Would AMDcpu be even better if Intel hadn't messed up so badly

[Rhi]

AMD is in the red due more to decreased demand/increased pressure, (from INTC), in the flash memory area. They've been gaining market share in the processor segment.

AMD riding coattails? The Athlon has an entirely different architecture than Intel's offerings. Sure they've taken Intel's products and improved upon them, but name any business model where this isn't done. Of course they can't invent something and force the industry to use it, (SSE...etc.), they're not nearly as big as Intel. According to what I've read, AMD doesn't want to be in the chipset business, is this bad? I cannot tell, but I've built several systems based on VIA's solutions, and have yet to have a chipset related problem.

Intel stable and reliable? Of course. AMD stable and reliable? In my experiences yes. Should I pay the price premium for the Intel name? Negative. As soon as Intel can match AMD's price/performance, that's when I'll have Intel Inside.

Oh, and I'm not sure about the Topic...I'd say that AMD might have been lagging farther behind, although you never know what they have up their sleeve...they are clamping down on prices a good bit, and instead of holding faster chips because they don't have anything more, maybe they're waiting for Intel to show their hand. Who knows.

Competition=Good

-Rhi

 

[imgod2u]

<< I don't think x86-64 will die. If anything EPIC will die first. I mean, how long has Itanic been in development? some 6, 7 years? The current design architechure just isn't practical. If it were, Intel would be pushing Itanics instead of P4s right now. If anything, I think we'll see intel counter with an x86-64 design of their own and THEN, EPIC will follow after that. >>



Well, keep in mind the 64-bit market when Hammer is released will be the server market, and that market is anything but practical. I don't think there are any x86-64 applications yet. So basically, EPIC and x86-64 will be off to the same start. In fact, EPIC has already gained grounds as Intel has already pushed developer support and provided many optimization tools. The thing to remember with anything new is that it takes time and it takes refining. Itanium was intended to be a powerhouse, it was intended to show that EPIC can work and allow developers time to play with the new design to make way for McKinley. As for x86-64 software, AMD will have to push really hard to get acceptance in the server market simply because most of the modern day server processors (and hence developers of software for them) are use to coding in multithreaded code and in a RISC like format. The reason x86 has stuck on so long in the mainstream market is because developers have gotten use to it and are unwilling to change. One thing I do applaud Apple for trying to do is move things foward. As with anything else, the architecture will first be accepted by the server market (hopefully) and then, eventually make its way down to the consumer market. I just hope the hammer can do fine in the 32-bit market, but I really don't want to see x86 go on any further than that, it's time to shed the old like M$ did when they released WinXP.

 

[SniperWulf]

<< The thing to remember with anything new is that it takes time and it takes refining. >>

Well geez, how long should it take? How long until they realize the majority of the project was a flop? I mean, nobody is selling em. Even Dell said to hell with Itanic a few weeks back?


<< As for x86-64 software, AMD will have to push really hard to get acceptance in the server market simply because most of the modern day server processors (and hence developers of software for them) are use to coding in multithreaded code and in a RISC like format. The reason x86 has stuck on so long in the mainstream market is because developers have gotten use to it and are unwilling to change. >>

While that is true, the advantage that AMD will have is that developers are already used to coding for x86 and it prolly wouldn't take much to get used to and incorporate AMD's 64-bit operands and the like.

 

[Kell]

<< << I saw today that a major flaw exits in some/all (not sure) Athlon chips with Linux Kernel 2.4. >>

I saw that on Slashdot yesterday. Something to do with AGP video causing crashes
if you compiled the kernel for Intel chips (actually should that be surprising?). To use the Intel term, errata was issued for this thing back in September 2000. I still don't know what the problem is exactly... I have an AGP video card and I've been using Linux 2.4.x kernel on my Athlon machine for more than half a year now and its never crashed my box. >>



I actually have one of the "problem chips." For months now, I'd been disabling AGP in the nVidia driver with a slight performance hit. Now I know why...

The other thing...if you look on LinuxToday.com, the whole situation seems to have been figured out. It is apparently not an Athlon bug--it's a bug in the way both Win2K and Linux manage virtual memory and AGP memory. The only reason the problem doesn't occur on Intel chips is because Intel chips don't have a little feature called speculative writes, so the VM fallacy can slide and actually has slid for a long time.

The Linux kernel will supposedly soon have a real fix, by the way of an updated VM.

 

[imgod2u]

<< Well geez, how long should it take? How long until they realize the majority of the project was a flop? I mean, nobody is selling em. Even Dell said to hell with Itanic a few weeks back? >>



Look at the P4, when it first appeared, it was laughable, but the technology was there, it just needed to scale and it needs refining, still does. Intel with its IA-64 architecture has brought about a new way of doing things and that takes time to get it right. You can always just add pure FPU power, but that's not very scalable. One thing Intel is known to do is look out for the future, not just make a solution that'll impress benchmarks. Itanium's main purpose was to introduce EPIC to the market, not to overwhelmingly outperform the current competition. For that, we'll have to see if McKinley can do it.



<< While that is true, the advantage that AMD will have is that developers are already used to coding for x86 and it prolly wouldn't take much to get used to and incorporate AMD's 64-bit operands and the like. >>



Only developers in the current consumer market. Hence why the Clawhammer will probably be a success. However, I doubt developers will code specifically for AMD's solution if the software wasn't compatible with Intel's current consumer solution (they still own what, 70% of the market?). Therefore, the 64-bit part of the Hammer design will probably go unused, at least unless Intel decides to make something similar. As for the market that currently does use 64-bit coding, that's the server market, and x86 has no business there. I would like to see the Hammer design do well in the 32-bit market if only to give Intel some competition and make them sweat. I wouldn't like to see x86 go beyond anything than 32-bits. It's really about time we got rid of the old.

 

[gerrick]

Damascus - maybe bottlenecks is the wrong word. Not too many software packages can really take advantage of the current chip speeds. Most if any gains are minimal. I mean who the hell need 200 FPS in quake?

 

[Endorion]

"Intel chips do cost more. They have better R&D and that costs money. Now since most of you will say Intel R&D sucks here is my proof. What company hit .25,.18, and .13 first. How about MMX, SSE, and SSE2? I could go back further with research. In my opinion Intel also has a higher quality chip. I can't unlock the multiplier with a hardware hack. Now I don't think the High end Intel chips are necessarily worth it. Intel has always set the high end prices pretty much the same. Intel has the best performing, most stable combination out there. If you doubt that what do most businesses buy? They don't just buy them because of Intel name. Intel has a rep of quality just like Honda or Toyota. If you look through these forums you will see more ATHLON issues then Intel issues. Most of these issues are chipset issues.
"

Intel chips do cost more since they are more expensive to produce.

Better R&D? Amd only needs a die size of 128mm^2 when Intel need 217mm^2 on the same process with equal or better performance to Intel. I think that pretty much explain where the good R&D is located...

What about 3DNow!, the first and only fully tripled pipelined x87 fpu, PowerNow!, HyperTransport..? This from a company 1/10:th of the size of Intel...

Why would the fact, that you can unlock an Athlon, make Intel cpu's of "higher quality"??? It is called feature...

I can agree though that AMD should make their own chipsets. I think that not producing a fast & stable chipset (Read AMD760; or even better a new chipset for DDR333 ) for Ahtlon is one major reason why it is so difficult for AMD to get into the corporate market.

 

[imgod2u]

AMD's developement team is no where near Intel's. Look at all the technologies on the P4. While it is not as effective with today's software, it is beyond anything AMD has come up with as far as technologically. The prediction algorithms on the P4 alone are a massive achievement. Not to mention SSE, SSE2, MMX, even the x86 architecture itself. 3DNow? I believe AMD hasn't improved on that since the original Athlon, in fact, it's using Intel's SSE now. Even the Slot and Socket technologies AMD is using was a rip off of Intel's packaging methods (since Intel won't let AMD use their socket technology). The palomino, with all its great improvements are all present on the P4, such as hardware prefetch and organic packaging. Hell, Intel's been using organic packaging since the original Pentium I think. The Hammer has been AMD's only real leading role so far. BTW, Hypertransport was not solely AMD's developement, why is everyone giving AMD credit for it?

 

[Sid03]

<< Intel chips do cost more since they are more expensive to produce. >>

wrong. intel chips cost more because intel can sell them for more. if amd could, they would sell theirs for more also.

while i really like amd, to say that they have better r&d than intel is just zealotry. 3dnow?... give me a break.

 

[KenAF]

<< Better R&D? Amd only needs a die size of 128mm^2 when Intel need 217mm^2 on the same process with equal or better performance to Intel. I think that pretty much explain where the good R&D is located... >>



That may sound like common sense, but it's completely wrong.

The P4 is so large because it has 20 pipeline (or 27/28 if you count another way) stages. Many of the transistors (and thus the extra size) are intended to get and maintain the high frequencies across the chip. It's this design which has enabled the 2.2GHz to 3.8GHz overclocks (with appropriate cooling) that you see on some of the hardware sites. It's this design that will enable the P4 (with air cooling) to scale to 3.0+GHz on .13m and 4.0 to 6.0GHz on 0.09m. The P4 was truly designed for future process technologies; the .18 process on which it debuted (Williamette), and to some extent, the current .13, are a waste of the chip's brilliant, highly-scaleable design.

Of course, the P4 core also has technology like SMT that is not yet enabled, because Intel did not feel the need.

 

[madmickey]

Intel is talking about coming up with a 486/64 bit solution after spending 7 year and billions in R&D on the IA-64 solutions, where they took a good HP solution and seemed to messed it up. Joining the list of the following dubious Intel accomplishment, the 820 chipset, PIII 1133 MHz, the poor performance with the original PIV, use of Rambus, the 845 chipset with SDR ram even though DDR ram was readily available, the problems with original Pentium, the patches you needed for the original PII, in 15 months of existence the PIV went from a 423 pins 100/400 MHz bus to 478 pin 100/400mHz bus Willamette to 100/400 MHz bus Northwood 478 pin to 133/533 MHz Northwood bus speed. During this same time AMD has been using 133/266 MHz socket A Athlon and this will be the top performing Athlon in bus speed until the end of this year if not longer. Do you say you like buying new motherboards?

 

[imgod2u]

It's not a matter of whether AMD has better choices, it's just that they're not as ambitious. AMD is in a position where it cannot afford to make bold decisions. They play it safe, releasing only what they already know is established. That is perhaps why they didn't follow into the 64-bit market with a new instruction architecture, because they want to play it safe. You can insult Intel's attempts at changing the market all you want but at least they're trying. They're the ones who were bold enough to support the RDRAM technology, and look at how well it does with the P4 today. They're the ones who came out with a high-scalable architecture that's in the P4, and while you may all laugh at its initial introduction, the scalability it is capable of even today is nothing to look down at. What about hyperthreading? Or SSE/2 or hardware prefetch or even the socket pin technology and soon the bumpless interface. You all talk as if Hypertransport is something incredible, but Intel's Northbridge-to-Southbridge connectors have been capable of 533MB/sec since the days of the P2 I think. AMD with their x86-64 is again playing it safe. And while I do admire the idea of the integrated memory controller, I really am disappointed that they chose to stretch the aging x86 architecture even longer.

 

[madmickey]

I can't seem to see last messages.