This week on Tom's Hardware Premium: October 17, 2026 — packaging roadmaps, Chinese DGX Spark rival emerges, and digging into ASML's earnings

This week, we've published a swath of news and features over on Tom's Hardware Premium. So, if you're not subscribed yet and wondering what you might be missing out on, we've got you covered. To read any of the articles listed here, be sure to arm yourself with a subscription to get access.

First up, we dug into Intel's EMIB and Foveros packaging roadmaps, as they continue to mature, and found new customers looking to cash in on building a more sophisticated offering. Intel's ultimate goal? To build system-in-packages the size of just 12 reticles using both EMIB and Foveros Direct 3D. We map out how the company plans to get there in our exhaustive report.

• Analyzing Intel's EMIB and Foveros advanced packaging roadmaps — how Intel plans to scale chips in every direction

Over the past month or so, we saw the debut of DLSS 5 and Multi Frame Generation on older graphics cards. Both features are officially locked down to Blackwell-based GPUs only, but amidst spiraling prices and the curious disappearance of the RTX 5090 from store shelves, modders have taken it upon themselves to port these newer features to older GPU architectures. We had the opportunity to sit down with two such modders to learn more about how both Multi-Frame Generation and DLSS 5's neural rendering capabilities made it over to older GPUs.

• Meet the modders porting ringfenced DLSS features to older GPUs — how DLSS 5 and Multi-Frame Generation were unlocked for older architectures

Earlier, we alluded to the fact that new consumer PC hardware is becoming much more expensive to purchase when compared to just a few years ago. So, our resident deals guru, Stewart Bendle, took it upon himself to take a look at the market overall and discovered that, in all likelihood, you're not going to find the best deals on online storefronts. However, the best place to buy your new PC components is limited to whether or not you'll have a suitable brick-and-mortar Micro Center location near you. We pore over the bundles and deals that may make a trip more worthwhile than you might initially think.

• The internet isn’t the best place to find upgrades for your computer — brick-and-mortar stores still reign supreme thanks to bundled deals

While those articles clear things out for all of the features we've published on Tom's Hardware Premium this week, we've also got a slew of deep news analysis stories that subscribers can enjoy.

To start things off, ASML posted a record quarter this week, posting $13 billion in revenue, claiming that demand continues to outstrip its own supply when it comes to delivering its advanced lithography machines. The company is working on expanding supply, as demand continues to soar.

• ASML reports record $13 billion quarter, vows to expand capacity as demand for lithography equipment soars — customers eager to secure orders in 2030 as demand is expected to outstrip 2027 and 2028 supply

Googlebooks have officially landed. But where do they sit in the wider ecosystem of devices in 2026? And is it even a good idea to purchase one? With the premium offerings nearing MacBook Air territory, we muse on the awkward timing, spec choices, and pricing of Google's new class of device.

• Googlebooks are as poorly timed as Valve’s Steam Machine, and may suffer a similar fate

Speaking of new classes of device, the DGX Spark and Ryzen AI Halo platforms both take direct aim at local AI enthusiasts. But, since these devices have launched, prices have skyrocketed. Now, Sipeed is entering the arena with a competitor of its own. Named Maixstation, the device promises Spark-like performance, based on chinese silicon and 128GB of LPDDR5X at a much lower price than the current DGX Spark equivalent. We have all the details, including lofty performance claims, all of the specs, and much more.

• Sipeed reveals Chinese-made DGX Spark and Ryzen AI Halo rival, expected to cost $3,999 — MaixStation pairs Axera’s AX8860P with 128GB of LPDDR5X at 273 GB/s, with promising local AI performance claims

SK hynix is reportedly set to place an in-house core, named a Hynix Transaction Processing Unit (H-TPU) in its SSD controllers to ease the load put on Arm cores. The benefits to using the new core, the company claims, is that they use less power, and free up the core Arm cores to work on other tasks, while the H-TPU works on repetitive operations.

• SK hynix to place custom tiny CPU cores in its SSD controllers to ease Arm load as drive speeds rocket — each H-TPU reportedly draws 2.5% of a CPU core’s power

Ukraine has targeted Russian Yandex data centers in the latest development in the conflict between the two nations in recent weeks. While bot traffic and online disinformation were distrupted, it also highlights just how fragile Russia's public services are, if data can simply evaporate due to less disaggregated infrastrucutre.

• Ukrainian drones hit Russian data centers, disrupting Russian apps and services — attacks highlight the fragility of the country's digital services

Of course, the world of AI continues to disrupt industries, and sometimes, itself. A few weeks ago we broke down TypeSafe's Jev, a decision model that spits out decisions based on the probability of its inputs. But, in the weeks following, companies have rushed to make its own decision models, and the provenance of the original idea of a decision model like Jev has been called into question.

Elsewhere, mathematicians are warning to not use AI when working and in research, as OpenAI posted 722 manuscripts and proofs in the field, which the community now has to qualify. This is an unprecendented event within the mathematics community, and the burden of non-artifically-generated proof lies in the research community itself.

• Jev-like decision models are flooding the AI market, after researcher claims TypeSafe copied their research — Microsoft, OpenAI among companies releasing lightweight 'System One' AI models
• AI causes uproar in mathematics as experts warn against sharing research with LLMs after OpenAI shares 722 manuscripts and proofs — professor argues that the field must focus on building community over raw solutions

Tom's Hardware Premium Subcription: $29 at tomshardware.com

Don’t miss out on this Tom’s Hardware Premium. Get a full year of access for just $29, or from $7 per-month. Get daily news analysis, deep dives into specialist topics in the semiconductor industry, as well as access to Bench, the largest benchmarking database around.

View Deal

To gain full access to the articles we mention above, as well as unlock Bench, you'll need to purchase a Tom's Hardware Premium subscription. Your support goes a long way in helping us to publish deeper, more insightful articles that might not otherwise be covered.