It’s been a longtime feature of the annual Defcon hacker conference that attendees come away not only with knowledge of new software vulnerabilities and hacking techniques but also an elaborately designed conference badge—often electronic masterpieces embedded with intricate puzzles, complex crypto challenges, hidden Easter eggs, and even the mechanical gear trains of a watch.
Each year’s badge creator endeavors to top previous designs and blow the minds of hard-to-impress hackers. But this year’s badges take a different tack. Instead of the badge designs being the star, it’s what is inside the hardware that will really stand out.
This year, Defcon asked legendary hardware hacker Andrew “bunnie” Huang to create the badges—revealed here for the first time—and they include an innovative open source chip that Huang designed and that aims to do no less than advance the state of security, transparency, and trustworthiness in computing.
NASA, SpaceX Advance Wind Tunnel Tests for Starship Rocket
Engineers at NASA Ames conduct transonic and supersonic wind testing on a 1.2% scale model of the Super Heavy Version 3 rocket that will launch the Starship human landing system.
Credits: NASA
NASA and its industry partners are preparing for next year’s Artemis III demonstration mission by completing new wind tunnel tests on SpaceX’s Super Heavy Version 3 rocket booster. The tests, conducted at NASA’s Ames Research Center in California’s Silicon Valley, focused on better understanding the extreme aerodynamic forces the rocket can experience during re-entry. The recent test series builds on previous testing completed at NASA Ames in 2024.
NASA is working with SpaceX to develop the company’s Starship Human Landing System (HLS) to safely carry astronauts from lunar orbit to the Moon’s surface and back. The upgraded Super Heavy rocket booster is part of SpaceX’s Starship launch system. Version 3 of Starship and Super Heavy is expected to be the basis for the Starship HLS for Artemis III in 2027 and a later crewed lunar landing in 2028.
Engineers at NASA Ames conduct transonic and supersonic wind testing on a 1.2% scale model of the Super Heavy Version 3 rocket that will launch the Starship human landing system.
NASA
Engineers at NASA Ames conduct transonic and supersonic wind testing on a 1.2% scale model of the Super Heavy Version 3 rocket that will launch the Starship human landing system.
NASA
Engineers at NASA Ames conduct transonic and supersonic wind testing on a 1.2% scale model of the Super Heavy Version 3 rocket that will launch the Starship human landing system.
NASA
Engineers at NASA Ames conduct transonic and supersonic wind testing on a 1.2% scale model of the Super Heavy Version 3 rocket that will launch the Starship human landing system.
NASA
Although Starship HLS is a lunar lander designed and built by SpaceX, NASA collaborates with commercial companies to provide access to specialized testing facilities, like the wind tunnels at NASA Ames, and technical expertise, such as the team that set up the wind tunnel testing and helped analyze the results.
“NASA has a lot of experience with unsteady aerodynamics,” said Manish Mehta, discipline lead engineer for the HLS Plume and Aero Environments team, NASA’s Marshall Space Flight Center in Huntsville, Alabama. “We used the agency’s broad experience base of conducting wind tunnel tests for the space shuttle, the SLS (Space Launch System) rocket, and Orion spacecraft to efficiently set up and analyze the wind tunnel testing for the Super Heavy Version 3. In fact, similar testing at the Ames Unitary Plan Wind Tunnel resulted in adding strakes to SLS for Artemis II, so what we learned for Artemis II is helping us get to Artemis III and beyond.”
Testing
SpaceX’s Starship consists of a 33-engine first-stage Super Heavy rocket, or booster, and the second-stage Starship. Version 3 of Super Heavy incorporates many new and upgraded systems, including:
New propulsion systems and new Raptor 3 rocket engines
No engine section skirt, individual engine shrouds, or integrated large-scale base heat shield
Three gridfins on the Super Heavy booster instead of four, with each fin now 50% larger
An integrated hot stage replacing the previous single-use protective interstage
With significant changes to Super Heavy, NASA and SpaceX engineers wanted more information about the steady and unsteady aerodynamic forces the rocket will experience during atmospheric re-entry as it returns to the launch site for refurbishment and re-use.
“When a rocket, or an airplane, flies through air at high speed, it’s subjected to steady aerodynamic forces and moments, and unsteady aerodynamic forces and moments,” explained Jayanta Panda, unsteady aerodynamics subject matter expert at NASA Ames and part of the Human Landing System Plume and Aero Environments team. “An example of a steady aerodynamic force would be when air smoothly flows over the surface of the rocket as it ascends. An unsteady aerodynamic force would be air ‘buffeting,’ or hitting, certain areas the rocket at less predictable times and potentially causing vibrations.”
Wind tunnels
NASA and SpaceX used a 1.2% scale model of the Super Heavy Version 3 in the transonic wind tunnel and the supersonic wind tunnel at NASA Ames for testing. The transonic tunnel blasts scale models of rockets or aircraft with air at speeds ranging from Mach 0.2 to Mach 1.4 (Mach 1 is the speed of sound, or about 761 miles per hour). The smaller supersonic tunnel fires winds at higher speeds, from Mach 1.55 to Mach 2.5. The wind tunnel tests on Super Heavy Version 3, conducted in late 2025, used both tunnels to measure steady and unsteady air flows on the surfaces of Super Heavy.
“Resulting wind tunnel data on steady forces and moments helps predict how the rocket will react to forces in the atmosphere during re-entry so the flight software can effectively guide the rocket during flight,” Mehta said. “The unsteady pressure data helps engineers understand the environment around the rocket as it re-enters Earth’s atmosphere. Engineers use that information as one input into software that analyzes loads on the rocket.”
Through the Artemis program, NASA is returning humans to the Moon for scientific discovery, economic opportunity, to establish an enduring human presence on the lunar surface, and build the foundation for the first crewed missions to Mars – for the benefit of all.
The Company posted quarterly revenue of $109.4 billion, up 16 percent year over year.
[…]
“Today, Apple is proud to report our strongest June quarter ever, with double-digit revenue growth across iPhone, Mac and Services, and in every geographic segment,” said Tim Cook, Apple’s CEO.
You can’t argue with the strong iPhone and Mac sales figures, and Apple specifically called out the MacBook Pro and the MacBook Neo, which Cook said “has been especially popular” and suggested that the company is still struggling to make them fast enough.
It’s admirable (and sensible) that Cook isn’t pointing fingers at anyone but customers who want Apple products, but the key phrase in his statement is “the supply chain just has less flexibility in it than normal.” This is an acknowledgement that while in the past, Apple might have been able to boost chip production from its partners, today the chip industry is maxed out thanks to the wild building of chips for AI applications.
[…]
As an aside, Cook also alluded to reports that Apple has tried to get the U.S. government to authorize purchase of RAM from some forbidden Chinese chipmakers, which has aroused political ire.
Se colocou uma música menos conseguida num post antigo do Instagram ou se o áudio foi removido por direitos de autor, já não precisa de apagar a publicação. Saiba como trocar a banda sonora mantendo os gostos e os comentários.
There’s been a shift toward using a single DIMM of DDR5 to skirt the worst of the RAM pricing crisis. AMD has suggested to us previously, at least, that it’s an option PC builders are exploring, and it’s a wrinkle that’s shown up in dozens of prebuilts; one that we’ve been careful to avoid when recommending products, particularly during peak shopping events. But just how much gaming performance are you giving up with a single DIMM when using one of the best CPUs for gaming?
The conventional wisdom is that you’re giving up quite a bit of performance, resting on the ideas of previous DDR generations, where going down a single stick will cut your bandwidth in half. The same is true with DDR5, but things are a bit messier, which we’ll dig into later.
At a high level, using a single DDR5 DIMM doesn’t compromise gaming performance by nearly as much as you might expect. We saw drops of around 8% to 10% across Intel and AMD CPUs, and less than a 5% difference with AMD’s Ryzen 7 9800X3D.
A proper dual-channel configuration is still the way to go for optimal performance. However, our testing shows that even a single DDR5 DIMM offers superior performance to dual-channel DDR4 on CPUs that support both memory generations.
Using a single DIMM isn’t ideal, and it brings up compatibility issues if you plan on adding another DIMM down the line (another area we’ll touch more on in-depth later). But as a stopgap measure amid surging DRAM prices, a single DIMM of DDR5 holds up surprisingly well in gaming scenarios, and it represents a nice on-ramp into a DDR5 platform while prices are out of control.
Testing single-DIMM DDR5 in games
The geomean below speaks for itself. Moving down to a single DIMM drops some performance, but it’s not nearly as stark as I expected. There are some important notes about how we tested before dissecting the results, however.
As usual, we tested at 1080p with a mixture of High and Ultra settings, using Nvidia’s RTX 5090 Founder’s Edition to minimize the influence of the GPU. That’s even more important here, as we’re not evaluating CPUs but the memory that feeds the CPU. We're looking at a separate GPU and CPU here, as well; dual-channel memory is vital if you're using an iGPU.
We also wanted to create a realistic testing scenario. Assuming you’re going for a 2x16GB configuration, that means starting with a single 16GB DIMM, not a 32GB one. That’s what we did here. Our single-DIMM configurations run with just 16GB of memory, while dual-DIMM configurations run with 32GB. Capacity isn’t a major limitation for our testing here, though it will pose performance issues if you’re running several applications alongside your game.
(Image credit: Tom's Hardware)
The Ryzen 5 7600X was our biggest loser, dropping 10.9% of its average performance across our 13-game geomean with a single 16GB DIMM. The Core Ultra 7 270K Plus, meanwhile, dropped 8.7% of its average performance, while the Core i5-14600K similarly dropped 8.3% of its performance. The 1% lows are especially notable here, scaling with average performance rather than falling off a cliff.
Unsurprisingly, the Ryzen 7 9800X3D fared the best, dropping just 2.8% of its performance on average. As you’ll see through our individual game tests, the Ryzen 7 9800X3D shows identical performance with two DIMMs and a single DIMM in multiple games. The large and speedy L3 minimizes the impact of using a single DIMM, giving AMD’s 3D V-Cache CPUs another notch in their belt (as if they needed another).
Tom's HardwareTom's HardwareTom's Hardware
Our other geomeans don’t say much. A single-DIMM configuration consumes a few watts less, leading to slightly lower efficiency when balanced against the lowered performance.
(Image credit: Tom's Hardware)
007 First Light, the newest game in our test suite, shows good scaling across our test pool, so it’s a good place to start. The Core Ultra 7 270K Plus dropped 8.9% of its performance with a single DIMM, while the 9800X3D dropped 6.5%. The 14600K is worth highlighting here. A single DIMM lost us 7.9%, but that’s still better than dual-channel DDR4, and by a decent 6.8% margin.
(Image credit: Tom's Hardware)
This trend is fairly consistent, too. Starfield is punishing for single-DIMM configurations, with the 14600K losing 10.6% of its average performance. Even then, a single DDR5 DIMM is 6.9% faster than dual-channel DDR4. Intel platforms can scale up to high data rates, and we use 7,200 MT/s for Intel chips. With slower speeds, the performance will equalize. Still, I’d strongly consider a single DDR5 DIMM if you’ve been eyeing an LGA 1700 platform with DDR4, assuming you’re primarily focused on gaming.
Starfield shows the benefits of 3D V-Cache in this single-DIMM configuration clearly. While the performance losses with other CPUs are brutal in this title, the 9800X3D dropped just a few frames, a 1.6% decrease in average performance.
(Image credit: Tom's Hardware)
Doom: The Dark Ages is similarly harsh on single-DIMM configurations. The Ryzen 5 7600X dropped 17.3% of its average performance, and the 270K Plus fell by 14%. The 9800X3D holds up well, however, with just a 4.2% average performance loss. Marvel’s Spider-Man 2 (in the gallery below) shows a similar dynamic.
(Image credit: Tom's Hardware)
However, the addition of 3D V-Cache mainly helps in these games where we see bigger losses with other CPUs. There are several games in our suite where the performance loss between a single DIMM and dual-channel configuration was minor, particularly in GPU-heavy titles. The Last of Us Part One shows that in action.
(Image credit: Tom's Hardware)
Similarly, the performance loss is minor in Counter-Strike 2, just a handful of frames even north of 600 FPS. Even the Core i5-14600K, which was hit the hardest by far, dropped just 5.9% of its average performance (and it’s still faster than a dual-channel DDR4 configuration).
You can browse the results of the other games we tested in the gallery above, though they tell much the same story as the titles we highlighted. The two biggest learnings from this testing are that a single, fast DDR5 DIMM often outpaces dual-channel DDR4, and that AMD’s 3D V-Cache can smooth over the rough edges of using a single DIMM, particularly in those memory-sensitive titles.
How we tested
We tested on our normal test bed that we use for CPU reviews, with the only difference being dropping one of our DIMMs. As usual, we enabled XMP/EXPO for testing, both on single- and dual-DIMM configurations, opting for higher speeds on Intel platforms. AMD’s sweet spot is around 6,000 MT/s.
The data rate is noteworthy for the DDR4 comparison, in particular. With slower DDR5, we expect the delta between dual-channel DDR4 and single-DIMM DDR5 to decrease.
In addition to enabling XMP/EXPO, we disable Virtualization-Based Security (VBS), enable Resizeable BAR, and disable any automatic overclocking features that aren’t covered by warranty, including AMD’s Precision Boost Overdrive and Intel’s Extreme performance profile.
DDR5 has much higher data rates compared to DDR4 (2x or more), and that’s the result of significantly more complexity on DDR5 DIMMs compared to DDR4. Notably for this testing are the two 32-bit subchannels of DDR5, as well as enlarged prefetch and bank groups. You are getting half the total bandwidth with a single DDR5 DIMM, absolutely, but it’s a little more nuanced than simply running your memory in a single-channel configuration.
DDR4 uses a single 72-bit channel, and 64 of those bits transfer data (the additional 8 are for ECC). With DDR5, you get two 40-bit subchannels, 32 of which are used for data transfer. Critically, these subchannels have their own Command/Address (CA) interface, allowing them to operate independently of each other.
In addition, DDR5 doubles the prefetch size to 16N, doubles the number of banks and bank groups, and increases burst length from 8 bytes to 16 bytes. The math here is straightforward enough. With DDR4, you can transfer 64 bits of data over the interface at a burst of 8 bytes, giving you 64 bytes of data and aligning with cache line sizes. Each of the subchannels in DDR5 can transfer 64 bytes of data, as well, with a 32-bit width and burst length of 16 bytes.
It’s pseudo-dual-channel memory on a single DIMM. There are two independent memory operations that can happen each cycle, both of which transfer 64 bytes of data. That is a big reason why even a single DDR5 DIMM (particularly one with a higher data rate) can outpace even a proper dual-channel DDR4 configuration.
The common wisdom about using a single DIMM comes mainly from previous DDR generations, where you can only get that single 64-byte transfer each cycle, necessitating a second DIMM to have two independent operations.
Let me be clear: two DDR5 DIMMs are still better. Instead of two operations per cycle, you can perform four. The main difference is that, for modern CPUs and games, just those two operations give you the vast majority of your performance, and you don’t run into a massive performance bottleneck as you do with just a single operation via DDR4. After all, there are still a lot of DDR4 systems out there.
On the issue of compatibility
The elephant in the room with upgrading to a DDR5 platform with a single DIMM is compatibility. If you get a single DDR5 DIMM now, you should plan to add another in the future. Unfortunately, there aren’t a lot of easy answers when it comes to combining two DIMMs that aren’t in a kit together. It should work, but mixing and matching can lead to stability issues and/or lowered performance.
Generally, combining two DIMMs of the same spec should work. For instance, if you bought this Patriot Viper Venom DDR5-6000 CL30 DIMM to start your new build, and then bought a second one later, it should work. “Should” carries a lot of weight on its shoulders, however. Maybe Patriot sources DRAM from multiple places, or your second DIMM’s timings don’t completely align with your first. That could cause stability or performance issues.
There have been some attempts to get past this hurdle, such as the ROG Certified program with select Asus motherboards, but it’s impossible to know for sure if two separate DIMMs will work together or not without testing. The best you can do is to make sure you buy the same brand of memory, with the same spec, if you add a second DIMM. Make sure to keep your receipt handy in the event you need to return it, as well.
Once you have your second DIMM, make sure to test it first at stock JEDEC speeds/timings (no XMP/EXPO). Run games or applications you normally run, and then use a utility like MemTest64 to check for stability. I’d recommend keeping your PC running in this state for a few days to see if any BSODs, application crashes, or poor performance issues pop up. After that’s done, repeat the process with XMP/EXPO enabled, assuming you’re using matching DIMMs (incompatible XMP/EXPO profiles won’t work together).
Assuming DRAM prices continue at an elevated rate before you upgrade, you may want to consider investing in a dual-channel kit down the line and selling your single DIMM. Even kits can run into compatibility issues, though the likelihood is far, far lower.
Login
