How will your PC perform at a specific task? Whether that’s manipulating a photograph, adding text effects to a 4K video, rendering a scene in a 3D package, compressing files for storage, or training AI on a new model, it’s not easy to get a meaningful answer to these relatively simple-sounding questions. This is because each of those scenarios will stretch your system in different ways: some will require more from your CPU, others from your graphics card, others again your memory, storage, or network connection.
This is further complicated by the differences that you’ll see using a variety of applications and programs to achieve the same result. For instance, there are a multitude of image editing programs out there, from Adobe Photoshop to GIMP, and they all put subtly different demands on your system. So, it’s rarely as easy as saying that this machine is great at photo manipulation or video editing.
Thankfully there are some constants and general principles that will help ensure that you have a machine that is right for the task at hand. And there’s also something else that can help you here: benchmarking.
Benchmarking has many uses, from checking and highlighting problems with subsystems, to stressing your machine to make sure it doesn’t overheat under load. We have, for instance, produced a comprehensive article on how to test your memory subsystem with MEMTEST86.
Benchmarking is also useful for seeing how your machine compares against others with a mind to upgrading, and that’s the context we’re looking at here.
A well-respected and industry-wide used benchmark will allow you to see how your system performs under certain loads. Such benchmarks will either stress specific elements of your PC or test the whole system in its entirety, depending on the benchmark you are running. This will allow you to see what difference an upgrade can make.
One good piece of news here is that for more serious work, increasing the memory capacity of your system will have a positive impact on the number of programs you can smoothly run at any given time. This is particularly useful for creative types that will often have several applications open at the same time performing different tasks – 3D modelers for instance will often have image editing applications open to tweak textures as they use in the main modelling software.
For more serious workstation-level PCs we’d recommend 64GB of RAM as a minimum to ensure that you have the memory capacity to do what you want without having to hit the page file or scratch drive, which will slow your work down considerably. Memory capacities of 128GB and greater aren’t unheard of in such machines either.
Below is a graph showing the difference that memory frequency can make to a range of benchmarks. This shows the uplift of running a 32GB memory kit at 7,200MT/s compared to a standard 32GB DDR5-4800 kit.
Back the question of benchmarking, though, which applications are worth running to find out how your machine is performing? That does depend on what you want to do. Ideally, you’d use the application or applications that you use for your job, but this isn’t always possible, as not all of them have built in benchmarks.
It’s also worth stating that in order to benchmark your system you need to make sure that you’re not doing anything else with the machine. Even browsing will throw the numbers off, and you want a true representation of your system’s performance.
Here’s a list of industry standard benchmarks worth considering:
PCMark 10 by UL Solutions is unusual, because it tests your whole system with respect to modern working. Along with specific tests for storage and battery life (useful for laptops), at its core it features a group of tests that simulate website editing, video calls, word processing, spreadsheet work, as well as image editing and word processing. At the end of the benchmark run you can see how your machine compares to others with similar core configurations.
PCMark 10 has been superseded by the Procyon Office Productivity Benchmark, which features more up to date workloads including AI inference, as well as updated versions of Microsoft Office applications. PCMark 10 is still very useful though, despite its age, because there are loads of system results to compare against and there’s a free demo version available too.
If you’re interested in your system’s gaming potential, then the 3DMark benchmark is also worth checking out. The demo is also available on Steam and includes a range of tests that show off what games are potentially capable of these days. 3DMark is widely used and there are loads of results out there that you can use for comparison purposes.
Unlike the other benchmarks we talk about here, Geekbench 6 is a synthetic benchmark. It’s still a useful test though, both in terms of assessing the capabilities of your machine and also highlighting problems with your configuration – if your final result is notably slower than what you should be getting, you know you have a problem somewhere.
Geekbench is primarily a CPU benchmark, and its small footprint, quick run time, and cross-platform support makes it a popular choice. The final result is an index, or actually a pair of indexes – one for single-core performance and another for multi-core benchmarking. In both cases, the bigger the better, but you need use the Geekbench Browser to see how your system compares to others for those indexes to make sense.
Artificial intelligence is all the buzz right now, so it only seems reasonable to want to include a benchmark that looks at such workloads when assessing the performance of your machine. AI workloads tend to be reliant on fast system memory as large models are assessed while held in system memory, and from our testing you can see big improvements from running high-capacity, high-frequency memory in your system.
The AI-Benchmark is a Deep Learning and Tensorflow benchmark that require you to install Python on your machine and then run the benchmark from there. You can find out more about the benchmark, and how to install it, on the official website.
The Corona benchmark has recently been updated to version 10, but the underlying idea remains the same – how quickly can your system render a ray traced scene? The new version recreates a cityscape that using elements typically found on a motherboard to striking effect.
You can download the benchmark for free and then it’s simply a case of extracting the files to somewhere useful and then running the benchmark from there. Depending on what CPU is in your system, it should take roughly a minute to run. Once completed, you can hit the Submit score button to see how your machine compares to your peers. Be prepared to be humbled by the result of the latest AMD Threadripper CPUs and some huge memory capacities – up to 512GB no less.
Blender is a 3D rendering suite that is free and Open-Source software that is owned by its contributors. It’s also incredibly powerful and strives to offer the best 3D CG technology for artists, covering the entire 3D pipeline, including modelling, rigging, animation, simulation, rendering, compositing, and motion tracking. It also includes powerful video editing and game creation capabilities for an impressive overall package.
The good news from a benchmarking perspective, other than the fact that it’s free, is that there’s a separate tool you can download that handles the benchmarking for you. You can grab the main benchmark and then run it to download the files needs – they weigh in at just under 1GB, so not too bad. Once these files have been downloaded, you select whether to use your CPU or graphics card and then hit ‘Start Benchmark’ to get things going. You’ll need to create an account to see how your machine compares to others.
7-zip is a free compression tool that includes a benchmark test. Compression and decompression are still important tools to have access to as the data that we routinely handle can be surprisingly huge and squeezing such data files to more manageable sizes make sense. The fact it’s free doesn’t hurt either.
The benchmark portion of 7-zip can be found in the Tools menu. Selecting the benchmark performs a Compressing and then Decompressing cycle before producing a total rating measured in GIPS, which equates to 1,000 MIPS (Million Instructions Per Second). Processors have a big impact on your system’s capabilities when it comes to compression and decompressions, but so does memory as you can see increases of up to 10% with faster memory.
That’s it. This is a list of some useful benchmarks that you can use to see how your system handles a variety of tasks. Note that this isn’t an exhaustive list, and there are always more benchmarks out there that you can use for various tasks. For instance, Cinebench R23 is another good 3D benchmark, HandBrake can tell you how your machine will handle video encoding, LAME is good for audio encoding, the list really does go on and on. But the above should be enough to get you started.
As far as memory is concerned, you should see upticks in performance in most benchmarks that test the overall performance of your system most. Specific tests that focus on just the CPU or just the graphics card are designed to cut out as much of the rest of the system as possible – useful for when looking to upgrade those components, but not too useful when looking at how your machine will perform at a specific task. And finally, if you can devise a benchmark using the tools that you use on a day-to-day basis, then that's an even better benchmark for you.
產品註冊