If you've been hunting for computer upgrades, it's been a rough couple of years. After the one-two-three punch of a worldwide chip shortage, a boom in cryptocurrency mining, and a scourge of scalpers (yes, “scourge” is the collective noun for scalper, I looked it up!), graphics cards in particular have been incredibly hard to find at any kind of reasonable price. But with new budget GPUs on the way from both Nvidia and AMD, your best chance to score a new card for less than a mortgage payment might be coming up very soon.

A bit of context. During CES at the start of January, Nvidia announced the GeForce RTX 3050 and AMD announced the Radeon RX 6500 XT, with retail prices at $250 and $200, respectively. We could go into a lot of detail on the respective technical power of these cards, but that's another article. Suffice it to say, they're built with budget gaming PCs in mind, targeting decent 1080p performance and bringing some high-end features like ray tracing and DLSS to a price point at which they've never been available before.

So what's different about these cards, that will make them any less susceptible to the market forces which have driven prices to double and even triple their MSRP? A few things. One, they're less powerful cards in general, especially in terms of memory, which means they'll be less desirable for cryptocurrency miners. The same number-crunching power that gives you a solid 60 frames per second at 4K also lets you tear through crypto hashes. In short: more power means more money, and the crypto miners would rather spend a few hundred dollars towards a more powerful card that will be more efficient for their purposes. Less demand from crypto miners means less interest from scalpers — trickle-down disinterest, if you will.

AMD

That will hopefully result in a lower demand from the non-gamer segment of the market. What about supply? Without getting too technical, less powerful processors can be manufactured in larger batches than the kind of screaming GPUs you see in the RTX 3070, RX 6700 XT, and higher models. The dies that create the processors are physically smaller, allowing the manufacturer to get more out of a single run — sort of like making more cookies out of the same dough by making each cookie smaller.

This means that, chip shortage or no, Nvidia and AMD should be able to get these less powerful cards out of the fabrication plants at a much higher rate. We're seeing at least a little evidence of that already: Rumors indicate that there's a full court press of RTX 3050 variants coming from Nvidia's various manufacturing partners, like EVGA, MSI, and Asus.

Nvidia

Of course, with the way things are going, you're still unlikely to snag a new graphics card simply by strolling into Best Buy on your lunch break. Gamers will probably still need to resort to pre-orders, waiting in line in front of stores, spamming the F5 button at their favorite etailer, and all the other tedious minutia of trying to get a faster gaming PC in the COVID era. (Incidentally, going for a pre-built PC might still be a better option, at least if you need to upgrade multiple components at once.)

The Radeon RX 6500 XT is scheduled to land on January 19th, and the GeForce RTX 3050 is scheduled to launch on January 27th. Good hunting.

At a glance

Expert's Rating

Pros

• Excellent real world performance
• PS5 compatible
• Affordable

Cons

• A couple of GBps slower on PCIe 4 synthetic benchmarks
• Uses host memory bus (HMB) rather that onboard DRAM, though seemingly without the usual performance penalty.
• PS5 doesn't support HMB

Our Verdict

The Atom 50 was second-tier on synthetic benchmarks, but tied for number one in our real life transfer tests over PCIe 4. Nice. It's affordable and PS5 compatible, but uses HMB (your PC's DRAM) for primary cache, which the PS5 does not support. Hence, there might be a small drop in performance inside Sony's console.

Price When Reviewed

$120 for 1TB, $250 for 2TB

Best Prices Today

Retailer

Price

Delivery

$119.99

Free

View

Price comparison from over 24,000 stores worldwide

The Adata Atom 50 is one of the more affordable PCIe 4 SSDs on the market, which would seemingly make it a nice option for the PlayStation 5. It's second-tier over PCIe 4, according to synthetic benchmarks, but it tied for 1st place in our real world 48GB transfers–a eye-opening performance considering the price.

Even more impressive, the Atom 50 managed that feat without a DRAM cache. It instead utilizes HMB (Host Memory Bus, aka your computer's RAM) primary caching, a technology we've seen less than stellar results from previously. Alas, Sony's FAQ says thePS5 doesn't support HMB, which means the Atom 50 may not be quite as fast inside that console.

Design and specs

Adata's Atom 50 bucks the trend of bargain SSDs based on Phison-based designs by employing a RealTek RTS5766DL to shuttle data to and from the SSD's 176-layer TLC (Triple-Level Cell, 3-bit/state) NAND. As you'll see below in the performance section, it proved very proficient.

As noted, there's no DRAM cache, Adata instead opting for HMB. Also as mentioned, this is the first time we seen HMB perform up to snuff.

The Atom 50 is currently available in 1TB/$120 (tested) and 2TB/$250 flavors. That's a hair pricy for a bargain PCIe 4 SSD, but it's still in that ballpark and doesn't consider any discounts you might see. The drives are warrantied for five years or 650TBW (TeraBytes Written over the life of the drive) per 1TB of capacity. That latter is about average for the price point.

Performance

The Atom 50 turned in good PCIe 3 numbers and decent second-tier numbers over PCIe 4 in CrystalDiskMark 6, CrystalDiskMark 7, and AS SSD 2. Those benchmarks rate the fastest PCIe 4 NVMe SSDs during sustained transfers at around 7GBps while the Atom 50 was just below 5GBps.

I should mention that second-tier NVMe is still quite good in the grand scheme of things and the norm at the Atom 50's price point. It was easily the match of the equally bargain Sabrent Rocket 4 as shown in the CrystalDiskMark 6 numbers below.

Where things got interesting were during our 48GB real world transfers shown below. The Atom 50 over PCIe 4 actually tied the Corsair MP600 Pro XT for the fastest aggregate time (1 minute, 39 seconds) we've seen.

Looking at the Atom 50's PCIe 3 48GB numbers, which aren't bad but hardly record-setting, I'm thinking that HMB might be a lot more effective over the faster PCIe 4 bus than it was over PCIe 3.

The Atom 50 turned in an average time writing our single large 450GB file over PCIe 4, though it never slowed down handily outpaced the Crucial P5 Plus. The 450GB write tests secondary caching algorithms and the speed of the NAND. Keep in mind that the Atom 50 we tested was only a 1TB drive while the Sabrent Rocket 4 it nearly matched was a 2TB unit.

I found little to complain about with the Atom 50's performance, even if it couldn't match the 7GBps that top-tier SSDs deliver on synthetic benchmarks or their super-fast sustained 450GB writes (191 to 215 seconds). How often do you write 450GB? That said, consider the FireCuda 530 or MP600 XTs if every second counts, otherwise save money with the Atom 50.

The PCIe 3 tests utilize Windows 10 64-bit running on a Core i7-5820K/Asus X99 Deluxe system with four 16GB Kingston 2666MHz DDR4 modules, a Zotac (Nvidia) GT 710 1GB x2 PCIe graphics card, and an Asmedia ASM3242 USB 3.2×2 card. It also contains a Gigabyte GC-Alpine Thunderbolt 3 card, and Softperfect Ramdisk 3.4.6 for the 48GB read and write tests.

The PCIe 4 testing was done on an MSI MEG X570 motherboard socketing an AMD Ryzen 7 3700X 8-core CPU, using the same Kingston DRAM, cards, and benchmark software. ImDisk replaces Softperfect as the RAM disk. All testing is performed on an empty, or nearly empty drive that's TRIM'd after every set of tests. Performance of any SSD will decrease as you fill it up.

Write performance will decrease as the drive fills up. In some rare cases, components may change for the worse. Adata promises that the components in the Atom 50 will not change over time. However, our standard warning with all drives is if , given similar hardware, does not perform as well as our test unit, let us know.

Excellent for the money

The Atom 50 is an excellent bargain and a good fit for your PC. It will work in your PS5, however, you might not see exactly the same performance. Still, this is a nice effort from Adata and the best HMB NVMe drive we've tested to date.

Looking for a portable 2-in-1 laptop for work? Well, you're in luck. Today, Newegg is selling the MSI Summit E13 Flip Evo for just $1099 after a $100 rebate. That's a whopping $500 off of the original price. The sale ends in about 18 hours, though, so you'll want to act quickly.

Let's talk about the guts and ports. The MSI Summit comes equipped with an Intel Core i7-1185G7 processor, Iris Xe graphics, 16GB of DDR4 RAM, and a 512GB of NVMe SSD. In other words, you should expect very zippy everyday performance. The port selection is pretty diverse, too. You're getting one USB 3.2 Type-C, one USB 4.0 Type-C (DP/Thunderbolt 4), one USB 3.2 Type-A, a micro SD card reader, a webcam lock switch, and audio combo jack.

The convertible aspect of this laptop is super appealing, especially if you're a business professional that travels a lot. You can it prop it up like a tent or swing the 1080p touchscreen display around and use it like a tablet. This deal includes a sleeve for the laptop and a MSI pen for the touchscreen display, too. The pen is a nice bonus, as this type of accessory is normally an additional cost. The minimalist aesthetic is perfect for a professional environment. The bronze trim really sells it for me.

This is a great deal that's going to end soon. If you're really jonesing for something that's both functional and swanky, look no further than the MSI Summit.

[Today's deal: MSI Summit E13 Flip Evo for $1099 at Newegg]

Intel's Core i9-12900HK indeed fulfills the company's proclamation of being the “The fastest mobile processor. Ever,” based on the results of our early, exclusive hands-on performance testing of several 12th-gen laptops.

Intel, which unveiled its 12th-gen Alder Lake laptop CPUs just last week at CES, let PCWorld take four nearly identical laptops outfitted with the Core i9-12900HK for a spin during a live stream of our Full Nerd podcast. You can watch us run the benchmarks live in the recording of our live stream below.

While the four preproduction laptops aren't quite final units, they're based on MSI's GE76 Raider with final drivers, firmware and tuning still to come. The laptops were outfitted with the 14-core Core i9-12900HK CPU, an Nvidia GeForce RTX 3080 Laptop GPU, 32GB of DDR5/4800 RAM, and Windows 11. The updated GE76 remains the same externally as the previous-gen model, but there have been some upgrades to it, including a new MSI “phase-change liquid metal” thermal pad that boosts the CPU's performance. MSI has also tweaked the 12th-gen laptop's GPU to 175 watts, versus the 11th-gen version's 165 watt TGP GPU.

We ran all but the gaming benchmarks on laptops set to MSI's balanced profile, rather than its extreme performance profile, and also set to hybrid GPU mode rather than discrete mode. That's actually how Intel set the laptops up in its lab, which we left intact. Normally we'd opt for the higher performance profile to get the best possible performance, but we left them as set and just wanted to point it out to shoot down assumptions that Intel might have “rigged” this performance preview by opting for the maximum cooling settings.

We picked the tests we wanted to run, and ran them ourselves live on The Full Nerd (subscribe to PCWorld's YouTube channel if you aren't already), but make no mistake—these are laptops from Intel's own testing labs so you should always read the results with some modicum of salt until independent reviews are performed. We do believe the CPU performance you see here is likely be fairly representative of performance the production laptop reaches. But no two laptops are the exact same—even among the same make and model—so you should expect a little variance in high-end Alder Lake gaming laptops, but hopefully not huge swings.

You should know, however, that a CPU and GPU's performance isn't static. The magic each PC maker brings to it in cooling, tuning, and chosen parts can have very large impacts on performance. Basically, there is no such thing as completely isolating the CPU in a laptop. You must evaluate performance based on the total platform, not just the CPU or GPU inside of it.

To give you a point of comparison, we've included two additional laptops:

• The previous-generation MSI GE76 Raider outfitted with an 8-core 11th-gen Core i9-11980HK, Nvidia GeForce RTX 3080 Laptop GPU with a TGP of 165 watts, and 32GB of DDR4/3200. This laptop set to its Extreme performance setting.
• An Asus Strix Scar G17 outfitted with AMD's 8-core Ryzen 9 5900HX, Nvidia GeForce RTX 3080 Laptop GPU with a TGP of 130 watts, and 32GB of DDR4/3200. This laptop was tested set to Turbo.

Unfortunately, due to our time constraints we were unable to update the 11th-gen Core and Ryzen 9 laptops to Windows 11. Fortunately, for many of the CPU benchmarks we ran, the OS difference can be minimal.

Intel Core i9-12900HK performance preview

We'll kick off our preview with Maxon's popular Cinbench R20 using its default multi-threaded test. Cinebench is based on the company's Cinema4D 3D modelling and rendering engine that's sold as a stand-alone application and also included in other products, such as Adobe's After Effects. It hugs cores to its body, and more cores means more speed. The results we see aren't much of a surprise when you remember that Intel's 12th-gen Alder Lake H processors feature a hybrid design, using six improved “performance cores” coupled with eight “efficiency cores” in this flagship Core i9-12900HK. That means 14-cores (even if eight of them are lighter weight than the Ryzen's and 11th gen cores) nets more performance in Cinebench R20.

Still, you're looking at the 12th-gen Core i9 outpacing the Ryzen 9 5900HX and the Core i9-11980HK by roughly 30 percent.

IDG

As much as we wish they would, the vast majority of applications people use don't tap into every core available. In fact, single-core performance may matter what for what you do everyday more than multi-core performance, so we also ran Cinebench R20 using a single-thread to gauge performance there.

Intel's 12th-gen “Alder Lake” cores greatly boost efficiency per clock and we see that manifest itself here, where the 12th gen Core i9-12900HK is roughly 16 percent faster than the Ryzen 9 5900HX and 9.3 percent faster than the 11th gen Core i9-11980HK. Mind you, neither of those rival CPUs are slow chips either. Last summer, in fact, it felt like AMD's newly released Ryzen 9 5900HX was a monster compared to all previously released CPUs, which was then itself somewhat eclipsed by the equally impressive 11th Core i9-11980HK. Obviously, there's a new goliath in town.

IDG

Up next is the very similar Cinebench R23, which unlike Cinebench R20, takes about 10 minutes to run. That point is important because modern CPUs all generate heat as they run and ease off the pedal when they heat up. While Cinebench R20 takes maybe a minute or so to run, the 10 minutes of Cinebench R23 is likely to depress performance. The good news for 12th-gen is it's still the winner, since 14 cores is still more than 8 cores. The performance difference between the CPUs close up a little but the Core i9-12900HK still offers a commanding 23 percent advantage over the Ryzen 9 5900HX and 16 percent over the 11th-gen Core i9.

One other advantage to running Cinebench R23 is Mac fans can also compare performance directly, since Maxon's Cinebench R23 has native version for Apples new M1 line. While we have not directly tested the MacBook Pro 14, our sister publication Macworld has, and recorded a score of 12,381 for the MacBook Pro 14 with its 10-core M1 Pro CPU. That would give the 12th gen Core i9-12900HK a 21 percent advantage over a MacBook Pro 14. Macworld didn't review the M1 Max, but others outlets have outlets have reported performance to range from 12,300 to 12,700 for the faster M1 Max.

IDG

Like Cinebench R20, we also record a score in Cinebench R23 using a single thread. Interestingly, we actually see the gap open up a little here, with the 12th gen Core i9-12900HK outpacing the Ryzen 9 5900HX by about 26.5 percent and the 11th gen Core i9-11980HK by 16.6 percent. For the M1 Pro in the MacBook Pro 14, Macworld recorded a score of 1,531. That's an advantage of 23.6 percent for the 12th-gen CPU over the M1 Pro.

IDG

Up next is BAPCo's CrossMark benchmark. It's essentially a scaled-down version of its sprawling SYSMark test, which uses real applications to gauge performance. But rather than commercial applications, CrossMark is built using open-source programs and intended to gauge productivity, creative (photo and video), and responsiveness, which it defines as switching between software or starting an application. It's compiled with open-source compilers and is designed to run on Windows, MacOS, iOS and Android alike. On the Windows PCs, the 12th-gen Core i9 is a beefy 36 percent faster than the Ryzen 9 5900HX and also impressive 21.5 percent faster than the 11th-gen Intel chip.

Again, we didn't have a score for CrossMark for MacOS, but the public database BAPCo publishes puts the fastest M1 Max laptop at 1,670, which lands the Core i9-12900HK at about 18 percent faster. Like all public benchmark databases, it's always hard to know the provenance of the results, but we'll hope for the best. Looking into the details, M1 Max does slightly outpace the 12th-gen Core i9 in the Creativity portion, with a score of 2,254 vs. the 12th gen's 2,132, but the newest Intel CPU beats it in productivity by 25 percent with a score of 1,874. Interestingly the Alder Lake H system also crushes the M1 Max in CrossMark's “Responsive” test to the tune of 59.9 percent, with a score of 1,836 for the Core i9.

IDG

All-in-one suites such as CrossMark attempt to measure everything you might do on your computer, but the vast majority of the world brings home the turkey bacon using Microsoft's Office suite. So to keep Bill Lumberg happy, we also ran UL's Procyon Office test, which uses Microsoft Word, PowerPoint, Excel, and Outlook to simulate what most people do all day while waiting for 5 p.m. to come. The 12th gen Core i9-12900HK wins, but it's really just a meh, with Intel's new chip outpacing the 11th-gen Core and Ryzen 9 systems by only 5 percent. Frankly, if you're using a Core i9 or Ryzen 9 with a GeForce RTX 3080 to primarily pound out a PowerPoint, you're probably doing it wrong.

IDG

Our next result is Principled Technologies WebXprt 3. It's a web-based benchmark designed to measure browser performance at photo enhancement, organizing an album using AI, stock option pricing, encryption notes, OCR scans, sales graphs, and online homework. These are fairly advanced features for a browser test. All of our results were generated using Edge on the Chromium 97.

The Core i9-12900HK again leads the field, turning in a score about 9.5 percent faster than the 11th-gen chip and 16.2 percent faster than the Ryzen 9.

IDG

We did manage to run games on the 12th gen Core i9-12900HK, but there are a lot of caveats with our comparison that we really want to highlight in yellow. Although all of the laptops feature GeForce RTX 3080 Laptop GPUs, their power ratings weren't the same, which can make any comparison in gaming really sketchy. That's especially true for the Ryzen 9 system, which was paired with a GPU rated at 130 watts vs. the 11th-gen's 165 watts, and the 12th-gen's 175 watts. That's basically a 27 percent larger thermal and power budget for the 11th-gen laptop and 35 percent for the Core i9-12900HK laptop. Between the 11th-gen and the 12th-gen, there's only about a 6 percent difference in GPU power.

That said, should also point out that one of the problems we saw last year was difficulty in getting a lot of gaming laptops with higher wattage GPUs. So if most 165 watt GPUs were only in Intel machines, then maybe it's actually a fair comparison?

Gaming performance isn't always about GPU wattage though and we see that below in our first result from Ashes of the Singularity: Escalation. We ran this benchmark at 1080p using the Crazy preset for graphics and opt for the CPU-focused benchmark run. The CPU-focused benchmark puts more units on the screen, which should push the CPU harder than the regular GPU-focused test. You can see the Ryzen 9 and 11th-gen systems are dead-even despite have a sizeable GPU thermal budget difference. That actually makes us even more impressed by the Core i9-12900HK's 11.7 percent lead over the others. The test basically doesn't seem to care about GPU wattage yet Intel's new chip is ahead by double-digits in the CPU test.

IDG

Our next game result is Far Cry New Dawn run at 1080p using the ultra preset. The caveat here is this game series has long given AMD's chips the raspberry, but that changed with the Ryzen 5000-series of desktop chips, which saw AMD actually eclipse Intel's best 10th-gen desktop chip at last. The laptop Ryzen 5000, however, features less cache, which may not give it the same boost. Still, given the 35 percent larger GPU thermal budget in the Core i9-12900HK laptop, we don't think it's necessarily fair to make direct comparisons with the Ryzen 9 system, so consider it there for reference.

We do think it's fair to compare performance against the 11th-gen GE76 Raider, however, since both GPUs are separated by just 10 watts of power. The 12th-gen GE76 with a Core i9-12900HK inside pulls up the win by double digits, with a frame rate about 11.8 percent faster.

IDG

Our last game result is from the popular Counter Strike: Global Offensive which is a graphically low-intensity game in this age. We assumed that would make it mostly a CPU-intensive game, but the results interestingly mostly mirror the GPU wattage differences between the different laptops. The Core i9-12900HK laptop is still faster than the 11th-gen notebook, but we're oly looking at roughly a 6 percent win. We'll just have to wait for final drivers to really make any determination in gaming.

IDG

We saved our last official result for the Official Benchmark of Twitter Shouting Matches: Geekbench 5. It's a popular, cross-platform benchmark based around a set of small programming “loops” that are designed to simulate various functions such as JPEG decompression, text decompression, encryption, AI and more. We find most of Geekbench's value to be in the sub-scores that let you dive into each individual result's performance. We also generally only compare results on the same platform. But that's not how Twitter hot takes are created, so we'll stick with just the overall score here.

The result gives the Intel Core i9-12900HK a massive 56.6 percent advantage over the Ryzen 9 5900HX chip. The 12th-gen chip is equally impressive compared to its predecessor, with a 40 percent increase in performance over the 11th-gen Core i9. Again, it's worth remembering that these are not slow CPUs, with the 11th-gen Core and Ryzen 9 chips delivering jaw-dropping performance just last summer—making the 12900HK's win all the more staggering.

We don't have an in-house score for the M1 Pro, so we'll crib from Macworld again. It reports a score of 12,590 for the 10-core M1 Max and a score of 12,544 for the 10-core M1 Pro. That puts the M1 Max just 6 percent away from the 12th-gen Core i9, and although Geekbench's utility as the final word on performance has some real limits, it's still an impressive showing nonetheless for both Apple's M1 and Intel's 12th-gen chips.

IDG

What happens when you try to deconstruct the performance of an 8-core Ryzen 9 or 11th-gen Core i9 versus a 10-core M1 Max and the 14-core Core i9-12900HK? You look at GeekBench 5's single-core performance benchmark. Against the Ryzen 9, the 12th-gen Core i9 shows an advantage of 27 percent, and a 14.6 percent gain versus the 11th-gen Core chip.

Referencing Macworld's review of the MacBook Pro again, we see near identical scores of 1,774 for the M1 Max and 1,778 for the M1 Pro—basically within the margin of error. That puts the Core i9-12900HK with a 6.8 percent win over the M1 Pro and M1 Max so, yes, head for Twitter folks.

IDG

For our last score, we ran PugetBench's Premiere Pro benchmark, but this is one test where we wonder if more tuning is yet to come. PugetBench is Puget System's popular benchmark that measures a fairly advanced set of actions and exports in Adobe's Premiere Pro.

We unfortunately don't have scores for the Ryzen 9 and the 11th-gen Core i9 laptops using the current version of Premiere Pro (22.1), so we dug up some published scores from PugetBench's database of user generated scores. Since we have no idea what conditions they were run in, we don't have the same faith in them as the scores from, say, Macworld or other reviewers. Still we did see a score in the database of 693 for a Ryzen 9 5900HX laptop with a GeForce RTX 3080 Laptop GPU, and a score of 892 from an Core i9-11980HK laptop with a GeForce RTX 3080 Laptop GPU. The Core i9-12900HK coughed up a solid 1,037, which would make you think it's case closed. The problem for the 12th-gen Core i9 is the reported score in PugetBench's database for an M1 Max MacBook Pro: 1,324. Like the Ryzen 9 and 11th gen Core i9, we don't know the conditions run for that MacBook Pro, but it's impressive nonetheless.

At the same time, we've seen very large performance swings in PugetBench's Premiere Pro benchmark based on the QuickSync capability in Intel's integrated graphics cores. It's entirely possible we'll see the Core i9-12900HK's PugetBench score move up with final drivers. And it's also possible it won't move at all either. We'll have to wait and see.

And that, ultimately, is the weakness of any performance preview. While it's been fun to see Intel's 12th-gen Core i9-12900HK strut its stuff and outpace its Ryzen 9, 11th-gen Core i9, and M1 Max rivals in many tasks, you need to wait until we have final numbers from reviewers of final products before making any buying decisions. Hopefully you won't have to wait long.