Damnation the new Pentium-M looks great

I was reading the newest issue of PC Magazine today, and they have some tests of the new Pentium-M based units. Impressive to say the least. Although the Winstone tests aren't exactly what we care about, they do illustrate how much potential this architecture has.

Acer TM 803LCi- Pentium-M 1.6GHz
Business Winstone 2002-31.9

Dell Inspirion 8500- P4-M 2.4GHz
Business Winstone 2002-23.2

Winbook J4 732 3.06- desktop P4 3.06
Business Winstone 2002-26.6

I was not surprised the P-M beat the P4-M like a rented mule, as the P4-M has always been underwhelming in the extreme. The Winbook however uses a normal desktop P4 3.06, and fared little better. P-M's are not going to outperform desktop P4's in all respects as badly as they do in this one performance metric, but it does point to the extremely efficienct architecture and the excellence of the new i855 supporting chipset. Sure would be nice to have a desktop board we could run one these puppies on...

what is the archecture of the pent-m?

is it similar to p3 or p4 or neither?

what is fsb, is it quad pumped?

dustybyrd said:

what is the archecture of the pent-m?

is it similar to p3 or p4 or neither?

what is fsb, is it quad pumped?

Click to expand...

The Pentium-M is based on the P6 architecture and I believe it uses a 100mhz quad pumped fsb so it's like a cross between a PIV and a P6

Its is a 686 hacked to work on the Netbus-architecture.
so yes it is quad pumped as all Netbus-architecture CPU's are.

that's why that chip rocks---

p3 architecture with p4's memory bandwidth

that's what intel should have done all along....

and that's a good chip...

i hate the crappy p4 efficiency...but love the memory bandwidth

pentium-m (banias) rule

... and have 1MB L2 cache!

this thing is making me actually want a laptop, (never really appealed till now) btw, dont get larva started on high cache levels but the successor to Banias, the Dothan will have 2MB L2 Cache

Heh, don't get me wrong, large caches are a good thing, my only point was the desktop P4 is well into the area of diminishing returns past 512KB for most applications. The architectural advantages of the Pentium-M are numerous, it is not all the fault of the cache size. Mainly though the large L2 reduces main memory accesses promoting battery life and allows good performance at low clock speeds. If notebooks had something like a dual channel PC1066 or dual channel 333+MHz DDR memory subsystem the effect on performance would be much less.

hey larva,

what do you think the Ghz limit on the 130nm p4 chip will be?

I say unless they switch to 90nm or a completely new architecture (like going from p3 to p4)---then Intel will not sell a chip past 3.6 gigs---and that will require new "D" stepping----

I say they don't go past 3.2--maybe---with the p4 C1 stepping...

It's really hard to say. These new multi-vid 2.4b and 2.53's are hitting 3.4-3.5GHz pretty regularly, but it's just guessing as to how far Intel rides the technology. A lot of has to do when the 90nm process is ready to go. Chips get cheaper to produce as you shrink the process, so the 90nm fabrication technology will be used as soon as it is viable regardless of the MHz situation.

Does anyone think the pentium m will make it to the desktop ever? I love the late model p3s, they are amazing cpus, if they ever bring the m type chip to the destkop, I will get one. Problem is its probably on a different socket etc.. Darn maybe I just have to get a pentium III s.

First of all, I saw someone mention the diminishing returns of increasing the cache size on the P4. I don't know why this was stated as the increase from 256KB to 512KB of L2 cache brought about quite a bit of increase in speed in a number of applications. Sometimes up to 50%+ in 3D rendering scenes.

Of course, how much the 1 MB of L2 cache helps on the Pentium-M architecture, we'll never know, it probably isn't as cache-dependent as the P4 design.

As for the benefits of 1 MB of L2 cache on the P7 core design, we'll know once Prescott comes out.

And as far as the scalability of the P4, Intel brought it to 4.1 GHz at IDF a while back (6 or so months I think). Usually, their super-cooled overclocks are a good indication of future processor releases. Of course, you'll probably run into thermal/power limitations before you'll run into architectural limitations (there's a first actually, previous processors had always hit a brick wall architecturally long before they run into thermal limitations).

Once again, I will elaborate. Look at the gains in going from 128KB to 256KB L2, then at the gains going from 256KB to 512KB. There is a drastic difference. Moving to 1MB will again produce a lower level of improvement. This is the definition of the law of diminishing returns.

Obviously if you are looking at a data procesing task that employs data object that would fit in a 512KB cache but not in a 256KB one, the difference in performance between the two will be drastic. This does not however characterize the overall difference in effectiveness of the two chips. I have stated repeatadly, certain specific tasks can and will improve markedly as the cache size is increased. The majority of tasks however, are experiencing notably diminishing returns past the 512KB level as evidenced by the markedly lesser increase present as we increase the cache from 256KB to 512KB as compared to going from 128KB to 256.

Sweet chip! I wish I'd had one of those. I wonder how much it overclocks?

I knew there was a reason I was holding off buying a new laptop.

It would be fun to overclock one in a desktop, but somehow I doubt that'll ever happen.

- JW

they will not overclock much, as the shorter pipeline makes attaining higher clock speeds more difficult, and makes the MHz attainable (on a certain process) a hard barrier (the speculation is about 2Ghz for the P-M on the .13 micron process). Additionally, the P-M was designed so all the tranistors run at the same speed, rather than many running faster (like in the P4) which further limits the max speed of the chip (its harder to increase the speed of 1 transitor than 1million)
________
No2 vaporizer