• After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.

GL benchmarks of the 3D's Processor...

All these comparisons to SGSII to the evo3d are very strange. Evo3d has qHD resolution and (correct me if i'm wrong) the SGSII is 800x480. Even if you play games 2 hours a day on your phone (and who does this?) wouldn't you rather look at 35% more pixels the rest of the time?

I think despite these "relatively poor" benchmarks, the resolution of e3d is a HUGE advantage. Besides, by july 1 there will be 17 rooted roms floating the net which will unlock the cpu up to 1.8ghz or something ridiculous, pwning everything out there for the 35 minutes it lasts before the battery runs out.

Just my $.02
 
All these comparisons to SGSII to the evo3d are very strange. Evo3d has qHD resolution and (correct me if i'm wrong) the SGSII is 800x480. Even if you play games 2 hours a day on your phone (and who does this?) wouldn't you rather look at 35% more pixels the rest of the time?

I think despite these "relatively poor" benchmarks, the resolution of e3d is a HUGE advantage. Besides, by july 1 there will be 17 rooted roms floating the net which will unlock the cpu up to 1.8ghz or something ridiculous, pwning everything out there for the 35 minutes it lasts before the battery runs out.

Just my $.02

Heh, I'm definately sold on the E3D. I can see why the comparisons are being made as the SG2 is a high end phone, but thats fine. Benchmarks aren't everything. The benchmarks shown have little influence on my buying decision. I have liked all that I have seen so far.

The higher resolution on the 4.3 screen is going to be perfect for what I will use it. Honestly, the dual core and gpu may be more than I need, but as I am in the market for a new phone, I wanted to go for what I felt had the best tech in the sweetest package. That, to me, is the E3D.
 
Your comment is perfectly valid if resolution was the same as quality. Alas resolution is but one aspect of the display quality.
-
Your comment is perfectly valid if games is the only thing impact of faster cpu/gpu. Alas many things run slower if the cpu/gpu are slower (such as scrolling on the browser, java script, flash, video playing, mail rendering, video chatting, in fact just about everything but voice communication and perhaps texting.
-
The question is does the reduction in speed at these levels impact the experience ?
-
Anyways you make some perfectly valid comments pity they do nothing to move the discussion forward.

All these comparisons to SGSII to the evo3d are very strange. Evo3d has qHD resolution and (correct me if i'm wrong) the SGSII is 800x480. Even if you play games 2 hours a day on your phone (and who does this?) wouldn't you rather look at 35% more pixels the rest of the time?

I think despite these "relatively poor" benchmarks, the resolution of e3d is a HUGE advantage. Besides, by july 1 there will be 17 rooted roms floating the net which will unlock the cpu up to 1.8ghz or something ridiculous, pwning everything out there for the 35 minutes it lasts before the battery runs out.

Just my $.02
 
Yeah, the msm8660 definitely runs at 1.5Ghz, though it's not clear to me why HTC opted for the 1.2 Ghz variety. Anandtech tested the 1.5 Ghz version (though on an WVGA screen):

AnandTech - Dual Core Snapdragon GPU Performance Explored - 1.5 GHz MSM8660 and Adreno 220 Benchmarks


From an earlier post -

AnandTech - Hands on and Benchmarks of two MSM8x60 Phones - HTC Sensation 4G and HTC EVO 3D

Removes the guesswork in the Anandtech benchmarking for this phone. ;)

My best guess is a cheaper price based upon better yields for a lower clock frequency part. It's too bad though. Only time will tell if the actual parts inside the Evo 3D have the physical capability of scaling up to 1.5 Ghz, or if the parts are so low quality that they max out a 1.2. It's also unclear to me whether HTC will retroactively apply its bootloader policy, which might hamper efforts to bring an overclocking kernel.
Given that the Evo isn't out yet and that the petitions were flooded with 3D references, I'm not sure how they couldn't keep their word, nor why you think that would be a retroactive move in this case. At the time of their announcement, they said all future phones - this is still a future phone.

It could be a yield issue for the clock, as you say, but with TSMC making them, I'd be surprised. I believe the lower clocking choice was sufficient to trounce 1 GHz phones when announced (in the minds of the press, the announcement target) and was chosen for battery life.

In a few weeks, we'll know.

~~~~~~

I've posted this often - but with the new flood, I'll repeat myself: benchmarks are useless unless you can correlate them to real-world results.

All they really benchmark are the performance when various service library functions are used - and there's more than one way do any job in graphics programming.

The benchmarks might help you decide what phone's right for you, but they don't give the results of a horse race - they give app developers clues as to which functions to use or avoid when developing for specific phones or the general market.

The benchmarks discussed here make the case that the 3vo has less than stellar graphics - yet the Smartbench gaming results put the 8660 above the Samsung Exynos - they even put the Hummingbird above the Exynos. (The Smartbench author clarified what his suite tests in the tech thread here.)

It's not that any one benchmark is wrong or the "real" one - it's that the true point of benchmarking is for devs to use to quality their approach to things.

~~~~

PS - Links from Acei, Smartbench author (with tip of the topper to you2 for the nudge)

http://androidforums.com/htc-evo-3d...-evo-3d-qhd-3d-dual-core-smp.html#post2656234

http://androidforums.com/htc-evo-3d...vo-3d-qhd-3d-dual-core-smp-2.html#post2661059

We also have an app dev in this forum backing his statements - will have to look for that link, can't find it off-hand.
 
I just finished reading through that other thread. It was really informative & helpful.

This discussion and that discussion reminded me of a conversation I had with a friend of mine. We were discussing the computers we were building 10 years ago and the ones we build today. While today the technical specifications are astronomically higher than the ones we built 10 years ago; on some basic level they are the same speed. For mundane tasks like Wordpad, web browsing, or using the Calculator... not much has changed except for the speed of the Internet connection. What I mean is, 10 years ago it took my computer maybe 1 second to load calculator. Maybe 5-10 to load Netscape (lol) and maybe about the same for Wordpad. After all of our technological advancements, not much has changed in that area of basic computer usage. It still takes about that same amount of time to start and run basic functions on my PC.

I'm trying to say that the requirements of modern software often match the specifications of the current CPU/GPUs on the market. This in some small way brings all this advancement to some sort of zero sum in my mind.

It's early and I am just getting my first pot of Coffee... so please pardon me if that made absolutely no sense...
 
My best guess is a cheaper price based upon better yields for a lower clock frequency part. It's too bad though. Only time will tell if the actual parts inside the Evo 3D have the physical capability of scaling up to 1.5 Ghz, or if the parts are so low quality that they max out a 1.2. It's also unclear to me whether HTC will retroactively apply its bootloader policy, which might hamper efforts to bring an overclocking kernel.

I really don't think HTC will lock the 3vo, especially retroactively. That would be very stupid of them and create a PR nightmare.

The Asynchronous portion of the dual cores is somewhat intriguing to me. During the benchmarking we have seen, I am wondering if one core doesn't simply shut down or minimize itself doing relatively trivial tasks. By just having one core focus on the benching, along with the GPU, could that result in the lower scores?

That would definitely explain the "low" scores. But tbh, the 3vo's Quadrant and Smartbench scores are on part with the Tegra 2 inside my Atrix, which isn't asynchronous.

So off the top of my head, I can think of 3 possibilities:

1. The 3vo's relatively low score is due to its asynchronous operation. So we see scores slightly lower than the 1ghz Tegra 2 inside the Atrix. But this would mean that the MSM8660 is ridiculously powerful, scoring almost as high as a Tegra 2 without fully stressing both cores.

2. The 3vo's relatively low score is due to Android 2.3 lowering the scores across the board, as the Wirefly review mentioned. So when the 1.2ghz Snapdragon should've scored a bit higher than the 1ghz Tegra 2 as that Anandtech article predicted, it actually scores a bit lower due to the software. But this does not explain why the SGSII is off the charts as it is also running on Gingerbread, even after taking the additional resolution into account.

3. Samsung did some sort of optimization for the common benchmarks for the SGSII. We see both AMD and Nvidia do this with their GPUs in the PC industry, where both parties would optimize their drivers for the set of tests in benchmarks like 3DMark Vantage. I don't know if Samsung did something similar via software or even if it would be possible, but it could be a possibility. I guess once we have both devices rooted and running on CM7 we'll have a better idea.

I suspect it's a combination of at least 1 and 2 though, and maaaybe 3 as well.

I suppose my main point is this: All viewing, tasks and experiences indicate things are flowing perfectly. The interface is rock solid, no hesitation, no choppiness, smooth transition. The E3D appears to handle everything thrown at it powerfully and without issue. For lack of better terminology, this appears to be a real life benchmark and it is fantastic, IMO. The video play was smooth, game didn't hesitate at all. I am not sure what more I could personally ask for on a phone.

Long and short of it is, I think the Async Dual cores will be a great balance of power usage and permormance. I can't wait for real life tests to see how true this is.

I think so too, now we are just waiting on official reviews. I'm really satisfied with how smoothly the phone runs in the videos, and what especially impressed me was how quickly he (the reviewer) backed out of that Spiderman game in the Wirefly review. Instantaneous reaction.

And that, is worth more than all the benchmarks in the world to me.

All these comparisons to SGSII to the evo3d are very strange. Evo3d has qHD resolution and (correct me if i'm wrong) the SGSII is 800x480. Even if you play games 2 hours a day on your phone (and who does this?) wouldn't you rather look at 35% more pixels the rest of the time?

Both are flagship devices launching on Sprint during the summer-fall period. What's strange about comparing them? :) It's just that many do forget about the extra 35% pixels the 3vo has to render when comparing benchmarks.[/QUOTE]

Everybody wishes the SGSII came with qHD, but now everyone has to choose between having AMOLED+ or a higher resolution. Some will lean one way and some the other.

I think despite these "relatively poor" benchmarks, the resolution of e3d is a HUGE advantage. Besides, by july 1 there will be 17 rooted roms floating the net which will unlock the cpu up to 1.8ghz or something ridiculous, pwning everything out there for the 35 minutes it lasts before the battery runs out.

Just my $.02

I completely agree :) Hopefully the battery will last more than 35min @ 1.8ghz though. Assuming the phone doesn't blow up before then :p

Heh, I'm definately sold on the E3D. I can see why the comparisons are being made as the SG2 is a high end phone, but thats fine. Benchmarks aren't everything. The benchmarks shown have little influence on my buying decision. I have liked all that I have seen so far.

The higher resolution on the 4.3 screen is going to be perfect for what I will use it. Honestly, the dual core and gpu may be more than I need, but as I am in the market for a new phone, I wanted to go for what I felt had the best tech in the sweetest package. That, to me, is the E3D.

Same here, I just couldn't imagine going back to 800x480 after getting used to the qHD display on my Atrix. I went to the Sprint store and played with the NS4G. The 800x480 resolution is okay on its 4" screen, but when stretched out to 4.3", I think it'll be too stretched out for my liking.

Plus I've never actually seen AMOLED+ in person, so at least I won't know what I'm missing out on :D

I've posted this often - but with the new flood, I'll repeat myself: benchmarks are useless unless you can correlate them to real-world results.

All they really benchmark are the performance when various service library functions are used - and there's more than one way do any job in graphics programming.

The benchmarks might help you decide what phone's right for you, but they don't give the results of a horse race - they give app developers clues as to which functions to use or avoid when developing for specific phones or the general market.

The benchmarks discussed here make the case that the 3vo has less than stellar graphics - yet the Smartbench gaming results put the 8660 above the Samsung Exynos - they even put the Hummingbird above the Exynos. (The Smartbench author clarified what his suite tests in the tech thread here.)

It's not that any one benchmark is wrong or the "real" one - it's that the true point of benchmarking is for devs to use to quality their approach to things.

I don't think any of us in this thread actually doubt the 3vo will be a great phone, or will let its "low" benchmark scores influence our decision to buy it or not (which is probably all "buy" for us). I think all of this conversation about the processor and benchmarks is to satisfy our curiosity on why it gets the scores it does when everyone expected it to score much higher.
 
Asynchronous operation should have no effect on benchmarks - if anything, it would create an improvement in simultaneous performance/battery life. With a synchronous SMP operating system, you'd have to choose which of those to optimize at the underlying Linux level. With aSMP, you don't.

I don't think that the benchmarks are optimized or not for various processors beyond what's built into the Android software developers kit (SDK) - at present that's the ARM 7 instruction set with neon extensions (and a few others) - the Tegra-2 lacks that.

The benchmarks aren't being lopsided by cores going to sleep. The aSMP job control happens below Android's part of the OS and Android apps cannot affect that. If an app is written with threads, per the Android SDK guidelines, it will distribute across any dual-core roughly the same (give or take kernel design). If the benchmark app distributed across cores on the Sammy, it's doing it on the 3vo.

The numbers are the numbers - they indicate not what the processor can do - they indicate what the whole phone does when stressed with a test program that makes specific operations via standard software libraries. The benchmarks are in no way making assembly level calls and assessing the actual processors directly.

If your apps use the same specific operations in the same way that the benchmark tests do, then the benchmark test will accurately predict which phones will perform better.

If your apps do the same type of work but using specific operations not specifically tested by the benchmarks, then all bets are off.

The battle cry in the SGS2 crowd is that Quadrant is right, Smartbench is wrong because Quadrant says the Exynos is best just like we expected hooray!

But the truth is that any apps written similar to the way the Smartbench test was coded will have the 3vo and SGS outperforming the SGS2.

I'm sorry to be a broken record on this, but it's not clear that the point is getting across.

Ultimately, for all we know, there is a clear ranking of processors to build the proper expectations.

If you think the benchmarks are telling you that, re-read this post, take it all with a grain of salt and remember when you get any superphone home and in your hands: your mileage may vary. :) ;)

/rant, promise
 
All these comparisons to SGSII to the evo3d are very strange. Evo3d has qHD resolution and (correct me if i'm wrong) the SGSII is 800x480. Even if you play games 2 hours a day on your phone (and who does this?) wouldn't you rather look at 35% more pixels the rest of the time?

I think despite these "relatively poor" benchmarks, the resolution of e3d is a HUGE advantage. Besides, by july 1 there will be 17 rooted roms floating the net which will unlock the cpu up to 1.8ghz or something ridiculous, pwning everything out there for the 35 minutes it lasts before the battery runs out.

Just my $.02

For me no... mainly because the SAMOLED+ still looks better than the qHD and still uses less power while doing it.
 
For me no... mainly because the SAMOLED+ still looks better than the qHD and still uses less power while doing it.

May be getting off topic, but I was under the impression that the various flavors of AMOLED screens were less power efficient than a typical LCD. I could be wrong of course.

While I'd have loved for the Evo 3D to have blown everything else away in every possible benchmark thrown at it, if it can run (and run well) everything that I want to use it for, I don't care if something else has better tech specs or benches better. My guess is that for things that I'll be doing, that extra speed will mean such small real world results that by the time it's really needed, I'll be a generation or two past the current handset.

So, in other words, waiting to see more hands on and review. Should be exciting when we get to say...a week out, at which point I would be expecting them to pour in from everywhere, and then even more peer reviews if those premiere members actually get their handsets on the 22nd or so...
 
For me no... mainly because the SAMOLED+ still looks better than the qHD and still uses less power while doing it.

The SAMOLED+ looking better is pretty subjective. To me, the LCD looks better (especially at higher resolution), but that's your decision.

Also worth noting is that the Evo 3D has the micro USB on the side which would be the bottom as you're playing games. So you can charge while you play if you're a hard core gamer. :)
 
Healthy skepticism when it comes to benchmarks is always appropriate, but in this case, all signs point to the dual core Snapdragon not being as competitive as some of its brethren. Generally, if you want that kind of information, you need to head to AnandTech (and here... and the meaty part here), where they are very serious about the art of performance testing.

The number of cores and the processor clockspeed are not the only indicators of performance, and the CPU architecture plays a large role in the computing efficiency of the processor. The nVidia Tegra 2, Samsung Exynos, Apple A5 (?), and Texas Instruments OMAP 4 dual core processors use a more recent and improved ARM Cortex A9 architecture. The Qualcomm Snapdragon, while not quite using the prior Cortex A8 architecture, is using something that generationally falls somewhere between the A8 and A9.

In short, even though the Tegra 2 runs at 1Ghz, it is not a huge surprise that it is either competitive with or somewhat exceeds a Snapdragon 1.2 Ghz SoC. There are questions as to preproduction units, optimization, and increase dual core software optimization, but we shouldn't delude ourselves too much as to what potential benefits they may bring.

The results are admittedly somewhat disappointing and Qualcomm needs to step up their SoC game a bit, though this isn't necessarily the first time us Evo owners have been in this position. Even upon release, the original Snapdragon was recognized to have terrible 3D performance (sigh Adreno 200) and the PowerVR SGX GPU in my year old Palm Pre was a stronger performer. Subsequent releases with the Samsung Hummingbird processor only further served to show how much we were lacking in the GPU department.

That said, I think the conclusion is that even when the Evo 3D launches, we will definitely not have the fastest horse in the stable, for the people for which it matters. However, the overall performance increases from the original Evo should be quite significant, and that will hopefully be enough for all of us.

(Personally, I'm keeping an eye out for comparisons with the current single core Hummingbird -- that thing has continued to impress me and I hope the Snapdragon gains over it are more than marginal).



Asynchronous dual cores isn't a performance issue, though it is notably a power consumption one. With the Tegra 2, both cores are running at the same clock, so if you need the performance, it's full blast. However, with the new Snapdragon and async cores, the amount of power is better tailored to each core's individual load. The power consumption should be better on the Snapdragon (though the fab processon the snapdragon is 45nm versus 40nm on the Tegra...)

...OTG is pretty cool, but it's not a performance issue, is it?

I took this from another post in another thread because after reading your post I think it needed some kind of response.

The dual-core 45nm gen 3 MSM8660 snapdragon with independent clock units and newer enhanced Scorpion MPCore draws lower sustained power across the board, even heavily overclocked (or simply at 1.5Ghz like they were designed for). In terms of handset differences, yes, HTC's snapdragon devices are only "barely on par" with the others because of many things, not just the battery size (which has tended to be a bit smaller). Sense 2 is very resource-hungry and the build (not to mention the various Sense widgets, additional Sense I/O, sense radio usage, etc) draw power like no other. It also is among the most graphically intensive overlays (due to the actual graphical output or to poor coding is debatable, of course).

"There is no clear winner. If you're a 3D gamer, ONLY get Tegra 2 or Exynos." And of course a conclusion drawn from nonsense is nonsense, like this one. Even for a gamer, they will all do well, and currently the Adreno 220 on the MSM8660 (in the Evo 3D and Sensation) is leading the pack by a very solid margin. By design, that is. Implementation-side optimization is always better left to the community, of course.

What people don't seem to realize is that beyond just a 'die-shrink', despite retaining the name, the third-gen Scorpion 45nm is now well over "5%" faster per-clock than the stock A8. Although the architecture is very similar, there have been remarkable improvements. As Qualcomm themselves presented on March 11 during their MPCore overview in Korea, the MSM8x60 is a huge amount over "5%" faster per clock, and a huge amount more power efficient than even the A9, not the least of which is due to the independent core clock scaling. As for the "console quality graphics" myth, no mobile chipset is there yet. During the same presentation (and now on their website), Qualcomm also admitted the Adreno 220 isn't the "console level" GPU they talked about at all -- that's the 3XX series with the 8X70 ("NEXT GEN", which is also reportedly 47% more power efficient than the A15 with 23% more performance headroom).

The MSM8660 (the one in the Evo 3D) has, in addition to its 128-bit FPU (with 1.5-6x performance of the 64-bit in the A9), an enhanced OOO unit, 8 outstanding non-cacheable loads in L1 cache (versus only 1 in the a9, meaning significantly better multimedia processing), and a tightly coupled 512kb L2 with much lower memory latency and better CPU snappiness/efficiency as opposed to A9's embarrassingly loose-coupled L2. "5%" better is the biggest load of bull I've heard since the 65nm scorpion was tested under a suboptimal arm linux build on lkml for a preliminary benchmark vs the already-optimized stock A8. Its dual core "real world performance" of the chipset in total at 1.2Ghz is roughly 30% better than the reference dual A9 with comparable governor and identically configured kernel (and distro/android build). The 2.6.32 stock kernel is now much better optimized for the 45nm chips and actually takes advantage of the additional FPU and larger NEON superset when compiled properly. Since then the last-gen 8x55 at 1.4Ghz has proven comparable to a dual core stock a9 at 1ghz, and with a similar power profile to boot. As clock speeds get higher, scorpion has an even greater power advantage. At 1.2Ghz, the snapdragon is quite a bit more power efficient than the 1.2Ghz a9, for instance, and beyond that the difference can breach 75%.

The Adreno 220 GPU is over double the performance of the Adreno 205, which itself was nearly on par or better than the likes of the first gen hummingbird's GPU (SGS 1's SGX) in almost all of the modern GL2.0 benchmarks (and well over double the performance of the Adreno 200).

Take the glorified pro-Samsung/Tegra nonsense you hear with a grain (or bucket) of salt. Just because people are stigmatized against Qualcomm due to HTC's legacy of providing underclocked devices with older graphics drivers (SE is the only company who bothered to ship their device, the Play, with updated Adreno 205 drivers, by the way), doesn't take away anything from the fact that even the Adreno 205 can essentially hold its own against even the Tegra 2. The 220 is double the real-world performance of that.



So the MSM8660 either sucks compared to its brethren or its on par or better? Who can cut through the BS?
 
2strokenut, can you provide a link to all that info you just posted?

*Edit:
nvm, you got it from this pdf here.
http://www.6ybh-upload.com/eb1vektirkdl/Qualcomm_MPCore_Overview.pdf

It's made by Qualcomm though, so I'd take any figure it has with a huge grain of salt.

Also on pg 19, it shows the 1.5ghz MSM8660 to be ~65% faster than a 1ghz A9 dual core. Scale it back to 1ghz and it's only ~5% faster clock for clock, just as Anand predicted.
 
I'll have to think about it and digest it some, but I'll note off the bat that there's a lot of talk about "efficiency" which tells us little about overall performance. I think it's an important characteristic to have (although the jury may be out in terms of overall battery life between the larger Evo 3D battery, the potential increased power draw from the parallax barrier, the (theoretically) more efficient Super AMOLED screen... there are just too many variables).

At any rate, a CPU can be "efficient" without necessarily being the fastest or even drawing the least power (the Intel Atoms draw little power compared to many of the Intel Core CPUs, but are frequently less efficient because they draw more power relative to the performance).

At any rate, I still think the Snapdragon will be a very strong performer, but not necessarily the strongest. I'm still waiting to see full tests on production units though.

Also, I personally don't care all too much for justifications involving drivers or Sense bottlenecks. I feel we should be taking phones as we get them -- if Sense is a performance bottleneck, then so be it. The same goes with the drivers -- phones are not yet like PCs in that we can just download the newest driver update every month or so. Phone software updates come few and far between so if the drivers end up being terrible, Qualcomm has no one to blame but itself. Only time will tell.
 
^This, what NeoteriX said times - oh, let's say - times a hundred.



So the MSM8660 either sucks compared to its brethren or its on par or better? Who can cut through the BS?

Snapdragons never really suck. They're rock-solid reliable performers and there's a reason many people prefer HTC, benchmarks or no benchmarks.

nVidia brought their A game when they broke into Android with the Tegra-2.

Samsung brought their A game last year with the Hummingbird and the SGS class phones and upped their game this year.

The 8660 is the result of Qualcomm's A game for this year so far. Those that have followed Qualcomm know that we expected a form of this processor last year, to be called the 8x72. That one never released (sfaik), and instead evolved into today's 8660. So - it's by no means a Qualcomm afterthought or scramble to get to dual-cores.

And never, never, never count out TI and their OMAP processor technologies.

Processors are a game of leapfrog, driven by the economy as much or more than by R&D.

When the smoke settles a year from now, I predict users in the fullness of time will look back and honestly say that their favorite phones had some sucky processor features compared to others and blew the others away in some areas. And they will say all of that post-benchmark, from real experience. Insert either Qualcomm/Samsung/nVidia/TI as the processor name.

If you're looking for this info to make a choice that's way cool, just remember the quad-cores are coming and will make all of this moot before you know it and you'll get to do it all over again.

In the meantime, what ebolamonkey3 and NeoteriX said: take Qualcomm's comparisons to others with a grain of salt, and take it as it comes.

Sense is fun,
Sense is great,
but Sense is Sense,
small and fast it ain't.

I could drop Sense and get better performance on my present phone. I could get basically the same features with aftermarket configurations. I do not care. I like Sense, I run with it.

I will run it on the 3vo and I will be likely to root and look for a fine independent kernel dev to give me a higher-performing replacement for that one part.

Trust your gut when you choose your phone. Features and overall performance is the thing, not just the processor - that's just one piece of silicon on a motherboard with LOTS of important silicon.

Pundits are fickle - their blogs show ever-changing opinions as the facts continue to come out. Their latest trend is that the Exynos will be the top dog of this crop. They may be right. Personally, like I said, I'll trust user reports when the hype dies down.

While all that's happening, I plan to be rocking a 3vo and I'll be complaining in no time that it's not a quad core - but that's just me.
 
2strokenut, can you provide a link to all that info you just posted?

*Edit:
nvm, you got it from this pdf here.
Download Qualcomm MPCore Overview pdf

It's made by Qualcomm though, so I'd take any figure it has with a huge grain of salt.

Also on pg 19, it shows the 1.5ghz MSM8660 to be ~65% faster than a 1ghz A9 dual core. Scale it back to 1ghz and it's only ~5% faster clock for clock, just as Anand predicted.

I got it from the Tegra2 thread here.
 
Correct me if I'm wrong... but this article says:
HTC Puccini might be almost ready for mass production - Phone Arena
and
HTC Puccini touted as a 10-inch LTE tablet with 1.5GHz processor, Honeycomb flavor -- Engadget
and
HTC Puccini Tablet Will Released With 1.5 GHz Dual-core Qualcomm MSM8660
A dual-core 1.5GHz Qualcomm MSM8660 offers plenty of processing power...
Is that not the CPU in the Evo3D? That would lead me to believe the Evo3Ds CPU has been under-clocked to 1.2ghz to save on battery. ;)

Which, if true, would mean after Rooting you could easily (and safely) run the phone at 1.5ghz. Correct?

EDIT: Or, we know how Sprint likes to make last minute shock-value announcements. Maybe this will be it and it'll launch with a 1.5ghz? Total speculation here obviously. I can't see anyone passing up a dual-core 1.5ghz 3D phone for another model. :D
 
Which, if true, would mean after Rooting you could easily (and safely) run the phone at 1.5ghz. Correct?

EDIT: Or, we know how Sprint likes to make last minute shock-value announcements. Maybe this will be it and it'll launch with a 1.5ghz? Total speculation here obviously. I can't see anyone passing up a dual-core 1.5ghz 3D phone for another model. :D

Maybe.

Mobile Processors | Snapdragon | Qualcomm

"... up to 1.5 GHz"

Depends on any number of things - maybe it can be overclocked like you can't believe, maybe only a little.

Won't know until we know.
 
Correct me if I'm wrong... but this article says:
HTC Puccini might be almost ready for mass production - Phone Arena
and
HTC Puccini touted as a 10-inch LTE tablet with 1.5GHz processor, Honeycomb flavor -- Engadget
and
HTC Puccini Tablet Will Released With 1.5 GHz Dual-core Qualcomm MSM8660

Is that not the CPU in the Evo3D? That would lead me to believe the Evo3Ds CPU has been under-clocked to 1.2ghz to save on battery. ;)

Which, if true, would mean after Rooting you could easily (and safely) run the phone at 1.5ghz. Correct?

EDIT: Or, we know how Sprint likes to make last minute shock-value announcements. Maybe this will be it and it'll launch with a 1.5ghz? Total speculation here obviously. I can't see anyone passing up a dual-core 1.5ghz 3D phone for another model. :D
One advantage to underclocking is battery life, but it's not the only one.

If you underclock, you can also improve your silicon yields as underclocking stresses the CPU less and allows for slightly lower quality chips to be used.
 
Well, I was more or less pipe-dreaming when I wrote that...

Just imagining Sprint/HTC coming out and going "We're proud to announce the Evo 3D will launch with a 1.5ghz Dual Core". How could anyone compete with that? A 1.5ghz dual-core 3D 4G phone. Insane.

If the Puccini and the Evo3D share the same CPU I don't see why not except for the battery life issues it would cause.
 
If the Puccini and the Evo3D share the same CPU I don't see why not except for the battery life issues it would cause.

I'll elaborate what I was saying earlier -- (I'm no engineer, but this is how I understand it)

So when you make semiconductors like say, Qualcomm's Snapdragon, you make a huge batch of them but they're not all the same. Due to imperfections that get introduced into the fabrication process or whatever other variations that exist, there will be higher quality and lower quality chips. By quality in this context, I mean whether a cpu can operate at a given clock speed without errors in computation. The higher you clock a cpu, the more you stress the cpu and the more likely the computational errors, so lower quality cpus can safely run at a lower clock speed, but not at a higher one.

Going back to our batch of chips, let's say 10% will have to be thrown out altogether, 60% are average quality, and 30% are high quality. In this hypothetical, if you were limited to high quality chips in order to run at 1.5Ghz, you'd have a 30% yield efficiency and would have to fabricate a lot of batches to meet your production target. However, if 1.2 Ghz can safely run on an average chip, then you've increased your yield significantly -- up to 90% yield efficiency. Thus the cost goes down significantly because you can fabricate less batches to meet your production target.

Computer CPU mfrs have been doing this for some time, separating the better CPUs for their "limited black edition" of CPUs that are intended to be overclocked with.

This is not an issue unless you're trying to overclock an average quality 1.2 Ghz cpu up to 1.5 Ghz. At this point, you're rolling the dice as to whether the processor is up to the task or not, so -- in a nutshell, just because the 8660 can operate at 1.5 Ghz and was designed for it says nothing as to how well our Evo 3D snapdragons can run at that speed.
 
I'll elaborate what I was saying earlier -- (I'm no engineer, but this is how I understand it)

So when you make semiconductors like say, Qualcomm's Snapdragon, you make a huge batch of them but they're not all the same. Due to imperfections that get introduced into the fabrication process or whatever other variations that exist, there will be higher quality and lower quality chips. By quality in this context, I mean whether a cpu can operate at a given clock speed without errors in computation. The higher you clock a cpu, the more you stress the cpu and the more likely the computational errors, so lower quality cpus can safely run at a lower clock speed, but not at a higher one.

Going back to our batch of chips, let's say 10% will have to be thrown out altogether, 60% are average quality, and 30% are high quality. In this hypothetical, if you were limited to high quality chips in order to run at 1.5Ghz, you'd have a 30% yield efficiency and would have to fabricate a lot of batches to meet your production target. However, if 1.2 Ghz can safely run on an average chip, then you've increased your yield significantly -- up to 90% yield efficiency. Thus the cost goes down significantly because you can fabricate less batches to meet your production target.

Computer CPU mfrs have been doing this for some time, separating the better CPUs for their "limited black edition" of CPUs that are intended to be overclocked with.

This is not an issue unless you're trying to overclock an average quality 1.2 Ghz cpu up to 1.5 Ghz. At this point, you're rolling the dice as to whether the processor is up to the task or not, so -- in a nutshell, just because the 8660 can operate at 1.5 Ghz and was designed for it says nothing as to how well our Evo 3D snapdragons can run at that speed.

That is all true.

Just chiming in that the tolerances where these are made tend be very, very good and although there will be a distribution due to anomalies in the manufacturing process, it's typically skewed already in favor of the rated speed.

Overclocking will be possible, no guarantees on overclocking amount per chip, and no idea on range until we all get the chance to try and report results.
 
More benchmarking fodder -- AnandTech - Motorola Droid 3 Pops up in GLBenchmark - A Verizon XT883 with OMAP4

The Droid 3 is predicted to run the 1GHz OMAP 4430 (SGX 540 GPU) on a qHD resolution screen (dual core, based off the ARM Cortex A9 architecture). This is perfect fodder for a closer apples to apples comparison between the dual core Snapdragon running on a modified ARM Cortex A8 and the new A9 based SoCs, as they are running similar resolutions on similar versions of Android (2.3 Gingerbread).

It was predicted by Anandtech earlier this year that Snapdragon would be slightly less efficient (processing power per clock speed) than the A9 based SoCs, with the 1 Ghz A9 and the 1.2 Ghz Snapdragon edging each other out depending on the circumstances:

In 2011 Qualcomm will introduce the QSD8660, a Snapdragon SoC with two 45nm Scorpion cores running at 1.2GHz. With a deeper pipeline, smaller cache and a largely in-order architecture, the QSD8660 should still trail NVIDIA’s Cortex A9 based Tegra 2 at the same clock speed. However Tegra 2 launches at 1GHz and it won’t be until Tegra 2 3D that we see 1.2GHz parts. Depending on timing we could see dual-core Qualcomm phones running at 1.2GHz competing with dual-core NVIDIA phones running at 1.0GHz. The winner between those two may not be as clear—it’ll likely vary based on workload.

Source: AnandTech - LG Optimus 2X & NVIDIA Tegra 2 Review: The First Dual-Core Smartphone

39027.png


39028.png


It now looks like that prediction was pretty accurate.

The OMAP 4430 edges out the Snapdragon in EGYPT, with the Snapdragon firing back and winning in PRO. Interesting stuff to say the least.
 
Back
Top Bottom