Qualcomm Snapdragon 821 is a clock-bumped chip for fall’s flagships

Speed bump is exactly as impressive as the model number conveys.

Read the whole story

 

[fuzzyfuzzyfungus]

Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

 

[Demento]

Yeah, my exact thought was "We're seeing the new Nexus". The past couple had to use the same SoCs as the spring releases before them. The original Nexus 5 got the boost of using the 800 when most of that year's phones were on the Snapdragon 600, so we can have high hopes for this one indeed.

 

[pqr]

Speed bump is exactly as impressive as the model number conveys.

Careful for some of us who had chance to Paradroid 821's pretty powerful number actually might even be best possible (killer dual laser)

 

[VertexMaster]

fuzzyfuzzyfungus[/url]":36fn88rj]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

OnePlus 3 has very little thermal throttling with the Snapdragon 820. Less than almost any other smartphone.

Anandtech did some efficiency measures between the Exynos version of the S7 vs the Q820 and found they were similar with a modest CPU efficiency edge to the Exynos. Qualcomm has the more efficient modem.

 

[mpat]

haar[/url]":dcnu0gkz]...max speed of 2.4Ghz. (it is a lie!, lol)

Great... screwup the wi-fi band while you are at it...
(or encase the phone in shielding so the phone weighs a "ton"...)
good luck it getting an FCC cert.
(Surprise... when being tested, it will never hit 2.4Ghz, more like 2.2Ghz.)
Wonder why the ipad pro a9x is limited to 2.2 Ghz? ,
also why intel/amd processor are never rated for 2.4Ghz (always above or below)
it is because of the RFI problem...

Ehm...

http://ark.intel.com/products/76529/Int ... e-2_41-GHz
http://ark.intel.com/products/53487/Int ... e-2_40-GHz
http://ark.intel.com/products/88190/Int ... o-3_00-GHz

And these are just a few which have that as the base frequency, many more will pass through that clockspeed while turboing up. Further more, just about every frequency band around normal clockspeeds is in use now - that 2.15 GHz in the old chip is in the middle of the regular 3G band, IIRC. No, the clockspeed of a small SoC does not measurably interfere with radio communications. We'd be in trouble if it did.

 

[ikjadoon]

What is the voltage increase? Like, is this just simple binning that took Qualcomm a weekend (a la i5-6600K / i7-6700k)?

Sounds like a slight hit for battery life.

 

[laststop311]

fuzzyfuzzyfungus[/url]":cdee1juc]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

 

[laststop311]

ikjadoon[/url]":1vtyvg57]What is the voltage increase? Like, is this just simple binning that took Qualcomm a weekend (a la i5-6600K / i7-6700k)?

Sounds like a slight hit for battery life.

The tiny extra energy that it may use is more than made up by getting back into idle mode. The extra time in idle mode more than makes up for the minute amount of energy used clocking higher.

 

[earthlingkc]

Yeah, this doesn't sound like smaller fabrication, so likely more heat and more power consumption. But if it is smaller fab, then a fist bump to Qualcomm.

 

[Voyna i Mor]

Back in the ancient days people awaited Mac speed increases because most of the time they were doing things like rendering at full power, and a 7% increase in speed meant you could go home twenty minutes earlier, or set up a bigger overnight batch job. On one occasion we were able to save a major software rewrite simply because a 10% speed bump enabled a program to execute in the time available. The cost of the CPU versus the estimated $120000 to recode was insignificant.

Now...it would be interesting to know how many people, in real world use, would actually notice the speed difference. Heat yes. Heat is really important.

 

[AxMi-24]

earthlingkc[/url]":lis0avym]Yeah, this doesn't sound like smaller fabrication, so likely more heat and more power consumption. But if it is smaller fab, then a fist bump to Qualcomm.

Could still be better binning and/or some small optimisations in layout.

 

[mpat]

earthlingkc[/url]":7a702vq2]Yeah, this doesn't sound like smaller fabrication, so likely more heat and more power consumption. But if it is smaller fab, then a fist bump to Qualcomm.

It is the same 14nm Samsung (LPP) process, but it has matured enough to reach a slightly higher clockspeed.

There's no word on whether the 821's two slower cores in the CPU will be faster than the 1.6GHz in the current chip.

They're apparently up to 2 GHz now.

http://www.anandtech.com/show/10482/qua ... 4-ghz-kryo

 

[Kerome]

Will be interesting to see how it compares to Apple's updated chips for the iPhone 7 in September. 10% speed bump is not a lot compared to Apple's consistent gains of 60-70% year on year for the last few years.

 

[theoilman]

Isn't clock speed just determined by firmware/software? I didn't think you needed new hardware just for clock speed, or is there some kind of optimization for higher clock speed in the silicon here?

 

[some_troll]

This sounds like a good opportunity to get a discounted previous gen model with minimal performance hit. Now if only my trusted Nexus 5 wasn't slowly dying one sensor after the other...

 

[Edward351]

theoilman[/url]":k2wet752]Isn't clock speed just determined by firmware/software? I didn't think you needed new hardware just for clock speed, or is there some kind of optimization for higher clock speed in the silicon here?

Clock speed is related to heat output and efficiency. You usually increase the clock speed by increasing the voltage and power in a chip. But there is a hard limit when the voltage is too high and the power leaks causing the chip to fail. Separate from that, higher clock speed means more work, more work needs more power, more power means more heat.

While the earlier chips could potentially hit the speeds the newer ones could, they couldn't do it in the efficiency and power envelopes the newer ones can. These are the same reasons you see liquid or Freon cooled PCs that hit 5GHz and higher. you need dramatic cooling to overcome the power and heat requirements to get the chips to hit those high frequencies.

 

[theoilman]

Edward351[/url]":16a35d6x]

theoilman[/url]":16a35d6x]Isn't clock speed just determined by firmware/software? I didn't think you needed new hardware just for clock speed, or is there some kind of optimization for higher clock speed in the silicon here?

Clock speed is related to heat output and efficiency. You usually increase the clock speed by increasing the voltage and power in a chip. But there is a hard limit when the voltage is too high and the power leaks causing the chip to fail. Separate from that, higher clock speed means more work, more work needs more power, more power means more heat.

While the earlier chips could potentially hit the speeds the newer ones could, they couldn't do it in the efficiency and power envelopes the newer ones can. These are the same reasons you see liquid or Freon cooled PCs that hit 5GHz and higher. you need dramatic cooling to overcome the power and heat requirements to get the chips to hit those high frequencies.

Gotcha. So we need to put freon in our phones

 

[daggar]

theoilman[/url]":15ug8iu0]

Edward351[/url]":15ug8iu0]

theoilman[/url]":15ug8iu0]Isn't clock speed just determined by firmware/software? I didn't think you needed new hardware just for clock speed, or is there some kind of optimization for higher clock speed in the silicon here?

Clock speed is related to heat output and efficiency. You usually increase the clock speed by increasing the voltage and power in a chip. But there is a hard limit when the voltage is too high and the power leaks causing the chip to fail. Separate from that, higher clock speed means more work, more work needs more power, more power means more heat.

While the earlier chips could potentially hit the speeds the newer ones could, they couldn't do it in the efficiency and power envelopes the newer ones can. These are the same reasons you see liquid or Freon cooled PCs that hit 5GHz and higher. you need dramatic cooling to overcome the power and heat requirements to get the chips to hit those high frequencies.

Gotcha. So we need to put freon in our phones

I am still waiting for some maker of phones with metal bodies to do the obvious thing and corrugate the case a bit to improve cooling. It could be marketed as both a style and functional thing. Give it some name like "radiant styling" or such.

 

[ssiu]

laststop311[/url]":3ix1kzk9]

fuzzyfuzzyfungus[/url]":3ix1kzk9]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

The mobile VR stuff needs high sustained performance. For example, Google said the Nexus 6P can be used for Daydream VR development, but "Caution: The 6P's thermal performance is not representative of the consumer Daydream-ready devices that will be launching later this year. In particular, expect the 6P to thermally throttle CPU and GPU performance after a short period of use, depending on workload."

 

[theoilman]

ssiu[/url]":6373azf0]

laststop311[/url]":6373azf0]

fuzzyfuzzyfungus[/url]":6373azf0]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

The mobile VR stuff needs high sustained performance. For example, Google said the Nexus 6P can be used for Daydream VR development, but "Caution: The 6P's thermal performance is not representative of the consumer Daydream-ready devices that will be launching later this year. In particular, expect the 6P to thermally throttle CPU and GPU performance after a short period of use, depending on workload."

I think at this juncture it's silly to think using phones for vr will be anything more than a novelty.

 

[maxwell]

laststop311[/url]":3nz5crz2]

fuzzyfuzzyfungus[/url]":3nz5crz2]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

Empires and Allies battles are 3 minutes long, and it's annoying when my phone overheats 90 seconds in and starts stuttering. Other RTS-y type game would be the same.

 

[Sandwichman]

I think we're seeing the new norm. Hamstring the full ability of the processor so you can give it a slight clock boost to give the illusion of a major upgrade. First flagship phones on a three year redesign cycle, now this.

 

[shadedmagus]

Has anyone looked at whether phone cases negate the cooling improvements the new chips and shell designs bring to the table? I suspect they do, but I'd hate to carry the phone around without one - I tend to drop them often, and my cases having a raised bezel have saved the screen more times than I care to recount.

 

[Voyna i Mor]

shadedmagus[/url]":26pg5cwe]Has anyone looked at whether phone cases negate the cooling improvements the new chips and shell designs bring to the table? I suspect they do, but I'd hate to carry the phone around without one - I tend to drop them often, and my cases having a raised bezel have saved the screen more times than I care to recount.

They do raise the phone temperature, and as this tends to shorten both battery and semiconductor life I try not to run phones in cases. But YMMV.

 

[ethd]

This is extremely pedantic, and I'm sorry in advance. But "821" is not the model number, it's the brand. The model number for a Qualcomm SoC starts with "MSM" and is followed by a four-digit number. It's similar to how the model number for something like a Core i5 is not just i5, it's like i5-3230M (in the case of my aging laptop).
EDIT: If I'm going to be a pedant, might as well be pedantic with myself. Processor -> SoC.

 

[AlexisR200X]

The snapdragon 821 @2.4 Ghz is kinda disappointing but not surprising, its hard to keep bumping the speed up and It will take some major improvements in other areas to make a flagship that actually stands out from the existing ones from last year. Most users might not be able to tell the difference in CPU speed and those that do, will likely see the thermal throttling start a wee bit sooner given that the heat is still a concern in smartphones. (Again barely perceptible due to how the minor difference is from the 820)

 

[woopla]

theoilman[/url]":375570hx]Isn't clock speed just determined by firmware/software? I didn't think you needed new hardware just for clock speed, or is there some kind of optimization for higher clock speed in the silicon here?

They probably did some design changes. Frequency is also affected by how you manage to propagate the clock signal through the design. The clock net is usually the most critical when it comes to frequency, since it's a huge net that has to spread through the whole design and keep everything in sync.

When implementing a design, you're not thinking directly in terms of frequency. You're looking at how the timing compares to your target - if the signal is too late or too early. When I say 'you', I meant the Place & Route engine is looking at this.

So they may have identified further areas for improvement after having taped-out the 820. They then modified their flow scripts and restarted the whole mess (think 1+ week-long runs on 16+ cores and 1TB RAM).

 

[AnonymousStapler]

Demento[/url]":1s9jipem]Yeah, my exact thought was "We're seeing the new Nexus". The past couple had to use the same SoCs as the spring releases before them. The original Nexus 5 got the boost of using the 800 when most of that year's phones were on the Snapdragon 600, so we can have high hopes for this one indeed.

The Nexus 6 used the 805, right around the same time the Droid Turbo and Note 4 came out.

 

[ikjadoon]

laststop311[/url]":3if0ugji]

ikjadoon[/url]":3if0ugji]What is the voltage increase? Like, is this just simple binning that took Qualcomm a weekend (a la i5-6600K / i7-6700k)?

Sounds like a slight hit for battery life.

The tiny extra energy that it may use is more than made up by getting back into idle mode. The extra time in idle mode more than makes up for the minute amount of energy used clocking higher.

Says who? You have data to back that up? I have data that says quite the opposite for laptop CPUs: the voltage (and power output scales quadratically with voltage) bump consumes more power than you get back in slightly longer idle. Please see here:

http://www.anandtech.com/show/7113/2013 ... i7-4650u/3

See this page and the next. Voltage bumps look like they steal much more power than you get back by "finishing" the task early. Who said it was "minute" power? I'm not sure where you're getting all these conclusions from. If you have data, I'd love to see it.

 

[jhu]

fuzzyfuzzyfungus[/url]":11ntsijp]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

I don't know. My SD820 never throttles, although it sits in a dev board, and I put a heatsink/fan on it.

 

[fuzzyfuzzyfungus]

laststop311[/url]":1lgnpqeg]

fuzzyfuzzyfungus[/url]":1lgnpqeg]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

Not too many(though, somewhat to my surprise, the 'Pip Boy' emulator app for Fallout 4 practically made my Nexus 7 cry); but the ones that don't hit the CPU terribly hard are also the ones where I'm least likely to notice the difference between a flagship-level CPU and a boring midrange one.

Lower end devices can be pretty unpleasant when they combine a feeble CPU, the worst eMMC the vendor could find, and a brutally maldesigned vendor skin; but my(admittedly anecdotal) experience has been that the difference between 'good' and 'top of the line' is barely noticeable because short, bursty, demands finish quickly enough to not try my patience either way; and sustained, demanding loads cause the top few SKUs to throttle to roughly the same speed.

 

[santos-l-halper]

Lots of unwarranted snark. Chip releases are known to function like this. 10% increase is more than 0%. Sometimes progress is incremental.

 

[AreWeThereYeti]

theoilman[/url]":cwbtkfxd]

ssiu[/url]":cwbtkfxd]

laststop311[/url]":cwbtkfxd]

fuzzyfuzzyfungus[/url]":cwbtkfxd]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

The mobile VR stuff needs high sustained performance. For example, Google said the Nexus 6P can be used for Daydream VR development, but "Caution: The 6P's thermal performance is not representative of the consumer Daydream-ready devices that will be launching later this year. In particular, expect the 6P to thermally throttle CPU and GPU performance after a short period of use, depending on workload."

I think at this juncture it's silly to think using phones for vr will be anything more than a novelty.

On the contrary. Several reasons. First, buying a $100 accessory to a phone you already need is a much lower bar to clear for consumers, and so you will naturally simply see a lot more buying at that much lower price point.

Secondly, lots of people (me included) are waiting out this generation of high-end VR, waiting for 4K. This gives people like me a nice holding pattern (and for media consumption in VR, the phones are actually better VR platforms than the Rift and Vive).

Finally, you overlook the fact that the biggest single VR market and user base right now is the Gear VR, not the Rift or Vive. More people are getting turned on to VR by a phone than any other way. It's a great gateway drug.

edit: forgot cardboard, which is probably the largest user base, but I don't consider that proper VR yet, without low-latency modes and decent gyroscopes. But that will change.

 

[theoilman]

AreWeThereYeti[/url]":33hp11ud]

theoilman[/url]":33hp11ud]

ssiu[/url]":33hp11ud]

laststop311[/url]":33hp11ud]

fuzzyfuzzyfungus[/url]":33hp11ud]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

The mobile VR stuff needs high sustained performance. For example, Google said the Nexus 6P can be used for Daydream VR development, but "Caution: The 6P's thermal performance is not representative of the consumer Daydream-ready devices that will be launching later this year. In particular, expect the 6P to thermally throttle CPU and GPU performance after a short period of use, depending on workload."

I think at this juncture it's silly to think using phones for vr will be anything more than a novelty.

On the contrary. Several reasons. First, buying a $100 accessory to a phone you already need is a much lower bar to clear for consumers, and so you will naturally simply see a lot more buying at that much lower price point.

Secondly, lots of people (me included) are waiting out this generation of high-end VR, waiting for 4K. This gives people like me a nice holding pattern (and for media consumption in VR, the phones are actually better VR platforms than the Rift and Vive).

Finally, you overlook the fact that the biggest single VR market and user base right now is the Gear VR, not the Rift or Vive. More people are getting turned on to VR by a phone than any other way. It's a great gateway drug.

edit: forgot cardboard, which is probably the largest user base, but I don't consider that proper VR yet, without low-latency modes and decent gyroscopes. But that will change.

Price is only one of many issues. Power, heat regulation, and battery life imo tip the scales for vr failing in mobile for everything but the occasional interactive video, at least for the foreseeable future.

 

[THX]

Kerome[/url]":2w7gnod5]Will be interesting to see how it compares to Apple's updated chips for the iPhone 7 in September. 10% speed bump is not a lot compared to Apple's consistent gains of 60-70% year on year for the last few years.

820 -> 821 isn't a fair comparison as the first SD820 phones started being released in March of this year. In Sept 2015 (around the time when new iPhones and Nexii are announced) the 810 was the top dog Qualcomm-wise, and the 820 is 35-54% faster. (Src: http://www.trustedreviews.com/opinions/ ... dragon-810)

By Sept 2016 the 821 will be standard and is 10% faster than the 820. So a comparison of the 810 -> 821 should be pretty even if you're comparing Apple's yearly SOC performance improvements.

 

[AreWeThereYeti]

theoilman[/url]":19tvd6lz]

AreWeThereYeti[/url]":19tvd6lz]

theoilman[/url]":19tvd6lz]

ssiu[/url]":19tvd6lz]

laststop311[/url]":19tvd6lz]

fuzzyfuzzyfungus[/url]":19tvd6lz]Any word on whether this includes any sort of process or packaging improvements, or should we expect the 821 to throttle to the same sub-820 clock speeds that the 820s in current devices tend to if you actually stress them?

(edit: just in case that came off sounding unduly negative: I don't think that the 820 is a bad chip by any means; it's just that, like most of its competitors, the advertised clock speed is effectively 'burst mode' and very few handset designs actually have the cooling to deliver that for any length of time. Throttling is a perfectly sensible response to thermal limits; but it makes paying extra for a slightly higher peak clock and then throttling down to basically the same speed as the last generation pretty unexciting, so I'd hope that the 821 either does have thermal improvements or is priced more or less exactly as the 820 was.)

What programs are you using on your phone that keep the cpu stressed long enough to throttle? Every workload I have on my phone is extremely bursty. The higher boost clock means the cpu can finish it's bursty task faster and "race" back to sleep faster. Sooner the cpu gets into idle mode the more energy is saved. So this cpu could easily be more efficient if it can do the same tasks but get back into idle mode faster. I have never noticed any thermal throttling using the youtube app, spotify app, or browse web on chrome. The cpu doesn't stay stressed long enough to throttle. Nobody really uses any programs that stress the cpu for longer than 20-30 seconds i don't think.

The mobile VR stuff needs high sustained performance. For example, Google said the Nexus 6P can be used for Daydream VR development, but "Caution: The 6P's thermal performance is not representative of the consumer Daydream-ready devices that will be launching later this year. In particular, expect the 6P to thermally throttle CPU and GPU performance after a short period of use, depending on workload."

I think at this juncture it's silly to think using phones for vr will be anything more than a novelty.

On the contrary. Several reasons. First, buying a $100 accessory to a phone you already need is a much lower bar to clear for consumers, and so you will naturally simply see a lot more buying at that much lower price point.

Secondly, lots of people (me included) are waiting out this generation of high-end VR, waiting for 4K. This gives people like me a nice holding pattern (and for media consumption in VR, the phones are actually better VR platforms than the Rift and Vive).

Finally, you overlook the fact that the biggest single VR market and user base right now is the Gear VR, not the Rift or Vive. More people are getting turned on to VR by a phone than any other way. It's a great gateway drug.

edit: forgot cardboard, which is probably the largest user base, but I don't consider that proper VR yet, without low-latency modes and decent gyroscopes. But that will change.

Price is only one of many issues. Power, heat regulation, and battery life imo tip the scales for vr failing in mobile for everything but the occasional interactive video, at least for the foreseeable future.

Power, heat, and battery life are problems that computing progress solves rapidly with time. It used to be we couldn't do this at all. Now we can do it for a little while. Soon we will be able to do it for much longer.

My favorite Gear VR game, Smash Hit VR, is a good proxy example. It is a relatively low-poly game, but looks great, although it came out a while ago. My old Note 4 can play it great. In fact, since my phone is overpowered for the game (here's the kicker), I can play it quite well even in power-saving mode. And guess what! I get massively better runtimes, without overheating. I can play Smash Hit VR for multiple hours straight, at least, without battery or heat problems.

Our phones are doing things that used to require a 300w desktop to do. Soon we will have 4k displays and phones designed with heat dissipation as a major factor, and although the graphics level will always lag the desktops, it will be "good enough".