Log in

View Full Version : What is the holdup for a newer processor


John Collins
10-13-2003, 04:30 PM
I realize that this might be a stupid question, but why have we not seen a faster processor for the handheld market in close to a year. By my recollection, the 400Mhz processor was introduced in the Ipaq 3900 series, and that was well over a year ago.

Is there a technology block or is it that 400 Mhz is still more than can still be utilized by any handheld (which seems vey unlikely). It just seems to me that up until we hit the 400Mhz limit, the processors were increasing every few months and now, nothing for the last year or so. Just curious to see what you think.

Jason Dunn
10-13-2003, 04:42 PM
It does seem strange, doesn't it? Intel is known for never sitting still when it comes to the Mhz race, yet you're right, here we sit at 400 Mhz more than a year later. I wonder if the biggest factor is a lack of competition? When Samsung releases their 533 Mhz ARM processor (I believe I heard it was "coming"), perhaps that will get Intel moving...?

Duncan
10-13-2003, 04:58 PM
Well - perhaps, since the processing needs of the PPC software have not increased, Intel sees no need to concentrate on faster processors but instead is focussing on more efficient processors e.g. PXA250 -> PXA255 ->Bulverde...?

HailFire
10-13-2003, 06:04 PM
Only one thing is going to drive the demand for faster processors - GAMES! Business professionals would rather get longer battery life than faster refresh rates. As soon as the world realizes that PocketPC is the ideal gaming platform, Intel will have to start applying Moore's Law to mobile processors. As a game developer, I think it's going to take an equivalent to id's DOOM to spark this revolution, and I know many people are working on one.

IMHO, if you want faster processors, buy more games!

PetiteFlower
10-13-2003, 06:26 PM
Faster processors use more battery power too....

Janak Parekh
10-13-2003, 07:01 PM
Is there a technology block or is it that 400 Mhz is still more than can still be utilized by any handheld (which seems vey unlikely). It just seems to me that up until we hit the 400Mhz limit, the processors were increasing every few months and now, nothing for the last year or so. Just curious to see what you think.
Well, apart from speculation as to Intel's motives or non-motives, it's hard to make power-efficient processors faster. Processor design is rather substantially complicated. You can't just increase the clock rate and expect the existing modules to work faster. Generally, to make a processor faster, you generally have to make it "deeper", e.g., more pipeline stages that might work in parallel. However, not only does it make the logic more difficult (branch prediction, etc.), but this involves more transistors and more power consumption.

This is one of the main differences between each major generational CPU. The Pentium 4 uses 20 pipeline stages, I believe, compared to 10 in the Pentium III. It's amazing that it actually works so well, but as a result the P4 is quite a bit more power-hungry (note the supplemental power ports on P4 mainboards!).

If you're curious about the lower-level detail on how all this stuff works, you might find the following articles informative.

http://arstechnica.com/paedia/c/cpu/part-1/cpu1-1.html
http://arstechnica.com/paedia/c/cpu/part-2/cpu2-1.html

--janak

maximus
10-14-2003, 01:57 AM
And there is also the economic factor ... why waste so much R&D expense and brainpower on something that is only marginally profitable ? They would rather concentrate their brightest engineers on the upcoming Pentium 64 and Xeon 64, that will generate more than $200 margin a piece, rather than wasting it on a chip that will only generate $5 a piece.

pootp
10-14-2003, 03:10 AM
i'm sure they're working on something and it may indeed be something very different after all this time without change.

cyclwestks
10-14-2003, 02:28 PM
I could go through my long line of PPC's, but to cut a long story short, the fastest I've owned yet is the 1940 with the Samsung. I don't understand why other mfg's aren't using it? I keep hearing talk of XScale optimization, & when or if this happens maybe everyone with the Samsung processor will be left out?

JustinGTP
10-15-2003, 02:46 AM
Faster processors use more battery power too....

Not exactly, the Intel Centrino technology is reaching higher speeds and takes up less battery power.

Saying this, they need to make a medium between the two, the best of both worlds when it comes to speed and battery usage.

-Justin.

Janak Parekh
10-16-2003, 07:22 PM
Not exactly, the Intel Centrino technology is reaching higher speeds and takes up less battery power.
That's because it was cleverly engineered -- a Centrino is actually a P3-M core with lots of cache. However, there's a limit as to how far you can go with that strategy. And in any case, the Centrino takes up less power than, say, a P4-M, but takes a lot more power than an XScale chip. The x86 and ARM cores are totally, utterly different.

--janak

JustinGTP
10-16-2003, 08:15 PM
Still, they can expand on the concept. That is what technology is always about, how much better can we get?

Jason Dunn
10-16-2003, 10:56 PM
I could go through my long line of PPC's, but to cut a long story short, the fastest I've owned yet is the 1940 with the Samsung. I don't understand why other mfg's aren't using it? I keep hearing talk of XScale optimization, & when or if this happens maybe everyone with the Samsung processor will be left out?

The Samsung processor is AWESOME on a Mhz per Mhz basis versus the Xscale. As to the Xscale optimization, that's just it, there isn't much of any - Microsoft optimizes for the instruction set (ARM4, ARM5) and not the specific processor brand (XScale). That's a very good thing.

Sheynk
10-17-2003, 12:19 AM
Is there a technology block or is it that 400 Mhz is still more than can still be utilized by any handheld (which seems vey unlikely). It just seems to me that up until we hit the 400Mhz limit, the processors were increasing every few months and now, nothing for the last year or so. Just curious to see what you think.
Well, apart from speculation as to Intel's motives or non-motives, it's hard to make power-efficient processors faster. Processor design is rather substantially complicated. You can't just increase the clock rate and expect the existing modules to work faster. Generally, to make a processor faster, you generally have to make it "deeper", e.g., more pipeline stages that might work in parallel. However, not only does it make the logic more difficult (branch prediction, etc.), but this involves more transistors and more power consumption.

This is one of the main differences between each major generational CPU. The Pentium 4 uses 20 pipeline stages, I believe, compared to 10 in the Pentium III. It's amazing that it actually works so well, but as a result the P4 is quite a bit more power-hungry (note the supplemental power ports on P4 mainboards!).

If you're curious about the lower-level detail on how all this stuff works, you might find the following articles informative.

http://arstechnica.com/paedia/c/cpu/part-1/cpu1-1.html
http://arstechnica.com/paedia/c/cpu/part-2/cpu2-1.html

--janak

Janak brings sense to the world!

Would building an "xscale" with more cache provide a good temporary solution?

Janak Parekh
10-17-2003, 03:28 AM
It's a good question, isn't it? Unfortunately, I don't know the hardware aspects of ARM that well -- as to how more cache would affect memory loads or instruction misses -- because I don't know the detailed performance profile of most implementations of an ARMv4 or ARMv5 core. When it boils down to it, I'm more of a software guy than a hardware guy.

--janak