Unlikely struggle: Intel's AMD problems

OK, I have a million dumb computer questions and here's another one:

Intel currently reaches the highest clock-speeds, but AMD has a higher IPC, and thus matches or sometimes even beats Intel in pure performance. This is even as Intel has all the numbers on its side: FSB, RDRAM, CPU, 0.13nm tech....

Do I understand it right that Intel shot themselves in the leg by reducing the cache-sizes of the P4 compared to the PIII and also removed one FPU? Can't be good for performance...

My question then, is this: if Intel is the larger manufacturer with most facilities, largest Si-wafers, and generally best processing technology, why aren't they way ahead of the game? Shouldn't they be smoking AMD out of the waters? Is the current struggle to fight off AMD basically a result of bad management / poor decisions made w.r.t. architecture, cache-sizes, deep pipelines, enhanced instruction sets, etc.?

I'm sure these questions could be extended to comparisons with Motorola, SUN, and so forth.
41 answers Last reply
More about unlikely struggle intel problems
  1. AMD's processors have a higher IPC because they were designed to be competitive with the Pentium III. The Pentium III had a similar IPC. However, to scale to future speeds, such as 10GHz, a longer pipeline is required. There is an initial performance cost to this, but the performance gain from the scaling in clockspeed more than makes up for this in the end. This has been discussed ad nauseum. Please try a search.


    = The views stated herein are my personal views, and not necessarily the views of my employer. =
  2. Intel isn't smoking AMD because they don't want to have a monopoly on their hands after what Microsoft went through in courts. That sums it up quite nicely. Intel could easily destroy AMD but they don't want to.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  3. that's a interesting theory, but i'm sure Intel would be allowed to out-compete AMD by using it's technological muscle, because the resulting monopoly would be legal, it's wouldn't be Intel's fault if AMD were to loose business for being behind Intel.

    Anyways, interesting to think of it that way mate!

    What happened with Microsoft was very different and besides hasn't microsoft "won" that case now?
  4. To what AMD_Man said, I think is very true. You have to remember that these companies are ones that are trying to make money. If Intel can hold a 80% share of the market, there is really no need to improve your chips. They are doing just fine and they know it. That may not be that great for us but for them as a company, it just allows them to stay in business and to keep researching and making new and faster chips.
  5. competition isn't making intel more money, it's taking away from them in my eyes. back when the p2 was the big cpu around, amd was pathetic and therefore intel was able to sell that p2 chip for what, nearly $1,000 (us dollars). maybe i exagerated a bit too much but you get the point. so when the p2 was the dominant force, everyone began to buy intel, and since there was no competitor, who needs to lower prices?

    didnt have one of em electronic pens so ill just type my name,<i>CoOoLMaNX</i>
  6. You're right. AMD changed the playing field with the release of the Athlon in 1999.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  7. OK, conspiracy theories aside: how about the decision to reduce the cache sizes and remove one FPU. Is this also a move to "scale for future speeds"? I don't think so.

    So Raystonn, basically what you're saying is that the current top-of-the-line P4 sucks (rudely summed up), and won't be good until they reach ~10GHz? Shouldn't Intel keep their research projects to themselves, and sell chips that clearly improves upon current technology?

    (somebody here made a calculation that reducing the IPC had a more detremental effect on the P4 than increasing the clock speed had...)

    You know, the 80% market share (is this as of now?) could turn very fast if Intel doesn't flex some muscle; people are fast to change loyalties, especially since we're not talking about Coke vs. Pepsi here. Oh, did I mention that I did a price check on a dual MP2000+ and a dual P4 2.0GHz with 2GB DDR/RDRAM and found that the Intel solution cost over 50% more...? So, whether or not a P4 is 10% faster or slower than an equivilant AMD at a matching PR, I could buy a third AMD machine and gain performance advantage that way!
  8. Then there's the issue of CPU sockets... isn't it so that AMD has been using the same socket since the K6? Intel on the other hand.... 2 sockets for the same generation chip!
    The upgrade cost doubles if you need another mobo, as well as a new CPU.
  9. Yes, but AMD did also change the core layout and the FSB speed so that older Socket A motherboards aren't compatible with the Athlon XPs.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  10. Well I'm not really into the chip architecture, but I remember reading here that the performance is the speed(MHz, GHz)*IPC. In turn by decreasing the IPC, they also reduce the complexity of their chips, which allows them to run faster and cooler. Intel is really using this GHz rating as a marketing ploy to win over people's wallets who know diddly-squat about computers. I for one know that to most people speed is everything in computers and the oly way they know how to mesure speed is in MHz. And yeah as of the end of 2001 AMD had a 22% market share of the desktop computer market and Intel had 78% or something like that? Not sure about servers though.
  11. There is also the size and heat factor. AMD's chips are at 0.18m, and were almost 2 times smaller than the Willamette's, plus it generated so much heat and power compared to the Tbirds, and yet performed very weakly. That in itself shows AMD more dedicated and somewhat moer sophisticated in their line of work here. Seriously Intel has 10 times the fab space and resources, and yet their P4 turned out huge? Laughable...

    For the first time, Hookers are hooked on Phonics!!
  12. Hmm, I guess what I'm getting at is that Intel appears to have the technological edge: 130nm, clockspeed and all that, but they're making some bad decisions that affects overall performance: they could be so far ahead of AMD, but AMD is tweaking the right things on their somewhat limited technology. Why isn't Intel pampering us with, perhaps a 1MB L2 cache and a larger L1 cache?
  13. Pretty much. Intel was hoping to innovate at a more liesurely pace (as they did with the P6 core), but AMD more or less broadsided Intel with the Athlon. Playing second fiddle in terms of performance has pushed Intel to make some pretty bad decisions; as a result, Intel still hasn't quite gotten its performance edge back.

    As for the P4 having a castrated cache...it was initally castrated compared to what Intel had planned for it to be (512K), but it was the same amount of cache as the desktop Coppermine (256K). The FPU was badly castrated, however.

    <i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?
  14. i know it might be pretty off topic in this thread, but since we're here. as far as dual setups go, xeons will be the only dual intel platform?

    I run duals because i multitask between notepad, outlook express and winamp :lol:
  15. But the Thunderbird and Coppermine were so similar it was scary. Yet the Thunderbird clocked 50% higher at the same die process. Heck, the .18 micron P4 did only 30% higher MHz than the Thunderbird, and ran 30% less performance. So what do we get, every processor is inferior to it's predecessor by an amount equal to it's clock increase?

    The Northwood, geeze, I'll have to go to at least 2400MHz to get double the performance of my PIII 1000EB. Probably not even double the performance at that!

    Having said that, it looks like I'm still stuck with the P4, as I'm an overclocker, and the Northwood has the best performance potential. Shame on AMD for delaying the Thoroughbred. Maybe next time, AMD. Shame, I already had a motherboard picked out and everything.

    What's the frequency, Kenneth?
  16. yeah, you're right. i guess that's what's bringing up the price of the Intel duals so much. Xeon: 256KB and 512KB RAM on the P4s, right?
  17. Quote:
    But the Thunderbird and Coppermine were so similar it was scary. Yet the Thunderbird clocked 50% higher at the same die process.

    Copper interconnects dude, copper interconnects.

    Plus the internal design of the tbird was more advanced than the p3, which allowed more ipc and a higher clockspeed.

    :wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
  18. I think it comes down to this mostly, and Raystonn, as I would imagine is correct, will probably be fairly plugged in with this.

    If you look at incremental ship releases, the difference is little. Marketing people and journo's place big hype in all that stuff. What really counts in silicon is what you are doing in 5 years, not so much what you do now.

    Corporations, and to a greater extent Wall Street are looking to technological strategy. They care little that Northwood is X% over Willy etc. they want to see overall direction. The state of play is that Intel have a very strong, long look-ahead strategy, they work with the big OEMs, provide total solutions.

    What they have on the table today may not always be optimal, but they hold market share and size by strategy and plans, not individual products.

    Don't get me wrong, a real lemon will hurt their p&l, however the lemons are only exceptions on the roadmap, and are soon removed.

    AMD are changing from a product to a strategy company, but their dependence on Athlon has ben great. It was a good chip, but the delays in TBred and XP hurt them somewhat. Now they are focusing towards X86-64 and new technologies they have a chance to be believed, but they are no-where near being an Intel, and if they got Intel's market share right now, they probably couldn't ship the parts to fill it anyway. That sort of volume change needs to be gradual since ramping production for a company AMDs size is very hard.

    -* <font color=red> !! S O L D !! </font color=red> *-
    To the gentleman in the pink Tutu
  19. Raystonn is correct. If you extend the pipe - you lose IPC. But you make up for it later is raw performance. Thats what AMD is doing with Hammer.. they just didnt take it as far as Intel did.
    At matching clock speeds, the hammer would be slower than athlon (if it didnt have a kick @ss chipset). Thats what happens when you add some pipe. So there isnt anything "wrong" with Intel, they just took their architecture another direction. The fact a 1.6 gig is able to keep up with a 2.2 gig isnt a big deal. In fact it reflects peoples bias and dependance on MHZ to determine performance.
    As mentioned before, we need a new metric..

    Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
  20. Quote:
    he Northwood, geeze, I'll have to go to at least 2400MHz to get double the performance of my PIII 1000EB. Probably not even double the performance at that!

    performance isn't based simply on mhz, we have discussed that. Doubling mhz will never double the speed of the pc exactly, because there are other components that are not having their speed doubled.

    <i>My life wasn't complete untill I tried sse-2 optimized pong</i>
  21. Quote:

    and generally best processing technology

    nope... that title usally belongs to IBM with motorola (which also makes AMD) a close second and Intel usally playing the catch-up...

    This post is best viewed with common sense enabled
  22. What I'm really getting at is this-the Thunderbird clocked better, but without performance loss as compared to the PIII. But yet the P4 had to give up a bunch of performance. Kind of sounds like sloppy engineering.

    What's the frequency, Kenneth?
  23. I don't think Intel really cares right now. They are getting money from their P4s, Celeron and so on. They are well known. They know that they are in a position where they are not in trouble. With about 78% market is damn good. But if they full down then they can pounce back. They have the resources to do that. They have all the stuff they need. Prescott looks damn good. So Intel ain't done for if that is what you would you expect and same with AMD.

  24. It's not sloppy engineering, it's progress. :wink: The Pentium Pro was the first CPU based on the P6 core! Look how slow it was in 16-bit programs. Everyone used to say that the PPro was designed for WinNT and only WinNT. Then came the PII, a souped up PPro with better Win9x performance.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  25. Are you saying the Hammer has slower performance clock per clock with Athlon???

    For the first time, Hookers are hooked on Phonics!!
  26. If you eliminate all the variables and other core improvements, then yes, the extra two pipelines will slow it down per clock.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  27. But indeed that is why they improved more than enough. Technically saying, with the more cache, Hyper Transport, On die mem controller, possibility of SSE2, improved prefetching for pipeline, I think we're more than above IPC.

    For the first time, Hookers are hooked on Phonics!!
  28. Quote:
    Intel isn't smoking AMD because they don't want to have a monopoly on their hands

    Are you kidding me? If Intel could waste AMD they'd do it in a heartbeat. Throughout the 90's Intel tried to put AMD under. Intel's shift from Socket 7 to Slot 1 is a primary example of how, by denying AMD access to it's technology, Intel has tried unsuccessfully to sink AMD. There are no laws preventing Intel from releasing products so superior to AMD's that the market doesn't buy AMD. Intel no longer has better technology than AMD so Intel's only option now is to spread FUD and fool people with blue alien commercials.

    What's <A HREF="http://www.vanshardware.com/" target="_new">this</A> all about?
  29. As we've said before, have you seen the ugly stuff Microsoft has gotten into? Intel still holds a commanding lead in market share and they really have no need to create some super chip to kill AMD. I think Intel is just guarding their hand very well and if Hammer really is a big success, they'll come back at it with something. I mean think about it, if you were making money and still could kill your opponent when you wanted, what would be the incentive to come out with amazing chips, unless you're getting threatened? Intel and AMD are both companies who are trying to make money and in the case of AMD they really have to increase their market share and to do that, they have to come out with better chips and sell them for less to entice us buyers into buying them. I think that AMD's PR rating is working to an extent as they have already increased market share by about 4%. As long as they continue what they are doing, they could do much better, but when they do start doing better, I'm gonna guess that prices are going to go up.
  30. Not really, AMD prides in low prices always below Intel's. I could not imagine an Intel system being even 10% below AMD's!
    Of course Hammer will be more expensive, but for us home users, the ClawHammer's architecture which is smaller than the current AthlonXP die, should make prices a little bit higher, but nothing astronomical like Intel's top of the line. (2.4GHZ at 610$, get em while you got the money!)

    For the first time, Hookers are hooked on Phonics!!
  31. This is true and I do believe that AMD chips probably will always be cheaper than the equal Intel chip, but even though AMD may pride themselves in selling chips cheaper than Intel, the money sense is going to (most likely) get the better of them and they probably will hike prices slightly when they see that they are doing fine.
  32. That is definitly true, I am not saying prices won't go up, however I don't see AMD ever putting home CPUs any higher than 500$ for the next year. Their top of the lines are at 300$ or less, and if economy and their market is doing great, I see a 20% up in prices, but nothing that can make it look gouging like Intel's 610$. If Intel had prices equal to AMD's not slightly higher (1.6A vs XP1600, 1.6A still is 20$ higher, which translates to a lot in Canadian pricing here) then there would be a very attractive system for a low price. RDRAM is very cheap now, but nobody should buy 128MB, they're actually more expensive than one 256MB chip (89$ for 1x128MB, 145$ for 256MB, so the 256MB's 128MB sticks would've cost 77.5$ each) if bought in sets of 2. Mobo prices are still high though, and although SiS has the best mobo prices, I guess until DDR400, RDRAM is better. However now we want to see the pricing on PC1066 RDRAM, to see how high it goes in Canada!

    For the first time, Hookers are hooked on Phonics!!
  33. Hey so you're a fellow Canadian too huh? Nice meetin you!
  34. Nice meeting ya too! I'm Canadian too. I'd guess about 30% of the members of this community are Canadian.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  35. I wish that were a valid comparison, but while the Pro was a 32-bit processor that had problems running 16-bit programs, the P4 is made to run the same software as the PIII. Sure it has SSE2 and a few other things, but those addons which are not normally used by many programs, nor should they be needed to get similar performance, since those are not "performance equalizers" but instead "performance enhancers".

    What's the frequency, Kenneth?
  36. My point exactly! Intel is using SSE2 now as a performance equalizer to compensate the missing FPU, thus acts an FPU shield, instead of providing its own better optimizations. Until the FPU is restored, and the SSE2 moved to ALUs, optimizing apps to use SSE2 on P4 is not too beneficial or will not put the P4 any better than Athlon's per clock performance in apps that even use SSE2.
    So the question is (from your POV a long time ago), if we tried to put the P4's bus to 200MHZ and saw little difference, that would mean that the P4's bus is indeed inefficient compared to Athlon's, and that 400MHZ which should've been awesome, isn't. However that's to think. If we took off 200 MHZ or put it at P3's bus with 1X multiplier, and the performance goes even worse, then indeed the bus helps a lot, but it also shows that without that, the pipeline increase has dramatically damaged a CPU. The bus and RDRAM are the only things holding the P4 alive now until Prescott!

    For the first time, Hookers are hooked on Phonics!!<P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 03/31/02 04:35 PM.</EM></FONT></P>
  37. Which is why I was saying those. In Canada a slight change in US pricing can affect our entire budget! Back when I was tight on money and wanted this new XP1600 system, the P4 was almost 100$ more to get a little closer to the XP's performance, and the price difference between a 1.6A and XP1600 is still dramatic to us Canadians, especially now that the XP1700 is only 200$, 1600+ being phased out, and the 1.6A is 240$! So buying a P4 system in Canada is very expensive, compared to US where the great low prices are there!
    Now you see why AMD's prices in Canada make a huge difference between economizing up to 350$ (XP1700+ vs 2GHZA) worth of similar performance and put it on even more performance like more RAM or a better video card!

    For the first time, Hookers are hooked on Phonics!!
  38. If Intel could kill AMD it would. If Intel could take 100% marketshare it would. Look at the FUD that Intel has recently targeted specifically at AMD. Isn't Intel happy to let it's technology speak for itself? Why must it lie to it's own dealers and pay for deceptive reports (Aberdeen)?

    What's <A HREF="http://www.vanshardware.com/" target="_new">this</A> all about?
  39. Somehow, I doubt the supreme court would allow Intel to have a complete monopoly.

    AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
  40. Well I don't really think any company can have a complete monopoly. Even Microsoft doesn't. There's MAC and Linux and those things out there. (My friend uses Linux and man if looks confusing esp. since it's the text version lol.) And even if they did they would have to follow so many rules, they may as well not have one.
  41. For those who havn't read it yet, this <A HREF="http://www.penstarsys.com/editor/Today/amd3/index.html" target="_new">article</A> (<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">originaly posted by Kemche</A>) addresses some questions/points brought up in this thread.

    - JW
Ask a new question

Read More

CPUs Performance Intel AMD