[Robotaz]

First off, I returned my Torch because of software problems that I couldn't wait to be fixed; namely the BT streaming glitch. That said, I have noticed some commentary regarding the Torch processor speed that I see as seriously ignorant and wanted to provide some insight.

Processing power is not a function of clock speed. Processing power is a function of the elegance of the instruction set, the design of the processor's peripheral registers and buses, and the programmer's competence.

What does this mean? A very simple processor will have lots and lots of instructions that you can use to build the machine code. Also, a simple processor will have registers to accept instructions and registers to respond linearly to what was input, with no peripheral registers. Think of a box with a chute where something comes in and a chute of identical size where something comes out. In contrast, an elegant processor design will have very few instructions, will have peripheral registers, and require a much more intelligent programmer to utilize. For this type of processor imagine a box with a chute where something comes in, a chute where something comes out, and chutes all around the sides of the box where specific products automatically come out that can be used to make something without more coming into the "input" chute. FYI, the elegant processor has fewer instructions because it already performs automatically many of the functions that require an instruction and extra clock cycle on simple processors.

A processor works by taking the product that comes out, storing it, waiting for more product to come out, then combining products to come up with usable data. Obviously the faster you pump out products, the faster you make something usable. But what if the elegant processor has products coming out the side chutes automatically, without being told by the programmer, that can be used to keep from having to input and output info over and over to come up with the same end product? Well, it can work more slowly and accomplish more in less time, because the design is sophisticated and the programmer knows how to instruct it. Think of the products coming out of the side chutes on a good processor as products that come out the end on simple processors. The simple processor needs extra instructions, and consequently extra clock cycles of output going to the input over and over, to make specific products come out that the elegant processor simply places on the side chutes without being asked. For this reason, good processors usually have fewer instructions to use.

Let's go back to the year 1997. You had Intel Pentium, AMD Athlon and Duron, and Motorola processors, in general. Intel has always had many instructions, few peripheral registers, and fast clock speeds. Motorola always had a very simple instruction set, peripheral registers, and relatively slow clock speeds. AMD fit in between.

What did everyone buy? Well, of course Intel because Intel was faster. But guess what? My AMD at 700 MHz smoked the crap out of an Intel at 1 GHz. Did the Intel fan boys care, or know what they were talking about? Of course not. We all know that old Macs with Motorolas killed PCs with Intel processors of identical speed and did not require ultra high processing speeds such as Intel.

Intels have historically been designed for newbs to utilize. They have (or at least had) an instruction set and overall design that requires an ignorant programmer to use clock cycles for every single calculation. The Motorola was designed for intelligent people to program however. The Motorola will yield usable data(on the side chutes) that can be used, or not used, for every clock cycle that can be combined with output to effectively stack data in one clock cycle that an Intel will need at least 2-3 cycles to produce by going from output to input over and over. Thus, the Intel has to be significantly faster than the competition to do the same work.

I suspect that Intel, over the last 10-12 years, has caught up with the overall quality of its processors because the top speed is changing little while their productivity is rising quickly.

So the bottom line is that clock speed is not an indicator of processing power. I'm tired of hearing stupid people talk about the Torch speed without talking about what matters. So, if you have no idea what you are talking about you should probably not speak.

That said, does anyone actually know anything about the overall quality of the processor?

[Mikey52]

Great post!.. And because I don't have a clue about processors, that's all i'm going to say!

[pirahnah3]

wow great write up. They should keep this around more in the explanation of things than just this thread...

[JohnMidnight]

Robotaz.
Congratulations on winning yourself a Top Tech, "I know the difference between 700mhz and 1ghz"!
I may not know the internal depth myself, more time is spent everyday simply troubleshooting, but I've always remembering my little AMD chip that could!

Posted from my CrackBerry at wapforums.crackberry.com

[zensen]

sounds like those extra chutes is using multi-chutes. i mean a multi core processor or at least very organised

your analogy did explain it well that more mhz isn't always about it being faster... amd were smart in trying to differentiate that with their AMD XP range of cpu's and only recently has the competition hit a point where mhz isn't the end all on desktop/notebook computers...

it's been a remarkably familiar trend this time round with mobile phones. I have no idea what this marvell chipset is like but the torch operating systems seems to handle the lower mhz well but would it have benefited greater as a multimedia device? Most definitely especially if the new chipset handles instructions better but again how much impact would it have been on battery life - since it is a portable device. would it have performed admirably better using a 1ghz processor? that i dont know but it is ignorant to think that the blackberry torch with OS6 struggles as a mobile phone with lower MHZ.

It's a shame there's not enough information out there but generally the negative response seems to highlight this more than what the actual phone can do as a phone with a physical keyboard/touch screen and how well RIM accomplish this

[JohnMidnight]

Zensen, we might not know, but as you can see, the OS is OPTIMIZED for the 642 Marvell. Going to a new processor, would also require the same Optimization.

Then again. We know RIM is capable of shooting themselves in the foot.

[Robotaz]

sounds like those extra chutes is using multi-chutes. i mean a multi core processor or at least very organised

your analogy did explain it well that more mhz isn't always about it being faster... amd were smart in trying to differentiate that with their AMD XP range of cpu's and only recently has the competition hit a point where mhz isn't the end all on desktop/notebook computers...

it's been a remarkably familiar trend this time round with mobile phones. I have no idea what this marvell chipset is like but the torch operating systems seems to handle the lower mhz well but would it have benefited greater as a multimedia device? Most definitely especially if the new chipset handles instructions better but again how much impact would it have been on battery life - since it is a portable device. would it have performed admirably better using a 1ghz processor? that i dont know but it is ignorant to think that the blackberry torch with OS6 struggles as a mobile phone with lower MHZ.

It's a shame there's not enough information out there but generally the negative response seems to highlight this more than what the actual phone can do as a phone with a physical keyboard/touch screen and how well RIM accomplish this

I'm talking about a single core processor, for the record. The chutes analogy refers to pins or buses where sophisticated processors put information automatically that simple processors do not. For example, Motorola used to make processors that automatically kicked out the "two's complement", which is a mathematical operation used by a processor to make a number negative or vice versa, on a side chute, but with an Intel you had to use extra clock cycles to compute the two's complement. The Motorola basically placed the two's complement, or negative version of what was crunched, out on one of the side chutes for the programmer to pick up without having to put it back in the input chute and crunch again to get negative. The Intel required the programmer to take the output, put it back into the processor and perform at least one more computation to get the two's complement of what originally came out. So the Intel took at least two cycles to do the same thing the Motorola did in one cycle.

Look, I don't claim to be a programming guru because I'm not. I took two microprocessor classes in engineering school and hated every second of both. This description of how processing power is not a function of clock speed is based on very basic principles of microprocessor design that anyone here would also know if you had studied this junk. I'm not an expert and don't claim to be, but I do hear some really stupid comments on here about the Torch's processor speed and couldn't resist commenting.

[Rickroller]

Great post..although i'll admit that most of it was way over my head. What has become evident to me in the last week or so, is that since the latest release of OS .214 the Torch has become what it was probably meant to be released as. Fast, snappy, non-laggy etc. Reading all the Torch reviews (which the majority were done within a few days of release) it was noted that certain parts of the OS lagged, or hesitated, and the browser also had these same issues. Most of the reviewers wrote this off due to the "slow" processor and basically stated that the Torch's processor couldn't handle the new OS6. But since .214, i think it's safe to say that in fact it's not the processor, but was rather some software issues. What i'd like to see (for RIM loyalists) is some updated reviews with the newer software, as well as maybe some new comparisons, because unfortunately i think the Torch got a bit of a bad rap in the beginning.

On a side note, i was reading some reviews of the latest and greatest Samsung Fascinate, which is sporting of course the same "top of the line" processors and whatnot of other Android phones (not bashing cuz i do like android), and it too was called laggy, choppy, etc..even with the 1GHz processor. Just goes to show that i think the majority of these issues are software related, not hardware.

[Robotaz]

Rickroller, I agree with your comment about most problems being software. It's quite possible that RIM winds up coming out with a 1 GHz processor that completely obliterates the competition, but just wanted to get the overall design of their new OSs fixed up before they kicked it into overdrive. As a BB nut, who's stuck with a stupid iPhone waiting for the Torch glitches to get fixed, I can only hope that RIM comes out heavy on the hardware now that they're working hard to revolutionize and fix their software.

[crester]

I agree with the great post remarks. You can't judge a phone only on the clock speed of its processor. It's about how the phone's software makes use of the hardware.

Posted from my CrackBerry at wapforums.crackberry.com

[xnwrx]

Well good post, but true and not true.
I'm working in electronic industry (not consumer) and used to �P, and if it's true that Motorola 68xxx processors were excellent �P (very good and efficient architecture and instruction set, pleasant to programm, very logical and straight forward) whilst Intel processors have always been crap (horrible architecture and instruction set), intel won the war because of strategy and marketing intelligence. That is unfortunatelly very common : the best technology is not often the winner (see RIM vs. iPhone vs. Droid today ;-) ). But at that time, we were talking of two completly different architectures.

Strictly talking about todays smartphone processors, they are all the same (talking about the CPU core) : ARM Cortex A8. That is Marvell PXA in RIM's phone, TI OMAP3xxx in Motorola's, Nokia, etc phones, Qualcomm snapdragon in HTC and others, Hummerbridge for Samsung's phones...etc. All are using the same CPU core licensed from ARM : the cortex A8 (Next generation for all will use Cortex A9 by the end of this year). Saying that, on-die implementation will vary from one chip manufacturer to the other, basically by the way peripherals are attached to the CPU core, type of peripherals, energy saving schemes...etc. But talking about CPU performance, all chip will perform more or less the same at the same frequency.
As you noted, RIM's OS6 implementation must be very clever as it doesn't give the impression of running on a 624 MHz CPU while Droid need the same processor running at 1GHz.

Now we should reasonnabily guess that next RIM's smartphones will go one step beyond as the Marvell PXA270 platform lacks of GPU (all other chip do have GPU for 2D/3D graphic acceleration) and lack of video CODEC (all other do have hardware H.264 CODEC accelerators). And that's what now should make the difference.

[Robotaz]

Interesting. I didn't realize that they're using the same CPUs. So is it a game of managing all of the board components that makes one maker's system more efficient than the other?

[infamyx]

Well good post, but true and not true.
I'm working in electronic industry (not consumer) and used to �P, and if it's true that Motorola 68xxx processors were excellent �P (very good and efficient architecture and instruction set, pleasant to programm, very logical and straight forward) whilst Intel processors have always been crap (horrible architecture and instruction set), intel won the war because of strategy and marketing intelligence. That is unfortunatelly very common : the best technology is not often the winner (see RIM vs. iPhone vs. Droid today ;-) ). But at that time, we were talking of two completly different architectures.

Strictly talking about todays smartphone processors, they are all the same (talking about the CPU core) : ARM Cortex A8. That is Marvell PXA in RIM's phone, TI OMAP3xxx in Motorola's, Nokia, etc phones, Qualcomm snapdragon in HTC and others, Hummerbridge for Samsung's phones...etc. All are using the same CPU core licensed from ARM : the cortex A8 (Next generation for all will use Cortex A9 by the end of this year). Saying that, on-die implementation will vary from one chip manufacturer to the other, basically by the way peripherals are attached to the CPU core, type of peripherals, energy saving schemes...etc. But talking about CPU performance, all chip will perform more or less the same at the same frequency.
As you noted, RIM's OS6 implementation must be very clever as it doesn't give the impression of running on a 624 MHz CPU while Droid need the same processor running at 1GHz.

Now we should reasonnabily guess that next RIM's smartphones will go one step beyond as the Marvell PXA270 platform lacks of GPU (all other chip do have GPU for 2D/3D graphic acceleration) and lack of video CODEC (all other do have hardware H.264 CODEC accelerators). And that's what now should make the difference.

The PXA930 (processor of the 9000/9700/9800) is not an A8 processor, Marvell does not have the license for it, only the instruction sets. The only processor license Marvell has is ARM9. The PXA930 is built off Marvells Sheeva PJ1 cores which is a hybrid of their Xscale days from Intel. Some clown on here a few weeks ago was adamant that the PXA930 outperformed A8 designs but didnt offer a shred of proof other than "trust me".

The PXA930 also lacks a GPU which is beyond hilarious for a touchscreen phone, but the fact RIM still uses its battery sucking 65nm core is even more hilarious since Marvell PXA935 has been out over a year now and has the 45nm core. One only needs to look at the iPhone 4 which dusts the 9800 in battery life in every regard, or the Droid X which has comparable-better battery life while packing much more powerful hardware and capabilities. Qualcomm has even revised the Snapdragon with a 45nm core to be better and cooler, and an enhanced Adreno GPU (x4 the current chip which pushes 20M polygons)

Anyways, OP is right, Mhz dont mean much when you run a different architecture but when your putting old chips running an older architecture than your competitors which also have faster clock speeds...well the differences are astounding.

[dchawk81]

Do we know for sure that the processor in BlackBerrys is actually one of these smarter, "elegant" designs though? Maybe it's slow AND stupid.

Personally, I don't care what's inside the phone. I care what it does and how it looks when I'm telling it what I want it to do.

My opinion is that it's neither fast enough nor smart enough for the fancy, fun apps - and the simple, hourglass inducing small not-so-overly-functional apps are just poorly coded. You can usually spot the latter because they'll make the phone lose memory like an Alzheimer's patient.

[xnwrx]

Oups!, I mentionned PXA270 instead of PXA930. But you're right, PXA93x core are Marvell's own implementation of ARM instruction set. One can say that the PXA930 has an equivalent performance/MHz than the Cortex A8.

We can also mentionned the Nokia N900 based on TI OMAP3 plateform (cortex A8 @ 600 MHz + DSP + CODEC), which while running Maemo OS is by far quicker than equivalent plateforms running Andro�d. Just to say that the OS and software coding make the difference. We are then not far from saying that Dro�d is the Windows dark side for mobiles...

IMHO, RIM faced a dilemna : provide for free a new OS to already OS6 capable plateforms (9700 and some others), but then no share return, or wait until they have an up to date hardware to launch the full package (OS + hardware) and get strong business returns. Their choice which is none of the above put them in a hazardous position.
That's enough now, RIM has used, reused and re-reused again the same hardware plateform since to long. The Torch is a great smartphone mixing the best of RIM's physical keyboard and touch screen, but they would have made a great and extraordinary business success by launching it with an up-to date SoC. Isn't it too late now...

[bjack56]

Great thread!

[Robotaz]

I'd like to hear a lot more about the architecture of the board and peripherals on this phone. Honestly, I don't know jack about current cell phone hardware. We need to get familiar with what we are dealing with so that the BB planet is more aware of where RIM should be as opposed to where it is. It's sounding to me like the available technology to phone makers is pretty standardized and streamlined. So, are we basically hoping that the Torch use parts that the iPhone has, or are we wishing for something far more advanced that nobody uses yet?

I'd like to hear from some people who really know about current cell phone hardware.