Announcement

Collapse
No announcement yet.

Intel vs. AMD - The Road Ahead

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Intel vs. AMD - The Road Ahead

    Feel free to post any concerns or questions you have regarding this article here.

    http://www.tweaktown.com/document.ph...review&dId=236

  • #2
    The thing we, the consumers, have to always bear in mind is that, despite what we want (be it higher bus freq. for the Athlon, eDRAM for 3D chips, etc), the business side of things will always be factored into the equation.

    As I'm looking at the current pricing scheme's for the AthlonXP/Pentium4, it's pretty obvious which one is logical choice...despite the shortcomings of the platform. So, even if Intel were to bring a quad-pumped 133 MHz FSB to the market today, you can also imagine that they would price those chips higher than the current P4 lineup.

    The current price ratio between their respective high-end chips is basically 2:1...By the time Intel releases their next major chip, it might be slightly higher (I would expect it to be higher)...At the end of the day, will it really prove to be a difference maker for consumers? Perhaps...

    But by that time, there's no telling how AMD's products will fare in comparison, again, despite their shortcomings. Personally, nothing is going to excite me at all until the Hammer is released. I have held onto my 1 GHz. Athlon (clocked to 1.33 GHz) since early 2001, and don't have any real need to upgrade, *despite* the fact that I use this system for doing reviews.

    Anyhow, I expect the advances that Hammer bring to the table will more than makeup for these things...

    Comment


    • #3
      I agree with you on the pricing. Regardless of how cool Intel's chips run I doubt most people are going to spend $550.00 for a top of the line Pentium 4. One could always just spend about $200.00 for a 1.8Ghz P4, but since it will perform slower than a 1.733Ghz Athlon XP you might as well just get the latter.

      Performace-wise, DDR above RDRAM and it is quite a bit cheaper to build a DDR-based AMD system than a DDR-based P4 rig. AMD does need to ramp up speeds with the Thoroughbred, but I don't think they are in too much trouble yet of just being delegated to the low-end value segment.

      As always, only time will tell though.

      Comment


      • #4
        What I find interesting is the fact that no where in the article was there mention of an RDRAM based system. DDR400 was thrown around quite a bit, but in analyzing the performance over the next year or so, I do think that it is important to consider RDRAM for the P4. As old as it is, the i850 still outperforms most competitors(arguable all). With the i850E coupled with PC1066 or even PC1200 RDRAM, I think that RDRAM deserves serious consideration. In terms of performance, there is a good chance that the P4 with 533FSB and PC1066 RDRAM will be the leader. So, why was it not mentioned in the article?

        Comment


        • #5
          The article was based around DDR and not other memory techniques, at the moment. Chances are we will keep this article updated on an on-going basis and include things like RDRAM. Things like RAMBUS 2 we'll talk about later.
          Cameron "Mr.Tweak" Wilmot
          Managing Director
          Tweak Town Pty Ltd

          Comment


          • #6
            With Intel arguably moving away from RAMBUS memory (see: http://www.ebnews.com/story/OEG20020226S0040 ) it will be interesting to see what happens in terms of future pricing/and acceptance. As it stands the P4 was designed for the high bandwidth that only RDRAM can provide, so this is where the best performance can be had (at a price).

            As mentioned the price verse performance ratio of AMD cpu's leaves allot in AMD's favour that I think will bear AMD well until Hammer makes it appearance (possibly 2002 Q4 unlike the 2003 mentioned in the article). The ability for AMD to produce CPU's with a very small footprint has allot to do with this, basically allowing AMD to get more per sheet of silicon. This interview here with Jerry Sanders is quiet a good overview of AMD past, present and future (See:http://www.upside.com/texis/mvm/news/story?id=3c98da541)

            What Hammer brings isn't just a new CPU, but an entirely new platform and selection of technologies - SOI (See:http://www.deviantpc.com/articles/SOI/index.shtml), integrated memory controller (lower memory latency) and HyperTransport interface to name some (See:http://www.aceshardware.com/read_news.jsp?id=55000382). The design of Hammer further allows for simpler motherboard layouts (4 layers) which hopefully will translate in to cheaper AMD motherboards in comparision to P4 parts (See:http://www.aceshardware.com/read.jsp?id=45000313).

            Final key point I'd like to raise is upgrade ability. By sticking with the current 266(133mhz) FSB AMD until Hammer creates an upgrade path for current socket A owners, something that Intel owners haven't had of late. Obviously AMD have the option for more to a 333(166mhz) FSB but from all reports with the current architecture - EV6 bus - this results in marginal (if not no) real world performance increase. (See: http://www.anandtech.com/cpu/showdoc.html?i=1595&p=12).

            So do I see the AMD being left behind with only value processors . . .no, most certianly not. From the first Slot A Athlon AMD has gone from strength to strength slowly eating market share from Intel. I don't see this good execution failing soon, where as Intel has struggled with competition in the market stumbling along the way only now really finding their footing again recently. I predict Hammer will, if brought to bear as promised, require something special from Intel in retaliation. At the end of the day it's us, the consumer who will always benefit from this market competitiveness, and that's what should really matter wether it be AMD or Intel.

            Comment


            • #7
              I agree with most of you in this reguards

              While AMD do have a strong platform at the moment i fear it will loose some groung very shortly with 533FSB, DDR 400 (RDRAM at this point is not worth mentioning as Intel are moving away from RDRAM) will make shorter work of the AMD CPU now.

              On the Value side AMD will either have to increase the Duron's bus as it can kill a P3 core Celeron but P4 core will leave it dead now with its lower FSB and lower cache

              You won't see RDRAM for P4 533FSB last long as its even more expensive than currant RDRAM (which is almost non-existant today). Intel has already developed its chipsets for the P4 533FSB and they are moving towards DDR-SDRAM rather than RDRAM as the I850 is the last of the line or RDRAM chipsets and even in high end server RDRAM is being shoved aside in place of DDR and Dual Channel DDR SDRAM sollutions

              RDRAM at PC1066 and PC2100 will not last long against 5.3GB/s Dual Channel DDR-333 and Dual channel DDR-400 at 6.4gb/s. more than enough for bandwidth for P4's new bus.

              As Tweak said we will be updating this article (most likely every end of quater) to reflect on our speculations and make future speculations

              That my rant for this evening

              Comment


              • #8
                I have a few problems with the article...

                While it addresses memory issues quite well, it tends to quickly gloss over (or not mention) many of the others...

                1. Along with deeper pipelines (and their associated problems of higher latency which was mentioned...), the P-IV suffers from a very poorly designed FPU. The only apparent solution Intel has for this is SSE2 (if your software has it...most don't)

                2. Duron is due to be discontinued by the end of this year (according to the Conference Call by AMD last week). This means that the AthlonXP will actually be the low-end part, and the Clawhammer will replace it as the flagship.

                3. The Hammer chips are a full generation ahead of the P-IV. Molman has some excellent points, but allow me to expand...
                a. Memory controller on die - allows for the increased memory bus speed as system bus increases
                b. MUCH lower latency (because of "a") which allows for longer pipelines and increased clock speed with no net decrease in IPC (instructions per clock)
                c. It will be able to address >4 Gig of Ram (I know that seems ridiculous today, but how long ago did HDDs top out at 500 Meg and processors at 90MHz?)
                d. It should be priced along the lines of today's P-IV, with 64 bit ability to rival the Itanium and simultaneous 32-bit ability of a 3.4 GHz P-IV (quite possibly in stores by Xmas...)
                e. Hypertransport connections to all points
                f. Motherboards shouldn't require much of a Northbridge (if any) because most of this function is built into the processor (much cheaper chipsets/motherboards...), not to mention Molman's point of 4-layer boards.

                One last point...I see it was mentioned that if Nvidia were to base their Nforce on the P-IV it would show far more significant gains than on the present Athlon. While this seems plausible, I would be stunned if it ever happened...
                1. Intel won't license Nvidia with the P-IV's bus as this would be a major and superior competitor to their own chipsets (unlike SIS who are strictly bargain chipsets...)
                2. NForce uses Hypertransport (owned by AMD...)

                Just my .02 worth...

                Cheers,
                Charles

                Comment


                • #9
                  ahmen to that Charles! hit that right on the head.

                  pAc

                  Comment


                  • #10
                    Well, I was going to reply to this article in regards to AMD's updated roadmap, which does phase out the Duron and moves the Hammer in, possibly as early as even late October!!! This is speculation as well, but some rumors are floating around that AMD will be making additional architectural changes to the Barton core. So there will be 512Kb L2 cache and possibly some other things though AMD hasn't said what. To me, it looks like Intel better step it up while it can cause I think AMD has something big planned for the entire computer market. (IE Hammer)

                    Comment


                    • #11
                      I think the other thing that wasn't accounted for in the article is the rumour (and it's more than a rumour...) that Intel dispatched a team to slap together *something* that's very similar to the Hammer...

                      If that doesn't smell of deparation, I don't know what is...In the field of engineering (of which I'm one...Computer Engineer to be precise), you don't just put together a team of engineers and instruct them to "do whatever you can to take some existing architecture, and change it to model/compete with the-other-guys offering...just in case"....If you catch my drift.

                      Even when Intel moves to .13u, look at how it will compare to Athlon's Thoroughbred...and then analyze the price/performance issue again.

                      It's just like with nVidia GPU's...Could nVidia produce some killer-ass 3D chip with gobs of eDRAM, 512-bit memory interface, multiple-chips, etc ? Of course they could...In the end, does the cost justify such a product? No way...Not yet.

                      With every successive generation of chips, people keep insisting that they move towards higher bandwidth architectures (don't get me wrong...I'm all for it)....As it now stands, the GF4 is approaching 70 M trans....Do you realize how much larger the chip size would be if/should nVidia do that right now? Or 2-3 generations ago? What about yields? What about profit? Until such time that they can pull it off from a technology standpoint AND a business one, they're not going to commit to such features...and the same goes for AMD.

                      If anything, if you forget about the P4 for just a second, and concentrate on the true next generation platforms...It's pretty darn clear to me that AMD is right where they want to be, and Intel is trying to figure out how to contend with AMD.

                      As was mentioned, the _very_ kool architectural changes that Hammer will bring about certainly add a lot of credibility to that assessment, not to mention the fact that they're going to beat Intel to the 64-bit Consumer platform game....

                      If you thought AMD was being a little bit creative/deceptive/etc with their XP marketing, wait until Intel does the same with respect to the Hammer.

                      Comment


                      • #12
                        I tend to agree with Charles in his comments....

                        we know the PIV produces less per clock cycle (about 2/3) what the current Tbird/Palamino core can produce because of its horrendous 20 stage pipe....and thats the only way it keeps up is by increased frequency...purely viable because of its lower work rate per cycle..... How long can the clock speed be scaled to cover poor core design????

                        We know the hammer will have 512K L2 cache (on their roadmap...later this year)

                        The bus speed is not really a product of the CPU....its the chipset that carries the weight.....thats how I can run my 1200 tbird at 12*100 or 8*133. There are some timing issues but these are minor.....Intel has put a lot into its chipsets to support the quad fsb....AMD and its chip make

                        rs will do the same....and as the Nforce shows it is achievable sooner rather than later.

                        I think we will see much of the same....a bit of a blip for AMD with Hammer but that is all dependent of OSs and apps which will use the 64 bits anyway.......it is very nice having a car that can doo 500kph....useless if the roads only let you actually do 100kph....

                        What is interesting is that they have a version of Hammer already running on a 64bit version of Wiondows...see here
                        http://www.theinquirer.net/23040208.htm

                        Looks good for all of us....if Intel and AMD keep at each other and stay about even then WE ARE THE WINNERS.....as soon as one or the other gets too far ahead then that company can start dictating the game (just like M$ do) and we would all be losers.....

                        Just hope for our own sakes that they stay about even

                        PS Here's another interesting article discounting the fsb fears http://www.theinquirer.net/19040202.htm

                        Comment


                        • #13
                          Charles, firstly the article was directed at the high-end market, not the value market which is why I focused totally on the AthlonXP and the P4.

                          Secondly, I didn't comment on the Hammer because as you said, it IS a whole new architecture and it is impossible to speculate on how it will perform, especially against Intel's 64-bit processor. What I did say is that, if the only change AMD will be making to their AthlonXP before the Hammer is upgraded cache and a shrinked die, Intel have a very good chance to pull far ahead as far as performance is concerned.

                          Think about it. If the 2.4GHz P4 outperforms the 2100+ now, how much better do you think it will go with a 533MHz bus, DDR400 and dual-channel DDR memory?

                          When I was talking about the nForce, I wasn't saying that nVidia should go out and make a chipset for the Pentium 4. I was merely trying to give an example of Dual-channel DDR and show how it could benefit the P4.

                          Comment


                          • #14
                            If you check the excellent review of Pentium 4 vs AMD Athlon XP on www.aceshardware.com, a dual Athlon MP 2000 system is still faster than a 2.4 GHz Intel Northwood. Results on database transactions per second are 151,919 transactions per second for an Intel P4 2.4 GHz system, and the dual AMD Athlon MP 2000 system does 231,333 transactions per second. Also the AMD 760MPX chipset has two slots at 533 Mb/sec, and it's not vaporware! I have one of these ASUS duallie systems running and was pleasantly surprised at how stable it is, how cool it runs CPU temperature 109 degrees F after running days on end, and how quiet it is. Based on data presented in the aceshardware.com review, the AMD MP2000 system even outperforms a dual Xeon 1700 system, and the Pentium 4 2.4 GHz system wins only one of the dozens of tests.

                            For my part, I will stay with AMD systems, because they offer a better combination of high performance and low price.

                            Comment


                            • #15
                              Originally posted by Unregistered
                              If you check the excellent review of Pentium 4 vs AMD Athlon XP on www.aceshardware.com, a dual Athlon MP 2000 system is still faster than a 2.4 GHz Intel Northwood. Results on database transactions per second are 151,919 transactions per second for an Intel P4 2.4 GHz system, and the dual AMD Athlon MP 2000 system does 231,333 transactions per second. Also the AMD 760MPX chipset has two slots at 533 Mb/sec, and it's not vaporware! I have one of these ASUS duallie systems running and was pleasantly surprised at how stable it is, how cool it runs CPU temperature 109 degrees F after running days on end, and how quiet it is. Based on data presented in the aceshardware.com review, the AMD MP2000 system even outperforms a dual Xeon 1700 system, and the Pentium 4 2.4 GHz system wins only one of the dozens of tests.

                              For my part, I will stay with AMD systems, because they offer a better combination of high performance and low price.
                              Thanks for that unrelated post, now tell me where in the article I was talking about the Server/Workstation Market? To quote myself:

                              I am only going to be talking about their desktop processors as that is the only concern we have in this article. The server/workstation and mobile segments are totally different issues altogether.

                              Comment

                              Working...
                              X