Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Qualcomm announces ARMv9 based successor to Snapdragon 700 series, the Snapdragon 7 Gen 1

 Share

Summary

 

Qualcomm has announced the new Snapdragon 7 Gen 1, its replacement for the upper midrange Snapdragon 700 series. This new 700 series successor is based on ARM's new ARMv9 CPU cores with 4x A710 cores (1x at 2.4GHz and 3x at 2.36GHz) and 4x A510 cores at 1.8GHz . The choice of CPU cores is interesting considering the Snapdragon 8 Gen 1 uses a single Cortex X2 core instead of the 2.4GHz A710 core.  This would mean peak single core perf would be much better on the 8 Gen 1 or 8 Plus Gen 1 than this new 7 Gen 1.

 

Compared to its predecessor the Snapdragon 780G has much faster 5G download speeds, much faster RAM speed, and a newer Samsung 4nm node (presumably Samsung 4LPE).

 

Edit: It should be noted that since this is using Cortex A710 cores this SOC will still retain the ability to run Aarch32 apps even if ARM's other ARMv9 CPU core designs don't support it.

 

Quotes

Quote

Within Qualcomm’s pantheon of smartphone SoCs, the Snapdragon 7 Gen 1 is the successor to last year’s Snapdragon 780G/778G parts. For the moment, Qualcomm is sticking with a single SKU, though last year the 780G and 778G were a few months apart – so Qualcomm may bifurcate things farther down the line.

 

Quote

In terms of organization, the new 7 Gen 1 is very similar to the 780G, implementing a 1+3+4 core configuration. Qualcomm is only using two core types here – the high-performance Cortex-A710 and high-efficiency Cortex-A510 – so rather than the prime core being a unique core type (ala the Cortex-X2 on the 8G1), it’s another Cortex-A710 here. Qualcomm is clocking this core at 2.4GHz, meanwhile the remaining 3 A710s are clocked a bit lower at 2.36GHz. Finally, the A510s come clocked at 1.8GHz, the same as on the 8G1.

The shift in CPU cores also brings an update to Qualcomm’s memory controller. The 7 Gen 1 adds LPDDR5 capabilities, supporting up to the same LPDDR5-6400 data rates as the 8 Gen 1. So with a 2x16-bit channel memory bus, the 7 Gen 1 can access up to 25.6GB/second of memory bandwidth, a 50% improvement over the LPDDR4X-capable 780G.

Unusual even for Qualcomm, the company isn’t talking about CPU performance with the new part. So absent benchmarking data, there’s little we can say here about performance expectations. The new Arm cores should allow for better performance, but there’s a lot of wiggle-room due to implementation details.

 

My thoughts

This is an interesting chip tho I really don't understand why Qualcomm didn't just make all the A710 cores in this chip run at 2.4GHz or, if that was unattainable on Samsung's node,  2.36GHz. I imagine the Snapdragon 6 Gen 1 will be basically this type of chip with 2 of the A710 cores disabled and a weaker GPU.

 

This chip seems cool and ngl I really like the idea of Quad core Cortex A7x series high performance big boi cores being in upper midrange class smartphones. Leaving dual core cortex A7x series big boi cores for lower midrange and budget phones cos ARM seemingly has a thing for making their low power A5x parts slow and nasty and this has traditionally dominated the low end budget phones with quad core or 4 + 4 configurations of low power cores clocked at high clockspeeds in one cluster and a low clockspeed in the other cluster. This leads to horrifically bad performance and is a big joke but ARM doesn't seem to want to fix this.

 

Of course as we saw with the Snapdragon 8 Gen 1, ARM is still worse at designing high performance cores (A7x series) and even maxinimum performance cores (X series) than Apple are so there's no doubt even a 3 to 4 year old Apple SOC will still trash this SOC in single core performance. As much as I dislike Apple, this is the one area they seem to be completely destroying people in, ARM chip design and not making bad CPU cores.

 

Sources

 https://www.anandtech.com/show/17397/qualcomm-announces-snapdragon-7-gen-1

Judge a product on its own merits AND the company that made it.

How to setup MSI Afterburner OSD | How to make your AMD Radeon GPU more efficient with Radeon Chill

Oneplus 7T (Mid 2021 to present), HP Envy 15" x360 R7 5700U (Mid 2021 to present)

Samaritan XTXH (Early 2021 Upgrade - AMD Ryzen 9 3900XT (12C/24T)  (2021) , MSI X370 Gaming Pro Carbon, Corsair 32GB Vengeance LPX DDR4-2666 (2020) ,  AMD Radeon RX 6700XT 12GB (Mid 2021), Corsair RM850i PSU, Noctua NH-D15 CPU Cooler (2021), Samsung 860 EVO 500GB SSD, Seagate BarraCuda 6TB HDD (2020) , NZXT S340 Elite, Corsair ML 120 Pro, Corsair ML120 x2 (2021)

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, AluminiumTech said:

 

Of course as we saw with the Snapdragon 8 Gen 1, ARM is still worse at designing high performance cores (A7x series) and even maxinimum performance cores (X series) cores worse than Apple are so there's no doubt even a 3 to 4 year old Apple SOC will still trash this SOC in single core performance. As much as I dislike Apple, this is the one area they seem to be completely destroying people in, ARM chip design and not making bad CPU cores.

 

Sources

 https://www.anandtech.com/show/17397/qualcomm-announces-snapdragon-7-gen-1

This is a large part of the reason I went iPhone in 2020, as I tend to hold onto my devices for a long time. Even now, my iPhone 11 still feels extremely fast. That, and not needing to use custom firmware to get software support beyond the 2-3 year mark. Though, I'll be needing a new battery pretty soon. Apple simply isn't afraid to expend a lot of die space, while with ARM, larger cores probably aren't particularly desirable for their licensees.

 

Still, always glad to see performance march forward, if not as fast as I'd like. If the Android space can achieve CPU performance and efficiency parity with Apple, and figure out a way to maintain consistent long-term support, I'll certainly be open to moving back to Android (giving my my cake, and letting me eat it too).

My eyes see the past…

My camera lens sees the present…

Link to comment
Share on other sites

Link to post
Share on other sites

20 minutes ago, Zodiark1593 said:

This is a large part of the reason I went iPhone in 2020, as I tend to hold onto my devices for a long time. Even now, my iPhone 11 still feels extremely fast. That, and not needing to use custom firmware to get software support beyond the 2-3 year mark. Though, I'll be needing a new battery pretty soon. Apple simply isn't afraid to expend a lot of die space, while with ARM, larger cores probably aren't particularly desirable for their licensees.

Apple's smartphone dies aren't even that big tbh (they're mostly sub 100m2 tho a few are more like 110 to 120mm2). It's mostly that Apple makes their high perf cores as fast as they can and their efficiency cores are designed more like ARM's older high perf cores but they're super efficient. ARM has a very different philosophy on how low power cores should be designed (Apple does Out of order execution, ARM does in order execution) that means low power ARM cortex reference core designs will be very slow for a very long time.

 

Also it's worth noting that a huge chunk of the ARMv9 architecture spec came from Apple. Apple implemented almost all (if not all) of ARMv9 in their latest implementations of ARMv8 CPU core designs. Most chip designers using ARM haven't been using updated ARMv8 architecture designs though (including ARM themselves for their reference cores) so Apple had to put them in their place and start fresh on a new ARM arch hence ARMv9.

 

At this point it feels like ARM is asleep at the wheel and Apple is the only thing stopping it from crashing and burning.

 

20 minutes ago, Zodiark1593 said:

Still, always glad to see performance march forward, if not as fast as I'd like. If the Android space can achieve CPU performance and efficiency parity with Apple, and figure out a way to maintain consistent long-term support, I'll certainly be open to moving back to Android (giving my my cake, and letting me eat it too).

Tbh prob not gonna happen until Qualcomm's new custom ARM core designs from their Nuvia purchase are announced in 2023. Even then, 2023 is optimisting for Qualcomm matching or beating Apple. It would take a miracle for Qualcomm to match Apple, let alone beat them.

Judge a product on its own merits AND the company that made it.

How to setup MSI Afterburner OSD | How to make your AMD Radeon GPU more efficient with Radeon Chill

Oneplus 7T (Mid 2021 to present), HP Envy 15" x360 R7 5700U (Mid 2021 to present)

Samaritan XTXH (Early 2021 Upgrade - AMD Ryzen 9 3900XT (12C/24T)  (2021) , MSI X370 Gaming Pro Carbon, Corsair 32GB Vengeance LPX DDR4-2666 (2020) ,  AMD Radeon RX 6700XT 12GB (Mid 2021), Corsair RM850i PSU, Noctua NH-D15 CPU Cooler (2021), Samsung 860 EVO 500GB SSD, Seagate BarraCuda 6TB HDD (2020) , NZXT S340 Elite, Corsair ML 120 Pro, Corsair ML120 x2 (2021)

Link to comment
Share on other sites

Link to post
Share on other sites

I have a Snapdragon 888 in my current phone and I hate it.

It gets so hot that its sustained performance is worse than my 4 year old snapdragon 835.

If they can't keep the SOC cool stop making it faster.

 

I might go out of the way to get a mid ranger (Snapdragon 7 Gen 1) class chip next time around if this keeps happening.

 

Do we know if the Snapdragon 7 Gen 1 is TSMC or Samsung? cos I'm aware Samsung is half the problem:

 

"Qualcomm Snapdragon 8+ Gen 1 unveiled: 30% more efficient, 10% faster" (GSMArena)

Link to comment
Share on other sites

Link to post
Share on other sites

24 minutes ago, TrigrH said:

I have a Snapdragon 888 in my current phone and I hate it.

It gets so hot that its sustained performance is worse than my 4 year old snapdragon 835.

If they can't keep the SOC cool stop making it faster.

 

I might go out of the way to get a mid ranger (Snapdragon 7 Gen 1) class chip next time around if this keeps happening.

Guessing your SD888 is really using that Cortex X1 core which only focuses on perf at the expense of power (This is by design), I'm not sure it's meant to ignore heat tho it probably doesn't have a heatspreader so it probably can't do much about heat besides throttle ghe X1 core or use the A78 cluster or the A55 cluster.

 

If it was using the A78 cores it would probably be cool enough to not be a toaster.

24 minutes ago, TrigrH said:

Do we know if the Snapdragon 7 Gen 1 is TSMC or Samsung? cos I'm aware Samsung is half the problem:

Samsung. Same as the 8 Gen 1 and most of Qualcomm's chips now.

 

The 8+ Gen 1 seems to be the exception.

24 minutes ago, TrigrH said:

 

Judge a product on its own merits AND the company that made it.

How to setup MSI Afterburner OSD | How to make your AMD Radeon GPU more efficient with Radeon Chill

Oneplus 7T (Mid 2021 to present), HP Envy 15" x360 R7 5700U (Mid 2021 to present)

Samaritan XTXH (Early 2021 Upgrade - AMD Ryzen 9 3900XT (12C/24T)  (2021) , MSI X370 Gaming Pro Carbon, Corsair 32GB Vengeance LPX DDR4-2666 (2020) ,  AMD Radeon RX 6700XT 12GB (Mid 2021), Corsair RM850i PSU, Noctua NH-D15 CPU Cooler (2021), Samsung 860 EVO 500GB SSD, Seagate BarraCuda 6TB HDD (2020) , NZXT S340 Elite, Corsair ML 120 Pro, Corsair ML120 x2 (2021)

Link to comment
Share on other sites

Link to post
Share on other sites

6 hours ago, AluminiumTech said:

Also it's worth noting that a huge chunk of the ARMv9 architecture spec came from Apple. Apple implemented almost all (if not all) of ARMv9 in their latest implementations of ARMv8 CPU core designs. Most chip designers using ARM haven't been using updated ARMv8 architecture designs though (including ARM themselves for their reference cores) so Apple had to put them in their place and start fresh on a new ARM arch hence ARMv9.

Are you sure about that? I know that Apple was very proactive in implementing ARMv8 extensions like memory tagging but I'm fairly certain ARMv9 was simply arm (the company) releasing a unified ISA for the next 10 years.

 

In fact, I believe arm explicitly disallow adding custom instructions on an arm CPU which is why Apple had to implement their custom extensions separately on the M1.

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, kvuj said:

Are you sure about that? I know that Apple was very proactive in implementing ARMv8 extensions like memory tagging but I'm fairly certain ARMv9 was simply arm (the company) releasing a unified ISA for the next 10 years.

Put together by ARM with pressure and backing from other companies including Apple and Mediatek.

1 hour ago, kvuj said:

 

In fact, I believe arm explicitly disallow adding custom instructions on an arm CPU

Not to my knowledge.

1 hour ago, kvuj said:

which is why Apple had to implement their custom extensions separately on the M1.

 

Judge a product on its own merits AND the company that made it.

How to setup MSI Afterburner OSD | How to make your AMD Radeon GPU more efficient with Radeon Chill

Oneplus 7T (Mid 2021 to present), HP Envy 15" x360 R7 5700U (Mid 2021 to present)

Samaritan XTXH (Early 2021 Upgrade - AMD Ryzen 9 3900XT (12C/24T)  (2021) , MSI X370 Gaming Pro Carbon, Corsair 32GB Vengeance LPX DDR4-2666 (2020) ,  AMD Radeon RX 6700XT 12GB (Mid 2021), Corsair RM850i PSU, Noctua NH-D15 CPU Cooler (2021), Samsung 860 EVO 500GB SSD, Seagate BarraCuda 6TB HDD (2020) , NZXT S340 Elite, Corsair ML 120 Pro, Corsair ML120 x2 (2021)

Link to comment
Share on other sites

Link to post
Share on other sites

6 hours ago, AluminiumTech said:

Guessing your SD888 is really using that Cortex X1 core which only focuses on perf at the expense of power (This is by design), I'm not sure it's meant to ignore heat tho it probably doesn't have a heatspreader so it probably can't do much about heat besides throttle ghe X1 core or use the A78 cluster or the A55 cluster.

 

 

 

Looking at the architecture details for X1, it looked like ARM's effort to throw the kitchen sink, with Apple square in it's sights. My guess is poor implementation by the OEM. Only other thing I could guess is that the X1 is kept awake to quickly take on workloads without pause, however, I don't think I buy the ramp-up time being that slow.

 

https://www.anandtech.com/show/16463/snapdragon-888-vs-exynos-2100-galaxy-s21-ultra/5

 

https://www.anandtech.com/show/14892/the-apple-iphone-11-pro-and-max-review/5

 

 

My eyes see the past…

My camera lens sees the present…

Link to comment
Share on other sites

Link to post
Share on other sites

15 hours ago, TrigrH said:

Do we know if the Snapdragon 7 Gen 1 is TSMC or Samsung? cos I'm aware Samsung is half the problem:

The Snapdragon 7 gen 1 is manufactured by Samsung.

 

Samsung's node is worse than TSMC, but I think people overestimate how much of an impact it has. The tests I have seen indicates that Samsung's 4nm node is about 10-15% less efficient than TSMC's.

That might sound like a lot, but we have to remember that the difference is only that big near 100% load, and there are plenty of other components in the phone that uses power, other than the SoC.

So in practical terms changing from Samsung to TSMC might result in your phone having 15% battery left at the end of the day, rather than 10%. It's an improvement, but I think most of the blame for the Snapdragon 8 Gen 1 being a power hungry furnace that can barely compete with the older Snapdragon chips is on Arm and Qualcomm.

The architectures (X2, A710 and A510) are simply not that good, especially when compared to Apple's cores. On top of that, Qualcomm cheaped out and went with lower cache than recommended by Arm, and the A510 cluster also shares resources when they could have given each core its own resources (like MediaTek did).

Link to comment
Share on other sites

Link to post
Share on other sites

18 hours ago, AluminiumTech said:

Put together by ARM with pressure and backing from other companies including Apple and Mediatek.

Not to my knowledge.

 

https://medium.com/swlh/apples-m1-secret-coprocessor-6599492fc1e1

 

Sorry, it seems I misremembered.

Quote

Developer Dougall Johnson has through reverse engineering, uncovered a secret powerful coprocessor dubbed AMX: Apple Matrix coprocessor inside the M1 chip.

 

(...)

 

ARM Ltd. creators of the ARM instruction-set architecture (ISA) has long resisted adding custom instructions to their ISA. This is one of the advantages of RISC-V

 

However due to pressure from customers ARM relented and announced in 2019 that they would allow extensions. EE Times reports:

 

The new instructions are interleaved with standard Arm instructions. To avoid software fragmentation and maintain a coherent software development environment, Arm expects customers to use the custom instructions mostly in called library functions.

 

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share


×