I’m using a Mac Book Pro but from what I understand Blender doesn’t work well with any Apple computers? I’m looking at PC options to use specifically with Blender.
How do I choose a computer to use Blender? Which hardware should I prioritize?
What graphics cards brands are better, AMD or NVidia?
Should I go with Intel or AMD processors? What amount of RAM is adequate?
According to rules, hardware questions are off topic here. You can ask about them over at Hardware Recommendations or other third party communities like Blender Artists Forum. To not leave you empty handed here is a tentative canonical post providing some general guidelines regarding choosing hardware for Blender.
Like in most situations, we can apply here what we apply to any other purchase decision, and basically always boils down to:
Either buy the best you can afford, or the cheapest that satisfies your needs
When doing so consider that
You should match your hardware to your expectations, or lower your expectations to match your hardware.
Time is money, if your are serious about your hobby, if you are interested enough in 3D modeling or rendering, if you are making money, or your work depends on it, or your livelihood relies on this, invest in better hardware. Otherwise save your earnings to do with them what truly makes you happy, spend as little as possible to fulfil your most basic needs.
Choose a good to mid level CPU, a good graphics card, higher end if you plan to use GPU acceleration for rendering, prefer NVidia over other brands for Cycles, and a decent amount of RAM that should match the complexity of you average scene. Everything else is mostly secondary, check the official hardware requirements.
Blender is relatively forgiving and modest in terms of minimum requirements among 3D Suites, especially considering the requirements of the competition. Most entry level or budget hardware these days will run Blender acceptably, when in doubt always check the official minimum recommended specs. This is for basic performance, if you want to "go faster" however, things may start to get more expensive.
The most essential of components for a good Blender experience are CPU (Central processing Unit, or Processor), GPU (Graphics Processing Unit, or Graphics Card), and RAM memory (Random Access Memory)
If you don't need mobility, you always work from a desk (at home, work or office), if portability is not essential and you can afford the difference (in both space and price) always go for a desktop computer.
They are lot more durable, robust, longer lasting, more upgradeable and more modular, allowing you to gradually update components as you go, spreading costs over time or delaying them to more convenient occasions. For the same price you often get slightly better performance on desktops over laptops.
On the other hand you will need more peripherals like external keyboards, monitors, speakers, etc which require (physical) space and can make things more expensive.
If mobility is in fact a requirement (you frequently move from home, work, school, office), then "gaming" laptops with dedicated graphics cards (i.e. AMD or Nvidia) often make an acceptable compromise. More performant laptops are often expensive, bigger, heavier, more prone to overheating and consume more power, thus needing bigger power adapters, all of which reduce portability.
The fact that it is marketed as "gaming" is not a guarantee of performance however, and is often a synonym for exuberant aesthetics, useless features and overpriced hardware. What you should look for instead is what should be a staple for gaming, which is a good graphics card. If you find a decent one in "non-gaming" line absolutely go for it, don't get fooled by marketing tricks.
When choosing laptops other secondary things to have in mind are screen resolution and keyboard size. Blender uses the Numpad for standard views, working without one is possible, but if available prefer having one to make things easier. Prioritize GPU choice over other features, which is almost always something you can't easily change afterwards; CPU, storage and memory are often user upgradable and can easily be swapped later if necessary, keyboards and displays can be mitigated with external peripherals. Choose them at your own discretion.
Linux is often said to be the best performing operating in general for its low resource requirement and overall lightweight architecture, this also benefits Blender. If you don't need any other particular software in your daily routine that is unavailable for Linux, (like proprietary programs, commercial CAD, gaming etc.) and can deal with the extra maintenance and setup go for it.
Windows is the most widely used and probably best supported and most polished platform, due exposure it receives from its sheer number of users, and real world road-testing it ends up getting.
Apple Mac platform is officially supported but performance and testing is some times not up to the same quality standards as the others. Performance may also not be optimal in some cases, and you generally get worse performance for your buck and limited graphics power due to unavailable dedicated graphics cads. Newer M1 or M2 Macs show some promise, and Metal Device support is underway to allow GPU acceleration on Mac, but they are relatively new, so don't yet expect the same stability or level of support.
Ok, so how do I choose the "best" hardware for Blender?
How do I determine what are good and bad components for my computer?
If in doubt look at the price, generally speaking more expensive products will tend to perform better. Save for some exceptions, better components will naturally costs more, caveats being paying for known brands, novelty features or aesthetics (RBG lighting, or exuberant designs, premium Brands).
Price to performance ratio tends dictate the final value of a product, if it doesn't market will naturally adjust by laws of supply and demand. Good products will sell too well, thus stock will be low and hard to find; otherwise they will sell poorly until prices match the perceived quality.
It also tends to decrease abruptly towards the higher end of the spectrum. A 300€ CPU may preform twice as fast as a 150€ one, but a 600€ CPU will probably only be 10% faster that the 300€ one, despite costing almost double, and things fall abruptly from there (take values as gross exaggerations for the sake of example).
General rule of thumb when looking at hardware components is to look at model numbers. A higher number will generally mean a better performing model. There are of course some caveats but the base rule mostly stands. CPU 12000 is higher end (or at least newer) than 11000, and GPU series 3080 is better than (or at least newer) than 3070.
For most hardware brands (CPU or GPUs) model numbers digits are often divided into two fractions, "Generation" and "Performance Tier".
The first digits generally indicate generation or year of release of a product. Models starting with the same digits will tend to be the same age and released in the same year (not necessarily a calendar year but the same "release cycle" or hardware architecture).
| Model Number | Generation | Performance Tier |
|---|---|---|
| 960 | 9 | 60 |
| 970 | 9 | 70 |
| 980 | 9 | 80 |
| 1060 | 10 | 60 |
| 1070 | 10 | 70 |
| 1080 | 10 | 80 |
| 3060 | 30 | 60 |
| 3070 | 30 | 70 |
| 3080 | 30 | 80 |
All 9## graphics all belong to the same generation of 900 series, like all the 1000 or 3000 to the 10 and 30 series respectively. The model numbers ending in 60 are all lower end, while de ##70 are mid range, and the ##80 higher end.
| Entry Level | Low End | Mid Range | High End | Enthusiast |
|---|---|---|---|---|
| 850 | 860 | 870 | 880 | 890 |
| 950 | 960 | 970 | 980 | 990 |
| 1050 | 1060 | 1070 | 1080 | 1090 |
| 2050 | 2060 | 2070 | 2080 | 2090 |
| 3050 | 3060 | 3070 | 3080 | 3090 |
| 4050 | 4060 | 4070 | 4080 | 4090 |
So generally, in the generic examples above the higher the number the newer the product is, so better performance is generally expected, but there is a nuance to the numbers. In the above example a 960 is newer than a 880 and a higher number would indicate a higher performance tier.
However, a 960 is a mid range hardware indicated by the 60 in the model number, while a 880 is a high end model because of the ending 80 digits. In many cases older higher end components are often still better performing than a newer generation models of a lower performance level.
That is not always the case, and is impossible to set universal reliable rules. Some times it goes back one generation some times two or more, the only way to be sure is to check on a case by case basis. Compare them, do some research, check search engines or some benchmark sites. Here are a few examples.
Of course newer models may also bring new features, introduce newer technologies, fabrication methods, better power efficiency and all the new bells and whistles. When choosing a model also have that in mind besides raw performance.
A frequent battle in the CPU market is to choose between AMD or Intel. For Blender the brand is as far as I know mostly irrelevant, and the brand choice in this case is quite irrelevant.
It all comes down to specific models, some Intel models perform better, some AMD will win, and some times it all comes down to price/performance ratio on a per model basis. When in doubt do some research, in search engines type model x vs mode y and see which does better, compare prices and availability, look for reviews or check the benchmarks recommended above.
The graphics card is easily one of the most important aspect of a new computer when working with graphics these days, considering that in modern software a lot of tasks are now leveraging the high parallelization capabilities of modern GPUs and taking advantage of hardware acceleration.
It is also one of the components that can get the most expensive, easily surpassing all others. A good top of the line graphics card can cost as much or more than other components combined.
In Blender it can be taken advantage of for rendering in Cycles, viewport performance while modeling, rendering, shading and lighting quality in EEVEE, and real time compositing.
I'd love to be able to say that in terms of GPUs brand is also irrelevant, but unfortunately that is not the case, at least not at the moment of writing, and not all GPU brands are born equal for Blender.
In an ideal world open technologies would dominate the market and offer the best performance and same features across brands, but that is not always the case. While for viewport previews and basic operation most brands perform well enough for most cases, the same can't be said for Cycles GPU acceleration. See the official requirements.
If you have the choice and the budget always go for NVidia above other brands. They are the best performing, best supported and most mature ones. The availability of CUDA (Compute Device Unified Architecture) guarantees the best experience when GPU rendering, with significant performance gains over other brands and over CPU rendering (with higher end or newer cards). Newer RTX graphics cards also support Optix acceleration with further benefits.
AMD Radeon would come second in the choice for graphics card brand for Blender. They sadly have has a less mature solution for GPU rendering acceleration, which has gained some considerable improvements in recent times. OpenCL based GPU rendering has been deprecated in newer versions of Blender and is no longer supported, being gradually replaced for the HIP device which shows some promise.
This is only available for cards newer than Radeon RX 5000 (from 2019) series for the moment, so you should be fine when purchasing brand new models, but should take care when choosing older models, second hand or used market ones. Even then beware that despite being available this solution is still relatively new and not as mature, polished or has heavily tested as NVidia's CUDA and Optix. Some some crashes and glitches are expected here and there, performance may not be on par.
Intel graphics cards are generally available only as integrated GPUs with low power consumption in lightweight or compact form factors. They generally have limited performance, and some times poor standards compliance which results in lower visual fidelity for EEVEE and viewport previews.
GPU rendering acceleration for Cycles in mostly unavailable or at an experimental level with One API for Intel Arc devices. It is the lower performing of the three, and the least well supported for GPU rendering. If you have the budget prefer a computer with a dedicated graphics card over integrated Intel.
In laptops, integrated GPUs have an upper hand in power consumption though, they are generally more compact and lighter weight, produce less heat, and allow thinner more portable laptops, with longer battery life, if mobility is a concern, at the expense of performance.
When buying Apple Mac computers there really isn't a discrete choice for GPU brand, but one newer Mac M1 or M2 laptops you will get an integrated graphics card with your processor. GPU rendering wise support is underway for Cycles acceleration through the Metal API. Performance is reported to be acceptable and working well, despite some crashes, due to newer technology support may be less polished than other "more common" brands.
If you have a high budget, it may be tempting to look into professional line of graphics cards for workstations, CAD and medical devices like Nvidia Quadro, ATI FireGL, or AMD FirePro.
My experience with these is that they are not worth their money for the extra performance you get in most situations, if any at all. For Blender specifically even less so.
Hardware wise these are essentially the same as the top of the line consumer products, they mostly only differ in drivers and firmware. Other than said artificially imposed software restrictions, they rarely perform better than their end-user counterparts in every day situations, maybe even for professional applications, though reliable numbers are hard to come by.
For Blender specifically they provide little real world benefit, and at worst may actually be harmful. Since they are generally expensive, exclusive, and hard to come by, developers have a hard time getting access to them so they don't get as much "first-hand" usage. This leads to less testing and optimizatio, resulting in their particular capabilities being potentially underused at best, or possibly experiencing worse performance than an average off-the-shelf model with high availability.
Your mileage may vary, but I'd only recommend these if you really need the extra video memory these cards generally offer, compared to the consumer counterparts.
Having more memory wont make you render faster, but it can allow you to load more complex scenes with more objects, higher poly counts, or more complex textures.
Blender won't generally need a lot of RAM, the amount you get should mostly be tied to how complex the average scene you expect to work with is. If you work with very large textures, models with many assets, high number of objects, very high poly meshes or very big scenes then it may justify buying more memory.
RAM speed is not crucial, faster memory is more expensive, if your budget is limited and you have to choose between faster clock speeds and more memory, then more memory is generally preferred over faster speed.
Beware system memory is not always the limitation, especially when rendering with GPUs Video memory should also match your computer memory, otherwise you'll find yourself with a scene that fits your computer but not your card, and you can't make renders of it or display it correctly in the viewport.
Blender is not particularly picky about storage, having a fast SSD will probably make it load faster on first startup, but beyond that benefits are probably minimal. Reading from disk and saving large files may be faster too with a fast disk, but bottlenecks are often elsewhere (like the CPU) so gains may not be very significant.
If you plan on doing a lot of video editing though, having a fast storage device will make a more significant difference. Video editing requires reading a big volumes of data from storage and having a faster read speeds can make working a lot more comfortable as you get faster previews and benefit from real time editing, rather than waiting for Blender to catch up with what you did.
Rendering animations as image sequences may also benefit from a fast SSD. Some image file formats are known to be "slow to encode" and the saving time writing the images to disk may add up to a significant overhead over longer works.
Most modern computers you'd buy new these days will have an SSD fast enough to handle most situations well. When choosing one, look at read and write speeds as the most relevant metrics.
Storage space is something that is often not critical, it will largely depend on how large you average project will be, and how much of them you need to "keep around". It is something you can easily expand later when necessary, and resorting to external storage to archiving old stuff to keep your device's available space in check is easy enough, so a lot of storage is often not necessary unless personal requirements demand it.
Exceptions are if you make a lot of animations and render them to video or HDR image sequences, which get big fast, caching animations, mesh sequence cashes, or physics simulations, or very large asset libraries with high quality textures.
If in doubt search and research, and then investigate some more. Look for reviews, benchmarks, articles etc. . Be sure before you buy, check recommendations and official requirements and supported hardware, so you don't regret your purchase or waste your hard earned money before it's too late.
Use search engines to your advantage, look for XXXX reviews, XXXX benchmarks, XXXX vs YYYY.
In the search box of you favourite search engine type just XXXX vs and see what suggestions the search engine shows up in the drop down list. This is generally a wonderful way to see frequent searches, what other people are looking for, similar products, common alternatives, direct competitors, what that product is often compared against, and a great way to find out that one alternative you didn't even know existed and ended up being the perfect choice for your particular constraints.