For your money, you get two 16GB DDR4-3600 modules with CL18 timings, aluminum heat spreaders, customizable RGB lighting and Intel XMP 2.0 support.
Today's best DDR4 RAM deal
32GB TridentZ RGB Series DDR4 3600 RAM: was $257.49 now $214.99 This memory kit provides 32GB of dual-channel DDR4-3600 RAM across two 16GB modules, with CL18 latency, 18-22-22-42 timings, aluminum heat spreaders, RGB lighting and XMP 2.0 support. For the full discount, use code LUYRF222View Deal
Why we recommend it
32GB is enough memory for demanding multitasking, gaming and creative applications. The DDR4-3600 speed is also a good fit for compatible DDR4 systems, particularly if you're upgrading an older machine that's currently running slower or lower-capacity memory.
The two matched 16GB modules provide a dual-channel configuration, with timings of 18-22-22-42 at 1.35V.
G.Skill's Trident Z design adds aluminum heat spreaders and customizable RGB lighting, making this a good choice if your PC has a windowed case and you want your memory to complement other RGB components.
Price context & historical value
Newegg sells the 32GB Skill Trident Z RGB kit for $257.49, and it's currently available for $249.99. Using coupon code LUYRF222 brings the final price down to $214.99.
While that's a decent price in today's inflated memory market, it's worth remembering just how much cheaper DDR4 used to be. Newegg sold this exact G.Skill kit for just $69.99 in August 2023, less than a third of its current discounted price.
Unfortunately, it's unlikely to drop down to that level again, so the current deal is worth considering if you need to upgrade your DDR4 system now.
Should you buy it?
✅ Buy the 32GB G.Skill Trident Z RGB kit if...
You have a compatible DDR4 desktop and want to increase its memory capacity without replacing your motherboard and processor to move to DDR5. The combination of 32GB capacity and DDR4-3600 speed makes it a solid upgrade for an older system.
❌ Skip the 32GB G.Skill Trident Z RGB kit if...
You're building a completely new PC around a recent processor and motherboard. In that situation, I'd seriously consider DDR5 instead, as you'll be investing in the newer memory standard rather than spending more than $200 on DDR4 that you won't be able to carry over to a future platform.
The Catch: What to know before you buy
You'll need to enable XMP, DOCP or A-XMP in your BIOS to reach the kit's advertised DDR4-3600 speed on compatible hardware. Otherwise, the memory will initially run at its default SPD speed.
Compatibility is also important. G.Skill recommends checking its motherboard QVL, and it specifically warns against mixing separate memory kits, even when their specifications appear identical, as doing so can cause stability problems.
One sleepless night can blur the brain, but just 20 minutes of exercise or a well-timed nap may help new memories survive for days. In a McGill University-led study, participants who exercised for 20 minutes or slept for 90 minutes after prolonged wakefulness performed about 22 percent better on a delayed memory test than those [...]
The human brain is remarkably adept at juggling different kinds of information, allowing us to switch rapidly between tasks and adapt our behavior as circumstances change. A new study by MIT neuroscientists takes a closer look at how the brain accomplishes this, providing evidence for flexible neural modules that may help support this cognitive versatility.
Using mice, scientists found that a group of neurons in the prefrontal cortex could switch between retaining sensory information and maintaining an action plan in working memory.
“We found that the brain doesn’t dedicate a separate group of neurons for every type of information,” says Yuma Osako, an MIT postdoc and the lead author of the new study. “Instead, it uses the same populations of neurons to perform the same computation on different kinds of information, which means the same subset of neurons can hold both an action and a sensory stimulus in working memory,”
The findings also lend support to a longstanding theory known as compositionality—the idea that cognition can be constructed by combining and reusing neural components.
“One of the solutions that’s always been proposed has been this idea of compositionality — that you can take pieces of cognition that perform part of a task and reuse them in another task,” says Timothy Buschman, a professor at the Princeton Neuroscience Institute, and the senior author of the paper, in a statement.
Another way to see this is to view it as “cognitive Legos,” Buschman suggests. Instead of building a new neural circuit when a new task starts, the brain reuses existing circuits and combines them in different ways.
The Test
To see how this neural network operates, the scientists trained mice to compare two tones and respond depending on whether the sounds matched. As the mice performed the tasks, the scientists recorded electrical activity from thousands of neurons.
“When mice do tasks that test whether memory computations can be reused, the answer is they are,” notes Mriganka Sur, the Newton Professor of Neuroscience at MIT’s Picower Institute for Learning and Memory. “There are subspaces of functional activity in the prefrontal cortex that can be the substrate of mixing and matching toward flexible cognition.”
Two important components of the task allowed the researchers to examine how different kinds of information were maintained. First, the mice had to remember the initial tone for a short period. They then had to retain their decision and form an action plan.
In the parietal cortex, neurons were dedicated to storing information about the tone. However, researchers found something different in the prefrontal cortex: a specific group of neurons could switch roles.
During one stage of the task, the circuit helped maintain a sensory memory. During another, the same population was involved in maintaining an action plan. This flexibility may help explain how a finite number of neurons can support an enormous range of behaviors.
“Our everyday life requires us to temporarily hold many different kinds of information. One big question is how the brain can represent an unlimited variability of information using only a finite number of neurons,” Osako says.
The researchers view these reusable neural circuits as potential building blocks for flexible cognition. More broadly, the findings suggest that the brain may be less like a machine filled with components dedicated to individual tasks and more like an interconnected system that continually repurposes its existing neural resources.
Chrissy Newton is a PR professional and the founder of VOCAB Communications. She currently appears on The Discovery Channel and Max and hosts the Rebelliously Curious podcast, which can be found on YouTube and on all audio podcast streaming platforms. Follow her on X: @ChrissyNewton, Instagram: @BeingChrissyNewton, and chrissynewton.com. To contact Chrissy with a story, please email chrissy @ thedebrief.org.
The MacVoices Panel tests a new live-production platform while discussing Apple’s reported $5 trillion valuation, and the end of Tim Cook’s earnings-call era. The pros and cons of Apple’s new device upgrade and leasing program is the subject of intense debate over factors such as balloon payments, ownership, carrier lock-ins, AppleCare costs, convenience, and flexibility. […]
Seeing the red 'limited-time deal' banner on some DDR5 RAM at Amazon should elicit some excitement given the rarity of price cuts lately. But when you click through and see the price for this 32GB Corsair Vengeance kit, you'll wonder if there's genuinely a deal here at all — and other wannabe buyers aren't wholly convinced, either.
"It's 'only' 4x the price it should be," wrote one user named Technikal on HotUKDeals, a community that shares deals from across the web and asks the community to vote whether they think they are hot or cold. "The current average is 5x," he continues, "so, in that sense it's a good deal...I guess."
A muted response, that's for sure. And while it may be true that the deal price of the RAM kit is technically one of the lowest prices available right now, it's so far from the much lower prices we saw before the RAM crisis turned buying memory for your PC into a depressing exercise.
The price for this specific RAM kit has increased dramatically over the last 12 months at Amazon. (Image credit: Future / Keepa)
In fact, even just a year ago, this very same RAM kit was available for £99.99 when it wasn't even on sale, according to price history information from Keepa.
Given that, it's no surprise that shoppers have reacted negatively to Amazon's flash deal. Unfortunately, though, it doesn't look like things are getting any better when it comes to RAM prices. A report from Samsung suggests the RAM crisis is just getting started, and it could extend to other components in the near future.
Sadly, the news is grim, and these prices could be with us for a while yet. So, yes, it may be difficult to feel happy about seeing supposed RAM deals like this, but from what I'm seeing and hearing, I don't think they're going to get any better, either.
Amazon has launched a limited-time deal on a set of much-sought-after Corsair Vengeance 32GB DDR5 6000MHz CL36 RAM — the sweet spot for most PC gaming builds. Even though you could technically argue this is a deal, as it's the lowest price we've seen in months, it's still astronomically more expensive than the record-lows of under £100 from last year. However, with the ongoing PC component price crisis showing no signs of ending, this is likely one of the better offers we'll see for some time.View Deal
It appears that the RAM crisis is getting worse than ever as we head further into 2026. I've been closely watching and reporting on memory price hikes — which were soon followed by other PC component cost increases — since this crisis first began, and the worrying thing is, I can't recall a gloomier stream of negative news than I've witnessed over the past few weeks.
This isn't just about RAM, of course, but other components, including CPU price rises, and devices themselves — PCs and laptops (and phones). But most notably over the past month it's been graphics cards in the crisis limelight. We've witnessed the resurrection of old GPUs to try to bolster stock levels of more affordable cards, and more recently we've been subjected to what seems like an endless parade of negativity about imminent price hikes.
While a good deal of this is individual pieces of regional activity with GPU pricing, it's obvious that these hikes are happening as part of a concerted shift, one that will surely be reflected globally — it's not like Asian markets are in a bubble of their own.
The overarching theme is a worsening of price hike misery, and that the pricing storm is likely to intensify this year, and probably in 2027, too. While previously the RAM crisis has been more of a light-and-shade affair, now it feels like the depression has been turned up a notch. Whereas before, there was certainly more darkness than light, we've had notable spots of relief where an exec in the memory industry stepped forward and theorized that maybe things aren't as bad as we think. But lately those embers of optimism appear to have burned out.
To me, it feels like there's a shift underway towards a full acknowledgement that we really are going to experience a lot more pricing pain in the foreseeable future. If it's so bad that even Apple is purportedly scrambling to secure RAM supplies, and is having trouble doing so, I think it's time to turn up the worry meter by a notch or two.
What should you do?
(Image credit: Future / John Loeffler)
Here's my advice on the current landscape with PC components and hardware (season it appropriately). Obviously, worrying isn't going to help you, but awareness of what might be sensible upgrades or purchases to make now will, based on the likelihood of hikes (we could call it the 'hike-lihood' – or maybe not). Nothing is certain about the future of components, of course, but it does feel like there's a clear area in which the potential price hikes are a bigger threat: GPUs.
As I discussed above, there's a lot of recent evidence that graphics card pricing is going up. Yes, it's a collection of rumors in the main, but there's a lot of it all pointing in a very similar direction — and a video RAM toll was always going to be exacted in the end. We've seen it already with higher-end graphics cards, and now I believe we're going to see it in the mid-range, and even budget models. Indeed, top-end GPUs are likely to get even more expensive as well.
If you're thinking about a GPU upgrade for this year or next, given all this, I think now really is the time to buy — especially with a lower-tier model, as you can still get more affordable graphics cards at their MSRP. Ditto for mid-rangers, although it's tougher to recommend higher-end Nvidia boards when they are already so expensive.
As noted, though, that could get worse. And granted, you may still want to wait for Black Friday — it's not that far off, and there might be some GPU deals then. However, I wouldn't bank on anything hugely compelling (or discounts that aren't offset by the price rises which are predicted to take hold in the next few months).
So, if there's one PC component that I think you should buy now, it's a graphics card. I think the signs are pretty clear on that (and see here for my recommendations on different models to consider). By extension, higher-end gaming laptops could also be a wise move for purchasing sooner rather than later. That's for the same reason, really — they have beefy (mobile) GPUs (with plenty of VRAM in some cases).
I don't think it's a bad move to buy any laptop, by the way, gaming or not (away from the higher-end), come your earliest window of opportunity. I think we'll see further price rises here, too, and RAM is going to cause more upward pricing pressures for notebooks — just look at Framework's recent revelation as mentioned.
(Image credit: Peter Hoffmann)
What about RAM and SSDs themselves? While system memory kits and storage have already seen huge price rises, it now looks like there's worse to come (somehow).
On the other hand, there's a ceiling as to how expensive this stuff can get before buyers withdraw from what they see as an increasingly unrealistic and out-of-touch market. This is a much more difficult call to make, but if you can find something that looks relatively reasonably priced, I don't think you'll regret the purchase in the next couple of years. But that said, I can't recommend a RAM upgrade in particular at current pricing levels, unless it's unavoidable, frankly.
Above all, though, consider a GPU upgrade if you're in the market for a new card, or you think you'll need one in the next couple of years (yes, I think it's wise to be looking quite far down the road here).
The other thing to bear in mind is that there's a danger there will be a rush for GPUs, a flurry of buying that puts further strain on supply and therefore prices. As VideoCardz reports, one Japanese retailer has warned of Nvidia RTX 5000 graphics card sales increasing "sharply", even just with the news of rumored price increases — let alone the confirmation. Prices are already rising and restocks of the RTX 5070 Ti and RTX 5080 are already being labelled as "unstable", hinting that inventory could start to dry up quickly.
That's just one report, of course, and it pertains to the high-end of the market, so it's not something to start panicking about yet. But it does make some sense that if pricing begins to creep up, more PC owners may start to act on GPU upgrades. I wouldn't blame them, frankly.
3D render claiming to depict Apple’s first foldable iPhone (image: Jon Prosser)
Apple and its suppliers are scrambling to secure enough DRAM (mobile memory chips) ahead of the expected September 2026 launches of the iPhone 18 series and the foldable iPhone Ultra, according to semiconductor analyst Tim Culpan’s exclusive reporting for Culpium.
With under six weeks to the anticipated debut, assemblers are rushing shipments of DRAM. Apple primarily sources from Micron, with additional supply from SK Hynix and Samsung. Severe shortages due to RAMageddon have left processor wafers sitting idle, awaiting packaging—creating pressure to avoid launch shortages.
The issue stems from Apple’s A20 Pro chips (built on TSMC’s 2nm process), which use Wafer-Level Multi-Chip Module (WMCM) packaging that integrates the processor and DRAM at the wafer level. This means DRAM must be available for packaging; unfinished dies cannot simply be stockpiled and completed later. Reports indicate TSMC is holding roughly $1 billion worth of unpackaged Apple processors delayed by the missing memory.
DRAM has been tight for months, driven in large part by memory makers prioritizing higher-margin chips for AI data centers over consumer devices. This has already affected other Apple products (such as certain Macs) with longer wait times. Apple and its partners (including assemblers like Foxconn) expect to meet initial launch demand, but retail stock could deplete quickly and online preorder wait times may stretch afterward. The foldable iPhone Ultra, with its more complex design and lower planned volumes (historically projected in the low millions for the second half of the year in related analyst notes), is viewed as particularly vulnerable to limited availability.
Culpan’s sources indicate Apple plans roughly 200 million combined units across the iPhone 18, 18 Pro, and Ultra over the product cycle, with the Ultra representing under 10%.
This memory constraint adds to broader industry dynamics of rising memory costs and competing demand from AI infrastructure, which have already influenced pricing discussions and production priorities for Apple hardware.
MacDailyNews Take: Buyers interested in the new Pro models or the Ultra at launch would be advised to pre-order very, very promptly!
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
RAM production capacity at Samsung Electronics, SK hynix, and Micron Technology is reportedly committed through the end of 2027, leaving consumer electronics manufacturers with little room to secure additional memory outside existing supply agreements.
Industry reports indicate that the three dominant manufacturers have allocated their available DRAM and high-bandwidth memory production for next year. Long-term contracts with cloud companies, AI chip developers, automakers, and device manufacturers have effectively reserved the output those suppliers expect to produce.
The claim does not mean every RAM module that will be sold in 2027 has already been manufactured or purchased by a consumer. It refers to future production capacity being assigned to customers before the chips leave the factory.
It also does not cover every memory producer worldwide. Chinese manufacturer ChangXin Memory Technologies, known as CXMT, and several smaller suppliers continue to produce DRAM. However, Samsung, SK hynix, and Micron control most of the advanced global DRAM market, making their available capacity the primary constraint for companies such as Apple.
RAM Production Is Already Committed
Selling out a year of production capacity means manufacturers have received enough contractual commitments to allocate the output they expect from their factories. Customers may have reserved specific volumes, product types, or manufacturing capacity under agreements that include minimum purchases and pricing protections.
Micron has been shifting toward this model through strategic customer agreements. The company disclosed 16 multiyear deals covering customers across data centers, consumer devices, automotive products, and industrial equipment.
Those agreements typically run for several years and include binding volume commitments. Some use fixed prices, while others establish price floors and ceilings. Micron has said that its signed contracts represent approximately $100 billion in expected revenue over their remaining terms.
Such agreements give memory manufacturers more predictable revenue than the traditional DRAM cycle, in which aggressive expansion frequently created excess supply and collapsing prices. They also give large buyers assurance that chips will be available for planned products.
The disadvantage falls on customers that did not reserve enough capacity early. A company needing more memory than its contract provides may have to pay higher prices, accept a different specification, delay production, or find another qualified supplier.
SK hynix CEO Kwak Noh-jung has described 2027 as potentially the worst year in the memory industry’s history from a supply perspective. He expects customer demand to remain above the company’s production capacity beyond 2030, despite major investments in new factories.
Samsung has issued a similar warning, expecting the constraints to become more severe in 2027 and continue into 2028. Analysts at UBS have projected that the DRAM market could remain undersupplied until at least the second quarter of 2028.
Image Credit: Bloomberg / Getty Images
AI Memory Consumes More Factory Capacity
The shortage originates primarily from the rapid expansion of artificial intelligence infrastructure. AI accelerators require high-bandwidth memory, or HBM, to move large amounts of data between memory and processors at extremely high speeds.
HBM and the LPDDR memory used in Apple devices are different products. An iPhone or Mac is not competing directly for the same finished memory package installed beside an Nvidia AI accelerator.
They still depend on overlapping manufacturing resources. HBM begins with advanced DRAM dies, and producing a given capacity of stacked HBM consumes considerably more wafer output and packaging resources than conventional memory. Suppliers have also concentrated spending, equipment, and engineering work on the more profitable AI segment.
Every production line converted or optimized for HBM can reduce the industry’s flexibility to make standard server DRAM, PC memory, or mobile LPDDR. New investment may increase total memory output while doing little to relieve the parts of the market serving phones and personal computers.
Cloud companies are willing to enter multiyear agreements and make large financial commitments because a shortage of memory can leave costly AI processors unused. Memory manufacturers therefore have a powerful incentive to prioritize these customers over consumer device companies seeking lower component prices.
The pressure is also moving beyond HBM. AI servers need large quantities of conventional DDR5 memory, while storage systems supporting model training and inference consume enterprise NAND flash. This spreads the shortage across several categories that were previously less exposed to AI infrastructure spending.
Apple Faces Expensive Supply Decisions
Apple purchases enormous volumes of LPDDR memory for the iPhone and iPad, along with memory used across the Mac, Apple Watch, Vision Pro, Apple TV, and HomePod product families. Its size gives it substantial negotiating power, but it cannot purchase capacity that suppliers have already assigned to other customers.
The company has acknowledged that memory prices are continuing to rise. Tim Cook recently characterized the pricing environment as a “hundred-year flood” and said Apple expects to pay more for memory during the September quarter.
Apple has already raised prices across several Mac and iPad models while reducing or removing some high-memory Mac configurations. Although memory availability is not the sole influence on every product decision, eliminating configurations can help a manufacturer concentrate limited components around the models expected to sell in larger volumes.
The design of Apple silicon makes memory allocation especially sensitive. Unified memory is integrated closely with the processor package and cannot be upgraded after purchase. Apple must decide how many chips to secure for each configuration long before customers begin ordering those machines.
A shortage can therefore affect more than the cost of an individual RAM component. It can influence base memory capacities, upgrade pricing, the number of configurations offered, production volume, and the timing of product introductions.
Apple can absorb part of the increase through its margins, negotiate longer agreements, adjust retail pricing, or redesign its product mix. Each option carries a cost. Absorbing higher prices reduces profit per device, while passing them to customers could weaken demand. Restricting configurations risks pushing professional users away from Macs that no longer offer enough unified memory.
The company is reportedly evaluating DRAM from CXMT for products sold in China. Adding a fourth major source could improve access to conventional memory, but qualification takes time because Apple must confirm performance, reliability, manufacturing consistency, and compatibility with its hardware.
The political environment creates another obstacle. US lawmakers have raised national security concerns about American companies purchasing memory from Chinese semiconductor manufacturers. Apple could consequently find that a supplier is technically suitable for China-market products but difficult to introduce across its international supply chain.
Sold Out Does Not Mean Stores Will Be Empty
The reported capacity commitments do not guarantee a universal shortage of finished devices. Apple and other major manufacturers will still receive memory under agreements they have already negotiated.
Products can also continue reaching stores while component prices rise. Manufacturers may reduce discounts, increase retail prices, limit premium configurations, or give priority to their most profitable devices before allowing a shortage to stop production completely.
Contracted capacity can change if customers reduce orders or if AI infrastructure investment slows. Memory makers could redirect production between customers and product categories, although converting lines and qualifying different chips cannot happen immediately.
CXMT could also expand faster than expected, particularly in mainstream DRAM. Its products have reportedly attracted interest from HP, Asus, Acer, Dell, Lenovo, and Apple as established suppliers devote more resources to AI memory. CXMT is unlikely to replace the combined output of Samsung, SK hynix, and Micron during 2027, but it may provide an alternative for products sold outside the United States.
New production facilities offer little immediate relief. Semiconductor factories require construction, equipment installation, process development, yield improvement, and customer qualification before reaching dependable volume.
Micron expects the first wafer output from its initial new Idaho facility around the middle of 2027, with a second facility following in late 2028. Those first wafers will still need to pass through the production ramp and customer qualification process, leaving much of their effect on available device memory for 2028 and later.
In a striking reversal of traditional supply-chain dynamics, China’s ChangXin Memory Technologies (CXMT) has rebuffed Apple’s requests for lower prices on mobile DRAM, according to industry reports. The move underscores a severe shortage in general-purpose memory chips and hands greater negotiation leverage to South Korea’s Samsung Electronics and SK Hynix.
Apple had been exploring CXMT as a potential new supplier of LPDDR5X and other mobile DRAM for upcoming iPhones and smart devices, aiming to ease manufacturing cost pressures amid soaring memory prices. Sources indicate the iPhone maker sought further price reductions during negotiations. CXMT refused, instead holding firm at levels matching or even exceeding those charged by Samsung and SK Hynix.
The Chinese manufacturer’s bold stance stems from robust domestic demand. Companies such as Huawei and Xiaomi have locked in large volumes of CXMT’s DRAM through high-priced, long-term contracts. This push for supply-chain self-reliance has been intensified by ongoing U.S. semiconductor restrictions on China. With strong local orders secured, CXMT has little incentive to accommodate Apple’s stricter quality requirements or discount demands.
The episode highlights a broader market shift. Major memory producers have redirected significant wafer capacity toward high-bandwidth memory (HBM) to meet surging demand from AI servers and accelerators. This has sharply reduced output of conventional DRAM products like DDR5 and LPDDR5X, triggering a pronounced shortage—sometimes called “memflation” — and rapid price increases across the sector.Big tech firms that once used Chinese low-cost alternatives as leverage in annual pricing talks with primary suppliers now find that strategy ineffective. The low-cost “bypass” route has effectively closed, transferring pricing power from device makers to memory manufacturers.
For Samsung and SK Hynix, the developments create a favorable cycle. CXMT’s absorption of substantial Chinese commodity DRAM demand relieves the Korean firms of pressure to supply lower-margin general-purpose chips. They can instead prioritize higher-value AI memory products, including HBM4, LPCAMM2, and enterprise SSDs. Industry observers note that this positions the Korean companies with stronger control over global pricing in negotiations with major customers in the second half of the year.
One semiconductor industry source described the situation as unusual: CXMT is effectively supporting a floor under general DRAM prices while Samsung and SK Hynix consolidate leadership in premium AI memory. As a result, the Korean firms’ operating margins and market influence are expected to strengthen further.
Earlier this year, Apple had lobbied the U.S. government for clearer approval to source from CXMT, which appears on a Pentagon list of companies with alleged Chinese military ties. Memory cost pressures had already prompted Apple to raise prices on some Mac and iPad models. The latest reports suggest any potential deal with CXMT would not deliver the hoped-for cost savings.
MacDailyNews Take: The memory shortage is widely expected to persist well into 2027 and possibly beyond, as new fab capacity takes years to come online and AI-related demand continues to outpace supply growth. In this environment, the balance of power in the DRAM market has tilted decisively toward suppliers, particularly those focused on advanced, high-margin products.
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
A fruit fly hunting a piece of rotting fruit or a mate navigates by smell, following plumes of odor. Out in the wild, turbulent air swirls these plumes into a chaotic, broken landscape—dense chemical filaments laced through long stretches of clean air. A fly trying to find the source gets the smell in stutters, from constantly shifting directions, with no guarantee that a next whiff is coming at all.
Scientist had little idea of how fruit flies manage this chaotic signal with a brain the size of a pinhead. For a long time, biologists stood by the “surge and cast” model, which posited that insects solved this with hardwired reflexes. The idea was that, when a fly registers the plume with olfactory neurons in its antennae, it simply flies upwind until it’s gone and then flies side to side attempting to catch it again. But now a team led by Vanessa Ruta, a neuroscientist at the Rockefeller University, has shown that fruit flies do something far more advanced.
A treadmill for flies
The trouble with the traditional surge and cast model is that it struggles to explain how an insect tracks a meandering plume across long distances. Chemical cues floating in the air in the natural environment are often sparse and unreliable. But those same features make the mechanism behind olfactory navigation notoriously difficult to test. "Odors are invisible," Ruta says, "and often they're carried along by turbulent airflow." We’ve got no way of knowing what the animal is smelling from one moment to the next.
A fruit fly hunting a piece of rotting fruit or a mate navigates by smell, following plumes of odor. Out in the wild, turbulent air swirls these plumes into a chaotic, broken landscape—dense chemical filaments laced through long stretches of clean air. A fly trying to find the source gets the smell in stutters, from constantly shifting directions, with no guarantee that a next whiff is coming at all.
Scientist had little idea of how fruit flies manage this chaotic signal with a brain the size of a pinhead. For a long time, biologists stood by the “surge and cast” model, which posited that insects solved this with hardwired reflexes. The idea was that, when a fly registers the plume with olfactory neurons in its antennae, it simply flies upwind until it’s gone and then flies side to side attempting to catch it again. But now a team led by Vanessa Ruta, a neuroscientist at the Rockefeller University, has shown that fruit flies do something far more advanced.
A treadmill for flies
The trouble with the traditional surge and cast model is that it struggles to explain how an insect tracks a meandering plume across long distances. Chemical cues floating in the air in the natural environment are often sparse and unreliable. But those same features make the mechanism behind olfactory navigation notoriously difficult to test. "Odors are invisible," Ruta says, "and often they're carried along by turbulent airflow." We’ve got no way of knowing what the animal is smelling from one moment to the next.
A recent report suggests Nvidia has informed its partners of GDDR6 and GDDR7 memory kit price increases
This means retailers may increase the prices for RTX GPUs using GDDR6 and GDDR7 memory
The potential price hike and one user's dual-GPU setup both highlight the importance of hardware preservation
The RAM crisis continues to impact PC manufacturers, with rising memory prices and shortages — ultimately leading to increased retail prices for consumers — and the trend may be set to continue with Nvidia.
According to a report on Benchlife, Nvidia is rumored to be planning an increase for prices of its GDDR6 and GDDR7 memory kits for RTX GPUs, and has informed AIB (Add-In Board) partners such as Asus, MSI, and Gigabyte, who make their own versions of Nvidia's GPUs. Unsurprisingly, the price increase for these memory kits is cited as a result of the ongoing RAM crisis.
While RTX GPU price hikes aren't guaranteed for consumers, because this will directly impact AIB partners who manufacture the GPUs, eventual retailer price increases are unfortunately likely. It's a trend we've seen with recent PC hardware.
Notably, Valve's Steam Machine was initially supposed to cost less than its eventual $1,049 / £879 / AU$1,609 price tag, which could now also be at risk of a price increase, based on hints from a Steam Machine engineer (as reported by Bloomberg). New handheld gaming PCs are also now often launching for more than $1,000, and the price of RAM and storage certainly plays a part.
(Image credit: Roman Samborskyi / Asus)
Consumer struggles
The RAM crisis is a sad state of affairs for consumers and underlines the importance of PC hardware preservation. GPUs aren't cheap by any means, especially with recent suggestions that AMD would also be increasing prices of memory kits for its own AIB partners.
We've also recently seen the return of the RTX 3060, due to Nvidia's attempt to help consumers with more options amid a rough PC hardware market. However, if these price hike reports are legitimate, the RTX 3060 will likely be subject to higher prices at retailers.
That's exactly why GPU setups like one Reddit user's RTX 3090 and RTX 3050 dual-GPU rig matter now more than ever. In this case, the RTX 3050 is responsible for allowing frame generation via Lossless Scaling to function.
A game then utilizes the RTX 3090's power for high-quality visuals and performance, with the two GPUs teaming up. The user states it resulted in 144fps at a 4K resolution over the initial 71fps without frame generation.
Frankly, a dual-GPU setup isn't exactly common, nor is it actually recommended for gaming, since modern games are only designed for single-GPU rendering (even though the frame generation use case is noteworthy).
Users would also need a beefier power supply, depending on the GPUs used, and likely a new motherboard with another compatible PCIe lane.
However, what I feel is more important here is that's it's a great example of why hardware preservation is so important, serving as backup in case GPUs melt or end up faulty, or for experiments like the one I've outlined above. With no end of the RAM crisis in sight, repurposing older hardware might be one way through this.
Archaeologists from the Israeli Antiquities Authority and Tel Aviv University, conducting excavations in the City of David, have unearthed massive wooden beams burned during the destruction of the First Temple in 586 BCE.
The research team behind the discovery notes that the huge wooden support beams, which were surprisingly well preserved for ancient wood structures, allow for much more precise dating of the site. The researchers also highlight the apparently deliberate efforts by the city’s rebuilders to leave the ruins in place for future generations to witness.
2,600-Year-Old Wooden Beams Preserved by Melting Plaster
According to a statement emailed to The Debrief, the ancient, charred wooden beams discovered within the Jerusalem Walls National Park were relics from a collapsed building destroyed by “the intense fire of the First Temple’s destruction” over 2,600 years ago.
“The beams appear to have served as the roof of an inner courtyard in a building from the First Temple period,” explained Excavation Director Efrat Bocher.
Burnt tree trunks exposed in the City of David. Photo by Reut Vilf, City of David.
Although not unprecedented, the team said that finding intact wood relics of this advanced age is rare, since organic materials like wood typically break down over such extended time periods. In this case, they suggest that the surprisingly intact beams were likely preserved by melting plaster, “sealing the burnt wood beneath.”
“During the destruction, they collapsed onto the floor,” Director Bocher explained. “We believe that the plaster cladding of the building’s walls, melted by the fire, covered the charred wooden beams, and ultimately enabled their exceptional preservation.”
Growth Rings Help Archaeologists Narrow the Site’s Dating Range
When discussing the historic and scientific value of the discovery, Dr. Johanna Regev of the Israel Antiquities Authority, who studied the wooden beams up close, noted the exceptional rarity of finding ancient wood with “such a large number” of visible growth rings.
“Here, in the City of David, we found thick beams with many rings, offering significant research potential,” Dr. Regev said.
The tree rings that were exposed in the City of David. Photo by Efrat Botzer.
For example, the researcher explains, more growth rings help scientists date sites and materials with increased precision and “higher dating resolution.”
“Until now, we could date finds within a range of hundreds of years,” Dr. Regev explained. “Now we can attain much greater precision, narrowing the dating range down to as little as ten years.”
City’s Rebuilders Left Ruins in Place as a “Testament to the Destruction”
While the historic site was witness to a tragic destruction, the research team said returning residents beginning to rebuild the city appear to have intentionally left the remains untouched for future generations to see.
“[They] probably made a deliberate decision to preserve the memory of the destruction by leaving behind remains that tell the story of the catastrophe,” the team explained.
Ruins of the destruction uncovered in the Israel Antiquities Authority excavations at the City of David. Photo by Reut Wilf, City of David.
Dr. Yiftah Shalev, who directs the excavation on behalf of the Israel Antiquities Authority alongside Prof. Yuval Gadot of Tel Aviv University, said the idea of a landscape of memory is “familiar from many cultures around the world.”
“Preserving memory is a human need, seeking to preserve what once existed,” Dr. Shalev explained. “Here we see it in its clearest form: a deliberate choice to leave ruins in place as a testament to the destruction and build the new layer over them.”
On a more personal note, Director Bocher noted the serendipitous timing of the discovery “just before Tisha B’Av.” According to Hillel.org, the holiday, which is recognized as the saddest day on the Jewish calendar, “is a day of mourning for tragedies across Jewish history, most particularly the destruction and loss of the first and second Temple and Jerusalem.”
“This discovery makes you feel that you are witnessing the very moment of destruction, the moment when it all happened,” the researcher concluded, adding that “it’s deeply moving.”
Christopher Plain is a Science Fiction and Fantasy novelist and has spent the last six years as Associate News Editor and Head Science Writer at The Debrief. Follow and connect with him on X, learn about his books at plainfiction.com, or email him at christopher@thedebrief.org
PrismML AI compression could give Apple a way to place far more capable artificial intelligence directly inside the iPhone. The Caltech spinout says it reduced a 27-billion-parameter model from approximately 54GB to less than 4GB, small enough to run locally on an iPhone 17 Pro.
That represents a reduction of about 93%, but the more useful number may be memory efficiency. PrismML says its technology can use up to 15 times less memory than conventional versions of the same model. Apple has reportedly held early discussions with the startup about possible uses for the technology.
Neither company has announced a partnership, acquisition or product plan. PrismML’s claims also need to be judged through independent testing, sustained performance and the compromises introduced by heavy compression. Even so, the demonstration points directly at one of Apple’s largest AI challenges: fitting stronger models inside devices without destroying battery life, consuming too much memory or sending every complex request to the cloud.
PrismML AI Compression Changes the Size Problem
Large AI models are difficult to run on phones because their parameters require substantial storage and working memory. A standard model may perform well in a data center equipped with powerful accelerators, large memory pools and constant electricity, then become impractical inside a device designed to last all day on one battery.
PrismML approaches that problem by representing model weights with far fewer bits. The startup has developed what it calls Bonsai models, including 1-bit versions designed to preserve much of the original model’s capability while sharply reducing size and energy use.
The reported iPhone demonstration used a compressed version of Alibaba’s Qwen model with 27 billion parameters. That is far larger than the compact language models normally associated with smartphones. Running it locally suggests that model size may no longer be controlled only by the number of parameters. How efficiently those parameters are represented becomes just as relevant.
Compression is not new. Developers already use quantization, pruning and other methods to make models smaller. PrismML’s claim is that its approach pushes that process much further while retaining useful reasoning and language performance.
That claim is exactly why Apple would be interested.
Why Siri AI Needs Smaller Models
Apple’s AI strategy depends on dividing work between the device and the cloud. Simpler Apple Intelligence requests can run locally, while more demanding tasks may move to Private Cloud Compute. That hybrid structure allows Apple to offer stronger capabilities without forcing the iPhone to carry every model it may ever need.
The cloud layer, however, introduces cost, latency and availability requirements. A request must leave the device, reach a server, be processed and return. Apple has designed Private Cloud Compute around strong privacy protections, but local processing remains faster and easier to explain.
A larger on-device model could improve Siri AI’s language understanding, reasoning and ability to complete multi-step tasks without a network connection. It could also reduce the number of requests sent to Apple’s servers.
That does not mean cloud processing would disappear. Server models can remain larger and support especially demanding requests. PrismML AI compression could instead move the dividing line. Tasks that currently need Private Cloud Compute might become possible directly on a future iPhone.
The benefits could include quicker responses, stronger offline operation and less server demand as Apple Intelligence reaches more users.
Image Credit: Apple Inc.
Privacy Fits Apple’s Preferred Direction
On-device processing has always been central to Apple’s AI positioning. Personal information can remain on the iPhone rather than moving through a conventional cloud service. The device can interpret a request, use local context and produce an answer without exposing the underlying data beyond the hardware.
A model capable of handling more work locally would reinforce that approach. Siri AI could potentially understand messages, notes, schedules or onscreen content while reducing the need to transfer information elsewhere.
Apple already uses Private Cloud Compute when local processing is insufficient, and the company has built significant safeguards around that system. Yet every task completed on the iPhone removes one additional processing step.
Privacy is not the only advantage. Local models can continue operating when connectivity is poor or unavailable. That could help Siri AI during flights, rural travel, crowded events or other situations where cloud access becomes unreliable.
A capable offline assistant would also feel more like part of the operating system and less like a remote service accessed through an interface.
The iPhone Still Has Physical Limits
Shrinking a model to less than 4GB does not automatically make it ideal for everyday iPhone use. The model must also run quickly enough, avoid excessive heat and leave enough memory for iOS and active apps.
A technical demonstration may complete tasks successfully while consuming more energy or time than users would accept from Siri. Apple would need to optimize the model for its Neural Engine, memory architecture and software frameworks rather than simply loading PrismML’s existing build onto every supported device.
Storage is another consideration. A model occupying several gigabytes would compete with apps, photos, offline media and system files. Apple could reserve more storage for intelligence features, deliver models selectively or use several specialized models instead of one large general system.
Battery use may create the hardest constraint. An AI model can fit in memory and still require too much sustained computation for frequent mobile use. The ideal system needs to be small, fast and efficient enough to disappear into normal iPhone behavior.
Apple’s chip design gives it an advantage here. The company controls the processor, Neural Engine, operating system and AI software stack. PrismML could provide the compression technique while Apple handles the deeper hardware integration.
Image Credit: Apple Inc.
Better AI Without Replacing the Cloud
The strongest use of PrismML’s technology may not be putting one enormous model on every iPhone. Apple could use compression across a family of models designed for different devices and tasks.
A smaller model could handle routine language requests. A more capable local model could be downloaded only to recent Pro devices with sufficient memory. Specialized models could support writing, image generation, app actions or Visual Intelligence. Private Cloud Compute could remain available when the request exceeds local capability.
That would give Apple more flexibility than an all-local or all-cloud design. Each device could carry the intelligence appropriate for its chip, memory and battery capacity.
The approach could also help Macs, iPads and future home products. A compressed model may run more comfortably on a MacBook Air, HomePod or smart home hub without requiring constant server access. Apple could scale the same intelligence foundation across products while adjusting model size to the hardware.
This is especially relevant as Siri AI moves beyond answering questions. App actions, personal context and ambient smart home control require intelligence that remains available throughout the day. Lower memory and energy requirements could make those experiences more practical.
Compression Always Includes Trade-Offs
A 93% reduction is a striking headline, but compression can affect output quality. A smaller representation may lose precision, weaken reasoning or struggle with uncommon information. The real test is not whether the model runs. It is whether it performs well enough across the tasks Apple expects Siri AI to handle.
PrismML says its technology preserves strong performance compared with full-precision models. Apple would still need extensive evaluation across languages, safety categories, tool use and personal-assistant scenarios.
Reliability is especially sensitive for an operating-system assistant. A model used for casual brainstorming can tolerate occasional weakness. A model handling messages, appointments, directions or app actions needs a higher level of consistency.
Apple may therefore use aggressive compression selectively. Some tasks can accept a smaller model, while others may remain on a higher-quality server model. The user would not need to know which model is active as long as the response remains fast and dependable.
Image Credit: Apple Inc.
Apple Wants the Technology, Not Necessarily the Company
Reports indicate Apple has spoken with PrismML, but early conversations can lead in several directions. Apple could license the startup’s technology, work on a technical integration, hire members of the team or explore an acquisition.
The company has a long history of purchasing smaller firms for specialized technology and engineering talent. Model compression would fit that pattern because it supports Apple’s existing AI strategy rather than creating a separate consumer product.
PrismML could also remain independent and provide its tools to multiple hardware companies. Its technology has potential uses across phones, computers, robotics, vehicles and industrial equipment where cloud access is expensive or unreliable.
Apple’s interest alone gives the startup attention. The larger significance is that efficient models are becoming as valuable as larger ones. The AI industry spent years treating scale as the primary route to improvement. Mobile computing forces a different question: how much intelligence can fit inside the power and memory limits of one personal device?
A Smaller Model Could Make Siri Feel Larger
Apple does not need to place the biggest available model on an iPhone. It needs a model capable enough to improve the interactions users repeat every day.
PrismML AI compression could help Siri understand more context, complete more requests locally and respond without waiting for a server. It could strengthen Apple’s privacy argument while reducing the infrastructure load created by millions of AI-enabled devices.
The technology remains early, and a startup demonstration is not the same as a shipping Apple feature. The 93% figure will mean little if quality, battery life or speed suffer under daily use.
But the direction fits Apple unusually well. The company has spent years designing smaller, more efficient hardware systems that perform beyond their apparent limits. PrismML is proposing a similar strategy for AI models.
If the startup’s claims hold up, the next major Siri improvement may not come from sending a larger model into the cloud. It may come from fitting one into the iPhone without making the iPhone feel the weight.
Login
