Did you know that each time you installed a Geforce driver update the old files get left behind on your system? This phenomenon has existed for years. However, with the introduction of the NVIDIA Geforce Experience it has gotten even worse. There are now three locations that get filled with unused files. GameplayInside shows you how to clean up old NVIDIA driver files to save disk space!

2017 update: This guide was originally released in 2014. Since then Nvidia introduced a major update of the Geforce Experience software. Unfortunately new Geforce Experiences still clutters your storage device, it just uses different file paths. Therefore this update was created.

Video tutorial

A video tutorial will be placed here later this week.

Location 1 – installer extraction directory

What is it?

When you run the NVIDIA Geforce Driver installer it will extract all files to a temporary directory. Each driver version has it’s own sub folder. Reinstalling the same version just overwrites the same directory.

Screenshot of the nvidia temporary driver installation folder.

The impact: 5.2 GB

As you can see I currently have 5214 MB of installation files on my C drive. It is actually a Samsung 850 Evo 500 GB so every MB does count!

File size of the temporary Nvidia driver installation folder

Removal instructions


All of these files and folders files are 100% safe to remove. They are only used during installation of the Nvidia graphics driver. To remove them and save disk space simply navigate to:


Now delete all folders inside it and you’re done!

Location 2 – Geforce Experience Download folder

What is it?

Geforce Experience also keeps a copy of all drivers it has downloaded automatically.  If you decide to install the driver it will still keep a copy forever. These copies are not cleaned and will stay on your system until you take action. Eventually you will end up with dozens of useless files because Nvidia releases updates monthly.

The impact: 3.7 GB

In this example there are 3721 MB of unnecessary files.

Removal instructions

This is one of the directories that was changed in GFE 3.x. To remove these files and save disk space navigate to:

C:\ProgramData\NVIDIA Corporation\Downloader

Most subfolders contain installer executables. Simply remove all folders with a random name, only leave config and latest. For example on my system the folder  0cfd1195e705a478237a4db99f7ce77c  contains GeForce_Experience_Update_v3.4.0.70.exe.

Location 3 – Geforce Experience driver installation repository

What is it?

With the introduction of Geforce Experience NVIDIA has decided it is a good idea to build a library of files. In theory these files can be used in a roll-back and SHOULD be deleted when uninstalling or updating your driver. However in reality you usually update your driver without the “perform a clean install” checkbox.

So guess what? You get a ton of files inside the actual driver installation directory that are completely useless!

The impact: 4.4 GB

The Installer2 folder is by far the largest folder in the Geforce Experience installation directory.

In this example I have 4445 MB of useless files collecting dust in the Installer2 folder.

Removal instructions

To remove these files and save disk space simply navigate to:

C:\Program Files\NVIDIA Corporation\Installer2

Now delete all folders inside this folder. Do not delete the Installer2 folder itself.

Disclaimer: Removing these files means you can no longer uninstall the Geforce Experience regularly. However, do not worry. If you ever need to uninstall Geforce Experience simply download the latest geforce experience setup and re-install the latest version. Hereafter you can uninstall the Geforce Experience as if nothing happened.

Bonus: Windows driver repository (14.4 GB)

The final location is not something Nvidia can be blamed for. Ever since Windows vista Microsoft started to keep a copy of every driver that was ever installed by the system. This caused the annoying phenomenon that  after a few years your 20 GB windows install was suddenly 60 GB. In those times most of us had no clue what was happening, we simply started to do a yearly reinstall of Windows. Microsoft kept using this system in Windows 7, 8 and 10. So today the same principle applies.

Impact: 14.4 GB

Do not delete these folders! Deleting them will mess up your OS. If you want to check your impact then you can find the driver store at


Again do not delete these folders!

Removal instructions



In 2014 I managed to reclaim 4 GB of storage using my own guide: 1726 MB from the Geforce Experience Installer2 directory, 1002 MB from the Netservice directory and 1287 MB from the temporary Nvidia folder.

A few years have passed and I have been using Windows 10 for about a year. Today I was able to reclaim a whopping 13.3 GB. Soon I will also clean the driver repository which will boost my personal savings to over 27 GB!

Conclusion: cleaning up after Nvidia drivers is now more important than even. Especially now that SSD storage is becoming more expensive due to NAND chip shortages.

Cleaning up old NVIDIA driver files的更多相关文章

  1. linux nVidia driver 304 319 . installation by hand

    It's so painful to install nVidia driver by hand on linux. If you remove it or you want to upgrade b ...

  2. Install Nvidia driver 367.18 or later

    Install Nvidia driver 367.18 or later from ppa:graphics-drivers/ppa as follows: sudo add-apt-reposit ...

  3. nvidia-smi failed because it couldn't communicate with the nvidia driver

    Ubuntu装好CUDA之后过段时间提示NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. NV ...

  4. Ubuntu 18.04 nvidia driver 390.48 安装 TensorFlow 1.12.0 和 PyTorch 1.0.0 详细教程

    最近要在个人台式机上搭建TensorFlow和PyTorch运行环境,期间遇到了一些问题.这里就把解决的过程记录下来,同时也可以作为安装上述环境的过程记录. 如果没有遇到类似的问题,想直接从零安装上述 ...

  5. 【linux基础err】NVIDIA-SMI has failed because it could't communicate with the NVIDIA driver.

    问题 安装nvidia driver和cuda关机重启之后出现不能进入系统的问题,进入命令行模式使用nvidia-smi检查驱动的问题. nvidia-smi NVIDIA-SMI has faile ...

  6. Ubuntu 16.04 Install NVidia Driver (download from nvidia official site)

    sudo apt-get update sudo apt-mark hold libreoffice sudo apt-get update && sudo apt-get upgra ...

  7. 查看显卡报错:NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

    当输入nvidia-smi时出现 NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make ...

  8. ubuntu 16.04 +anaconda3.6 +Nvidia DRIVER 390.77 +CUDA9.0 +cudnn7.0.4+tensorflow1.5.0+neural-style

    这是我第一个人工智能实验.虽然原理不是很懂,但是觉得深度学习真的很有趣.教程如下. Table of Contents 配置 时间轴 前期准备工作 anaconda3 安装 bug 1:conda:未 ...

  9. centos7 intall nvidia driver

    此教程是介绍于 CentOS 7 以上的 Linux 系统中安装 NVIDIA 显卡驱动和 CUDA Toolkit .此文中以 CentOS 7.4 64 bit 为例,显卡型号为 NVIDIA T ...


  1. 探测.yml

    liveness.yml #探测apiVersion: v1kind: Podmetadata: labels: test: liveness name: livenessspec: restartP ...

  2. NOIp2017D2T2(luogu3959) 宝藏 (状压dp)

    时隔多年终于把这道题锅过了 数据范围显然用搜索剪枝状压dp. 可以记还有哪些点没到(或者已到了哪些点).我们最深已到的是哪些点.这些点的深度是多少,然后一层一层地往下推. 但其实是没必要记最深的那一层 ...

  3. ZABBIX 3.4 监控服务器TCP连接状态(六)

    TCP的连接状态对于我们web服务器来说是至关重要的,尤其是并发量ESTAB:或者是syn_recv值,假如这个值比较大的话我们可以认为是不是受到了攻击,或是是time_wait值比较高的话,我们要考 ...

  4. 使用kubeadm部署kubernetes1.9.1+coredns+kube-router(ipvs)高可用集群

    由于之前已经写了两篇部署kubernetes的文章,整个过程基本一致,所以这篇只着重说一下coredns和kube-router的部署. kube version: 1.9.1 docker vers ...

  5. [HEOI2013]SAO ——计数问题

    题目大意: Welcome to SAO ( Strange and Abnormal Online).这是一个 VR MMORPG, 含有 n 个关卡.但是,挑战不同关卡的顺序是一个很大的问题. 有 ...

  6. 牛客练习赛29 F 算式子

    https://www.nowcoder.com/acm/contest/211/F 经典题. 1.分区间 2.向下取整的值变化 & 合并相同值 #include <bits/stdc+ ...

  7. plink, vcftool计算等位基因频率(allele frequency,vcf)

    计算等位基因频率有两种方式,第一种用vcftool计算: /path/to/vcftools --vcf file.vcf --freq --chr 1 --out filefreq 很简单的一个命令 ...

  8. 第二节,TensorFlow 使用前馈神经网络实现手写数字识别

    一 感知器 感知器学习笔记:https://blog.csdn.net/liyuanbhu/article/details/51622695 感知器(Perceptron)是二分类的线性分类模型,其输 ...

  9. 反射attr以及模块动态导入

    一.实现自省的四个函数 1.hasattr判断一个对象中有没有一个name字符串对应的方法或属性 class BlackMedium: feture="Ugly" def __in ...

  10. HTML格式化标签

    除了div.p.h1~h6.a.span这几个极常用的标签外,HTML还有一些不常见的标签(10个,5对:加粗.斜体.大小.上下标.特殊),默认效果如下: 当然,我们习惯用css编写效果来替代这些效果 ...