Exactly!..what's the point? Would rather have 6 A55s (in two different power domains) and dual core M3...I'm guessing the performance per watt on the big cores is not as good as the Monsoon cores. Given that the SOCs are power limited in mutl core mode, the big cores have to be clocked way down compared to the 2.9ghz max. On the A11 the big cores can run closer to their peak performance.
If this is the case, it makes you wonder why they even bothered with 4 big cores. Perhaps it can stretch its legs more in a laptop form factor.
That would give less MT performance. An M3 core at full perf is over 4x faster than one A55. It's better to go low clock quad M3. Also the Zephyr/Mistral cores were like 2x the size of an A53 and probably the same perf advantage, no surprise Apple has better MT.Exactly!..what's the point? Would rather have 6 A55s (in two different power domains) and dual core M3...
So Apple has cores similar in performance/area to the Cortex A73 as their little cores xD
Can't wait to see more about Samsung's new M3 cores performance and power consumption.
Someone knows what is the state of UFS 3.0 and LPDDR5 for smartphones? @Andrei.
Yes, both have listed support only up to UFS 2.1, but with Samsumg 8nm in the middle of the year they will probably launch another processor this year.SD845 uses UFS 2.1, like SD835. I suppose it's the same with the Exynos 9810, and that we'll probably not see UFS 3.0 and its speeds before 2019.
Yes, both have listed support only up to UFS 2.1, but with Samsumg 8nm in the middle of the year they will probably launch another processor this year.
Exynos 9810:
Single-core: 3850
Multi-core: 8870
Snapdragon 845:
Single-core: 2400
Multi-core: 7870
As far as i remember, A75 is +25% in Geekbench at the same clock frequency, while the Exynos has the frequency uplift already priced in.
Do you? Qualcomm very specifically say on their site the Snapdragon 845 is clocked at 2.8 GHz (which most likely is lower in real life, just as with the Exynos 9810), and that it overall brings "up to 25% performance uplift" over its predecessor. They (and Samsung) have always been accurate about the performance improvements their products bring, as they're not really advertised towards consumers.
AI needs and wants dedicated IP not faster generalist CPUsIt seems to me the a75 is stuck between slim a73 that is perfect for midrange on cheaper 14nm and 9180 that outright is much faster. Arm needs a solution here. Either they follow apple and samsung or they dont and focus on smaller cores. In the long run doing the last is a risky path due to AI whatnot and they would slowly disappear.
But M3 cores are much larger than mistral...and there are 4 of them...That would give less MT performance. An M3 core at full perf is over 4x faster than one A55. It's better to go low clock quad M3. Also the Zephyr/Mistral cores were like 2x the size of an A53 and probably the same perf advantage, no surprise Apple has better MT.
The 9180 makes segmenting into midrange and lowend much better.
It seems to me the a75 is stuck between slim a73 that is perfect for midrange on cheaper 14nm and 9180 that outright is much faster. Arm needs a solution here. Either they follow apple and samsung or they dont and focus on smaller cores. In the long run doing the last is a risky path due to AI whatnot and they would slowly disappear.
I doubt it. Samsung has been using the same SoC for Galaxy S and Note for several generations now. Also, the Exynos 9810 bring such huge improvements in CPU performance alone that I doubt they really feel the need to release another, specific mid-year SoC for the Note 9 later in the summer. We probably won't see 7nm SoCs before next year (excluding the A12).
All I'm hoping for right now is that Google actually uses the Exynos 9810 in their Pixel 3 later this year, after the recent announcement about Samsung starting to sell Exynos SoCs to other OEMs. Would be pretty fucking stupid to wait till end of 2018, only for the Pixel 3 to have an SoC (SD845) that is so vastly inferior to the Exynos 9810 -- especially when you're just months away before next-gen SoCs in 2019 are released at that point. Knowing Google, however, they'll probably do that. They have a tendency to put lackluster hardware in their products (like their recent LCD OLED displays) and ask a premium price for it.
But M3 cores are much larger than mistral...and there are 4 of them...
The increase in performance of A55..the L3 cache? The likely new memory controller and fabric, the new 10nm LPP (+15%?) - Plus the twice as wide M3 that can be down clocked for high efficiency...combined I expected more than 40% MT increase over 8895.
Could it be Apple has more available bandwidth?..doesn't make sense to me otherwise.
They have used a new SOC or a new revision in 4 out of 7 years for their Note line.
I tend to disagree. Looking at Snapdragon 845 in particular, it is most likely roughly on par on the multicore side, better on the GPU side
Coming back to ARM, they will introduce the successor to A75 soon, which again will increase IPC 20-25% over A75. S.
Personally i would value a better GPU performance higher, than better single core performance.
Of course M3 cores won't use anywhere near max frequency, but that's besides the point.
The 4 large cores probably don't all have the same (high clock speed) at any time -- I would be guessing only onr ot two of them do.
That's a pretty vague way to make your point. Instead of looking 7 generations independently of each other, look at the past three years (which is clearly the most important cases, due to them being closer to the present). The past 3 years, the Note devices have used the same SoC as the S ones. Note 8 and S8, Note 7 and S7, Note 5 and S6...
The Mali-G71 (E8895) is either as good or better than the Adreno 540 (SD835). From the numbers we have from both companies, the Adreno 630 in the SD845 will perform 10% better than the Mali-G72 in the E9810. On the other hand, the E8995 have somewhat better multithreaded perf than SD835, and the 40% perf increase of the E9810 is clearly higher than Qualcomm's vague "25%" statement. So how can you say it's similiar for one, where's there's only a 10% gap, and say the performance is similiar for the other, where there's a gap of %?
Except they won't. This is pure assumption from your side. The A73 brought hardly any performance improvements to the table, over the A72. What is to say that their next architecture, which for all we know can come in 2020, won't do the same?
Yes, 10% higher GPU performance is clearly better than 70% higher singlethreadead performance...
Except they won't. This is pure assumption from your side. The A73 brought hardly any performance improvements to the table, over the A72. What is to say that their next architecture, which for all we know can come in 2020, won't do the same?
Yes, 10% higher GPU performance is clearly better than 70% higher singlethreadead performance...
Since you are so affirmative by claiming ARM won't "introduce the successor to A75 soon, which again will increase IPC 20-25% over A75", can you tell what your sources are?Except they won't. This is pure assumption from your side. The A73 brought hardly any performance improvements to the table, over the A72. What is to say that their next architecture, which for all we know can come in 2020, won't do the same?