Mediatek Helio X20 benchmarks revealed

You need to slow down and take a bit more time before just reacting.

The Geekbench difference in integer and FP is not assumed, you can go and search for it, i was specific , you failing to notice that, is on you. I haven’t mentioned the interconnect or memory controller as potential factors that impact that result since i can’t quantify their impact yet (no data) and that would be too in depth here anyway.

Power consumption, depends how you count it, max, or same task over an equal time interval? A75% claim is very bombastic and very misleading and you know it.

You seem to have a special relationship with DSPs and that’s fine, hell in a perfect world i would love to see lots of specialized cores in SoCs but cost makes that problematic. Marketing that is rather problematic too. Against what you compare DSP power needs to be well defined and you don’t do that. Ideally,and i’m sure we can all agree on that, one should use the best compute unit for the task no matter what compute unit it is. Some tasks will do better on CPU, some on DSP, some on GPU and so on. Overutilizing some compute unit is conterproductive. DSP perf is not getting much testing and marketing dollars when it comes to consumers and that should be fixed but mobile benchmarking is beyond terrible for now .And no idea why you seem to assume that Mediatek hates specialized accelerators , that’s actually the path they are following too, many specialized heterogeneous cores.

The overall power claim despite the extra cluster and process advantage is a bold move and would assume a scenario where somehow everything runs on some magical DSP. Users do browsing, IM , social , some video and audio playback, some gaming. It is true that we got tool little info on GPU to judge gaming power consumption. For LTE power consumption Mediatek got rather good at it but lets not assume who is winning instead of waiting for data.

SD652 vs SD650 for me depends on price.If SD650 is 20% cheaper or more ,would say it’s the better deal for most users. As for X20, if it’s worth 50% more or nothing at all, depends on what a user values. I would have really liked to see UFS support for the X20 since NAND speeds have been getting better but i still see them as insufficient.

This doesn’t mean that i don’t like SD65x , Qualcomm got outfoxed so many times in the last years that it’s nice to see them outfoxing the others this time. The SD650 in 150$ phones is awesome and it does put the MTK P10 in an awkward position.

lol you are going in fanboy mode.

Already asked before but it didn’t register, 75% based on what? On same task? Obviously not. In peak power,maybe since they choose to go 2 cores instead of 4 to be able to reach those clocks and perf. What’s the impact in actual usage, well you don’t want to talk about that since you really like your 75%. What’s the power for the same task , you don’t want to talk about that either. Max power is a TDP limitation ,none of these can put 100% load on all cores in a phone TDP for long and when it comes down to it MTK has far higher efficiency, if you are gonna talk about max load numbers.

But hey lets just bark 75% instead, because we don’t think we just react.

The perf of the core doesn’t depend only on the core. The memory subsystem ,the interconnect ,plenty of other things and in this case the newer core revision impact the perf and you know that very well but for w/e reason you just decided to flat out lie.And again the Geekbench results in integer and FP are not estimates, just rounded numbers based on actual and available results, a verifiable fact.

The SD65x uses V56 not the newer 680, that one is in SD820 and the problem was that your were making it sound like a DSP is some kind of general purpose computing unit when it is just a DSP., it’s not gonna magically replace the CPU.

As for the future, depends how far ,what form factor and a lot more. The future are glasses and less , close to 0 volume available, 0 power and heat. How things get there fast, too early to tell. For now an OS that uses a neuromorphic computing unit in at least 10 good ways, would be really nice.

On the LTE side, depends how you define it, the conversation here was about power, having cat 12/13 in a cat 6 world isn’t a real advantage. And power wise, MTK is pretty good now ,would wait for actual numbers before declaring a winner. And ofc don’t forget that this is SD65x on 28nnm not SD820. Not that LTE matters that much to users since the bulk of the data uses wifi.

820 will win over x20 in most metrics and will cost more too. On the NAND side, SD820 supports UFS 2.0 and that’s a big plus, assuming phone makers take advantage of that.

We’ll see what X30 is and maybe they have more than just x30. I do hope wifi ad gains traction even in the lower end,not just devices with SD820.

SD820 does support UFS 2.0 and phones with it will tend to have 2×2 wifi , those would be significant advantages in real usage. Would be nice if we see more than 1 stream in X20 phones too but wifi isn’t getting any marketing dollars so phone makers don’t invest much there.

For Geekbench i do wonder if SD820 runs all the tests on the CPU but no real clue if anything iffy is going on. It does have a crazy memory score and if that’s for real, it should impact real usage.

On the GPU side , we’ll see but i don’t expect it to throttle so badly that it would lose to MP4 at 700MHz or even 900MHz. I also wonder if they spent more on the GPU size this time around and went with more conservative clocks. I do expect SD820 to be well ahead in GPU even in sustained perf.

X20 wasn’t really aimed at SD820, it’s on 20nm, small enough GPU, just DDR3 so it should have a significant cost advantage over SD820 and fill a different segment.