true">https://workfor.live/posts/wtrmax-increase-gpu-vram/</guid><description>Increase the GPU VRAM inside a Proxmox Ubuntu VM, with Aoostar WTR MAX 8845HS 780M iGPU</description><pubDate>Fri, 23 Jan 2026 00:00:00 GMT</pubDate><content:encoded>&lt;h1&gt;Issue&lt;/h1&gt;
&lt;p&gt;AMD 8845HS internal GPU can use RAM as VRAM, but the linux kernel by default will only allocate &lt;strong&gt;50%&lt;/strong&gt; of total RAM due to safety concern.&lt;/p&gt;
&lt;p&gt;This is too conservative and a waste of RAM &lt;em&gt;just&lt;/em&gt; sitting there.&lt;/p&gt;
&lt;p&gt;We would like to find a balance between &lt;strong&gt;system stability&lt;/strong&gt; and &lt;strong&gt;utilization&lt;/strong&gt; of the RAM.&lt;/p&gt;
&lt;h1&gt;Before We Start&lt;/h1&gt;
&lt;p&gt;&lt;img src=&quot;./images/pve-1.jpg&quot; alt=&quot;pve vm view&quot; /&gt;&lt;/p&gt;
&lt;p&gt;Calculate maximum RAM you want to used as VRAM&lt;/p&gt;
&lt;p&gt;Typically we keep &lt;strong&gt;8GB&lt;/strong&gt; RAM as system RAM to keep it stable, you can adjust based on your needs.&lt;/p&gt;
&lt;p&gt;So in this case&lt;/p&gt;
&lt;p&gt;$$64GB - 8GB = 56GB$$&lt;/p&gt;
&lt;h1&gt;Execution&lt;/h1&gt;
&lt;h2&gt;Calculate setting value&lt;/h2&gt;
&lt;p&gt;$$([size~in~GB] * 1024 * 1024) / 4.096$$&lt;/p&gt;
&lt;h2&gt;Update grub&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;open grub in editor&lt;/li&gt;
&lt;/ol&gt;
&lt;pre&gt;&lt;code&gt;sudo nano /etc/default/grub
&lt;/code&gt;&lt;/pre&gt;
&lt;ol&gt;
&lt;li&gt;update &lt;code&gt;GRUB_CMDLINE_LINUX_DEFAULT&lt;/code&gt;&lt;/li&gt;
&lt;/ol&gt;
&lt;pre&gt;&lt;code&gt;GRUB_CMDLINE_LINUX_DEFAULT=&quot;amdttm.pages_limit=&amp;lt;setting_value&amp;gt; amdttm.page_pool_size=&amp;lt;setting_value&amp;gt;&quot;
&lt;/code&gt;&lt;/pre&gt;
&lt;ol&gt;
&lt;li&gt;update grub&lt;/li&gt;
&lt;/ol&gt;
&lt;pre&gt;&lt;code&gt;sudo update-grub
&lt;/code&gt;&lt;/pre&gt;
&lt;ol&gt;
&lt;li&gt;reboot machine&lt;/li&gt;
&lt;/ol&gt;
&lt;pre&gt;&lt;code&gt;sudo reboot now
&lt;/code&gt;&lt;/pre&gt;
&lt;ol&gt;
&lt;li&gt;validate VRAM&lt;/li&gt;
&lt;/ol&gt;
&lt;pre&gt;&lt;code&gt;sudo dmesg | grep -E &quot;amdgpu: .*VRAM|amdgpu: .*GTT&quot;
&lt;/code&gt;&lt;/pre&gt;
&lt;p&gt;You should see something like this (my dmesg log has be cleared out, use &lt;code&gt;journalctl&lt;/code&gt; instead)
&lt;img src=&quot;./images/vm-1.jpg&quot; alt=&quot;validation&quot; /&gt;&lt;/p&gt;
&lt;h1&gt;Final&lt;/h1&gt;
&lt;p&gt;Now you have&lt;/p&gt;
&lt;p&gt;$$ 2GB + 58GB = 60GB$$&lt;/p&gt;
&lt;p&gt;of usable VRAM&lt;/p&gt;
&lt;p&gt;you can use it to run large LLM model, or run multiple LLM model, etc&lt;/p&gt;
&lt;h1&gt;Reference&lt;/h1&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href=&quot;https://www.jeffgeerling.com/blog/2025/increasing-vram-allocation-on-amd-ai-apus-under-linux/&quot;&gt;Increasing the VRAM allocation on AMD AI APUs under Linux&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
</content:encoded></item></channel></rss>