Qualcomm silently introduces Snapdragon 6 Gen 3

02 September 2024
The chipset is just like the Snapdragon 7s Gen 2 but with underpowered efficiency cores.

Sort by:

  • ?
  • Anonymous
  • bI1
  • 01 Oct 2024

Disappointing for sure.

Switching to 4nm from 6nm bit itself should improve battery performance, I bet by 20% on average ( depending on display, drivers etc...)

At same time QC will surely give a price increase. Those 6gen3 will be for phones of around 300 or more.

Not many alernatives for mid/lower range procs.. tensoshit's been out of competition for years. MediaTek had some good ones but there's a lot of renaming scam there too.

Really disappointing.

    Anonymous, 02 Sep 2024To help OEMs to fool clients Exactly. There are so many variants of 4x A78 + 4x A55 at this point and the GPUs aren't that different nor do they even matter. Just know if you buy this you are buying 4 year old technology at a new price point. You will be better off with a used or priced reduced flagship or even one of those Pixels that people clown the performance of.

      jgblanco88, 08 Sep 2024Chill out. Anyone that knows a thing or two about this has ... moreYah am delusional, Qualcomm is delusional, Geekerwan delusional, wikichip delusional, even Tsmc themselves delusional, everyone knowledgeable is delusional but you guys who have zero data until now to back up your claims are right.

        ashrobb, 08 Sep 2024Yeah but not as big as you claim, and doesn't invalida... moreThen wait for 6 gen 3 results and test it with 778G perfect frequency 2.4Ghz.

          ashrobb, 08 Sep 2024Yeah but not as big as you claim, and doesn't invalida... moreChill out. Anyone that knows a thing or two about this has learned the lesson not to use any phones with Samsung made chipsets. Leave him alone in his delusion.

            Hemedans, 08 Sep 2024from article you linked There’s just a 7% typical perfor... moreYeah but not as big as you claim, and doesn't invalidate the korean tests I linked before. You're right we're deviating though, the point still stands that 4nm LPE wasn't competitive enough even against TSMC older nodes 👍

              • ?
              • Anonymous
              • LS$
              • 08 Sep 2024

              Hemedans, 07 Sep 2024Bin won't cause more than 20% deficit, we know sd 778G... moreAll high performance mobile products are being made by tsmc isn't it? I read even google is going tsmc for the Tensor g5 and new qualcomm midrange 7 series stuff

                ashrobb, 08 Sep 2024The actual uplift is 4% efficiency over the previous genera... morefrom article you linked

                There’s just a 7% typical performance improvement from the revised micro-architecture. However, that comes with a 4% reduction in power consumption,

                if you gain 7% while reducing power by 4% thats not 4% efficient. its more than 10, am lazy doing calculation.

                  Hemedans, 08 Sep 202420% from arm claim, also from geekerwan you can compare D81... moreThe actual uplift is 4% efficiency over the previous generation A77. For marketing purposes they claimed 20% for a 1W power budget and different fab process.

                  https://www.androidauthority.com/arm-cortex-x1-cortex-a78-1119666/

                  You're very gullible.

                    ashrobb, 08 Sep 2024But where's the 20% deficit here? The test difference ... more20% from arm claim, also from geekerwan you can compare D8100 vs 865 difference is close to 40-60% even if you remove advantage of 5nm vs 7nm still A78 has lead there, i dont even know why we discuss this, its common knowledge A78 is more efficient than anything except newer A715 and A720.

                      Hemedans, 07 Sep 2024Bin won't cause more than 20% deficit, we know sd 778G... moreBut where's the 20% deficit here? The test difference is smaller and also remember the flagship chipset has larger cache than 778G which pushes performance a bit further.

                        OpposedScroll75, 08 Sep 2024The SD865 is not superior to the SD870 💀 The SD870 is li... moreit is, use konabess to overclock you will understand, 865 can easily reach 940mhz, earlier version of 870 struggle to reach that clock and later version were blocked by Qualcomm so we cant overclock.

                        also Andrei did review these soc in spec 2017 int

                        prime core
                        -870 score 3.94 use 2.4W (1.64 point per watt)
                        -865 score 3.59 use 1.84W (1.95 point per watt)

                        so you may think 870 is 865++ but reality 870 is those soc which didnt qualify as 865 were overclocked, there is a reason why 870 devices were much cheaper than 865

                          Hemedans, 07 Sep 2024Samsung foundry is close to 100% capacity, last time I saw ... moreThe SD865 is not superior to the SD870 💀

                          The SD870 is literally the SD865++

                            ashrobb, 07 Sep 2024Yeah, since TSMC is overbooked and thus, more expensive. I ... moreBin won't cause more than 20% deficit, we know sd 778G really well it was one of the best midrange efficient wise,

                            Tsmc is booked in highend node aka 3nm they have plenty of capacity in their older node.

                            These figure are from digitimes
                            "ET News quoted sources reported that the utilization rate for Samsung's foundry business for 5nm is about 80%, and the combined utilization rate for 5nm and 7nm processes reached 90%, up from 60% in 2022."

                            That was before switch 2. Samsung foundry was 90% utilised

                            " 2023 is being dragged down by a rapid slowdown in 7nm and 6nm chip orders, but the foundry has managed to keep its overall fab utilization rates at 70% or higher, according to sources at fab toolmakers."

                            Tsmc 3nm while it's more utlised at 95% but overall Tsmc utilization is Around 70-80% while Samsung advanced node has 80% utilisation overall their fab has 90% utilization.

                              Hemedans, 07 Sep 2024Samsung foundry is close to 100% capacity, last time I saw ... moreYeah, since TSMC is overbooked and thus, more expensive. I like competition, but everyone and their mother knows TSMC does the better product and they would use them if money and capacity isn't an issue. Samsung isn't cheaper because of their good will, they're cheaper because they're worse and the demand is lower. Qualcomm can get away by rehashing a mediocre chipset fabbed with them and getting better profits for it. All their good stuff is from the best foundry. You defend the 4 Gen 2 with your life as if isn't a terrible product with a weak and outdated, more than six year old GPU architecture.

                              And the last bit is fair enough, specifically about some chips having better bins and yields than others, that also could have applied to the results I linked you before, about an A77 core from a flagship chip slightly beating an A78 from a midrange chip. You could lower your ego down a bit and consider those variables as well.

                                ashrobb, 07 Sep 2024So a whopping 5% increase from the 4 Gen 2, best case scena... moreSamsung foundry is close to 100% capacity, last time I saw it was more than 90%, they even got switch 2 deal recently, which is huge for a device which will sell millions of unit, because it can't defeat Tsmc in highend doesn't mean it's useless. Without Samsung foundry you wouldn't get budget devices with cortex A78, many sub $200 devices with A78 are manufactured by Samsung.

                                Also as i said for millionth time now in many thread sd 695 is only exception by your logic why sd 695 is more efficient than 4 gen 1 and Dimensity 900/920? Does 4 gen 1 and D920 use inferior Tsmc 6nm over superio Tsmc 6nm? No in some case some products are better binned and only those with high yield are choosen that can be case in lot of soc sd 865 is superior to 870, 695 superior to 4 gen 1, 8s gen 3 superior to 7+ gen 3 etc.

                                  not so silent now that everybody knows.

                                    Hemedans, 07 Sep 2024Take more power while provide more perfomance is that hard ... moreSo a whopping 5% increase from the 4 Gen 2, best case scenario, is worth the extra 30% consumption from an inferior node process... Got you there, no wonder Samsung fabs are doing great.

                                      ashrobb, 07 Sep 2024Geekerwan shows 4 Gen 2 takes 30% more power than 695 under... moreTake more power while provide more perfomance is that hard for you to understand? You want me to draw lines and teach you like 5 years old? And what about 4 gen 1 and D920?

                                        Hemedans, 07 Sep 2024Me and you who fail to prove anything lol? So far I gave... moreGeekerwan shows 4 Gen 2 takes 30% more power than 695 under load. Where is the lie there? How is 4nm LPE on par or better than TSMC 6nm then? Simple questions you will not be able to answer from your biased view.
                                        Have a nice day as well.

                                          1 of 6