ARM announces v9.2 X925, A725 and A520 cores to power next-gen flagships

30 May 2024
The Cortex-X925 replaces the Cortex-X4, as well as the new Cortex-A725 and Cortex-A520.

Sort by:

Ok, 31 May 2024But you compared it to the 8g1, this is nothing like 8g1 at... moreIdk man, whichever battery drain tests I saw, the 8 Gen 3 phones were far hotter than the 8 gen 2 and 8+ gen 1 counterparts were. That makes me think the 8 gen 3 is more inefficient than the 8 gen 2 overall. The 8 Gen 2 doesn't get THAT hot when pushing it but the 8 gen 3, YIKES. The stress tests take us back to SD 888 days. The heat temps remind me of the SD 810. I remember everyone clowning on it back in the day cus of its heating.

    • Y
    • Yea
    • CbI
    • 01 Jun 2024

    Kangal, 01 Jun 2024I think people should avoid referencing chipsets built on i... moreSure i read all that bs

      AtomicMarrow147, 30 May 2024Smartphones are now becoming next gen handeled console. Ma... moreYes, Facebook will never be so fast of rendering pictures in the split seconds 🤣

        Ok, 31 May 2024But you compared it to the 8g1, this is nothing like 8g1 at... moreI think people should avoid referencing chipsets built on inefficient silicon.
        QSD 855
        QSD 865
        MTK D1300 (avoid mentioning the Exynos 2100, 2200, QSD 888, 888+, QC 8g1)
        MTK D9000 (or QC 8g1+)
        QC 8g2
        QC 8g3
        QC 8g4 (or MTK D9500)

        On another topic, this whole announcement from ARM was disappointing. No raw figures. Using incorrect graphs. And throwing a lot of marketing buzzwords. Did I mention AI ??!
        Looks like the labs in ARM Facilities haven't produced anything substantial this past 1-2 years. All the performance and efficiency gains seem to be coming from the lithography jump. And that new naming scheme is something straight out of Intel.

        I think the upcoming architecture from Apple/Nuvia/Oryon itself may have been really impressive and innovative, but that was 2019. The industry has moved on. ARM has probably caught up to it in some sense with these "Cortex X5" cores. So I don't expect much progress from Qualcomm this year either. Overall it should still be decent, but that's owing mostly to TSMC fixing up their 3nm chips. The Apple A17-PRO and M3 lineup are pretty horrible. But the M4 is good, and so the A18 Pro is likely to be good too. MediaTek's next Dimensity 9500 should be good as well, now that we've seen their new designs (no small cores) works well, so they will further refine that with the new silicon node. As for Google, the Tensor G1 was competitive because the competition was bad, but the G2 was not really an upgrade back when the competitors got their act together, and the G3 is a good improvement but falling behind the competition. The G4 promises a big leap forward, but it will probably only surpass the QC 8g2 in terms of efficiency and fall behind the QC 8g3, let alone it's direct competitor which is the QC 8g4. Different case with Exynos, where it tried to remain competitive performance wise but came up short efficiency wise. The latest Exynos 2400 made big leaps, but still it's more or less a year behind. It will likely make another big jump for next year, and only remain something closer to being half-generation behind instead. So overall, there's nothing too exciting or different expected to come.

          • q
          • qool
          • 0nI
          • 31 May 2024

          Awesome. AI is the future. I want my phone to make sandwiches

            • O
            • Ok
            • CbI
            • 31 May 2024

            jiyen235, 31 May 2024I didn't say the 8 gen 3 isn't efficient, I said ... moreBut you compared it to the 8g1, this is nothing like 8g1 at all. Pushing past what the phone can sustain is just to show some big benchmark scores. This has been the case since the 888, but that and 8g1 needed those high power to overtake 865 and even 855 in the case of 8g1.

            8g3 is nothing like that.

              So, 30 May 20248g3 is much more efficient than 8g2 when doing tbe same thi... moreI didn't say the 8 gen 3 isn't efficient, I said it's an efficient chip that is pushed to inefficient levels. What I mean is, when there's a task requiring more power, the 8 Gen 3 is pushed to levels beyond diminishing returns, more power is required for small performance gains, compared to the 8 gen 2.

                ashrobb, 30 May 2024Would that performance boost put it finally ahead of Apple ... moreOn a tablet or laptop, as that peak performance probably uses up to 10W (or higher) for a single core.
                Not sustainable for a phone.

                  RaviSS, 30 May 2024I believe that the baseline in this comparison is X4 (8G3) ... more"Mediatek 9400 will be using such a high clock. 3.4 Ghz seems more sensible and probable to me. Therefore, I'm expecting the next gen flagship based on ARM design to score around 2700 points in GB6."

                  Rumors say precisely that: X925@3.4Ghz, around 2700 single.

                  That 3.8Ghz at the end of the curve should use stupidly high power for a phone, as the X4 already uses pretty much the same power as the A17 but with much lower performance (more akin to A14).

                    • A
                    • ARMin Arlert
                    • eCg
                    • 31 May 2024

                    ARM using much confusing names again for the X core lol.
                    Why not just name your cores like:
                    Cortex X5
                    Cortex A5 or P5 (I know there've been one in the past lol)
                    Cortex E5

                      that core name is no longer consistent. X4 to X925?

                        • L
                        • LEGv12-T
                        • Sju
                        • 30 May 2024

                        I don't understand why ARM still prefers to use older architecture designs. ARMv9.4-A is already out and if I'm not mistaken, the previous gen was using ARMv9.2-A as well.
                        It's been about 2 years since ARMv9.4-A was first announced.

                          • S
                          • So
                          • CbI
                          • 30 May 2024

                          jiyen235, 30 May 2024Wow, this should allow the next gen of android SoCs to be m... more8g3 is much more efficient than 8g2 when doing tbe same thing at the same speed. Watch the geekerwan chinese channel, he emulates gta 5 on 8g3 and 1600x720p 30 fps only uses 4.4w, and same is 5.8w on 8g2.

                            Well, looks like we know now what Exynos 2500 will be.

                              ProG4merPlus, 30 May 2024"36% increase compare to a 2023 premium android phone&... moreI believe that the baseline in this comparison is X4 (8G3) and not X3(8G2). 8G3 was announced in October 2023 based on architecture announced by ARM in 2023 itself, therefore it is more likely to be used as a baseline. Now the 36% increase is inclusive of the increase in clock speed (3.8 GHz for ARM reference vs 3.2 Ghz for 8G3), therefore the actual IPC calculates to 14.5%, which is similar to the 15% IPC improvement claim made by ARM.

                              Another reference point has been used is "2023 best in class @3.8 Ghz" which clearly seems to be referring to Apple A17 Pro (which is clocked at 3.78 Ghz). Now this second reference has been depicted as around 27% faster than the first reference which is indeed quite close to the 28-30% higher single-core performance of A17 Pro vis-a-vis 8 Gen 3 (2900 vs 2250 in GB6).

                              In conclusion, as per the claim of ARM, the single-core performance would be around 3050-3100 points.

                              Two caveats, however. Firstly, in the slide where 36% is represented, it says "with system and software optimisations", whatever that may be, and the 'normal' increase is depicted as 33% or so. Secondly, it is doubtful that final commercial chips such as Mediatek 9400 will be using such a high clock. 3.4 Ghz seems more sensible and probable to me. Therefore, I'm expecting the next gen flagship based on ARM design to score around 2700 points in GB6.

                                Anonymous, 30 May 2024i bet you didnt believed we will get current gpu+cpu power ... moreLMAO first deal with the chemistry problem! The energy density need to be 10x higher but we haven't researched a better material yet.

                                  • ?
                                  • Anonymous
                                  • nDy
                                  • 30 May 2024

                                  MrSTTEK, 30 May 2024at 10w TDP and around 6-7w heat dissipation capability for ... morei bet you didnt believed we will get current gpu+cpu power in smartphones 15 years ago
                                  So yes, it is doable with proper thermal materials and more efficient SoC..probably starting with 2nm

                                    • ?
                                    • Anonymous
                                    • gvN
                                    • 30 May 2024

                                    another year and ARM still refuses to redesign their efficiency core, which still is both very slow and very power hungry

                                      GS88, 30 May 2024Would be something if they make a chip that is just as fast... moreWell, that's on the chip manufacturers. It's still possible.

                                        "36% increase compare to a 2023 premium android phone" 🙂🙂🙂🙂

                                        So Compare to X3 with around 1950 Single core performance this one should gives about 2650 point

                                        I am really disappointed by X925 Performance 🥲, I thought it's performance would be something like 3000 point, I mean both Oryon and apple's new cores performance is higher

                                          1 of 2