Ryzen AI 9 HX 470 Leads CPU Renders, Intel 386H Uses Less Power

The useful finding is how often the buying decision changes when the test changes. A Ryzen AI 7 laptop can compete well in the reported graphics workloads despite trailing in CPU rendering; an older Ryzen AI 9 can nearly match its successor; and a quieter operating profile can preserve most of a notebook’s maximum CPU performance.

Andrei Girbea’s comparison for Ultrabookreview draws on reviews of Lenovo Legion, Lenovo Yoga, Asus ROG Zephyrus, and Asus ProArt notebooks. Those are measurements of complete products, with different power limits, cooling systems, graphics configurations, and displays. They offer useful purchasing evidence, but they do not isolate the processors under identical conditions. The performance measurements below remain single-source results, rather than findings independently reproduced by multiple publications.

Ryzen AI 9 HX 470 versus Core Ultra 9 386H starts with the laptop​

The main comparison puts the Ryzen AI 9 HX 470 in a Lenovo Legion 7a against the Core Ultra 9 386H in two different machines: an Asus ROG Zephyrus G16 and a Lenovo Yoga Pro 9i. Ultrabookreview lists approximate CPU power levels of 85W, 75W, and 65W, respectively, alongside its benchmark table. That is already enough to rule out interpreting the results as an equal-power contest between AMD and Intel silicon.

The supporting systems widen the decision rather than merely filling out a chart. According to Ultrabookreview, the older Ryzen AI 9 HX 370 appears in a Zephyrus G16 at approximately 80W, while the Ryzen AI 7 450 appears in a Legion 5a at approximately 80W. A Strix Halo Max+ 395 in an Asus ProArt PX13, listed at approximately 70W, provides a reference for buyers whose work might justify a different processor class. These figures describe the reviewed implementations; they are not universal operating requirements for every laptop carrying those processors.

The architecture specifications explain why neither core count nor the largest number on the box settles the contest. Ultrabookreview lists the HX 470 with four Zen 5 cores and eight Zen 5c cores, providing 24 threads. Intel’s official Core Ultra 9 386H specification lists four Performance-cores, eight Efficient-cores, and four Low Power Efficient-cores, providing 16 cores and 16 threads. These are different arrangements of processing resources, not interchangeable units that can be ranked by counting them.

Intel’s primary specification also corrects two details in Ultrabookreview’s platform table. The comparison gives the 386H an Efficient-core maximum frequency of 3.9GHz and LPDDR5x support up to 8,533MT/s. Intel lists 3.7GHz and support up to 9,600MT/s, respectively, alongside DDR5 support up to 7,200MT/s. Those corrections concern the processor’s published capabilities; they do not establish what memory speed a particular reviewed notebook actually used.

Core Ultra 9 386H capabilities are not a complete notebook specification​

Intel lists a 25W processor base power and an 80W maximum turbo power for the 386H. Its reviewed notebooks operating around 65–75W therefore need to be understood separately from that base-power figure. Likewise, a processor’s maximum supported memory capacity, interface speed, or PCI Express revision does not establish the configuration a laptop manufacturer sells. The distinction becomes important when comparing an attractive processor specification with a fixed-memory retail configuration.

For memory, Ultrabookreview reports that many of these notebooks use onboard, non-upgradeable LPDDR5x, while some mid-tier designs provide replaceable DDR5 modules in SO-DIMM slots. The practical choice is therefore between actual configurations: enough installed memory at purchase, or a machine with a documented expansion path. A buyer should not accept a smaller fixed-memory configuration merely because its processor wins a CPU benchmark by a few percent. That recommendation follows from the permanence of the configuration, not from any memory-capacity benchmark in this comparison.

Storage presents a similar distinction. Ultrabookreview identifies PCIe 4.0 support on the AMD platforms and PCIe 5.0 capability on the Intel side; Intel’s official 386H record confirms support for both PCIe 5.0 and PCIe 4.0. That makes a PCIe 5.0 SSD configuration possible on the Intel platform, but does not guarantee that every Intel notebook exposes such a connection at its storage slot. The comparison supplies no storage-workload results, so interface support alone cannot justify a claim that the Intel machines load applications or finish file operations faster.

CPU benchmarks favor the HX 470 selectively, not universally​

Ultrabookreview’s broad finding is that the Ryzen AI 9 HX 470 and Core Ultra 9 386H usually finish within roughly 5–10% of one another in its sustained CPU comparison, with an AMD advantage. Its individual benchmark results are more revealing than that summary because they show both larger gaps and reversals. For a buyer running a specific application all day, the relevant result is the one closest to that application—not an average impression assembled from unrelated tests.

The following measurements are Ultrabookreview’s reported results. Higher scores are better in the Cinebench and Geekbench columns; a shorter time is better in the Blender CPU-rendering test. The source identifies Cinebench R23 as a best run and Cinebench 2026 as a 10-minute run, so those columns also represent different test conditions, not merely different scoring scales.

In Cinebench 2026, the HX 470 Legion scores about 5% above the Intel Zephyrus and about 9% above the Intel Yoga, calculated from Ultrabookreview’s figures. That fits the publication’s general description well. In Cinebench R23, however, the corresponding advantages are approximately 12% and 14%. The choice of benchmark changes the apparent size of the lead, even before accounting for the different notebook power levels.

Geekbench 6.6 reverses the multi-core ranking. Using Ultrabookreview’s reported scores, the Intel Zephyrus finishes about 9% above the HX 470 Legion, and the Intel Yoga about 8% above it. This is direct evidence against describing the HX 470 as universally faster in multi-threaded work. It leads in the reported Cinebench tests and Blender render, while the 386H systems lead in this Geekbench multi-core result.

Blender provides a useful time difference, with a narrow scope​

The Blender results make the purchasing consequence easier to picture. Ultrabookreview reports that the HX 470 Legion completes the Classroom CPU render in 225 seconds, compared with 249 seconds for the Intel Zephyrus and 264 seconds for the Intel Yoga. That is a saving of 24 or 39 seconds for this particular scene and test version. A user repeatedly doing similar CPU rendering has a concrete reason to give the AMD implementation more weight.

Percentage language needs care here. Calculated from those times, the HX 470 takes roughly 10% less time than the Intel Zephyrus and 15% less time than the Intel Yoga. Expressed as rendering throughput instead, the advantages are approximately 11% and 17%. Those are different ways to describe the same measurements; neither supports promising that every Blender project will finish 15–20% sooner.

The CPU-compute setting is also part of the result. These timings do not establish the relative performance of Blender rendering on each notebook’s discrete GPU. Nor do they directly establish compile times, virtual-machine responsiveness, spreadsheet recalculation, or another application’s performance. For software developers and IT professionals, the responsible use of this evidence is to identify promising candidates for a workload-specific evaluation, rather than converting a rendering result into a universal productivity forecast.

Single-threaded results are similarly mixed. Ultrabookreview’s Cinebench R23 scores put the Intel notebooks at 2,127 and 2,115, ahead of the HX 470’s 2,059. Cinebench 2026 also favors Intel, at 521 and 502 against AMD’s 482. But Geekbench 6.6 single-core gives the HX 470 a score of 2,952, above the Intel machines’ 2,907 and 2,865. The evidence supports “Intel leads in these Cinebench single-thread tests,” not an unconditional Intel single-core victory.

This is why a modest benchmark lead deserves proportionate purchasing weight. Someone buying primarily for the documented Blender CPU workload can favor the HX 470 Legion on that evidence. Someone whose application resembles neither Blender nor Cinebench cannot derive an equally precise recommendation. With the flagship processors this close, a meaningful difference in memory capacity, price, portability, or the behavior of the exact notebook can reasonably decide the purchase.

Ryzen AI 7 450 and HX 370 challenge the flagship upgrade​

The older Ryzen AI 9 HX 370 is the clearest challenge to paying for a new generation solely to obtain more CPU performance. Ultrabookreview reports a Cinebench R23 multi-core score of 23,719 for its HX 370 Zephyrus, against 24,113 for the HX 470 Legion. The newer machine’s advantage is about 1.7%, calculated from those scores. The single-core results, 2,049 and 2,059, are closer still.

The other reported tests preserve that general picture without making the processors identical. According to Ultrabookreview, the HX 370 scores 15,238 in Geekbench multi-core against the HX 470’s 15,671, and completes the Blender render in 3 minutes 53 seconds against 3 minutes 45 seconds. The latter is an eight-second difference on a render lasting almost four minutes. Cinebench 2026 has no HX 370 result in the comparison, so there is no basis for filling that missing cell with an estimate.

For an existing HX 370 owner, these measurements provide little CPU-only justification for replacing a satisfactory machine. For a new buyer, they keep an older configuration in contention if the actual price and specifications are favorable. That is a conditional value judgment: Ultrabookreview suggests that previous-generation Strix Point notebooks can cost substantially less, but provides no like-for-like price table here from which to calculate a universal saving.

The Ryzen AI 7 450 offers a different compromise. Ultrabookreview lists four Zen 5 and four Zen 5c cores, with 16 threads, and characterizes its sustained CPU performance as roughly three-quarters of the Ryzen AI 9 systems. Its reported Cinebench R23 score is approximately 75% of the HX 470 result, while its Cinebench 2026 score is about 73%. That is a substantial reduction for work that tracks those measurements.

The reduction is smaller elsewhere. From Ultrabookreview’s figures, the Ryzen AI 7 reaches about 88% of the HX 470’s Geekbench multi-core score. Its Blender result of 4 minutes 40 seconds takes 55 seconds longer than the HX 470, which means approximately 24% more elapsed time—not 25% less performance in every conceivable task. The practical question is whether the buyer will repeatedly encounter the workloads where that difference is visible.

Max+ 395 offers a different performance class, with exceptions​

Ultrabookreview’s Max+ 395 results demonstrate that the two headline processors are not the ceiling for portable CPU performance. Its ProArt PX13 completes the Blender test in 2 minutes 36 seconds, compared with the HX 470 Legion’s 3 minutes 45 seconds. That is 69 seconds less waiting for this render, or approximately 31% less elapsed time. Its Cinebench 2026 multi-thread score of 6,816 is about 35% above the HX 470’s 5,031.

But the comparison does not support applying a blanket “at least 50% faster” description to every Max+ 395 result. Its listed Cinebench R23 score of 20,841 falls below the HX 470, both Intel implementations, and the HX 370. The source does not explain that reversal in the table, and it would be inappropriate to invent a thermal, software, or configuration explanation. Its Geekbench multi-core lead over the HX 470 is approximately 21%, another result that resists a single universal percentage.

For buyers doing the documented CPU rendering, Max+ 395 is nevertheless a meaningful alternative rather than an incidental entry. The measured time saving over the HX 470 is larger than the HX 470’s saving over either Intel notebook. That suggests a sensible order of consideration: first decide whether the workload warrants the higher-performing configuration, then decide which mainstream flagship offers the better complete laptop if it does not.

Taken together, the supporting processors divide the buying decision into three distinct cases. The HX 370 challenges the need to upgrade generations; the Ryzen AI 7 450 challenges the need to buy a flagship for graphics-oriented use; and Max+ 395 challenges the assumption that either headline processor is the best destination for sustained CPU work. None can be reduced to a position in a simple newest-to-oldest hierarchy.

Lower-power profiles preserve much of the CPU performance​

The largest practical opportunity in Ultrabookreview’s comparison may be available after purchase, through the notebook’s supported operating profiles. The publication reports that the tested AMD and Intel platforms retain approximately 90% of their high-power CPU capability around 50–60W, with quieter fans and lower internal temperatures. That makes maximum-performance mode a choice with a visible trade-off, rather than an obvious setting to leave enabled permanently.

The reported relationship is more useful than the wattage alone. If a notebook retains 90% of a throughput score, completing the same amount of work would, under a simple proportional assumption, take about 11% longer. That is an illustration of what the reported percentage means, not a new measurement of any application. For a short task, the extra waiting may be small; for repeated long renders, it accumulates and may justify the louder mode.

Ultrabookreview also reports approximately 70–80% of maximum capability at lower power levels associated with Quiet or Silent operation and thinner implementations. This is a larger performance compromise, but still leaves substantial CPU capacity available. The comparison therefore provides grounds to consider a quieter notebook profile for ordinary work without assuming that the machine becomes unsuitable for demanding tasks. The exact result remains dependent on the reviewed design and workload.

These are not universal Windows power settings. The report uses names including Quiet, Silent, Whisper, Adaptive, and Battery Saver across different products, without establishing a common menu, software interface, or wattage mapping. It does not supply a verified universal procedure for imposing a 50W or 60W limit. Buyers and owners should therefore interpret the results as evidence about supported notebook profiles, not as instructions to enter an arbitrary power limit in firmware or a tuning utility.

For the Core Ultra 9 386H, Ultrabookreview specifically reports about 65% of maximum tested capability around 25–30W, citing the Asus Zenbook S14 as an ultraportable implementation. That result is useful when comparing a thinner laptop against the larger Zephyrus or Yoga machines. The same processor name does not imply the same sustained score, and a large notebook’s benchmark should not be used as the promised performance of the thinner one.

CPU power limits and battery power answer different questions​

The 50–60W discussion concerns CPU performance under load. The much smaller wattage figures later in the comparison concern average notebook power draw during text editing, streaming, and browsing. These are different measurements. A machine can permit substantial CPU power during a demanding plugged-in workload while consuming much less during light battery use; the source does not establish a single number that describes both circumstances.

This boundary also prevents an unsupported energy-efficiency conclusion. Retaining most of a benchmark score at a lower CPU power level is encouraging, but it does not by itself establish the total energy used to complete a task. That would require the relevant whole-system draw over the task’s duration. The comparison reports useful performance and light-use power observations, but does not provide a complete energy-per-render accounting for every profile.

Ultrabookreview says the Strix Halo Max+ platform loses more performance when its power allowance is reduced, while still retaining a substantial advantage in the sustained comparison it describes. That means the higher-tier alternative also needs to be evaluated at the mode a buyer will actually tolerate. Its best result alone does not establish its advantage at a particular noise level or in a particular small chassis.

The buying implication is straightforward: compare performance in the operating mode you expect to use, not merely the fastest mode a reviewer can select. A quieter profile that preserves most of the measured capability can be a better fit for shared offices or sustained everyday use. The comparison does not include the numerical noise and temperature evidence needed to declare one notebook universally quieter or cooler, so that final choice must remain specific to the product and profile.

RTX 5060 results weaken the case for buying by CPU tier​

The graphics section gives the Ryzen AI 7 450 its strongest argument. Ultrabookreview compares RTX 5060-equipped Legion 7a, Yoga Pro 9i, and Legion 5a configurations, and reports broadly similar results in several graphics-oriented workstation tests despite the processors’ different CPU scores. This supports a narrower, useful conclusion: spending more for the flagship CPU did not produce a correspondingly large advantage in those measured graphics workloads.

The configurations still need to remain attached to the results. Ultrabookreview lists RTX 5060 power allowances of 95W and 75W for the Legion 7a, 100W and 65W for the Yoga Pro 9i, and 105W and 75W for the Legion 5a. Those are GPU power figures, described as total graphics power, or TGP. They should not be confused with the CPU power levels in the earlier benchmark table, and they are not identical across the three laptops.

At the higher listed GPU power settings, Ultrabookreview reports SPECviewperf 2020 3DSMax scores of 106.98 for the HX 470 Legion, 108.37 for the Intel Yoga, and 108.52 for the Ryzen AI 7 Legion. Its Maya scores are 406.04, 400.80, and 411.38, respectively. These closely grouped results show why a sizeable CPU-rendering gap need not translate into a similar gap in the tested graphics viewsets.

That does not establish equal performance throughout 3ds Max, Maya, or another professional application. The reported SPECviewperf viewsets are specific tests, while the Blender table explicitly measures CPU rendering. A workstation buyer can reasonably use the contrast to ask which part of the workflow consumes time. The source does not support treating the better-value processor choice for one operation as the right choice for every operation inside the same application.

The graphics scores also show how a processor hierarchy can become misleading. According to Ultrabookreview, the Ryzen AI 7 Legion records the highest Time Spy graphics score among the three higher-power configurations, at 12,650, versus 11,829 for the Intel Yoga and 11,499 for the HX 470 Legion. Its GPU also has the highest listed power allowance. The result is evidence about those complete configurations, not proof that the Ryzen AI 7 CPU is inherently better at graphics.

The gaming table cannot isolate an AMD-versus-Intel winner​

Ultrabookreview explicitly warns that its gaming comparison is compromised by differences in cooling and the presence of a graphics MUX in the Legion systems. The publication says it had not tested a directly comparable Core Ultra 9 plus RTX 5060 configuration with a MUX for this exercise. That warning belongs beside any interpretation of the frame rates, rather than being relegated to a footnote after declaring a winning processor.

One reported result illustrates the problem well. At the higher listed GPU power settings, Ultrabookreview records 81 frames per second in Red Dead Redemption 2 for the HX 470 Legion, 59 for the Intel Yoga, and 78 for the Ryzen AI 7 Legion. Read without the configuration warning, the table could look like a decisive CPU-brand result. Read with it, the table shows how differently the reviewed notebooks perform while leaving the cause unisolated.

Cyberpunk 2077 adds another reason to avoid a one-number verdict. With the demanding settings, DLSS, and frame-generation configuration reported by Ultrabookreview, the HX 470 Legion averages 58fps against the Intel Yoga’s 56fps. Yet the listed 1% low figures are 26fps and 46fps, respectively. The Ryzen AI 7 Legion averages 62fps with a 26fps 1% low. The ordering therefore changes depending on which reported frame-rate statistic the buyer emphasizes.

These measurements are also tied to the games’ listed resolution and visual settings. They cannot be carried across unchanged to another resolution, a different rendering preset, or operation without the reported upscaling and frame-generation options. Ultrabookreview expects comparable Ryzen AI 9 and Core Ultra 9 implementations to have no major real-world frame-rate difference, but that expectation is not a substitute for the matched comparison it says it lacks.

For a gaming purchase, the actionable lesson is to prioritize the reviewed notebook configuration over the CPU badge. The Ryzen AI 7 450 should not be dismissed because it loses in Cinebench, and the HX 470 should not be crowned a gaming winner because a better-cooled Legion beats a Yoga. This evidence is about Windows laptop selection, not a ranking of the games themselves, and its strongest contribution is identifying which comparisons are safe to make.

Core Ultra 9 386H shows the clearest light-use power advantage​

Battery-oriented buyers get a more consistent directional result. Ultrabookreview reports lower average power draw from its Core Ultra 9 386H notebooks during text editing, Netflix playback in Microsoft Edge, and Edge browsing than from the tested HX 470 Legion. Its newer AMD systems also generally improve on the older Strix Point notebooks in those light-use observations. The measurements favor Intel, while their conditions prevent assigning the entire difference to the processor.

The table below preserves the publication’s system-level power readings. They are watts consumed during the reported activities, not battery-life durations and not isolated CPU package measurements. Ultrabookreview keeps Wi-Fi enabled in these tests, but varies the notebook profile, brightness setting, and—in the streaming tests—the stated video format.

The source’s naming is inconsistent around the older AMD systems: its main comparison identifies the HX 370, while the battery section abbreviates the processor to “Ryzen AI 9 370.” The readings are therefore best kept attached to the named Zephyrus notebooks rather than silently treated as a perfectly controlled generational experiment. The broader comparison still makes clear that these entries represent the older AMD implementations being contrasted with the 2026 systems.

For the HX 470 Legion and Ryzen AI 7 Legion, Ultrabookreview specifies Quiet mode and 50% screen brightness. Its Intel Yoga readings use 40% brightness, with Battery Saver for text editing and Netflix, and Adaptive mode for browsing. The Intel Zephyrus uses Silent mode at 50%; the older AMD Zephyrus machines use Whisper mode at 50%. The Zephyrus Netflix tests explicitly specify 4K HDR playback, whereas the Legion and Yoga entries simply say fullscreen Netflix.

Those distinctions materially limit the claim. A brightness-slider percentage is not a documented common display luminance across different notebooks, and different operating modes cannot be assumed to impose identical behavior. The results nevertheless give buyers useful evidence about how the reviewed products behave in their reported light-use configurations. The appropriate conclusion is that the tested Intel notebooks used less power—not that the 386H alone accounts for every watt saved.

Lower power draw does not establish a fixed number of battery hours​

The Intel advantage is especially clear in the reported streaming readings. Ultrabookreview’s Intel notebooks fall around 6–7W, compared with 9–10W for the HX 470 Legion and 10W for the Ryzen AI 7 Legion. The older AMD Zephyrus readings are higher again, at 14–15W. That supports giving the newer Intel implementations serious consideration for users who spend substantial unplugged time watching video, subject to the playback and display conditions above.

Browsing is less cleanly separated. According to Ultrabookreview, the HX 470 and Ryzen AI 7 Legion systems both range from 12–20W, while the Intel notebooks range from 9–13W, 10–12W, or 9–12W depending on the model. The newer AMD readings overlap the older AMD machines’ browsing ranges. Consequently, a broad statement that every Gorgon Point browsing result dramatically improves on every Strix Point result would overstate what the table shows.

Battery capacity is the other indispensable variable. In general, runtime depends on the usable energy in the battery and the rate at which the complete machine consumes it. Two notebooks with different battery capacities can therefore deliver different hours of use even when one has a more favorable power-draw reading. Ultrabookreview explicitly chooses average power draw for this comparison because battery sizes vary, so the figures should not be converted into promised runtimes without the corresponding capacities and test conditions.

This also separates light-use efficiency from sustained-load behavior. A notebook drawing 7W during streaming has not thereby demonstrated superior energy efficiency during a long CPU render, and the fastest Cinebench machine has not thereby demonstrated better battery life. Buyers who divide their day between desk-based rendering and unplugged browsing have two distinct pieces of evidence to weigh, rather than a single overall efficiency champion.

The Intel systems’ consistency across three reviewed implementations makes them a strong starting point for a battery-focused shortlist. The HX 470 still offers a more attractive light-use profile than the older AMD Zephyrus measurements in several activities, so choosing AMD need not mean accepting the previous generation’s reported draw. The final comparison should use the actual notebook’s battery capacity and measured runtime, not a projected number of hours derived solely from the processor name.

Choose the Windows laptop around the work you actually do​

Start by identifying whether the purchase is constrained by sustained CPU work, graphics performance, unplugged use, or the cost of the complete configuration. The comparison provides different answers for those priorities. It does not provide a single score that can reliably combine them, and its pricing discussion is too general to establish a universal value winner.

For repeated CPU rendering resembling the documented Blender test, the HX 470 Legion has a supported advantage over the two tested 386H notebooks. A buyer with enough of that work should also consider the Max+ 395 result, because its time saving is materially larger. Conversely, an existing HX 370 owner whose machine already meets requirements has little reason in these CPU measurements alone to replace it.

For graphics-oriented professional work and gaming, keep the Ryzen AI 7 450 Legion in contention. Ultrabookreview’s results show that its lower CPU tier does not prevent competitive RTX 5060 performance in the tested configurations. The money question is then whether a particular flagship notebook offers enough additional capability, portability, or other required features to justify its actual price—not whether “Ryzen AI 9” sounds more appropriate for a demanding user.

For developers and IT buyers, memory configuration deserves an explicit place before ordering. The comparison establishes that onboard LPDDR5x and upgradeable DDR5 designs coexist in this market, but it does not provide application-specific memory requirements or a universal expansion procedure. Choose the required capacity and documented upgrade arrangement for the exact configuration, then use the CPU results to refine the shortlist. Neither a processor’s maximum supported capacity nor a benchmark lead repairs an unsuitable fixed-memory purchase.

For unplugged office work and streaming, prioritize a notebook-level battery evaluation among the Intel candidates and the newer AMD alternatives. The reported power readings give the 386H systems an advantage, but the eventual hours of use still depend on the machine’s battery and operating conditions. A cheaper older HX 370 configuration may remain sensible for predominantly plugged-in use; the comparison does not establish that its possible purchase-price saving compensates every buyer for higher reported light-use draw.

The most concrete purchasing rules from this comparison are these:

• Prefer the Ryzen AI 9 HX 470 implementation for the documented CPU-rendering advantage, but do not extend that lead to every application or every multi-core benchmark.
• Keep the Ryzen AI 9 HX 370 on the shortlist when its actual price and configuration are favorable, because its reported CPU performance remains close to the newer AMD flagship.
• Consider the Ryzen AI 7 450 for RTX 5060-based graphics workloads when its lower sustained CPU performance does not constrain the work you do.
• Give Core Ultra 9 386H notebooks priority for a battery-focused evaluation, while comparing actual runtime, battery capacity, display conditions, and operating profiles.
• Compare supported quieter profiles before deciding how much maximum CPU performance you need, because Ultrabookreview reports substantial performance retention at reduced power.
• Verify installed memory, upgradeability, GPU power configuration, and storage-slot capabilities on the exact notebook rather than treating processor-level specifications as guarantees.

The next purchasing decision should therefore be between complete configurations, not AMD and Intel names in isolation. Ultrabookreview’s measurements make the HX 470 a credible CPU-rendering choice and the 386H notebooks compelling light-use candidates, while preserving good reasons to buy an HX 370, a Ryzen AI 7 450, or a higher-performing Max+ system. The buyer who matches those documented strengths to a real workload has a stronger basis for choosing than the buyer who simply orders the newest flagship.