Sign in with
Sign up | Sign in
Your question

In Depth Review on P4 .................

Last response: in CPUs
a b à CPUs
December 30, 2000 12:04:56 PM

Please read this article and and comment....

More about : depth review

a b à CPUs
December 30, 2000 3:34:53 PM

I think Craig Barrett accidentally ran over that guys dog!! He is one angry fellow! In terms of his review... what a crock of... rubbish. The P4 is better at some things and worse at others. As software is optimized (contrary to the article, there are C and Fortran compilers ready to take advantage of the P4 SSE2 instructions) the P4 will pull away in more applications. I don't think it ever will in office apps and the like. It isn't meant to. Let's face it, the operator is the weak link in office applications, not the cpu (how fast can you type?). Lastly, I have a couple of points to make regarding what he thinks of Intel's future. First, why isn't AMD getting bashed now that DDR memory is more expensive than RDRAM? Why aren't they getting bashed when they can't deliver a DDR platform on time? Let's face it, the world loves an underdog. My last thought on the issue is looking forward to next year. I have a big question for everyone looking forward to the Palomino... Why are they dumping copper interconnects for Aluminum? They already went through the trouble of setting their fabs for copper, now they take a step backwards? That is going to hurt their speed. There are only two possibilities: 1) the Palomino changes to the T-bird core are so good they can take a speed hit with no worries 2) they have a manufacturing problem they couldn't resolve with copper and were forced to shoot themselves in the foot. Competition is good for everyone and neither company is likely to go anywhere next year (or the year after).
December 30, 2000 3:50:02 PM

Not only is DDR expensive (but that may change in the near future), it also doesn't deliver as expected.
Related resources
a b à CPUs
December 30, 2000 6:56:26 PM

Rambus has had over a year to prove itself, and it hasn't performed near what was expected. DDR has been few select computers for about two months on a few select platforms. Hasn't it performed very well in video cards?!?! The price will fall sometime in February, when large shipments begin. It will be cheaper than Rambus because it doesn't have as high of production costs as RIMMS. The cost right now is due to economics: there is a high demand and very little supply. The resellers would be fools not to make a grip on what they have...remember in America, we have that Captalism thing.
In regards to AMD delaying their chipset, who cares...I want the best product possible not a rushed incomplete product. This goes for Intel too, although I would never buy there
As for the article on emulators, I am not a machine language programer, and neither of you said you were. But as I understand, and as it was reinterated by the author of the article on emulators, the people that know the true potential of a processor and the processors flaws are the machine language programers, they don't deal will the pretty side of programming. Yes there was a slight amount of bias in the article (AMD's stock is trading at 13 13/16 which was not mentioned) but the purpose of the article was not so much to build up AMD (he would have spend more time explaining why the T-Bird was so great if that was his purpose) but the purpose was to point out the glaring flaws and technological degression in the P4.
Anyways, please don't take anything I said too seriously, (No hate mail please, save that for DRPC, hehe). I am just happy that we can have this discussion.


Take your PILL, and get some sleep.
a b à CPUs
December 31, 2000 5:45:41 AM

I think all the Intel problem is they didnt anticipated that AMD would come out so fast (with Athlon and Duron). All the while they have been dominating the market, so they just relax and enjoy high profit. When AMD comes out so strong (cheap and fast CPUs) suddenly, they just panic as wasnt expected that so they just rush thing out (PIII 1.13 recall etc.). They also have to crippled the P4 as to reduce cost (maintain high profit) as they cannot afford to lower their profit margin as it would affect their stocks. Intel is a great company in the past (I have used many Intel CPUs from 8086 to PIII), but they have been too relax and overlook the threat from AMD.

As for the pricing Intel cant afford to lower their profit margin (I think this would be the major problem Intel would face). As lower the profit margin would created a chain reaction in their whole organisation. As for AMD the story is totally different, their shareholder didnt expect high profit margin, their organistion is not as big as Intel (more flexible and less overheads), they spend less on advertising. Therefore AMD can "Afford" to sell cheaper.

The future will be interesting as to how Intel comes out something that is Cheap and Good to hold AMD back. While we as a enduser will able to enjoy something cheaper as the result of the competition.
December 31, 2000 8:19:50 AM

DDR performs well on video cards indeed, but the same kind of speed improvement is not seen on DDR mainboards.

...and of course prices will go down when the memory becomes main stream.
December 31, 2000 4:56:47 PM

Have you seen the benchmarks with the new DDR motherboards? Obviously not. They DO make a difference.
a b à CPUs
December 31, 2000 5:03:51 PM

So a couple of things in reply to the messages above. First, writing code in assembly is done to optimize code as much as possible. So, when you get new instructions and architectures, naturally you get changes in how the assembly will perform. The difference with the P4 (as opposed to the Athlon) is that Intel made a choice to abandon the older accepted norms in favor of something radically new. People like the guy who wrote the article are going to be pissed because it amounts to 15 years of work that won't run on the new processor (last change of this type was 1986). That would tick me off too! However, progress requires change. New assembly code can be written which will perform just as well or better in some cases when compared to other processors. People just need to be clever.

Second, in regards to crippling the P4 to get it to market, what did you expect? It wasn't a choice of getting it to the market fast enough. It was a choice of die size!!! It is already very large (being designed with the 0.13 micron and smaller processing in mind). Now add in more cache (I forget the original design - more L1 or 2) and include many of the sections hacked out to make room. It is my feeling that those have not been removed for good. They will reappear when the process goes to 0.13 micron. Some changes will appear in the Northwood, but expect the cream of the crop to show up in Foster.

Third, DDR memory is good. It doesn't have the bandwidth of RDRAM and never will! However, it doesn't have the latency either. The P4 set up manages to get around some of those issues with more than one channel going to the RIMMS (like the 840). The bottom line is how fast can you clock it?! The memory that is able to clock the fastest will win. I think you will find that RDRAM will take that crown (pure speculation). A true test will come when we have the same Via motherboard (using an Intel chipset) that can support both memory types. I am looking forward to that comparison.

I still haven't heard any thoughts on Palomino. It doesn't matter what people think of the P4 if no decent competition is forthcoming from AMD. The T-bird core is just too hot to be clocked any further. Why do you think we haven't seen a 1.3 GHz Athlon? They can't make them in volume, that is why. All the speed path corrections in the world aren't going to lower the power generated by that hot blooded beast. Enter the Palomino... but I think their manufacturing is failing them. They have good ideas, but if you can't make it manufacturable, then you have nothing. They got their current copper process from Motorola, who can they turn to to get the processing technology they will need to advance further? They don't have the in-house expertise to make it happen on their own. However, practice makes perfect.

a b à CPUs
December 31, 2000 8:30:01 PM

"They got their current copper process from Motorola, who can they turn to to get the processing technology they will need to advance further?"

Well lets see the .13 fabrication process that "may" be used by AMD will come partly from IBM. IBM's .13 micron process in conjunction with IBM's SOI technology should be more than enough to combat anything intel has @ the .13 fabrication process.

Throw in current copper technology from motorola, a redesigned Althlon(20-30% reduction in heat- if reports are accurate), and AMD may be using pure silicon wafers to reduce heat even further.

From the looks of it, AMD has a lot of technology to choose from. They can cherry pick the best . Throw in the fact that AMD can sell a processor for a lower asp than intel and still make a huge profit. AMD is looking pretty good methinks.
a b à CPUs
January 1, 2001 6:15:09 PM

A quick heads up for you: AMD is dropping the copper interconnects on their backend for the Palomino! NO COPPER! It will be aluminum. I was hoping someone could tell me why they switched.

In terms of using isotopically pure Silicon, I wouldn't hold my breath. They claimed a 60% increase in heat transfer properties. Think about it. A 3% change in isotope content (97% Si-28 versus 3% Si-29/30). They said phonon scattering... whatever. The crystal structure will be subtly different, but that won't account for a 60% increase in heat transfer out of the processor. To make matters worse, the company that was going to manufacture that silicon type has announced a 3 or 4 month delay in delivery... bye, bye for that idea on the Palomino.

Then you said they were reducing heat by 20-30% on the Palomino. That is a HUGE chunk, especially considering the Palomino isn't starting out on a 0.13 micron process as far as I know and is using Aluminum interconnects. However, it can probably be done or they wouldn't make the claim. Given that... a 30% reduction of 70 watts is still 50 watts. How many times can you up the speed before they are back at 70?

I am not trying to bury the Palomino right up front, but like I said earlier: neither Intel nor AMD are going anywhere anytime soon. They will trade punches through next year and we the consumers will benefit from faster processors cheaper.

Happy New Year!
January 1, 2001 8:26:32 PM

hahaha! happy new year to all amd puppies, who are you going to turn to now for your food,...Cyrix?
a b à CPUs
January 1, 2001 11:14:20 PM

I did see where AMD demonstrated a Palomino with aluminum interconnects but I have never seen anything implying that production Pally's would revert back to it. Since all Athlon's come from Dresden where the copper process is the only process used, why do you keep asserting that aluminum will be used? Do you realize what an undertaking it is to convert back to aluminum? Is AMD done with copper? I don't think so. Can you please provide a link in which AMD states Pally will be aluminum?
a b à CPUs
January 1, 2001 11:16:45 PM

IntelMeltdown, the only thing worse than going to Cyrix is going to Intel. Now wouldn't that suck?
January 1, 2001 11:27:20 PM

Interesting theories there- but like was said dresden fab is only copper. To go back would cost way more than they'd save sticking with aluminum. Also maybe you don't understand crystallin structures as well as you think you do. A little contamination can go a long way in the travel of electrons and light. If they could get a pure silicon then the friction of the electrons would be less because the flow would be more streamlined. It's like a river, throw a few huge bolders in there, still only accounting for a small bit in the whole scheme of the river (say 3%) and you have turbulent water and white foam and water backup. Take those few away and the whole river smoothes out, speeds up, and makes less odd currents all over since one current flux makes many more in the water and electrons are the same. 3% doesn't equal 60% is not a good argument when dealing with this because it isn't a simple trade. If they really do have that pure silicon, then 60% heat reduction is very probable. I would like to see intel get this silicon too- just a matter of time I'm sure, but don't underestimate it- It's a hell of an accomplishment.

"Are you saying that I can dodge bullets?"
January 2, 2001 12:57:45 AM

yeah, you amd puppies will always go for the underdog, very admirable.
a b à CPUs
January 2, 2001 1:20:22 AM

Duck I believe Jg38141 has rebuttled your view about pure silicon.
This is the first time I have heard that AMD is going to be using Aluminum Interconnects instead of copper. Connie has a valid question, that I would like to see answered.

As far as AMDMELTDOWN goes: it's not about rooting for the underdog it's about rooting for the future winner(AMD) we are concerned with not the future loser(INTEL). You are totally entertaining. Nothing like comic relief on a message board.
January 2, 2001 1:30:01 AM

Now I started with AMD and love the underdog myself. But most of all I want a stable system. When is AMD and VIA going to make a stable sys? I hope the new Intels don't go tis route. I am seriously thinking about an Intel sys (some year down the road) if it comes to stability.

I was also pleasantly surprised by the RDRAM performance. I have been biased towards Rambus and all their rants and attacks but looking at the performance it can offer as shown with the PIV. The smuttering of an advance with DDR (so far) is not too attractive. I hope I'm wrong here with the future of DDR as I was with RDRam.

<b> Fragg at will!!! </b>
January 2, 2001 2:30:54 AM


I was also pleasantly surprised by the RDRAM performance. I have been biased towards Rambus and all their rants and attacks but looking at the performance it can offer as shown with the PIV. The smuttering of an advance with DDR (so far) is not too attractive. I hope I'm wrong here with the future of DDR as I was with RDRam.

Umm ok.... Doesn't DDR perform way better than RD? Just look at the benchmarks! (And prices)
January 2, 2001 5:41:09 AM

oh, yeah I read the article about fortune magazine. Thanks for the linky.
January 2, 2001 8:49:57 AM

The bandwidth of the RDRAM looks as if it can do much more than DDR. yadaa yadaa yadaa . . . . .

Hey check this out --> Now if your a gamer and have the best video card you must play at a min. of 1024x768 16mil colors. Right? Prob. 32mil colors? Well who needs a CPU over 500 MHz? It will match a 1.0 gig and higher!!! Now how many of youdo other processor intense apps?

Funny as hell ---> Heard a guy in one of these rooms talk about getting his sys to boot in 17 secs. Hmmmm Disabled all drivers for vid, audio and down the line. And the point is???? I can get into dos in 6 secs. :o P

<b> Fragg at will!!! </b>
January 2, 2001 4:37:17 PM


But I can get my computer to boot in 24 seconds - drivers, everything is normal. It kicks :) 
a b à CPUs
January 2, 2001 7:03:18 PM

You are correct Connie, the article refers to the test Palomino's. However, why would they develop the process on Aluminum when it is planned for Copper? That is a big change and not something that is easily done late in the game. Presumably you people know something about manufacturing and realize there are too many differences to just switch over. My main point is if they planned to use copper why on Earth would they bother to deveop the test chips on Aluminum?? It just doesn't make sense. I am willing to bet the farm that the first Palomino processors will be Aluminum. Finally, as I have also said earlier, if they do go with Aluminum it is because the Palomino design is that good, or their copper process is sucking wind and they were forced to. Who knows which?

For those of you interested, the link is below:;P ID="edit"><FONT SIZE=-1><EM>Edited by DuckDodger on 01/02/01 01:27 PM.</EM></FONT></P>
a b à CPUs
January 2, 2001 7:12:38 PM

"Interesting theories there- but like was said dresden fab is only copper. To go back would cost way more than they'd save sticking with aluminum." - see my reply to Connie jg38141

As to the rest of your statements about isotopically pure silicon... you are off the mark on impurities. This isn't light OR ELECTRON travel we are talking about here. Heat transfer is done almost entirely by molecular/atomic vibrations. Remember, we are not talking about 3% of some foreign material. We are talking about 3% of a material with one or two more neutrons, the same number of electrons and incredibly similar properties (including crystal structures, lattice constants, etc.). In addition, there are other factors AMD is going to have to work through to make this work. I don't doubt it can improve heat transfer properties (the more regular the lattice, the better it should be), but I doubt that quote of 60%. We won't know for sure until we see the numbers. In addition, I agree that switching to that type of silicon is not as straight forward as you might think. Even slight changes in crystal structure can cause major manufacturing headaches resulting from adhesion and stress issues.
a b à CPUs
January 2, 2001 7:17:40 PM

Grizzly - I don't think you can blame RDRAM for issues associated with the P4's new architecture. The proof is in the pudding as they say. As I mentioned before, I can't wait to see RDRAM and DDR memory on the same chipset with the same processor.
January 3, 2001 5:01:14 AM

Ahh, I see your point and understand why you are scepticle about the 60%. We'll find out eventually I guess. As for the aluminum chip though- I'm pretty sure they are developed in Austin. That would mean they have some smaller machine that can make developmental chips on a limited basis. So the models are aluminum because they haven't gone to the german fab and set up the dies for the palamino on copper. The process isn't substantially different for copper or aluminum, it's just expensive to buy all new machines for one or the other, but what can be made right now with copper can be made with aluminum as well, albeit not as efficient a chip. But anyway- Development labs for palamino are in Austin so they figure it out with what they have most easily available and if they can get it to work with aluminum they know they can with copper so they go to germany and impliment it there. The athlon/ thunderbird started out the same way and then moved to copper. Some athlons are still made in austin with aluminum.

"Are you saying that I can dodge bullets?"
a b à CPUs
January 4, 2001 3:07:52 AM

Hmmm... if the Palomino can hit 1.5Ghz on Aluminum then it is going to scream on copper. I am looking forward to this year (hands rubbing together), even if my stocks do go to hell.
January 4, 2001 4:08:28 PM

LOL, I'm still waiting for AMD to hit $12 or $13 a share!