Photographers probably spend more time behind a computer than a camera. Yet, the technological progress of this industry is becoming slow, and manufacturers hide the issue with marketing tricks. Here is what's going on.
CPU Hot Mess: Same but Different
Until 2015, processor manufacturers used to introduce a new generation of processor every two years or so and the performance gain were massive between each release. Since then, Intel hit a technological wall with the 10-nanometer chip and is still stuck with the old 14nm lithography process. The latest Intel 9th Generation Core family (Coffee Lake) is nothing more than a reheated Skylake iteration from 2015 and Intel marketing experts desperately try to hide their development issues with endless variations, "++" suffix, and "refresh" generation of the aging Skylake architecture. Indeed, the new Intel 9600K processor is only a few percentage points faster than the previous 8600K released in 2017.
AMD is confronted with similar difficulties but managed to shrink its CPU from 14nm to 12nm last year with the Ryzen Threadripper 2 based on the Zen+ design. In reality, these chips are rebranded server processors (EPYC) with insane prices and power consumption. Nothing groundbreaking except for the wallet and the power bill.
But this derivative of server performance was enough to upset Intel's salespeople last year when they decided to pull a last-minute trick at the Computex trade show in Taiwan. Faced with the imminent AMD announcement of the 32 cores Threadripper processor, Intel came up with a counter fire to avoid the dishonor of being left behind with its "little" 28 cores Skylake-SP processor from the server industry. Since AMD took the leadership of the core counts, Intel replicated the frequency and put together a "new" 28 cores CPU able to reach 5GHz and break the speed record. Some engineers overclocked an old Xeon CPU and kept this monstrosity under thermal control with the help of a 1000 Watts industrial chiller hidden underneath the table. Intel managed to steal the show, but this scam was uncovered the next day by a suspicious journalist from Tom's Hardware. Eventually, an Intel representative explained that "in the excitement of the moment," the company "forgot" to mention this little detail: the CPU was extremely overclocked.
This embarrassing anecdote illustrates the severe crisis faced by the company. Historically, Intel's architecture used to be ahead of the competition by one or two years, but the manufacturer is now trailing AMD which might release the Zen 2 architecture produced in 7nm sometime this year whereas Intel is still struggling to step down from 14nm to 10nm. The decline of the computer industry and the poor management of the previous CEO can explain the situation. Unfortunately, Intel is not done yet with little lies as we discovered recently with the Core i9 9900K processor introduced a few months ago. Officially, the thermal design power (TDP) of this CPU is listed at 95W, but all benchmarks revealed a much higher thermal profile. To avoid instability, the real thermal envelope is set to 210W by Intel. Therefore, motherboard manufacturers follow this value to scale the correct voltage regulation stage of the Coffee Lake CPUs.
Essentially, Intel can't figure out a way to reach the 10nm architecture and only propose an endless variation of the 2015 Skylake CPU. Unfortunately, no amount of creative rebranding and inaccuracies on the specification sheet can hide the fact that these processors are nothing more than overclocked chips. Logically, the power consumption reaches new heights despite deceptive TDP values that fool no one. As for AMD, the situation is similar if not worst with the high-end Threadripper monsters that can pull up to 180W of power. Multiplying the number of cores generates additional electrical consumption, thus heat. As such, water cooling which used to be an exotic accessory for overclocking nerds a few years ago is becoming a standard requirement on performance computers.
To be fair, these processors are potent and capable, but they are not innovative. In the end, they are the equivalent of a doping syringe or a nitro-boost button: an unsustainable and short term trick to reach a certain level of performance.
The Expensive but Useless and Crippled Motherboards
Once upon a time, motherboards were critical parts of a computer. But nowadays, most functionalities such as storage, USB, audio, and network are embedded directly in the chipset and/or the CPU. Therefore, motherboards are nothing more than a Southbridge base plate designed to receive the main parts of the machine.
For this reason, the CPU's socket (physical connector) and chipset (software controller) govern the selection of the motherboard. Faced with the commoditization of the industry, motherboards and chipsets manufacturers (Intel & AMD) reacted with various strategies. First, they try to justify premium prices by designing "aerodynamic" boards filled with LED, useless heat sinks and few extra connectors. Don't fall in to this marketing trap because these gimmicks only inflate the price without providing any performance gain unless you plan to overclock your CPU.
The constant change of socket and chipset is another classic way to force the upgrade and justify the acquisition of a new motherboard. Indeed, AMD and Intel have this horrible habit of changing the type of socket and chipset with every new generation of CPU. After each processor release, the choice of compatible motherboards is limited to a few options costing around $300-400. Then, the price usually goes back to a reasonable level after a year or so. Therefore, the choice of CPU dictates the motherboard selection, and both components must be considered together in terms of budget. For instance, Intel just released the Z390 chipset along with the latest 9th generation Coffee Lake CPUs (9600K, 9700K, 9900K).
On the hardware side, this chipset is identical to the previous ones dating from the 6th generation Skylake platform (2015). Essentially, the Z390, replaced the Z370 which replaced the Z270, and the Z170 before that. We must salute the performance here: four identical but incompatible chipset release in 4 years. In fact, the Z370 chipset can handle the new Coffee Lake processors via firmware update of the BIOS, but the voltage regulation module of Z370 motherboard might be too limited for the real power requirement of the 9900K chips.
To be honest, these chipsets are not entirely identical. Despite the minor software changes, chipset makers tend to artificially limit critical functionalities of the component by locking some features in the BIOS such as overclocking capacity and connectivity. The user is invited to purchase the new chipset to unlock the full potential of the motherboard.
Graphic Card: Premium Price for Everyone
The GPU market is finally leaving the cryptocurrency nonsense thanks to the diminishing return of mining. However, the market is currently dominated by Nvidia on the high-end segment. Radeon cards offer similar performance and price on the mid-range segment, but they tend to be less efficient and consume more power, hence more heat to evacuate with noisy fans.
The consequence of this lack of competition is a general price increase. The traditional middle of the market x70 CPU series from Nvidia used to be found in the $350 range but the latest GeForce RTX 2070 cost about $550 at the moment. Hence, the affordable mid-market GPU is now priced as a premium product, and this situation will continue as long as AMD Radeon is not able to propose a competitive alternative.
A Lucrative Market Concentration
The lack of innovation and inflated prices are clearly linked to the market concentration in the computer industry. The majority of sub-components are only produced by duopolies or oligopolies where the dominant player naturally tends to raise the prices and slow down innovation as soon as its competitor can't keep up with the pace of technological development. As we just saw, this situation is currently happening with Nvidia and Radeon (AMD), but the CPU market is also affected. For instance, the LGA11XX socket has been artificially limited for many years by Intel to the Quad Core offering, but the company finally unlocked this socket to 6 and 8 cores CPUs to counter the resurrection of AMD with its Ryzen processors. Before that, buyers had to operate a transition to costly motherboards fitted with "advanced" sockets and chipsets if they ever wanted to install CPUs with more than four cores.
This example illustrates the positive effect of competition in the CPU market. Finally, the storage industry suffers from the same issue. WD, Toshiba, and Seagate dominate the hard drive industry while the flash memory (SSD, DRAM) sector is concentrated among Samsung, Hynix, and Micron with strong suspicions of price fixing, especially on the DDR memory.
Software Optimization: Coding with Your Feet like Adobe
Market dominance is also the cause of the disastrous performance of Adobe software on modern computers. As if the hardware issue was not enough, Adobe programs are poorly optimized to take advantage of multi-processors architecture and powerful graphic cards. Except for a few effects, the GPU will stay idle most of the time despite the huge reserve of power available.
As Lee Morris and many users realized, entry-level CPUs can outperform expensive 10+ cores CPUs. Why? It's all about parallelism, or the lack of it. Parallelism is the ability to distribute the processing load among several CPU cores. But more than a decade after the spread of multi-core CPUs in consumer computers, Adobe applications still can't handle multi-threaded tasks correctly. Instead, they mostly rely on CPU clock speed (frequency) to execute the computation. Unfortunately, the frequency race hit a thermal wall several years ago which is why AMD and Intel now propose high core count to increase the level of performance and escape the frequency dead end.
We find ourselves in an absurd situation where the main creative software company ignores the technological evolution and release programs that massively underuse the processing power of modern CPU and GPU.
Ironically, one of the marketing arguments expressed by Adobe to justify the transition to the Creative Cloud subscription-based model was the "continuous improvements through frequent iteration" as the Product Vice President explained in a blog post in 2012. Six years later, Adobe finally revamped Lightroom to use more than a few CPU cores at the time. Other than that, the programs from this company are painstakingly slow, unable to use the hardware correctly, and plagued with bugs and instability.
In reality, as the CFO of Adobe said, the motivation of the Creative Cloud introduction was financial: "the move to subscriptions just drives a bigger and bigger and bigger recurring revenue stream." And the strategy paid off big time with record boost in revenue. Sadly, this stream of cash didn't translate into "continuous improvements" for the user as it initially promised. For Adobe, shareholders must be pleased, and that's what matters.
Once again, market dominance is the reason behind this lack of innovation. Why would Adobe bother to optimize the Creative Cloud suite when it dominates the market and experiences a massive increase in revenue? Software development is expensive, and the re-coding old software core takes time. Introducing incremental side features during keynote shows is much easier than tackling years of negligent coding.
Conclusion: What Can You Do?
The so-called Moore's Law was never a law but an observation made in 1965 by the founder of Intel who merely described the rate of growth of the semiconductor industry. Ten years later, Gordon Moore revised his observation as the speed of progress was getting slower but still doubled every two years. 50 years later, the computer industry is confronted with many difficulties related to miniaturization as we approach the atomic scale. Traditionally, the main road to technological progress in the electronic industry was to shrink component allowing to cram more transistors on processors and bits on media. But this scaling strategy is now hitting physical walls, and each generation of products demands significant investment to yield modest results, which can explain the concentration of the industry, especially in the memory business.
Sadly, certain companies rely on blatant lies and deceptive communications campaign to hide an abyssal lack of innovation. They also resort to artificial market segmentation and product crippling. When the competition is trailing behind, then there’s always some manufacturer that takes advantage of the situation to raise the price to unreasonable levels. Finally, some software developers enjoying a market dominance don't even try to optimize their apps, which results in poor efficiency and a waste of computing resources.
Here are a few bits of advice to get the best out of your money in this environment:
Processors and Motherboards
First, consider the CPU and motherboard acquisition together, as the choice of CPU dictates the socket and chipset type. Try to find the sweet spot between price and performance and don't hesitate to consider the previous generation of processors and motherboard as the progress have been extremely slow the past few years, especially with Intel. For instance, the Core 9600K is only a few percentage points faster than the old Core 8600K. Generally, avoid high-end processors above $500 as they tend to be a waste of money and offer poor performance gain per buck. I would also ignore 8+ cores CPU with their exotic sockets and chipset for the same reason.
As a rule of thumb, Intel CPUs tend to perform better for single threaded applications due to their higher frequency while AMD Ryzen will shine in multi-threaded tasks and cost less than Intel. At the moment, the eight cores Intel i7 9700K and AMD Ryzen 7 2700x are solid performers with excellent price-performance ratio.
But keep in mind that performance depends on the software optimization and usage type. There is no such thing as the best CPU. The key is to find the right one for your need and priorities. Therefore, proceed with a precise evaluation of your user profile. Which software are you going to use most and which task do you perform in priority? What bottleneck are you trying to address first? If your main editing program doesn't take advantage of multi-core processors, opt for a high-frequency CPU or vice-versa. To complicate the matter, a given software can behave differently depending on the task: real-time editing, playback, pre-render, final export, and encoding each takes a different toll on the processor. Some tasks will benefit from the higher frequency while others will spread the load on multiple cores. In other instance, the program might prioritize the graphic card over the CPU.
Thankfully, there are plenty of benchmarks and reviews available on the internet and YouTube, even for niche applications. However, be careful with broad benchmarks as they only give a general indication of CPU performance. The reviews based on dedicated benchmarking tools are usually optimized for multi-threaded applications while video game benchmarks are skewed in favor of high-frequency processors. One of the references for workstation reviews is Puget Systems.
Nvidia leads the graphic card market, and Radeon GPUs are not competitive at the moment. Things might change with the next release cycle, but there aren't many alternatives for now. But as with the processor, you need to assess your need in terms of graphic processing power. What software are you using and which task are you trying to improve first? Then, is this particular software and task optimized for GPU rendering? Some video task can benefit from high-end GPU, but others completely drop the load on the CPU. Generally, regular photo editing software don't use the graphic card that much while some effects in video editing such as color grading, transitions, and 3D effects can benefit from a powerful GPU.
Contrary to the hardware world, there isn’t any oligopoly in the software industry. The offer of Adobe’s substitute is expanding. One of the most serious Lightroom challengers is the great Capture One with its advanced studio and tethering functionalities. DxO PhotoLab is also a good option. In the video department, Avid Media Composer has already been adopted by many productions. Final Cut Pro X is very popular with the editors working on Apple computers. DaVinci Resolve is another rising star in the industry, especially for its advanced color correction features. Blackmagic Design also proposes a good After Effects alternative with Fusion. These two pieces of software are free and can be downloaded directly from the company's website. The advanced versions cost only $299.
The situation is more complex for Photoshop. This one still reigns as the undisputed king in the professional industry. But you may want to take a look at GIMP (free), Pixelmator Pro, or Affinity Photo to name a few. Please, feel free to share other alternatives in the comment section below.