I am really bored of x86 arhitecture. Having a lot of Mhz is just marketing, it is NOT a good arhitecture. I wish to see something new - perhaps something like the Power4. It has great performances. It is from IBM; personally I have more trust in IBM that in Intel. If Apple is going to switch to Power4, well I am in for it!!
The Power4 is a great part. Lower clocked Itanium2's have better FP #'s. And I don't see how you trust IBM over Intel. Especially since their Hard drives went from best reputation to worst in a two year period. They've got a solid server reputation.....maybe that's where you're coming from. But in that case, Intel isnt in their league yet. Theres no comparison w/ high end servers since Intel is now trying to break in.
true true... but you do remember going from P3 to the first P4 too right? every manufacturer has their ups and downs. and ibm IS the biggest influence as to where we are now with PCs. and i don't think their hard drives have much to do with the cpu.
<b><font color=orange>*my favorite bum on 21st street*
-please help-
will code HTML for food.
-god bless you-
I was sure that somebody would say something about the IBM's HDD division. That is true; they have done a really bad job with 60Gxp and 75Gxp. That has changed lately - the 120Gxp line seems to be pretty good. It seems to be pretty reliable and they are damn fast. I have a 120Gxp and I am happy with it. Well, for now anyway.
About HDD division you are right. I am aware of this problem, but besides this problem I am not aware of other dirty things done by IBM. Now let's go on to Intel: I think that one can write a book about their ugly things that they have done only in the last year. If you don't agree with me on this point I will give you the list - but tomorrow morning. It is already 1:36 o'clock in France so I have to go to sleep.
Why I trust IBM: OS2. No other company would sustain an operating system for such a long time. And that is not all: they have done a pretty good job with OS2. OS2 has the stability of XP for more that 10 years.
Also I like the IBM's policy: they say that their products are not going to be influenced by current public opinion/reaction. Any product launched by IBM in backed up for many years; only if it will fail then - it will be retracted. That means they provide very stabile solutions.
Let's think about Intel: some 6 months ago they where all RAMBUS. Rambus all the way. Some 3 weeks ago they said - we are going DDR. We will not develop any Rambus solution. The clients where puzzled - those who have invested a lot of money in Rambus. After the client's pressure they said that i850 will be further developed at least for RDRAM1066. That is Intel: going with the wind...
>They've got a solid server reputation.....maybe that's where you're coming from.
True. One personal matter: I have an IBM keyboard aged to more that 10 years. It weights more that 3 kilo. Yet it is working like new. I also have an IBM mouse - the same age. Still working very well. This computer is of no use today - of course!! - but when it was useful it was not spared. I don't know many companies that can do such good - quality items.
Also, as a student I remember that all the stations from the labs where IBM. I don't need to tell you what treatment where receiving from students. The stations where working 16-20-24h/day for years. Besides the floppy (which where always dead) everything was OK. I must say that I was pretty impressed.
>and i don't think their hard drives have much to do with the cpu.
True. They are separate divisions; almost like distinct companies. The bureaucracy is a big problem in IBM - partially because of its size and also because the divisions are very independent.
The present PC is a product of IBM's bureaucracy. More on that tomorrow.
>>>Now let's go on to Intel: I think that one can write a book about their ugly things that they have done only in the last year. If you don't agree with me on this point I will give you the list
Ok Raz, I work for Intel, and have for the past year. I'd like to see the list of things we've done that are so awful, because I don't recall going to work thinking what awful things I was going to help do that day. I thought I was more observant than that.
>>>Let's think about Intel: some 6 months ago they where all RAMBUS. Rambus all the way. Some 3 weeks ago they said - we are going DDR. We will not develop any Rambus solution. The clients where puzzled - those who have invested a lot of money in Rambus. After the client's pressure they said that i850 will be further developed at least for RDRAM1066. That is Intel: going with the wind...
Hmmm, as I recall practically <b>everyone</b> here on Tom's was also verbally reaming Intel for trying to push Rambus - despite the fact that ALL the benches showed it performing better. Now as soon as a company tries to give people what they want, you see what happens? How quickly they turn....
I thought IBM sold their HD division. I read this in another thread somewhere in this forum, but then again why are they comming out with new hard drives(180gxp). Someone clear this up for me plz.
"If you sign up for AOL now, we'll give you 1024 hours of slow service and disconnects free of charge"
Well, the RAMBUS... I think it was priced to high compared to the DDR... The Rambus royalties and all... It didn't smell right, with Intel really pushing it. I can see Intel not wanting to waste money on developing a DDR solution with the Rambus, which might of hurt the rambus memory sales.
I love Intel CPUs. The Rambus thing 6-8 months ago caused me to go to AMD because it was like 300-400 bucks cheaper due to the higher memory cost and the slight slight cpu cost added together.
Well, 75% of all the private sector has done illegal actions. So hard to say which company is actually "clean" and which one is an ex-con.
Remember Nvidia and 3DFX lawsuits. (if i remember correctly). Nvidia had better laywers and more money, heh. We still love Nvidia (well 75% of us).
A 5 foot work fan, plastic sheeting for the funnel, some tubing, and duct tape is all you need for a cooling system.
I'd like to see the list of things we've done that are so awful, because I don't recall going to work thinking what awful things I was going to help do that day. I thought I was more observant than that.
You didn't do anything wrong. I think you know that very well. You are just trying to protect Intel because you are working there. Saying that you have done anything wrong is like saying that all the people that live in Iraq are cruel criminals - just because Sadam is a criminal.
Again: I have nothing against Intel employee. I try to be rational. How could I have something against Intel employee?! IMO Intel's employees (especially the highly trained engineer) have created Intel. I have respect for them. I think that anybody would be proud to write in his own CV that he was worked for Intel. I assumed the problem is clear: nothing against the hard - working Intel engineer.
Let's see what I am remembering now - things that have disappointed me:
1. The P3 1,13Ghz launches; a total failure. It was not working. Some failure of this type from time to time is unavoidable, so why I am disappointed? Well, some 6-8 months before this paper launch Intel was making a prediction about the future in which they have compared a 650Mhz Athlon with a 1Ghz P3. They have assumed that the competition would not be capable of bringing anything against P3 1Ghz - AND THE COMPETITION was there first (Athlon was the first proc to reach 1Ghz). This is only happening because those who are managing Intel now are too proud; too proud to realize that the competition has good engineers too. Too proud to realize that in the last 2-3 years Intel has lost some of its image - image that was created by other managerial teams long before they get to work for Intel.
Here I would like to mention Intel's attitude with regard to this site. When Tom had discovered that P1.13 GHz was not stable he informed Intel on this matter and asked them to clarify the situation, because otherwise he would write about it. Intel's reaction was something like:
<b>
if you have nothing better to do, then go ahead and write your article, because we are so BIGGGG anyway and we don't care. Loosing some customers is no big deal. They WILL come back because they have no option.
</b>
2. remember the bug that was found in Intel's infamous 'i820' or 'Camino' chipset?
Read this: (I will have nothing more to say on this matter)
Bad Information Policy = Official Intel Policy?
Since the MTH-announcement was made and the press releases went to the press, Intel has been trying to do a nice job to look good officially. We, the press people, know that Intel is planning to handle it this way: "The replacement option will include an Intel(r) VC820 Desktop Board and 128MB of RDRAM. Details of replacement availability will be communicated to your place of purchase in the near future." We have also been assured that every owner of an MTH- motherboard will get a replacement or refund if he should want that. You, the readers, read a different information at Intel's website "Intel Corporation today announced that it would replace motherboards that have a defective memory translator hub (MTH) component that translates signals from SDRAM memory to the Intel(r) 820 Chipset." How can you find out if your MTH is one of the 'defective' ones? Well, you basically can't, and you wouldn't have to, because in fact every owner of an MTH motherboard is entitled to a replacement or refund, even though Intel's official announcement sounds a bit different. I got tons of mails from upset owners of Intel CC820 motherboards, who told me that Intel-officials deny the replacement or refund categorically. The answers of those official Intel-spokespeople go from "We won't replace any memory" to "Only affected CC820 boards, which is about 15-20% of them, get replaced". Both statements are simply WRONG. Now if those kinds of comments were the sad exception, I wouldn't really say anything.
3. the P4 high frequency: tell me honestly: because you are working for Intel you may answer this question:
Who decided that P4 would have to reach high MHz numbers? The engineers decided that it is better to create an architecture with low IPC and high frequency, or the marketing department? Really, who decided that?
The latest proc from Intel - Itanium hasn't the same architecture (low IPC, high freq). Well... WHY? If this is the future: low IPC, high freq then WHY?
Well, it is enough for now. If you want to read more on the subject then go ahead and read my article from my web page:
despite the fact that ALL the benches showed it performing better.
?!! Rambus is performing better on P4. On P3 this was NOT true. I repeat: it was NOT true. Search this site's archive and get convinced. Rambus was not worth your money on a P3 platform because it was SLOWER that the SDRAM solution. There are many explanations for this behavior, but this is not the point. The point is that it was slower and more expensive, yet Intel claimed the OPPOSITE. What do you think?
Intel was using Rambus long before P4 was launched. Don't forget that.
Quote :
Now as soon as a company tries to give people what they want, you see what happens? How quickly they turn....
Are you suggesting that Intel did that because they care about people's opinions? I'm sorry for this expression, but: quit the crap. They did it because they where loosing money; they did it because their evil monopoly that they tried to create with Rambus was a failure. If they would have succeeded in their intentions and Rambus would be all over the place WE (you, me and all the other comp users) would have to pay royalties to this company whenever we wanted or not. Because if Intel plans where to succeed we would have only one choice: Rambus. Quite an ugly future, in my opinion.
Quote :
despite the fact that ALL the benches showed it performing better.
Just one more idea. Even if you are right: let's suppose that Rambus was a better performer - right from the start. Why forcing the market to use only one option? Perhaps I don't care about 10% of lost performance and I want to save money. This should be MY choice. I should choose that and not Intel's marketing dep. They should at least let other companies build chipsets that are supporting DDR. In the beginning Intel has not licensed anybody for the P4 bus tech and it has sued VIA for developing such a solution. It was clear that the users where needing that solution, yet Intel was against it. And now you are coming here and tell me that Intel is trying to give me what I want!? Unbelievable...
3. the P4 high frequency: tell me honestly: because you are working for Intel you may answer this question:
Who decided that P4 would have to reach high MHz numbers? The engineers decided that it is better to create an architecture with low IPC and high frequency, or the marketing department? Really, who decided that?
The latest proc from Intel - Itanium hasn't the same architecture (low IPC, high freq). Well... WHY? If this is the future: low IPC, high freq then WHY?
Seeing how this is the only piece that isn't "the blame game". I think I will respond. The reason why Itanium takes a different approach? From Intel marketing's point of view, because the server market doesn't give a rats @#@ about clockrate. From an engineering point of view? Because it's IA-64. The reason that modern x86 MPU's have been forced to go to such high frequencies is because the x86 ISA is simply very difficult to extract parallelism from. Consider this, the Athlon has devoted 3 times as much die space to decoding x86 as the P4, it has devoted 1.5 times more die space to try to boost peak instruction throughput as to make up for instruction halts and data dependencies. And the gain? About 20% higher average IPC throughput in your average code and that margin shrinks as code is adjusted for the P4's little.......quirks. Itanium's core is less than 20 million transistors and yet it is capable of more total work than a 45 million transistor core P4 at the top of its yield. It's incredibly wasteful to try to extract any level of parallelism from x86 on a thread-level. So the solution? Hyperpipeline it and pump up the clockrate. You may not be able to do 2 instructions in parallel most of the time but you can do 1 after the other really fast.
IA-64 was designed for parallelism. It depends a lot on the compiler but the potential is there. It's just simply too wasteful to do something like that on x86.
"We are Microsoft, resistance is futile." - Bill Gates, 2015.
From Intel marketing's point of view, because the server market doesn't give a rats @#@ about clockrate.
My point too. In other words: those who decide what server to buy are more informed than the average Joe; they are not easy to trick. So.. why bother to trick? Just give them a fair solution.
Quote :
The reason that modern x86 MPU's have been forced to go to such high frequencies is because the x86 ISA is simply very difficult to extract parallelism from. Consider this, the Athlon has devoted 3 times as much die space to decoding x86 as the P4, it has devoted 1.5 times more die space to try to boost peak instruction throughput as to make up for instruction halts and data dependencies. And the gain? About 20% higher average IPC throughput in your average code and that margin shrinks as code is adjusted for the P4's little.......quirks.
This seems to be a rational explanation. Hmm... nothing more to say.
Quote :
So the solution? Hyperpipeline it and pump up the clockrate. You may not be able to do 2 instructions in parallel most of the time but you can do 1 after the other really fast.
Yes, that is a solution. I am hoping that sooner or later the x86 platform will be replaced so that such compromises will not be necessary any more.
People hate macs. Wanna know why? Because the architecture changes so often that once you buy a mac and the new mac comes out you have to buy a new mac to make use of the new software because the new software makes use of the new mac architecture.
thats why the pc is so succeful because they haven't changed in over 20 years. You buy a software for your pc now you are gunateed that it will still work 5 years from now. Unlike a MAC.
The architecture changes as often as MS releases new operating systems.
when i say the architecture changes it is implied that the software you bought for the mac will not work on the new architecture.
that is why people hate makes. Ya great, new power4 cpu. What you don't know is that the software isn't compatible with it. You have to buy all new software!
If i'm wrong all the more power to ya!
You go buy a mac and i'll stick with my pc which i can upgrade without fear of "will this be compatible with my mac" heh!
Life is irrelivent and irrational.
<A HREF="http://www.anandtech.com/mysystemrig.html?id=9933" target="_new"> My Rig </A>
Hey imgod2u, what is IA32, and where did it come from aside from the IA64?
I had heard once IA32 is something we are on now, though I think that's not true.
IA64 does seem to be interesting but it does a halt on a lot of the silicon technologies which strive for higher clock speeds and less power usage.
Another thing, while I have a hard time understand how the IA64 parallelism makes it better in IPC, there was something I wanted to ask. Ya know how you always said that the output is what matters, and that the most operations that come out in each tick, the better the IPC result?
Well when I see it now, how the IA64 has complex instructions, well assume both an IA 64 CPU vs an x86 CPU calc 2+2. Where exactly on the IA64 will it benefit if it had the same amount of execution units, and why would IA64 be more powerful?
I'm having a hard time seeing how parallelism works and how a code is more effective, despite if the output amount determines the performance and if both architectures calculated the same code.
--
What made you choose your THG Community username/nickname? <A HREF="http://forumz.tomshardware.com/community/modules.php?name=Forums&file=viewtopic&p=19957#19957" target="_new">Tell here!</A>
