In a nutshell
Buy refurbished. Don't build or buy new.
The most cost-effective Blue Iris computers are refurbished business-class systems such as "HP EliteDesk
" and "Dell OptiPlex
". You can usually get a good deal by searching Ebay
for specific CPU model numbers, and selecting the "Desktops & All-In-Ones" category. Links here are for the USA. Pricing and availability will differ in other countries.
Models recommended here are chosen for favorable pricing. These are not the most powerful computers by modern standards, but are still very capable if you follow the Blue Iris optimization guide
and especially if you use sub streams.
Some models to search for:
6th generation - For H.265 hardware acceleration
4th generation - Often much cheaper, but nearly the same performance
- 4 GB of memory is sufficient for a smaller system (4 or fewer cameras) but 8 GB is recommended as a minimum for most.
- Systems with a "T" suffix on the CPU (e.g. i5-6500T) are underpowered versions designed to fit in smaller cases and meet stricter energy-usage requirements. These are significantly less bang for your buck.
- It is best to choose a system with Windows preinstalled to be sure you don't have to pay for a license.
- Some used PC sellers offer a warranty. Some do not. The original manufacturer warranties on these systems have mostly worn off by this point.
Complete systems based on an Intel i5 desktop CPU can be found for $100-$200 USD. These make great Blue Iris computers for medium-size workloads. i7 systems can be found between $200-$500. These are better for large workloads, such as if you have more than about 12-16 cameras.
Older systems, such as those based on i5-4590
can be found for $100 or less. These still make very capable computers for Blue Iris if you configure things properly
Each manufacturer also has an "outlet" store where they sell their own refurbished computers, although these have often been picked clean of all the best deals already.
Real-World Performance Data
Since late 2017, users of the tool "Blue Iris Update Helper
" are regularly contributing performance data to a public database. Data is collected in an automated fashion, ensuring that it remains unbiased and as accurate as possible. This is all in an effort to provide greater understanding of the capabilities of different hardware configurations running Blue Iris
Check it out, here: Blue Iris Performance Data
The most important aspect of a Blue Iris computer is the CPU. The CPU determines how much video your system will be able to process, and how much electricity it will consume.
There are several important factors in choosing a CPU:
* High Multi Thread performance
helps your system scale up (more cameras, more concurrent remote viewers, better AI analytics response time, etc).
* High Single Thread performance
is mainly important for working with 4K video at high frame rates. But low single thread performance (below about 2000-2500) may cause stuttering or jerky playback even with lower-resolution video than 4K.
* Integrated graphics is highly recommended for its cost-effectiveness compared to a dedicated GPU.
* Intel Quick Sync Video support can be used for hardware accelerated video decoding, which improves energy-efficiency and performance. Quick Sync is only available on Intel CPUs with integrated graphics, and it doesn't work within virtual machines unless you pass through the Intel integrated graphics to the VM, which does not work on all systems.
Since 2020, proper use of sub streams
will enable nearly any quad-core desktop CPU made since 2012 to handle nearly any number of cameras up to Blue Iris's limit of 64 cameras. As a minimum, I recommend a CPU with a CPU Mark
of at least 6000, and single thread rating
of at least 2000. However these are not hard limits. Lesser CPUs will still work fine with proper software configuration, just perhaps not with as high of resolutions and frame rates. Higher-spec CPUs may be able to provide better frame rates at 4K resolution, better remote viewing performance with multiple users connected, and can have enough power left over for Deepstack AI if you choose to use it.
If you want to buy or build a new machine in 2022, I recommend Intel's 12th generation Core series with integrated graphics. The entry-level i3-12100
would offer excellent performance for Blue Iris, comparable to a top-of-the-line system from 2018. Stepping up to the i5-12600K
gets you about double the multi-threaded performance.
Quick Sync Video capabilities vary by the age of the CPU:
* Sandy Bridge (2xxx series) offers H.264 hardware acceleration only for resolutions up to 1920x1080.
* Ivy Bridge (3xxx series) removes the 1920x1080 limit and works at 4K.
* Skylake (6xxx series) is the first generation to include H.265 hardware acceleration capability (4K supported of course).
Intel CPU models with a "T" suffix (e.g. i5-6500T) are reduced-power versions designed to fit in smaller cases and meet stricter energy-usage requirements. Unless you are extremely space-constrained, you are better off with a normal, non-T version.
Intel CPU models with an "F" suffix (e.g. i9-10900KF) do not include integrated graphics, and therefore lack Quick Sync Video while offering only a negligible discount in price. Do not buy these for Blue Iris.
Server CPUs (Intel Xeon, AMD EPYC, etc) and Multiple Sockets
Generally speaking, if you can appreciate (and afford) an enterprise-grade server platform, your needs will be better met by enterprise-grade video management software instead of Blue Iris.
That said, here is some guidance. If your needs would be met by an i5 or i7 CPU, but you want ECC memory or some other feature only found on server platforms, consider a similarly priced and specced Xeon E (or older: E3) series model that has Quick Sync Video so you can take advantage of efficient hardware accelerated video decoding. Otherwise, look at general CPU benchmarks and use those to decide what the best CPU is for your money.
Beware of old used servers claiming lots of cores and memory at low prices. These will be loud and inefficient by today's standards, and in many cases outperformed by a cheaper workstation that is years newer. Also note that Blue Iris is not optimized for multiple-socket servers. If you are considering running such a heavy load that Blue Iris would be unable to handle it with a single CPU, then you should be using different software.
4 GB of memory is sufficient for a smaller system (4 or fewer cameras) but 8 GB is recommended
for most systems. The largest systems (20+ cameras or 800+ MP/s) benefit from 16 GB.
Memory speed matters
. If you are seeking maximum performance (and not using sub streams), then it is important to utilize all the memory channels
available on your chosen platform. Most systems are capable of running dual channel memory
, and I highly recommend doing so if you are going to be pushing the limits with your system.
For platforms that support either DDR4 or DDR5 memory, get whatever is cheaper (almost certainly DDR4).
For most users, a dedicated graphics card will provide no meaningful benefit to Blue Iris, and will only serve to consume additional electricity. Onboard Intel graphics are more than enough. In fact, it can sometimes be complicated getting Quick Sync hardware acceleration to work if you have installed a dedicated graphics card.
If additional display outputs are required, or you find that Blue Iris requires too much CPU time to draw live video to the screen (common with 4K monitors), then an Nvidia graphics adapter may help. The cheapest 4K-capable Nvidia card I know of (as of Sept 2017) is the GT 1030 (around $70 USD). You could spend more money on a faster card, but it won't make any difference for this.
Since 2018, Blue Iris supports H.264 and H.265 hardware acceleration via Nvidia CUDA using recent Nvidia graphics cards. Performance of Nvidia CUDA acceleration scales with the compute power of the GPU, such that a GTX 1060 for example can handle more video than a GT 1030. This method of hardware acceleration is not recommended for most systems because it is very expensive and inefficient with power.
A modern graphics card can be used to [speed up Deepstack AI processing](https://ipcamtalk.com/threads/deepstack-case-study-performance-from-cpu-to-gpu-version.58425/).
How much storage space you need depends on the average bit rate you will be recording. We have a storage calculator
, or you can read-up on how to do the calculations manually
Storage for Video Clips
For video recording, you want a mechanical hard drive (HDD) as these are more cost-effective than SSDs and have better write endurance. For best performance, buy a surveillance drive
such as Western Digital Purple
or Seagate SkyHawk
. Configure Blue Iris to put all recording folders on the HDD(s), via Blue Iris Options > Clips and archiving.
Storage for Clip Database
Blue Iris keeps a database containing metadata about the video clips it has recorded. The size of the database is small (approximately 1 GB for every 35,000 video clips stored), but the database is frequently accessed so you should put it on your fastest storage device. If you are so inclined, Blue Iris and its clip database will easily fit on a 120+ GB solid state disk (SSD)
alongside the Windows operating system. For systems with limited physical space or budget, everything will still work fine without an SSD.
Multiple Hard Drives / RAID
RAID is not usually recommended, as most of what a video surveillance system records is worthless. If you wish for your system to be protected from disk failure, use a RAID type which offers redundant storage, such as RAID 1, 5, 6, or 10. Avoid configurations which offer increased I/O performance at the cost of redundancy (e.g. RAID 0), as these do not provide any benefit for video surveillance.
If all you need is additional storage capacity, consider keeping each drive as its own separate volume, and configuring specific cameras in Blue Iris to record to specific drives. This way, if you lose a drive, you only lose video from some of the cameras, and still have an uninterrupted archive of video from the other cameras. Compare this to a spanned array of disks, where the loss of one drive means losing a large chunk of time from all cameras.
Blue Iris only runs properly on Windows
. Windows 10 or 11 is recommended. Windows 7 is not recommended, as there are difficulties with hardware acceleration not working with BI in service mode. Windows 8.1 should be fine, but you might be the only person who uses it. Recent Windows Server editions also work, if that is what you have available.
But what CPU do I actually buy?
To determine which CPU you need, first add up the total megapixels per second (MP/s) you intend to run. Then multiply by 10 to get a rough idea of the CPU Mark score
that would be recommended as a minimum. For any CPU that does not support Quick Sync Video, add about an additional 40% to the requirement.
IP Camera Megapixels Per Second
This table lists MP/s (megapixels per second) of common video resolutions at various frame rates.
|Resolution||MP/s at 5 FPS||MP/s at 10 FPS||MP/s at 15 FPS||MP/s at 30 FPS|
|2688x1520 (4 MP)||20||41||61||123|
|1920x1080 (2 MP)||10||21||31||62|
|1280x720 (1 MP)||4.6||9||14||28|
|704x576 (D1 Sub Stream - PAL)||2||4||6||12|
|704x480 (D1 Sub Stream - NTSC)||1.7||3.4||5||10|
For eight 4K cameras each at 15 FPS, without using their sub streams, 8 * 124 = 992 MP/s. Multiplying by 10 gives us 9920 as a CPU Mark score to aim for. This suggests you would need at least a 6-core Intel CPU like the i5-8600K (~10,000 CPU Mark) or AMD Ryzen 2600X (~14,000 CPU Mark). And honestly I would recommend more than that, because those CPUs would be continuously near their limits to handle a load like that.
This is where sub streams come in. If you used a D1-resolution sub stream with each of these cameras, your MP/s total would only be around 80-90 MP/s, and a much weaker CPU would be acceptable. Except there is a "gotcha". Even if you configure sub streams to be used on all the cameras, there will be times where you are solo-viewing a camera, or viewing a clip, where the sub stream will not be used. This will temporarily increase your CPU requirements by a huge amount! This is why I recommended a minimum of 6000 for the CPU Mark earlier. It is not because I think everyone will use 600 MP/s of video continuously, but because the system must be sized to handle the peak load, not just the minimum load.
Very High-Load Systems
For loads greater than 1500 MP/s, all bets are off. In my experience you will be running up against the limits of dual channel DDR4 memory bandwidth, and a more expensive platform with more memory channels will only go so far before it, too, finds a performance bottleneck. It is typically a much better decision to just use sub streams to reduce the video decoding load, or choose a different VMS software that has lighter system requirements than Blue Iris.
For very high-end systems where sub streams will not be used, memory bandwidth plays a big role in performance. Running only a single channel of memory could be a performance bottleneck. Also, if you are buying memory, faster is better. Read more about my tests that led to this conclusion, here: https://ipcamtalk.com/threads/ryzen-3950x-for-blue-iris.43719/
In particular, sub streams
can be utilized to greatly reduce the number of MP/s your system needs to process. For example two "D1" resolution sub streams at 15 FPS each would be only around 10 MP/s (versus 248 MP/s for 4K main streams!). Just be aware that sub streams are not always used. The main stream is decoded when viewing single maximized cameras, and your system must be able to handle the peak load during these usage spikes. You could run, for example, dozens of 4K cameras on a midrange system with a CPU Mark of less than 10,000 as long as you are careful with how you use Blue Iris.