Danh mục: Tin tức

Japanese computer accessories company Far East Gadgets launched its so-called Typesticks earlier this month (h/t PC Watch). Typesticks are key-spaced platforms made from hard plastic and silicon. The strategically positioned standoffs on the underside of Typesticks are supposed to work with any laptop keyboard using “a key gap of 2.5 mm or more, and a key height of 2 mm or less.”

Have you ever wished you could use a mechanical keyboard with your laptop? Well, you can simply plug in your favorite keyboard, and use it, but this would usually mean pushing back your laptop to fit your favorite text input device on the desk in front of it. By placing some Typesticks over your laptop keyboard, you can again type in the screen-to-user position intended by the device designer. With some configurations, the laptop touchpad will remain easily accessible.

Official Typesticks images show a compact mechanical keyboard perched directly atop of a laptop keyboard, affording a superior typing experience. If you choose your Typesticks positioning carefully you can even use the height adjustment feet on your plug-in keyboard for enhanced ergonomics.

The Typesticks designers appear to enjoy using Apple MacBook laptops with the cutely compact HHKB (Happy Hacking Keyboard). However, other laptops have been tested and verified as working with the Typesticks, namely:

• Lenovo ThinkPad series laptops inc X28
• Alienware X17 gaming laptop
• iPad Pro 12.9″ M2 & Magic Keyboard
• VAIO Pro PG laptop
• MacBook Pro M1 16 inch
• MacBook Pro 14″ 2021
• MacBook Pro 13″ 2020
• MacBook Air 13″ 2020
• MacBook Air 15″ 2023

The above list certainly isn’t exhaustive, instead it seems like these may be the laptops the keyboard accessory maker and his colleagues had available for testing.

Likewise, the Typesticks work with a wide range of keyboards you might use. Remember, you can vary Typestick placements to run in line with the rake-adjusting feet on the keyboard base. As well as the HHKB, discrete keyboards tested and verified include the Keychron K2 Pro、NuPhy Air 75, and iPad Magic Keyboard.

Typesticks are compact and portable weighing just 15g. Pairs stick together in transit due to built-in magnets and are about the same size as a USB memory stick or pack of gum (actual size: 72.5 × 23 × 9 mm), to easily fit in a pocket of your laptop tote.

One of the drawbacks of the Typesticks, mentioned in the official product pages, is that an elevated keyboard could obscure the lower part of your computer screen, depending on various factors. Another drawback to some readers will be purchasing products from Japan, and the price is 2,480 yen ($17) plus shipping etc.

Lastly, buyers are warned not to close their clamshell laptops with the third party keyboard and or Typesticks still in place.

DIYers Will DIY

Computer DIYers with access to the best 3D printers, CNC or best laser cutters might find it pretty easy to make their own custom Typesticks. This could be a good option if the official Typesticks are incompatible with your particular laptop, or importing seems too much effort.

Intel intends to increase the L2 capacity of its upcoming codenamed Arrow Lake processors to 3 MB per core, according to Golden Pig Upgrade (via @9550pro), a renowned leaker who tends to have accurate information about future Intel products. If the information is accurate, then Arrow Lake CPUs will offer higher performance in memory bandwidth-dependent applications.

Intel’s 13th Generation Core ‘Raptor Lake’ processor features a 2 MB L2 cache per high-performance Raptor Cove core and 512 KB L2 cache per energy-efficient Greacemont core, thus has 32 MB of L2 cache in total as well as 36 MB of L3 cache in total (3 MB L3 cache per P core, 3 MB per four E cores).

Assuming that Intel’s Arrow Lake processors will retain eight high-performance cores, its total L2 capacity for performance cores will increase to a sizeable 24 MB. Meanwhile, it is unclear whether Intel also plans to expand the size of the L3 cache of Arrow Lake’s performance cores. Keeping in mind that Arrow Lake CPUs will be made on Intel’s 20A (2nm-class) fabrication process, the company might increase the size of all caches as it may not have a significant impact on die size and cost. Yet, only Intel knows what makes sense to do to increase performance without significantly affecting costs. 

Increasing the L2 cache capacity for high-performance cores is done to boost performance. One of the primary benefits of increasing cache size is to improve the hit rate. If the working set of a given workload fits better within the enlarged L2 cache than it did before, this will reduce the need to access the slower L3 cache or main memory. This potentially leads to a reduction in average memory access time and potential energy savings, especially beneficial for workloads whose data sets can better fit within the enlarged L2 cache.

On the downside, a larger L2 cache can introduce slightly longer access latency. Keeping in mind that Intel’s processors already have large L3 caches increasing L2 cache size gets diminishing returns on performance benefits. Additionally, a large L2 cache might consume more power, produce more heat, and increase die area. 

Elegoo has entered the race for speed with a much faster version of their popular Neptune 3 Pro, The Neptune 4. It has improved speed thanks to a hidden install of Klipper firmware, plus everything that made the Neptune 3 Pro our favorite 3D printer for beginners. 

Currently priced at $259, the Neptune 4 is quite a bargain in the field of fast printing. However, the way Elegoo chose to integrate Klipper is confounding. It’s hidden in the background, granting the user speed while making its other perks difficult to access. The hands-off approach might work fine for complete beginners, but if you want to poke under the hood you’ll have to find a long enough Ethernet cable to reach your router or take a leap of faith on finding a compatible WiFi dongle. 

There’s still a lot to love about the Neptune 4. It arrives 90% pre-assembled, making it a quick build. It has dual Z axis, a grippy direct drive that handles TPU like a champ, a high flow nozzle rated to 300 degrees, a giant cooling fan and an easy to navigate touch screen. It even has an LED light on the gantry. 

Leveling has taken a step backwards. Though it’s still very good, it’s adopted a more tedious manual + auto leveling system. It’s not too surprising that Elegoo didn’t include an accelerometer for tuning input shaping, but there’s no visible way to hook one up either. A beginner won’t see this as a failing, as the factory settings are really quite good. However, this is the kind of feature a more seasoned maker will sorely miss.

Specifications: Elegoo Neptune 4

Elegoo Neptune 4: Included in the Box

The Neptune 4 comes with everything you need to get your printer set up. You get tools to build and maintain the printer, side cutters, a plastic scraper, 2 spare nozzles, and a USB stick. There’s also a small sample of white PLA to print your first model. 

The USB stick has a very helpful short video on assembling the printer. You also get a PDF copy of the manual, a copy of Elegoo Cura and a sample model in both pre-sliced .gcode and .stl format.

Design of the Elegoo Neptune 4

The Neptune 4 looks exactly like the Neptune 3…with the addition of a huge cooling fan bolted to the back of the X gantry. The gray aluminum with white detailing and “Create the Future” motto is still fairly distinctive from any other brand on the market.

There are few improvements over the third iteration – the fans now spin on ball bearings and leveling can tap 121 points. The dual-gear direct drive is lighter weight with a bigger gear: a 5.2 to 1 ratio rather than 3 to 1. The hotend also boasts a copper-titanium all-metal throat for more efficient heating.

The giant cooling fan on the gantry is made up of four 4020 ball bearing fans. It’s pointed right below the nozzle and does a good job of cooling the just printed layer extremely fast. If you don’t need to print at supersonic speeds, you can switch off the bonus fan. It’s also fairly loud, but I’ve yet to find a high speed printer that wasn’t loud.

In case you’re not familiar with the Neptune 3 Pro, this one shares the same dual Z axis with a synchronizing belt at the top. The large removable touchpad with a magnetic base is the same, as is that curly landline style cord. Personally, I never take the touchpad off, but if you’ve got a problem with glare I can see it coming in handy.

What’s more intriguing than the screen is what’s on it. It looks very much like the Neptune 3 Pro, a printer with traditional Marlin firmware, not Klipper. I’ve reviewed several printers running Klipper and the best ones have adopted KlipperScreen, which gives you more features. It seems that Elegoo decided the best way to run Klipper is to not see it at all.

The Neptune 4 does not come with WiFi, making it tricky to even access Klipper. You can plug your machine directly into your router if you have an Ethernet cable long enough. I only hooked our test unit to the router for a short time because there was no good place to put the printer except the floor. You may be able to find a WiFi dongle that will work, but since Elegoo does not provide one, you’re on your own.

Accessing Klipper through Fluidd is the only way you can send files remotely or tinker with the machine’s settings, like input shaping or acceleration. Fortunately, the factory calibration is pretty decent, which should satisfy Elegoo’s desired beginner audience.

Assembling the Elegoo Neptune 4

The Neptune 4 comes mostly assembled, only needing a few bolts to put the machine together. The upper frame is attached to the base with bolts that come up through the bottom.

The touch screen holder screws into the side and plugs in with a curly RJ11 cord, making it look like an old landline phone. All the electrical connections are labeled – or extremely obvious where they need to go.

Leveling the Elegoo Neptune 4

The Neptune 4 has returned to pairing manual leveling with an inductive auto level probe – a backwards move as the Neptune 3 and 3 Pro use a hard mounted bed. This system is not only more complex, but inferior. I saw the adjustment wheels shake loose and fall off the machine during an afternoon of speed testing. I would highly advise stopping by your local hardware store for some lock nuts to keep your knobs in place.

To start leveling, you’ll press the Level Icon on the main menu then find the Z height adjustment in the center of the screen. It’s not labeled, but it’s the number flanked by up and down arrows. Slide a piece of ordinary paper under the nozzle and tap the arrows until the nozzle just scrapes the paper.

Now press Auxiliary to enter manual leveling mode. Tap the corner icons to move the nozzle to each corner. Place the paper under the nozzle and adjust the knob until it just scrapes the paper. You’ll need to do this several times as adjusting one corner will throw another corner off balance.

Return to the leveling screen and press Auto Leveling. This will perfect your manual leveling efforts. The bed will heat up, and the probe will bob over 36 points on the bed.

Return to the leveling screen and double-check the z height as it may have moved. Use the paper again and repeat the first step.

IMPORTANT: press the save icon in the upper right corner before leaving the leveling menu to save this data into the config file. The printer will use this data before every print and you won’t need to relevel unless you move the printer, or the wheels fall off.

Loading Filament on the Elegoo Neptune 4

The Neptune 4 is a direct drive printer with its extruder and hotend all in one spot. Loading filament takes a few more taps on the screen than previous Neptunes, but the mechanics are the same.

To load filament, tap Prepare on the main menu, then Temperature and select the one of four presets that match your filament. Once the hotend is warmed past 190 degrees, you can insert the filament into the top of the extruder and let the wheels pull the filament through.

To unload or change colors, simply reverse the process.

Preparing Files / Software for Elegoo Neptune 4

The Neptune 4 comes with a copy of Elegoo Cura, which has profiles for all its machines and several materials. It also has the added ability to create a thumbnail of your model to display on the printer. Though Elegoo claims the Neptune 4 has a maximum speed of 500mm/s, the default profile provided is 250mm/s. There is no profile for higher speeds, so you’ll have to experiment with that yourself.

Standard versions of Cura or Prusa Slicer don’t have profiles for the Neptune 4, but you can use the profile for a Neptune 3 and bump up the speed.

Note: The Neptune 4 doesn’t have Klipper “START_PRINT” and “END_PRINT” macros, but rather uses the Marlin start and end codes.

Printing on the Elegoo Neptune 4

You’ll quickly learn that speed isn’t everything, especially when printing filament that is normally glossy. High speed printing wrecks havoc on the finish, often producing surfaces that are flat and dull. If it’s a practical print, you may not care if your part isn’t shiny – but if you’re printing decorative parts you’ll want slower speeds. Fortunately, you can find balance by printing the infill and inner walls at high speed while slowing the outer and top layers for a nicer finish.

Below is a model printed in the same filament, Polymaker Starlight Mercury PLA. The one on the left is dull grey after being printed at 250mm/s. The one on the right retains its purple sparkle and was printed with 250mm/s for the infill and inner walls, but 75mm/s on the outer wall and top/bottom layers.

The other issue with high speed is that many prints are not long enough to reach the manufacturer’s claim of 500mm/s. Once you factor in acceleration (capped at 5000mm/s on the Neptune 4) and the need to slow down for corners, you’re doing 250mm/s at best.

Still the Neptune 4 is much faster than the Neptune 3 Pro and similar last-gen bedslingers. The printer achieved a very good 20 minute Speed Benchy. This is the traditional Benchy test print using standards adopted for speed contests. It has 2 walls, 3 top and bottom layers, 10% grid infill, a .25 layer height and .5 layer width. Printed in ordinary gray Inland PLA, this boat is a little rough on the hull, but there’s no ghosting and the overhangs are perfect.

Elegoo sent a spool of Rapid speed PLA+ to try out with this printer. The results are marginally better than our tests with Inland PLA. You don’t want to limit yourself to one kind of filament, so check out our guide to the best filaments for 3D printing for examples of our favorite materials to add to your supply.

The printer is more than capable of producing smooth, quality prints when you slow it down a bit. This X Wing pen holder was printed using the default settings, which uses 250 for the inner parts and 130 for the outer walls and top/bottom layers. The detail is crisp, without any ghosting, but the color is a bit washed out from the speed. Printed in ProtoPasta Wonder Black Rainbow Glitter, Polymaker Starlight Mercury PLA for the stand, and Prusament Galaxy Silver for the engine highlights. Total print time is 3 hours and 9 minutes, using 3 walls, 250mm/s inside parts and 75mm/s outer layers & walls.

The Neptune 4 did a remarkable job with TPU when slowed down to 50mm/s. These bike handle covers may not be the best print to show off the machine’s quality – the pattern did get rougher the farther away it got from the bed. But considering these slender tubes are 157 mm tall and quite squishy I’m amazed they worked at all. These printed in 1 hour and 53 minutes each using Elegoo’s default settings for TPU.

I did several prints in PETG, but this Maker’s Muse Clearance Castle really shows off the Neptune 4’s quality. The castle is a torture test with tight clearances and bridging, which all printed quite well with crisp lines and little stringing. This used Elegoo’s default settings for PETG which slowed the printer to 50mm/s on the outer walls, 80 on the inner walls, but kept a brisk 150 mm/s on the infill. It was able to complete the print in 2 hours and 6 minutes. The Neptune 3 would have taken three hours more, so this is still a significant time savings.

Bottom Line

The Neptune 4 is a confusing little 3D printer. Is it for beginners? Maybe. Is it for fans of Klipper? Not really. It comes with an unrestricted copy of Klipper, a huge plus, but makes it difficult to access without WiFi. You can certainly run a cable to it, but how many people have their router on the same workbench as their 3D printer?

The variables that come with high speed printing make me reluctant to recommend this printer for beginners, who may not understand why their shiny new printer can’t run full tilt at the manufacturer’s claim of 500mm/s. On the other hand, those who have experience with Klipper may feel frustrated that their access is physically restricted by a simple lack of hardware.

Who is it for? With a tantalizing price tag of $259, the Neptune 4 is perfect for makers at the midpoint of their journey. People who understand the ins and outs of Cura settings, but aren’t control freaks and need to tweak every aspect of their Klipper config file, people who are comfortable with “good enough” input shaping and relish the bargain of an affordable fast printer.

I’m still going to recommend the Neptune 3 Pro for absolute beginners, especially with its reduced price of $199.99. If you want a simple-to-use Klipper printer with no restrictions, then check out the Sovol SV07. Priced at $339 it has both the expected WiFi plus a really sweet Klipper Screen that puts all the data at your fingertips. If your budget allows, the $699 Bambu Lab P1S cannot be beat for pure speed, range of filaments and ease of use.

MORE: Best 3D Printers

MORE: Best Budget 3D Printing

MORE: Best Resin 3D Printers

Renowned hardware leaker @momomo_us has published what seems to be specifications of AMD’s upcoming EPYC 8004-series ‘Siena’ processors. The new Zen 4-based CPUs are projected to hit the market later this year or early next year and address lower-power servers that may not need the highest performance, such as those used for edge and telecom applications.

If the specifications published by the leaker is accurate, then AMD’s EPYC 8004-series ‘Siena’ lineup will consist of at least six models featuring eight, 16, 24, 32, 48, and 64 cores operating at 2.20 – 2.55 GHz and consuming from 90W to 200W depending on the model. Meanwhile, AMD’s EPYC ‘Siena’ CPUs will feature an up to 128 MB L3 cache, which is two times smaller than L3 cache of AMD’s EPYC ‘Rome’ processor with 64 cores.

The origin of the specifications is unknown, but leakers like @momomo_us typically obtain their information from documents by chip developers or their partners. While the information may come from a legitimate source, it may be preliminary or outdated, so take it with a grain of salt.

In general, specifications of AMD’s EPYC 8004-series ‘Siena’ processors look rather logical. These CPUs are meant for low power servers with reduced total cost of ownership (TCO) that do not have to be performance champs, but have to be energy efficient and easy to build. Indeed, Siena will top at a 200W TDP, according to the newly published information. 

AMD’s Siena is expected to use the company’s new SP6 platform that is considerably less complex than the SP5 platform used for AMD’s EPYC 9000-series processors that are designed to feature the highest core count possible and deliver unbeatable performance for those who need it. For example, SP6 is expected to feature an eight-channel DDR5-4800 memory subsystem that will deliver enough bandwidth for a 64-core CPU, but which is considerably less complex than a 12-channel memory subsystem of AMD’s Genoa, Genoa-X, and Bergamo processors.

Respected hardware leaker @wxnod has published a picture of what seems to be an Intel processor in LGA1851 packaging. The new chip could be Intel’s Meteor Lake-S or Arrow Lake-S and its main difference with the current Raptor Lake-S is a slightly altered integrated heat spreader (IHS). Given that the image is published without any explanations, this could be a prototype of an Alder Lake processor that never hit the market.

Intel’s sockets formally called LGA1700 and LGA1851 are essentially the same socket with 0.8 mm pitch, but the latter has more active pins. This allows makers of motherboards and cooling systems to maintain their designs. Meanwhile, LGA1851 will increase the height of IHS from 6.73-7.4 mm to 6.83-7.49 mm, according to earlier reports. To prevent installation of LGA1851 processors into 1700-pin sockets, the upcoming CPUs will have a different cutout configuration.

Perhaps the most interesting part about the processor marked as ‘Intel Confidential NA QDF4’ is their slightly larger heatspreader. It is unclear whether the taller heatspreader was made slightly larger to prevent bending that Intel’s LGA1700 processors are known for, but this is certainly a possibility. 

If the CPU is Intel’s Meteor Lake-S, then it will never reach the market as the platform has been cancelled. Meanwhile, the processor still gives an idea how Intel’s LGA1851 CPUs will look like when they come to market in 2024 in the form of Intel’s Arrow Lake products.

Just when we thought the trail went cold, AMD’s upcoming Ryzen Threadripper 7000WX (Storm Peak) has emerged in a new shipping manifest. While there was no indication of an estimated launch date, the fact that the Zen 4-powered chips are in transit suggests they could hit the market soon to compete with the best CPUs for workstations.

The alleged shipping document, courtesy of Harukaze5719, listed the Ryzen Threadripper 7995WX, 7845WX, and 7945WX. The leaked model names cause a bit of confusion, though. The trio of processors seemingly lacks the “Pro” moniker, which AMD utilizes for its workstation parts, such as the current Ryzen Threadripper Pro 5000WX (Chagall) lineup. On the other hand, the chips carry the “WX” suffix which  points to the Pro lineup. It’s probably a human error, so in any event, these are the Pro chips that target the workstation market and not the HEDT market.

The next-generation Threadripper processors will wield AMD’s latest Zen 4 cores. The leaked information claims that the three Ryzen Threadripper 7000WX processors will find their place in Socket SP6 (LGA4844), a socket expected to accommodate AMD’s EPYC Siena chips. 

Regarding dimensions, Socket SP6 isn’t as big as Socket SP5 (LGA6096) for EPYC Genoa processors. On the contrary, Socket SP6 shares similar dimensions with the older Socket SP3 (LGA4094), which coincidentally houses the current Ryzen Threadripper Pro 5000WX series.

Ryzen Threadripper 7000WX Specifications*

*Specifications are unconfirmed.

Unfortunately, the leaked specifications don’t tell us anything about core counts. However, we can piece that together with previous leaks. The 100-000000884 OPN code had surfaced as early as November 2022 in the Einstein@Home database. The processor, which we now know corresponds to the Ryzen Threadripper 7995WX, reportedly features 96 cores and 192 threads. That would be a massive upgrade since AMD’s Threadripper chips have topped at 64 cores for the last two generations. The Ryzen Threadripper 7995WX would be a conflicting SKU, as it would be pushing into EPYC territory.

The Ryzen Threadripper 7985WX has already appeared in several benchmarks, so we know it’s a 64-core, 128-thread part. That would mean it’s the successor to the Ryzen Threadripper Pro 5995WX. We suppose AMD didn’t follow the conventional naming scheme to save the 7995WX for the top SKU and probably used the 7985WX for the 64-core part. The Threadripper 7945WX, however, is perhaps the direct replacement for the Ryzen Threadripper Pro 5945WX. If that’s the case, we’re looking at a probable 12-core, 24-thread design as its predecessor.

According to the document, the Ryzen Threadripper 7000WX processors have a 350W TDP, accounting for a 25% increase over the existing Ryzen Threadripper Pro 5000WX series. On another note, the Ryzen Threadripper 7000WX’s TDP is only 10W lower than the highest-performing 4th-Generation EPYC Genoa chips, lending credence to the Ryzen Threadripper 7995WX potentially having 96 cores.

Noctua plans to release two new CPU air coolers for the next-generation Ryzen Threadripper lineup in October, insinuating that we could see the Zen 4-powered workstation chips on the market very soon.

A Chinese reviewer has obtained a pre-production sample of Intel’s Core i7-14700K processor, which belongs to the Raptor Lake Refresh family, and tested it in a number of performance benchmarks for his review at Bilibili (via @9550pro). The processor, which got more cores and higher clocks, expectedly beat its predecessors, but what is even more noteworthy is that it can offer performance that is very close to that of the current flagship Core i9-13900K.

Intel’s upcoming Core i7-14700K processor will pack eight high-performance Raptor Cove cores clocked at 3.40 GHz – 5.60 GHz and 12 energy-efficient Gracemont cores, which is four cores more than the Core i7-13700K has. The new Core i7 ‘Raptor Lake’ refresh not only got slightly higher boost clocks, but also more cores, which will positively affect its performance in multi-threaded workloads. Meanwhile, the new CPU consumes more power.

 

Indeed, the Core i7-14700K, which the reviewer calls ‘Core i7-13700KS’ to avoid problems with Intel, in CPU-Z multi-threaded benchmark by around 20% and in Cinebench R23 multi-threaded test by approximately 14%.

 

Due to higher clocks and increased number of cores, the upcoming processor also outperforms its predecessor in games.

While the reviewer does not compare the Core i7-14700K to the current flagship Core i9-13900K, it has the same configuration of cores (eight high-performance and 12 energy-efficient cores) and very close frequencies (the Core i9-13900K has 200 higher boost clock of 5.80 GHz). Therefore, performance of the new Core i7-14700K will be very close to that of the Core i9-13900K.  

Meanwhile, Intel’s next-generation flagship Core i9-14900K and special-edition Core i9-14900KS will have boost clocks of 6.0 GHz and 6.20 GHz, respectively. As a result, performance difference between the most expensive Core i9 parts and moderately priced Core i7 CPUs will likely be negligible in most cases.

Nvidia’s codenamed Blackwell family of graphics processors will contain five different GPUs and will lack a direct successor to Nvidia’s highly-successful AD104 chip, according to leaks by Chiphell and kopite7kimi (via VideoCardz). The information is unofficial and may be inaccurate, but if true, Nvidia will have to address market segments currently addressed by AD104 with two different GPUs.

Apparently, Nvidia’s Blackwell family of graphics processors contains five chips codenamed GB202, GB203, GB205, GB206, and GB207. Nvidia’s Ada Lovelace family released to date also contains five processors (AD102, AD103, AD104, AD106, AD107), just like the company’s Ampere lineup of GPUs (GA102, GA103, GA104, GA106, GA107) that still powers some of the best graphics cards. Meanwhile, Nvidia’s Turing family comprised of three members, whereas the Pascal lineup contained five GPUs. 

It is unclear why Nvidia’s Blackwell family is said to feature GB200-series GPUs, but not GB100-series graphics processors. Typically, 200-series represents re-spinned GPUs.

Historically, Nvidia’s XXX04 served performance-mainstream segment of the market and contained 50% – 66% transistors of the top-of-the-range part. The gap between the high-end and the performance mainstream part was quite noticeable. To fill the gap between its GA102 and GA104 GPUs in the Ampere era, Nvidia introduced GA103 part and did the same with the Ada Lovelace family: there is the AD103 sitting between the AD102 and AD104.  

While the now Nvidia’s lineup no longer has wide gaps, a cut-down AD103 would overlap with the full AD104, which means that Nvidia has to either throw away AD103 GPUs that have one or more defective streaming multiprocessors or even CUDA cores so as not to compete against AD104, or keep them and then use them quietly to substitute AD104, which means cutting them down substantially and not using the whole potential of AD103.

Apparently, the company wants to avoid such a situation in the future. As a result, its GB202 will keep addressing the highest-end of the market (e.g., GeForce RTX 5090, GeForce RTX 5090 Ti), its GB203 will address high-end and performance-mainstream segments (e.g., GeForce RTX 5080, GeForce RTX 5070 Ti, and GeForce RTX 5070), while GB105 will address mainstream part of the market (GeForce RTX 5060 Ti, GeForce RTX 5060). This will enable Nvidia to use all GB203 silicon that it has even with some defective SMs. Of course, we are speculating here.

Nvidia’s next-generation Blackwell GPUs are expected to hit the market in late 2024 or early 2025, so the company’s plans may change a lot between now and then. Therefore, take this information with a grain of salt for now.

Right now at Amazon, you can find the WD_Black SN850X 1TB SSD for its lowest price ever. This high-speed SSD has been going for $75 lately but a coupon at Amazon will take the price down to $54.

We reviewed the WD_Black SN850X when it debuted and recognized it as having some seriously high performance. One of our biggest drawbacks was that the drive was quite pricey at the time. However, today’s discount makes for a much more appealing offer. 

The WD_Black SN850X comes in a few capacities including 1TB, 2TB and 4TB but this discount applies only to the 1TB version. All of the drives in this line have an M.2 2280 form factor and connect using a PCIe 4.0 x4 interface. They use WD proprietary controllers and 112-Layer BiCS5 TLC flash memory. The 1TB SSD is capable of reaching read/write speeds as high as 7300/6300 Mbps.

The WD_Black SN850X is supported by a 5-year warranty that voids if the 1TB edition reaches 600TBW. You can also take advantage of Amazon’s 30-day return policy. It’s also worth nothing this offer applies to the edition without a heatsink.

Visit the WD_Black SN850X 1TB SSD product page at Amazon for purchase options. Don’t forget to use the coupon box before adding it to your cart to redeem the offer.

The Raspberry Pi is obviously one of the most deliciously named boards of our time, but this Raspberry Pi project might be the tastiest use case we’ve ever seen. It was created by a maker and developer known as Simon over at Hackster. Using our favorite SBC, he’s developed a system that scans his refrigerator for food and recommends recipes based on what’s inside.

According to Simon, he trained his project using YOLOv5 to recognize potential ingredients from the contents of his fridge. The food is scanned using a simple camera module and processed using image recognition. The Pi is programmed to organize the information and research for possible recipes and cooking videos related to its findings.

The recipes are selected using ChatGPT with Open AI. A script sends a list of detected food with a prompt to suggest recipes using the ingredients. Once a list of possible recipes is put together, the Pi then notifies the user using Telegram. These messages include recipes and tutorial videos related to the recipes it’s gathered—saving you tons of time spent manually researching for a possible dinner menu.

There are actually two Raspberry Pis used in this project. A Raspberry Pi 4B is the main board handling most of the processing but a WIZnet W5100S-EVB-Pico is thrown into the mix, as well, connecting to the Pi 4 via UART. Simon is using a Raspberry Pi Camera Module but any camera should work fine in its stead.

The software for this project involves a few layers of software, all of which is detailed in the project page at Hackster for anyone who wants to check it out for themselves. OpenAI handles the ChatGPT portion while some Python helps integrate the Telegram API. MicroPython is used to program the WIZnet W5100S-EVB-Pico board.

Visit the Raspberry Pi fridge scanner project page at Hackster to get a closer look at how this Raspberry Pi project goes together and be sure to follow Simon for future updates.

Some of the most advanced Roomba robot vacuums are on sale right now. Wellbots has the poop-detecting iRobot Roomba j7+ for an all-time low price of $549. Or, you can opt for our favorite vacuum / mop hybrid, the upgraded Roomba j7+ Combo, for $799. In both cases, you can enter the coupon code ENGROOMBA200 to receive the $200 discount exclusively for Cunghoctin readers.

The Roomba j7+ includes a bundled cleaning station and iRobot’s poop detection tech. The company says the vacuum’s advanced sensors will avoid pet feces — and its “Pet Owner Official Promise” (P.O.O.P.) policy assures you it will replace the device for free if it runs over pet waste within its first year. The self-emptying vacuum also includes an intelligent mapping feature that lets you choose which rooms it covers, and you can schedule multiple cleanings per day in advance. The Roomba j7+ is usually $800, so its $549 discounted price (with code ENGROOMBA200) is worth noting if you’ve been on the lookout for a cleaning machine.

Wellbots’ coupon also applies to that model’s upgraded sibling, one of Cunghoctin’s picks for the best robot vacuums. The Roomba j7+ Combo takes the j7+’s base features (including self-emptying tech and poop detection) and adds wet mopping, letting you transition seamlessly between carpet and floor cleaning. (It can do both simultaneously on hard floors.) It also includes a feature that can help keep your rugs dry by lifting the mop out of reach when it detects it’s on a carpet. The j7+ Combo typically costs $1,100.

Both devices work with the iRobot mobile app. It lets you manage the devices’ smart mapping capabilities, remote startup and scheduling. We find the app to be simple, intuitive and a significant selling point for Roomba robot cleaners.

Follow @CunghoctinDeals on Twitter and subscribe to the Cunghoctin Deals newsletter for the latest tech deals and buying advice.

Taiwanese component company Lian Li is well known for its premium PC cases and fans, but the company also offers power supplies and AIO coolers. Today we’re looking at Lian Li’s latest 360mm AIO, the Galahad II Trinity Performance. It features upgraded designs for virtually every part of the cooler, from the radiator to the fans to the liquid pump.

Over the past two years, the market for liquid cooling has grown more intense as manufacturers have released upgraded liquid coolers to more effectively dissipate heat to keep up with the demands of today’s hottest CPUs. Are the changes made by Lian Li for the Galahad II Performance enough to earn the top spot on our best AIO Coolers list? We’ll have to install and test the cooler to find out, but first let’s take a look at the cooler’s specifications and features.

Lian Li Galahad II Trinity Performance Specifications

Packing and Included Contents 

Lian Li’s Galahad II Trinity arrives in a box similar in size to other coolers in its class. 

The inner contents are protected with plastic coverings, foam, and molded cardboard. 

Included with the package are the following:

• 360mm Radiator with pre-installed 120mm fans
• Thermal Paste & Spreader Tools
• Leather Cable Management Strap
• Mounting for Intel and AMD Platforms
• Alternative CPU block covers
• Manual

Installation on LGA1700

Installing Lian Li’s GA II Trinity Performance on a LGA 1700 system is surprisingly simple, arguably the easiest installation I’ve ever encountered. 

1. The fans are pre-installed to the radiator for simplicity of installation, so the first thing you’ll want to do is to secure the radiator to your computer case. 

2. Apply the backplate to the back of the motherboard.

3. Attach the mounting standoffs.

3. Mount the CPU block on top of the standoffs after applying thermal paste and secure it with a screwdriver.

4. Next you’ll want to decide what cords to attach to the CPU block. You can choose to control the device’s PWM speeds and lighting traditionally by connecting them to your motherboard’s ARGB & PWM headers, or you can connect the fans to the CPU block and let Lian Li’s software control pump and fan speeds and RGB lighting options.

Features of Lian Li Galahad II Trinity Performance

Every aspect of the Galahad II Trinity performance – including the fans, the radiator, the pump, the tubing, and more –  have been upgraded to provide a premium cooling experience. The combined improvements made to the Galahad II are not minor compared to previous products. This cooler looks, feels, and (as we’ll soon see in testing) performs much betterthan many competing coolers.

⋇ 3x 120mm high performance LCP fans 

There’s more to a cooler than just the heatsink or radiator. The bundled fans have a significant impact on cooling and noise levels and the unit’s aesthetic. These fans were designed specifically for liquid cooling with an impressively strong 7mm H20 maximum rated static pressure and up to 108 CFM of airflow. 

Like the fans included on the previously reviewed EKWB CR360, Lian Li’s fans feature unique connections which simplify cable management by converting PWM and ARGB connections into a single header connected by micro-fit 8-pin connectors that have a similar form factor as CPU power connections. This results in a more secure connection with less cable management to worry about, even if th connectors feel a bit bulkier than necessary. 

By default, these fans are solid black with no ARGB. However, those who prefer an ARGB can upgrade the fans with a Lian Li side ARGB strip. 

⋇  Full Performance/Low Noise Toggle

Included with the AIO is a toggle switch that lets you quickly switching between full performance and low-noise modes. In full performance mode, the fans will spin at up to 3000RPM; with low noise mode this is reduced to 2300RPM. 

Most users will not notice any performance loss with the low noise mode. I did not observe any meaningful differences when paired with Intel’s i7-13700K. You’ll need a very thermally demanding CPU to gain performance from the full performance mode. But for those chasing top overclock speeds, it’s nice to know it’s there when you need it.

⋇  CPU block improvements 

⋇  Designed to be mounted in all directions 

The 45 degree fittings connecting the liquid tubing to the liquid pump are rotatable, making it easy to use in any installation direction. Some AIOs lose performance when installed in a side mount. Lian Li’s Galahad II Trinity Performance is designed for full performance whether top mounted, front mounted, or side mounted.    

⋇ Three lighting configurations

The decorative top of the CPU block can be customized one of three ways. The default style is an infinity mirror. The second style available offers diffused RGB lighting. The third option is a combination of the two, with a diffused RGB outside and a mirror on the inside.

⋇  Large copper contact plate

The Galahad II Trinity Performance features a large CPU contact plate, made of copper with significant revisions compared to the previous-generation product. The skived fins have increased spacing to prevent buildup of debris, and the height and thicknesses of the fins have been improved. Additionally, a trench down the middle of the skived fins has been implemented for better heat dissipation of the coolant over the copper plate. 

The seal for the copper cold plate has also been improved, with straighter pathways for reduced turbulence and increased airflow.

⋇ Fine Sleeved Tubing with increased width

The inner diameter of the water tubes has been increased from 5.8mm to 7mm, upping the total volume of liquid by roughly 20%, which allows for better heat absorption.

The rubber tubes of the unit are reinforced and protected with fine sleeving, which also makes the unit look and feel more premium. Lian Li also includes a leather cable management tie to keep the tubes tidy.

⋇  Complete RAM compatibility

As the compact CPU block of the AIO does not overhang or interfere with RAM in any way, all sizes of DDR4 and DDR5 DIMMs are supported.

⋇  32mm-thick radiator 

The Galahad II Trinity Performance sports an extra thick 32mm radiator, featuring a double-wave fin design for premium heat dissipation.  

⋇  (Optional) Software Control 

Manufacturers of AIO cooling systems tend to fall into one of two extremes when it comes to cooler control: Either they force you to use bloated, often buggy software to control the device, or they don’t include any software and rely only on motherboard PWM and ARGB controls. 

With Lian Li’s Galahad II Trinity Performance, you have a choice. You can use the default motherboard controls if you prefer, but the company also offers an optional download of the L-Connect3 3 software for a more customized experience. This software offers in-depth lighting and pump/fan speed customization options, with the ability to control other compatible ARGB devices as well. Interestingly, you can also return control of these options to the motherboard with “MB Sync” option. 

Cooling Considerations

Modern high-end CPUs, whether Intel or AMD, are difficult to cool in intensive workloads. In the past, reaching 95C+ on a desktop CPU might have been a cause for concern. But with today’s processors, it is considered normal operation. Similar behavior has been present in laptops for years due to cooling limitations in tight spaces. 

Despite assurances from the CPU manufacturers that there is no concern in running the CPU at or near its maximum temperature, many enthusiasts still prefer to have their CPUs run at lower temperatures. Most coolers won’t be capable of achieving this in the strongest and most power-hungry of workloads, but the strongest of 360mm AIOs are able to handle the heat with Intel’s i7-13700K.

LGA1700 Socket Bending

There are many factors other than the CPU cooler that can influence your cooling performance, including the case you use and the fans installed in it. A system’s motherboard can also influence this, especially if it suffers from bending, which results in poor cooler contact with the CPU. 

In order to prevent bending from impacting our cooling results, we’ve installed Thermalright’s LGA 1700 contact frame into our testing rig. If your motherboard is affected by bending, your thermal results will be worse than those shown below. Not all motherboards are affected equally by this issue. I tested Raptor Lake CPUs in two motherboards. And while one of them showed significant thermal improvements after installing Thermalright’s LGA1700 contact frame, the other motherboard showed no difference in temperatures whatsoever! Check out our review of the contact frame for more information. 

Testing Methodology

All testing is performed at a 23 degrees Celsius ambient room temperature. Multiple thermal tests are run on each CPU to test the cooler in a variety of conditions, and acoustic measurements are taken with each result. These tests include: 

1. Noise-normalized testing at low noise levels

2. “Out-of-the-box”/default configuration thermal & acoustics testing.

     a.)  This means no power limits on Intel’s i7-13700K.

      b.) Because CPUs normally hit Tjmax in this scenario, the best way to compare cooling strength is by recording the total CPU package power consumption.

3. Thermal & acoustics testing in power-limited scenarios.

      a.)  I’ve tested with limits of 175W and 125W enforced.

The thermal results included are 10-minute testing runs. To be sure that was sufficiently long to tax the cooler, we tested both Thermalright’s Assassin X 120 R SE and DeepCool’s LT720 with a 30-minute Cinebench test with Intel’s i9-13900K for both 10 minutes and 30 minutes. The results didn’t change much at all with the longer test: The average clock speeds maintained dropped by 29 MHz on DeepCool’s LT720 and 31 MHz on Thermalright’s Assassin X 120 R SE. That’s an incredibly small 0.6% difference in clock speeds maintained, a margin of error difference that tells us that the 10-minute tests are indeed long enough to properly test the coolers.  

Testing Configuration – Intel LGA1700 Platform

This review will focus specifically on how Lian Li’s Galahad II Trinity Performance compares against the strongest AIOs on the market : DeepCool’s LT720, EKWB’s CR360, MSI’s MEG S360, and Cooler Master’s 360L Core. If the cooler isn’t included here and I’ve reviewed it previously, that means it failed my unlimited power Cinebench tests and, as such, is not included.

The Compute Express Link technology allows for building various devices to address a vast array of workloads, from expanding memory subsystem capacity and performance to offering ultra-fast persistent storage. Being one of the world’s leading makers of NAND memory, Kioxia can address the latter. Recently, it demonstrated its 3D NAND and XL-Flash-based CXL solutions at the Flash Memory Summit 2023, as reported by ServeTheHome. 

Kioxia revealed plans to offer two lineups of its CXL products: CXL + XL-Flash-based devices for performance and reliability-centric applications such as in-memory databases and AI inference workloads, as well as CXL + BiCS 3D NAND-powered devices for capacity-hungry applications like Big Data and AI training. In both cases, storage devices use a special controller and CXL.mem protocol for reading and CXL.io protocol for writing to minimize respective latencies.

Regarding demonstrations, Kioxia showed a sample of a 1.3 TB CXL 1.1/CXL 2.0 BiCS 3D NAND-based device in an E1.S form-factor that can be installed into an E3.S chassis for higher performance/thermal capacity. The device uses a PCIe x4 interface (PCIe Gen5, we presume), though Kioxia does not disclose its performance characteristics, perhaps because the development has not finished.

While usage of a low-latency 3D (TLC) NAND-powered device over a PCIe interface with the CXL protocol on top seems like a very plausible idea, usage of an XL-Flash-based storage device promises to be even more fruitful due to the higher performance of XL-Flash compared to commodity 3D NAND.

Kioxia’s proprietary 1st Generation XL-Flash is essentially single-level cell (SLC) NAND spread over 16 planes, whereas 2nd Generation XL-Flash is multi-level cell (MLC) flash memory spread over a higher number of planes, which by definition offers lower latency and higher parallelism for read/write operations, thus guaranteeing massively higher performance compares to mainstream 3D TLC NAND. 

Last year Kioxia said that its CXL storage devices would use its 2nd Generation XL-Flash, which promises to outperform the 1st Generation XL-Flash while being more cost-effective and thus enabling higher capacities. 

For now, Kioxia remains tight-lipped when it plans to ship its CXL devices featuring commodity 3D NAND and proprietary storage-class XL-Flash memory. Still, based on the fact that it is displaying some of the former products and does not showcase the latter (at least openly), we can assume that 3D NAND-based devices will be available a bit earlier.

AI and climate change represent two ways humans may ravage life as we know it on Earth, but the former can also help with the consequences of the latter. The California Department of Forestry and Fire Protection (Cal Fire) revealed a new program today that uses AI to detect wildfires. Created in partnership with the University of California San Diego, the Alert California AI program takes feeds from 1,032 360-degree rotating cameras and uses AI to “identify abnormalities within the camera feeds.” It then notifies emergency services and other authorities to check if a potential blaze warrants a response.

The program, launched in July, has already quelled at least one potential wildfire, according to Reuters. A camera reportedly recorded a fledgling fire burning at 3 am in the remote Cleveland National Forest east of San Diego. The AI spotted the inferno and alerted a fire captain “who called in about 60 firefighters including seven engines, two bulldozers, two water tankers and two hand crews.” Cal Fire says the flames were extinguished within 45 minutes.

Cal Fire / University of California San Diego

The Alert California technology website says it uses LiDAR scans taken from airplanes and drones to create “equally precise, three-dimensional information about scanned surfaces.” It combines this with the physical traits of tree species to learn more about California’s forest biomass and carbon content. Cal Fire says the ML model leverages petabytes of data from the cameras to differentiate between smoke and other airborne particles.

The system was developed by UCSD engineers using AI from the California-based company DigitalPath. Cal Fire has invested over $20 million in the program over the past four years and promises an additional $3,516,000 in the near future.

“We’re in extreme climate right now. So we give them the data, because this problem is bigger than all of us,” Neal Driscoll, geology and geophysics professor at UCSD who serves as the program’s principal investigator, told Reuters. “We need to use technology to help move the needle, even if it’s a little bit.” However, Driscoll adds that the current sample size is too small to determine the program’s overall effectiveness.

You can check out Alert California’s “camera quilt” on your computer or mobile device. The website displays a grid of the remotely operated live camera views from across the region.

Jabra is preparing some earbud refreshes with high-tech features, according to The Jabra Elite 8 Active earbuds look to be a refresh that combines elements from the Elite 7 Active and the The company also seems to be skipping an entire iteration, as it’s also reportedly prepping the Jabra Elite 10 earbuds, just like how Apple skipped over the iPhone 9 in favor

The Elite 8 Active true-wireless stereo (TWS) earbuds will offer voice assistant support and adaptive hybrid Active Noise Cancellation (ANC) which is a pricier and premium alternative to more traditional ANC methods. The earbuds will reportedly have an IP58 rating for both water and dust-resistance, with the charging case offering an IP54 rating.

The earbuds boast Jabra’s proprietary secure in-ear ShakeGrip technology, which is basically a liquid silicone rubber exterior that stays put when placed in the ear. Reports indicate that these buds will also allow for fast pairing, Spotify tap playback and more.

MySmartPrice

Not as much is known about the Jabra Elite 10 line, though it looks like the earbuds will offer ANC, fast pairing, Dolby Atmos support, voice assistant support, Spotify tap playback and the same water and dust-resistance ratings as the Elite 8 Active earbuds. They’ll also be available in gold and beige, with a possibility for more color options.

Some major specs, like battery life, are still being kept under wraps, as these earbuds have yet to be officially announced. It’s worth noting that the Jabra Elite 7 earbuds got around 30 hours per charge. There’s no pricing or availability information yet for either of these products. We’ll keep you updated.