Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
AMD

AMD's New Thunderbird Articles & Benchmarks 69

nd writes: "The NDAs for AMD's new Athlon Thunderbird reports just expired, and the benchmarks have been pouring in. Tom's Hardware's coverage (in German) is here , a translation to English is here - Anandtech also covered the new CPU release. For those of you who want to learn more about the Thunderbird, here is an interview with AMD on the processor release. Overall, the Thunderbird is performing quite well, and will be sold at the same price as current Athlons. "
This discussion has been archived. No new comments can be posted.

AMD's New Thunderbird Articles & Benchmarks

Comments Filter:
  • As Tom stated in the end of the review, it is somewhat disasppointing. I was hoping the new Athlon would completely trounce the intel chip. Instead, intel beats it by a fraction on a couple of benches, and Athlon comes out significantly ontop in just a few scenarios.

    Amusingly enough, if AMD wins, it will be because of price and availability, not sheer performance. I imagine there are benchmarks out there that Tom didn't present that might cast the athlon in a more favorable light.. theres no doubt in my mind that it's simply a better designed processor.
  • by levendis ( 67993 ) on Monday June 05, 2000 @02:03AM (#1025702) Homepage
    Here's [tomshardware.com] a far better English version of the Tom's Hardware article

  • Some sort of hack that will post to slashdot for you with the same link.
  • The Thunderbird was meant to smash Intel, but looking at the benchmarks is hasn't really. At the high end > 850Mhz AMD is now slightly ahead of the CuMines from Intel.
    But, Intel has new Willamette chios due by the end of the year and hasn't had to start using copper yet. As I see it, AMD maybe ahead now, but they probably won't be by the end of the year as they don't appear to have anything to put up against Willamette.

  • Some sort of hack that will post to slashdot for you with the same link.
  • by Acronym ( 7913 ) on Monday June 05, 2000 @02:05AM (#1025706) Homepage
    OK, this is a troll which I just fell for.

    It's a malicious CGI script which makes you follow up to the article, as anyone just seeing the article (or later on surfing at -1 when we all get modded through the floor) will find out.

    Frankly I don't give a damn about karma, but this is just *annoying*.

  • Agreed, Intel has had there problems delivering lately. AMD seems to have a little better track record of late with delivering on what is says it will.
  • by Eviltar ( 175008 ) on Monday June 05, 2000 @02:17AM (#1025708)
    If you appreciate this CGI script, then you can contact the original author at zk65@hobbiton.org [mailto]. You can also visit the author's web page here [hobbiton.org].

    :)
  • Pretty interesting little CGI script. To me this would constitute a trojan of sorts. Obviously it has, and looks like it will continue to dominate the thread. But maybe that isn't all that off-topic, due to the nature of the article. These are usually flamefests of Intel vs. AMD anyway, so having a troll take over seems somewhat natural.

    I wonder how many hit's this has received. Very Steven Wotson-ish in flavor, you can hear the cash registers ringing up the clickthru revenues. At the rate it's replicating, someone sure is having a laugh. At least the script seems harmless enough, doesn't appear to be singing anyone up for Pr0n of the week subscriptions or anything, just replicating itself.

  • by Black Parrot ( 19622 ) on Monday June 05, 2000 @02:23AM (#1025710)
    Skim the visible posts first. It looks like the "More Informative" link is a trap.
    --
  • by Linkmastah ( 195740 ) on Monday June 05, 2000 @02:26AM (#1025711)
    On a thread full of evil links, you can trust the Linkmastah to give you good links...

    The Register is running a story [theregister.co.uk] that AMD have unveiled the Thunderbird.

    CPU Review has a review of the Thunderbird here, [cpureview.com] and Sharky Extreme review it here. [sharkyextreme.com]

  • by cybrthng ( 22291 ) on Monday June 05, 2000 @02:35AM (#1025712) Journal
    Now what benchmarks actually make use and what motherboards actually support these CPU's to its fullest?

    I don't think AMD's cpu's are being optimized fully by any software/driver and hardware combination yet. The world is yet still waking up from Wintel and i'm sure with more optimizations, better memory to cpu bus and better motherboards on the horizon the same chip will outperform intel counterparts.

    Just my 2 pence

  • Please disregard the above message, it's a link I dubiously clicked on and it points to some CGI that ends up redirecting your browser to a POST url on slashdot. Sorry about that.


    "This .sig temporarily out of service"
    SkyHawk
    Andrew Fremantle
  • I click that damn link, go to another virtual desktop, then come back and see two slashdot windows up and think 'damn.. ns must have opened another window and not gone the the link.. guess I'd better click it again'..

    ..it must be fuckin monday morning. Bitchass trolls.

    //phizzy
  • Guess they fooled me.

    Does this mean /. has a virus now?
  • Heh, thanks. Someone got me once [slashdot.org]. Now I've become a little more cautious.

    This one uses a cgi, so you can't get the source, but here [multimania.com] is another way of doing it (save it locally and open it with vi, if you're paranoid like I am now).

    I'm not too well versed in http, but couldn't the slashdot comment submission script be patched to check for the referrer field of the browser and reject it if it isn't coming from slashdot.org?

  • I wouldn't trust Tom's review. He's a whore for bad benchmarking (remember his GeForce flub which completely alienated him from other sites?). Here's Sharky's [sharkyextreme.com], and I'm waiting for the FiringSquad [firingsquad.com] to do theirs.

    Don't rush out to buy your T-Birds just yet. To take advantage of the chip (or in many cases for it to work at all) you need to get a motherboard based on the KT133 chipset. That's KT, not KZ folks, because it was renamed at the last minute. KZ was an abbreviation for German concentration camps - now that's a naming flub, forget about Intel's "E" and "B" debacle!

    To be honest, I'm disappointed. Previously the cache divider had held Athlons back behind CuMine chips at higher speed, and now that it's integrated I would have figured the T-Bird would have been kicking ass all around the block. I'm sticking with my P III 700E overclocked to 1008 MHz for now...

    I'm watching Jerry Sanders, AMD's CEO, on CNBC TV right now talk about the T-Birds and I'm nonplussed.

  • by Phizzy ( 56929 ) on Monday June 05, 2000 @02:46AM (#1025718)
    ..pisses me right the hell off. He makes the Thunderbird look worse by taking any chance at all, no matter how insignigicant, to point out anything this chip has against intel. Everywhere you look on thie review, the benchmark graphs show the coppermine on the new BX133 even to the Thunderbird or ahead of it, and all Tom can muster is 'Thunderbird is able to leave its predecessor as well as Intel's Coppermine behind it, as long as this processor does not run on the BX133-chipset' Yeah.. So the Thunderbird can run with the new Intel's, as long as your don't run it on the best motherboard. Now I'm as big of a fan of AMD as anyone else.. I'm about to buy an athlon in the next week or so, but Tom is totally biased against Intel and it shows in his articles. He shouldn't have to be making excuses for the Thunderbird not beating the Intel like it should, he should be showing us that it doesn't, and questioning why.

    Fscking benchmarkers.

    //Phizzy
  • The parent post has another CGI "trap" link in it. It does not appear to be the original, because its timestamp is later than the original problem.

    --
  • Thunderbird is able to leave its predecessor
    as well as Intel's Coppermine behind it, as
    long as this processor does not run on the
    BX133-chipset

    I might be terribly wrong here, but there is no such thing as a BX133 chipset. The BX133 is a BX chipset overclocked to 133 mhz. AGP bus is overclocked to 89 mhz.

    You won't be able to buy a computer with the BX133 chipset.

    Johan V.
  • by puppet10 ( 84610 ) on Monday June 05, 2000 @03:08AM (#1025721)
    I agree Tom does have an AMD bias (maybe with some legitamacy recently) but this is not a symptom of it however. The reason he separates out the BX133 as being different is that this is NOT an accepted configuration by Intel, in other words it is overclocked. The closest thing to this is the Solano board which comes out behind the Thunderbird. Additionally Intel and motherboard manufacturers tell you that running a BX chipset at 133MHz is not recommended (although Tom actually has gone into detail in previous reviews on which boards can do this well).

    A different point can be argued, which benchmarks should be run to measure performance. Here we can see some difference, at Sharky Extreme [sharkyextreme.com] different benchmarks are used and slightly different conclusions are reached (Sharky's sometimes seems to bias toward Intel). However, the overall point of both is that the two chips are now basically the same in terms of performance for almost all purposes, and the Athlon is generally cheaper (and available at the moment).

  • Whoops, this took over the whole thread, shows how stupid people are...
    As somebody wants to view the source it is here [hobbiton.org] (this is the source to the cgi programme)
    and if you want to see how many people have visited go to the log [hobbiton.org]. Maybe I should disable it now....

    zk65@hobbiton.org
  • wow.cgi has now been edited to put its comments in sid=wow which you can access by going to http://slashdot.org/comments.pl?sid=wow [slashdot.org]
    If you want to run it click wow.cgi [hobbiton.org] or view wow.txt [hobbiton.org] (the source to the cgi)
  • I find it odd that the Athlon can be shown to be 33-38% faster in pure FPU benchmarks than Coppermine, but none of that performance seems to show up unless you make a test specific for it. Is this an indication that the Athlon CPU is great but that the chipsets are awful for it? What would an Athlon do with a chipset as good as Intel's?
  • I'm not sure why it shows how stupid people are.
  • Sure, with time the T-Bird will outperform its Intel counterparts but the problem with that is before long its counterpart will be Willamette which will be a tough act to top. In addition to that I've yet to hear what AMD is planning to counter Willamette - hopefully DDR enabled chipsets/mobos can pick up the slack because the T-Bird in its this incarnation doesn't appear to be up to the task. A disappointed AMD supporter...
  • Some reviewers are agreeing that the power of the AMD CPU's won't be fully used until the new chipset has arrived: the new chipset (760 ?) is due to Q3/Q4 2000, it will enable "good" AGP 4x (hopefully) and will support PC 266 DDR SDRAM.

    But I think that AMD should also add support for Firewire bus on their chipsets, this would differentiate them from Intel's offering (as Intel is a die hard supporter of USB/USB 2, I doubt very much that they would add also Firewire support on their chipsets).

    I think that people would appreciate much more to plug easily their DV camera to their computer (without having to buy an expensive card) than going from 88 to 95 fps in Quake III or things like that...

  • Hmm...I wonder why the results are slightly behind Intel's (or about the same). If you consider pure CPU power benchmarks which are 10% (30% FPU) it seems just weird. None of these so-called-non-biased-benchmarkers have written a single word about it.

    How would st. like gcc's -mcpu=athlon parameter for compilation change Athlon's performance? (I was told 30% -- could it be?)

    Anyone else tried to compare AMD's officials results to Sharky's or Tom's. What do you say?

    I have an alter-ego at Red Dwarf. Don't remind me that coward.

  • by lbrlove ( 164167 ) on Monday June 05, 2000 @03:53AM (#1025729)
    Tom's benchmarks (as people have already pointed out) are based on an overclock, and are only valid for the overclocker. The "BX-133" is just a BX chipset overclocked to 133-MHz FSB, and thus not something that Intel or most manufacturers recommend. Typical of Intel technology, the BX is an excellent chipset that was abandoned for market reasons.

    If compared on a testbed of Intel's choosing, it would be a 820 or 840 with 800MHz Rambus which lacks the performance of an overclocked BX according to general testing I have seen. Thus, the minor disparity between the Coppermine and Thunderbird in Tom's tests would likely evaporate into a strict cost or brand loyalty decision.

    Basically, AMD has caught up to Intel for most practical purposes, and need only the better chipset designs to make them their equal. Some SMP sets, and perhaps better optimized single-chip sets, would be much appreciated. IMO, they have an advantage in the area of memory by not being tied to a proprietary architecture like Rambus. When DDR-SDRAM becomes widely available, AMD will unveil even more "thunder" while Intel pays their penance.

    -L
  • I'm not too well versed in http, but couldn't the slashdot comment submission script be patched to check for the referrer field of the browser and reject it if it isn't coming from slashdot.org?
    Yes, it could easily... if (HTTP_REFERRER ne "slashdot.org") { print "Go away naughtly little boy."; }

    Or something like that... :-)

    I thought this was pretty standard in Perl scripts - nobody wants their scripts to be accessed from an outside source!

    On topic now: The Thunderbird looks very nice. 50W power requirement though, and according to Sharky Extreme it beats the PIII in Quake III by a fair margin except at 1600x1200 (i use that reolution all the time for playing games, honest). With a bit of driver optimisation and final design work this should be great - remember the motherboards are not really production quality yet, and they won't be for another month.

    At least AMD can deliver on their promises. Intel can't! I wonder how many 1.1GHz, 1.2GHz and 1.3GHz processors they are stockpiling at the moment - like they stockpiled 1GHz Athlons for months before they released them. I wouldn't be surprised if there were a few Thunderbirds capable of running at 1.5GHz at AMDs fabs - apply one of those cooling solutions to that and you could have a 2GHz machine - ideal for running Windows 2000 on! :-)

    Still, I will wait a few months for the technology to stabilise, and for the prices to drop. I wonder if any of the motherboards will come with L3 cache on them - the logical extension. 1 or 2 Mb of on-board L3 cache would just bitchslap Intel into touch... drool.

    Shame about the trolling though. I hope that those in power sort it out soon, and give themselves a good slapping over missing out that HTTP_REFERER thing in the Slash code.

  • It doesn't even need to be done with a CGI script. A simple redirection in a static web page is all you need to do the trick. I demonstrated this last month here on Slashdot--got about 900 posts in 48 hours (my link gave a specific warning what it would do though.) At any rate, there was some discussion of the problem here [slashdot.org]. Plus links to PHP source to do it, etc.

    Near term solution--run your mouse over the link to see if it's suspicious. This offers little protection really, but it's the best you can do.

    numb
  • by hattig ( 47930 ) on Monday June 05, 2000 @04:11AM (#1025732) Journal
    The problem is that most software is written with Intels FPU in mind - 2 units not fully pipelined. Not 3 fully pipelined FPUs like the Athlon has. So most software won't run at full efficiency on the Athlon and it won't be getting enough instructions to keep its FPUs running full tilt. I am sure that before long games and renderers will start taking advantage though - that extra framerate could be essential!

    So software that has been optimised a little for the Athlons CPU does indeed show 30% better performance!

  • I'm not too well versed in http, but couldn't the slashdot comment submission script be patched to check for the referrer field of the browser and reject it if it isn't coming from slashdot.org?

    If the slashcode just forced the post to be previewed before submitting then it wouldn't be a problem. Forcing people to preview first would also have other beneficial side effects.

    numb
  • No.

    This is indicative of the benchmarks. All these benches are either Winhoze luser performance (whatever that means) and/or Winhoze gamez performance (whatever that means).

    There is no real benchmark data: no lmbench (context switching, etc), no linpack (real FPU), no database benchmarks, etc.

    Nada. None...

    So what quite a lot of slash readers are interested in - namely how does this beast shovel pages under linux, BSD, Slowarix or even NT is not present. So I guess you will have to buy a cat in a bag.
  • I must be special, it's showing up over threshold.
  • Other people have already pointed out good reasons for very limited overall performace gains. There is one more thing though. In most of these tests (except for very low resolutions) the performace was limited by the *video card*, and not the CPU. Check the graphs. At 640x480x16 you can notice some performance difference, while at 1024x768x32 most CPUs perform pretty much the same -- because they work faster than the video card.

    When the original Athlon was first released, Tom also ran a 3D Studio rendering benchmark, which is purely FPU-intensive. It did show Athlon to be about 40% faster. I wonder why he didn't do it this time...

    ___
  • by leiz ( 35205 ) <leiz@ j u n o . com> on Monday June 05, 2000 @05:09AM (#1025737)
    There is a Linpack benchmark available http://www.aceshardware. com/Spades/read.php?article_id=156 [aceshardware.com]. Ace's hardware ran the test against a K7 athlon, a PIII, a celeron, and an UltraSparc II... Thunderbird kicked ass, of course, it's not exactly fair to match a 1 Gz thunderbird against a 733 PIII.


    Zetetic
    Seeking; proceeding by inquiry.

    Elench
    A specious but fallacious argument; a sophism.
  • What distribution are you using? I can't think of one that lets a regular user nuke /usr/lib/*

    Or maybe you are using the root account for everything (snicker), this isn't Windows you know.
  • by Tassach ( 137772 ) on Monday June 05, 2000 @05:22AM (#1025739)
    There's an article on ZDNet [excite.com] that (for once) has some good information. To quote the article:
    The new Athlons will be marketed under the same brand name but will be marked as "performance enhanced." AMD will offer them in speeds of 750MHz, 800MHz, 850MHz, 900MHz, 950MHz and 1GHz clock speeds. Pricing will range between $319 and $990.

    And...

    While the Dresden-made chips will be wired with copper, AMD said same-speed Athlons produced at the company's Austin, Texas, fab will use traditional aluminum wiring. Nothing on the products' packaging will note whether the chips, both made using the 0.18-micron process, contain copper or aluminum wiring, an AMD official said.

    I'm not too thrilled about this last part, the fact that there won't be an easy way to tell the difference between an aluminium T-bird and a copper T-bird. I'd imagine that copper vs aluminium will make a big difference in terms of heat and overclockability. I would imagine that the copper T-Birds are going to run cooler and overclock higher than an aluminium chip.

    I seem to recall seeing a web site somewhere that gave directions on how to decode the Athalon's serial number; and that part of the information available therein was which fab line it came off of. Does anyone have that link? Then, all I'll have to do is find a dealer who'll let me look at the serial # of the chip before I buy it. $319 for a 750MHz sounds like a sweet deal to me (The article didn't say if the prices quoted above are estimated retail or AMD's price for 1000 chip lots)


    "The axiom 'An honest man has nothing to fear from the police'

  • Tom has been pushing the 133Bx for some time (don't know why intel never made one). But the follow up to 810E will have 133 fsb/sdram and agp4x, so there is hope.
    ---
  • Excellent idea. I second that.

  • AMD can scale their chips up to Willamette speeds now, and even further by shrinking to 0.13uM. Willamette will initially be a 0.18uM process and be in very limited quantities until at least Q1 2000.

    The Thunderbird doesn't require copper interconnects yet, and gains no benefit from it, but it will when transitioning to 0.13, same as Willamette will.
  • by Ozzy ( 119339 ) on Monday June 05, 2000 @06:06AM (#1025743)
    If you really must have a copper T-Bird, the aluminum ones have a blue die, and copper has a green die.

    But, there is no difference between copper and aluminum in terms of performance or overclockability (yet! wait for 0.13uM)
    see FiringSquad Review...

  • Don't rush out to buy your T-Birds just yet. To take advantage of the chip (or in many cases for it to work at all) you need to get a motherboard based on the KT133 chipset.

    Where did you get this nugget? I hadn't seen anything (yet) that indicated that I wouldn't be able to slap a T-Bird onto my existing motherboard (a FIC SD-11 [fic.com.tw]). I wasn't planning on buying a new motherboard until the 2-way SMP socket-A boards hit the market.

    From what I've seen so far, all the new KT133-based mobo's are socket-A and not slot-A. The T-bird is supposed to be available in either slotted and socketed form factors. I don't see any point in AMD selling slot-A t-birds if they won't work in existing first-generation Athalon motherboards. I could see where you might not get optimal performance out of a t-bird without the new chipset, but for it not to work at all with the older chipsets is really bad - it's not like they've added any new functionality that needs chipset support.

    The only major differences, AFIK, between t-bird and first generation Athalon is on-die L2 cache, 0.18 micron process (vs 0.24 on 1st gen IIRC), and some tweaks in the core. I didn't notice anything in the literature that said t-bird is running at a different voltage or anything like that. If you have specific information on why T-birds won't work with the AMD-751 / VIA 686A chipsets, please let me know.

    As to your dissapointment in Athalon's performance, I think most of it is due to the fact that most pre-compiled software is optimized for Intel chips. IIRC, the lastest version of GCC has a switch to optimize for Athalon. From what I've seen so far, this makes a noticable difference. (Obligitory karma whoring: Yet another reason to use open source software!)
    "The axiom 'An honest man has nothing to fear from the police'

  • This shows the problem with special optimizations. People often talk of their "cool ASM hacks" which give them another 5 to 10% on some piece of hardware. The problem is that a year or so later, when that hardware is dreadfully obsolete and we all have new hardware, the software performs worse, or not much better at all because of the hack. This is why clean coding would be good.

    If all software was written in a general way for a "generic" x86 processor, we'd see how truly nice the Althon does compared to the PIII (which would stall out constantly, I'm guessing)

    Although on a static platform (console unit which can never be upgraded), such performance hacks make sense.
    ---
  • Not to disagree with you about Tom's bias, but you do realize that the "new" BX133 is just the old BX chipset running at a 133Mhz FSB, right? Since the BX chipset doesn't technically support the 133 Mhz FSB at all, and Intel has been trying to phase it out for a long time, it is in fact something of an exception, even if it runs just fine.

  • From what I've been reading AMD does not recommend using their new Thunderbird Socket CPUs with KX133 slotA boards, such as the Abit KA7 (even though Abit say it's T'bird compatible). Well I've been delving a little deeper & it seems that while the T'birds from the Dresden fab30 (copper interconnects) are at best iffy& at worse won't work with a KX133 board (via a slocket), many seem to think the Austin fab25 (aluminium interconnects) T'birds do seem to be compatible with KX133 boards, such as the Abit KA7 (via a slocket). Well I just read at www.sharkyextreme.com that the Austin T'birds will have a blue core, while the Dresden T'birds will have a green core. Plus as they are 'flipchip' type socket CPUs, like their previous K6 designs, however not with the aluminium 'primary heatsink' cover, you'll be able to see the colour of the core, unless the fansink is pre-installed & epoxied on. So it seems that if slockets become avaliable for the T'bird, & you already have a KX133 board, such as the Abit KA7, you might be able to upgrade to the T'bird if you buy a green one (& cross your fingers). As their are a few P6 boards that have both a slot & a socket, hopefully a few of the board makers will come out with VIA KT133 boards that also have both a slot & a socket on them too. Apparently, also, AMD will be supplying slotA T'birds to OEMS for use with such Irongate (AMD 750 chipset) boards as the MSI K7Pro, which are said to be quite compatible with the T'bird. Hopefully some of these OEM SlotA T'birds will filter down to the reseller market.
  • Just compare the performance of a early FIC SD11 with its immature BIOS & the later SD11 board revisions with later BIOS revisions - there's a huge difference, relativelly speaking. Also compare a VIA 133A board with old 4in1 chipset drivers to the same board with the latest 4in1 drivers - again a huge difference, relative to what changes have been made.
  • Need I say more. Also the Austin ones are said to be compatible with the earlier generation Athlon boards, via slockets. The Austin ones will also be avaliable in slot form (as well as socket form) for OEMs.
  • AGP4x is faster than you can push data to it based on the speed of CPU and memory, today.

    AGP4x should be useful with a 266mhz FSB and DDR133 memory.

    And finally, Firewire doesn't really belong in the chipset, though I do agree it has its place on the motherboard.

  • If the slashcode just forced the post to be previewed before submitting then it wouldn't be a problem.

    Couldn't the bad-guy auto-post script just fake the HTML state that indicates you've previewed the comment at least once? You'd have to keep that state in a server-side database, which would add overhead to the existing system. How much overhead, I don't know.
  • ...all Tom can muster is 'Thunderbird is able to leave its predecessor as well as Intel's Coppermine behind it, as long as this processor does not run on the BX133-chipset' Yeah.

    'bx133' is a name Tom invented for a 440BX chipset running at 133 MHz FSB. I'm willing to bet the only reason it outperforms T-Bird at all is because of the overclocked AGP (89 MHz, 2/3 the FSB clock). BX can't run AGP at 1/2 the FSB clock.

    The i815 Beta comparison is probably more realistic. The AGP is running within spec at 66 MHz.


  • The comment could be stored when they click "submit", then when it is previewed back they could have the option to "keep", "delete", or "go back and edit." They might have to add an IsValid field to the comment database to do it that way.

    There's another way I thought of also that wouldn't force previewing. The slashcode appears to generate a random formkey to make sure the comment doesn't get double-submitted...it won't accept 2 posts with the same formkey. They could keep track of valid formkeys possibly.

    Either way you're right--the server-side needs a way to keep track of state. How much (or little) overhead it adds depends on how creatively it's implemented.

    numb
  • I thought the dual processor workstation market was the market the thunderbird was supposed to grab some parts - not ?
  • You can't trust benchmarks because it is so easy to skew their results. I think you should buy the AMD chip because it is cheaper and the performance is as good or better.

    There is also another important thing to consider: acording to http://www.netcraft.com/whats/, www.amd.com is running on apache under linux. Likewise www.cyrix.com is runing apache under bsd. However, www.intel.com is running some sort of micros~1 crap.

    I am not sure it that is good enough to sway your chip buying practices, but it is food for thought.
  • I've created a page on my site [alereon.org] that has links to all press releases, reviews and news bits regarding the Thunderbird. You can find this page here [alereon.org].
  • Although still in the works, below is the AMD announcement over the AMD-760 chipset which enables DDR-SDRAM. This stipulates that 266MHz DDR SDRAM can potentially deliver 2.6 times the memory access bandwidth of traditional SDRAM (due I think to lower latency rates in concert with higher bandwidth). Read it here [amd.com].

    I have heard release dates as early as late July, but you know how that goes.

    -L
  • Quite right on the AGP. And even if we could push data across it fast enough, it's not much of an issue. With 32MB and texture compression on most cards, they rarely have texture bottlenecks where they're swapping out textures all the time. A decent PCI Gefore 2 would probably only score 10% (tops) slower than an AGP 4x card.

    And, I think you're wrong on the Firewire issue... stuff like that will never get adopted until it's on chipsets. And it's good for a lot, it's basically serial scsi, great for digital video, external harddrives, etc.

    USB is aimed at mice and other low bandwidth or temporary devices (scanners). USB2 is fatally flawed and even if working perfectly, not terribly exciting seeing as how it requires CPU interaction on all transfers (device -> cpu -> device) which limits it to maximum of half its rated speed. Then you get all the old USB1 devices (like mice) slowing the whole bus while they transfer data, not to mention that the handshaking between fast devices must be done at slow speeds... Ugh. Not only does the handshaking eat bandwidth, but it eats essentially 10* as much because it's slower.

    Firewire is an essential connector for the future, leave USB where it belongs, as a replacement for serial, and let firewire take over the high performance niche.
  • Ace's hardware ran the test against a K7 athlon, a PIII, a celeron, and an UltraSparc II... Thunderbird kicked ass, of course, it's not exactly fair to match a 1 Gz thunderbird against a 733 PIII.

    Ace's intent wasn't to provide a fair benchmark, but rather to generate a graph which would clearly show the way differing cache architectures affect Linpack results at different data sizes. Frankly, it's a brilliant graph and a perfect example of how a well constructed benchmark run can show what's really going on inside a processor. What it wasn't meant to do was to say "Athlon beats PIII" or anything like that.

    In any case, this benchmark shows just how misguided the parent to this thread is. In fact, Linpack doesn't show "real FPU" performance, but rather data bandwidth. Depending on the size of the data set one chooses, you can "show" that an Athlon "Classic" 750 is 1.5x "faster" than a PIII 733 (data set = 64k), or that it's 1.5x "slower" (data set = 180k). Or you can "show" than an UltraSparc II 400 is 3x "slower" than that PIII 733 (data set = 180k) or that it's almost 2x "faster" (data set -> infinity).

    The point is that Linpack, while useful as a synthetic benchmark, must be taken as a whole graph, not as a single number, and that it is at least as dependent on cache and memory bandwidth than on CPU core. Furthermore, it has almost no relevence on the performance most people (i.e. anyone not running scientific models) can expect with the way they use their computers.

    Quake 3, on the other hand, is a marvelous benchmark, which can be easily understood, yet tuned to test most major bottlenecks in a PC. Most importantly, it much better models some of the conditions that most PC buyers put their computers through.
  • Just gave the go-ahead for a friend to build me an Athlon 700 system w/512MB RAM, G400 32MB, two huge IBM HDs, video capture, SB Live, CD-R, etc...

    Think it's worth waiting for the Thunderbird? Any idea when I'll be able to buy motherboards/CPUs for it? I really don't want to wait long. Been waiting long enough already! (First planned to buy the thing last October then kept puting it off for a couple months at a time...)
  • k7 thunderbird is a better cpu than a p3. That doesn't mean systems based on it, are necessarily better though. There are no shipping KT mbs yet anyway, so the pre-release numbers have a chance of improving

    What tom does is most relevant. Comparing systems as a whole. He along with every other mainstream techy site thinks his readers just want to play games. Tbird's only advantage is faster L2 cache speed, and games don't care about cache.

    You shouldn't be asking him to root for the same processor you would like to win.

    Anyways, if you want to just focus on the processors, and not the mbs, then just compare the KT to the apollo. Your pecker enhancing athlon measures as long as you think it does.

    The viewperf benchmarks show 20-30% difference over the apollo P3.

    Thunderbird should be good when the 760 comes out. You have to wonder how committed mb makers can get to the KT chipset, when 760 is 2-3 months away.

    I might get a Duron.

  • Don't worry. Your SD11 should work fine -- only the KX133 chipsets have problems with the Thunderbirds as far as I've heard. (and I HAVE heard that the AMD 751 based chipsets will work with the Thunderbirds.)
  • Thunderbirds will only be available in Socket form to the DIY market. Slotted processors will, supposedly, only be available to OEMs.
  • > kept puting it off for a couple months at a time...

    Clearly, the optimal time to buy a computer is when you're on your death bed.

    I posted a "when to buy" query to some Ask Slashdot article a month or so ago, and based on the responses decided to wait a few more months to build. Mostly we will see the expected gradual increases in top-of-the-line speed and decreases in price for a given speed.

    However, there appears to be a substantial discontinuity in progress when DDR memory shows up, and the posts to the aforementioned thread left me with some hope that it would be here by late summer or early autumn, and would not be much more expensive than conventional RAM. (Indeed, I read a couple of news articles over the weekend that seemed to imply that people are getting DDR engineering samples already.)

    If anyone can give updated information on the expected discontinuity, I'd certainly be glad to hear it.

    --
  • One thing to consider is that Tom is not only benchmarking an overclocked Chipset, he is also benchmarking an overclocked Graphics card, Hard Drive, etc. Remember that the old Intel BX Chipset have syncronous clocks!!!.

    Many of the test are dependant on the Video Card, so they are realy pushed by the 33% overclocked Intel BX. And you need a realy new Video Card to support this kind of overclocking.

    So I think that the Intel BX have to be taken only as a reference...
  • I can't believe you assholes fucked me like this. I'm reading an article, I click the fucking link, and you trigger happy moderating assholes RATE ME offtopic and troll, into oblivion! Well FUCK YOU TOO! I didn't even know it was going to do this - I had no idea it wasn't an informative link, and thanks to you assholes, my Karma droped below 25!

    Rob said it was a real shame, and it was messed up, but he isn't going to start giving Karma back to those of us who were victimized by this, so tough. It's too much of a headache you see, too much like work. So I have my own solution:

    EVERYONE WHO GETS ONE OF THESE IN METAMODERATION - DISAGREE WITH IT - COST THE MODERATORS A POINT PER POST - FUCK THEM LIKE THEY FUCKED ALL OF US WHO FELL FOR IT.

    Hey Rob, Thanks for that tarball!
    Scott
  • hmmm... this certainly leaves a love-letter type taste in your mouth.
  • Why doesn't AMD make a Xeon killer? A Thunderbird with a one meg on chip l2. And where are the SMP motherboards. I want to build a killer server.
  • Actually, the Abit KA7 (which has KX133) will work with a Thunderbird Slot A, if you make certain mods in the BIOS. I am not sure what you would want to put a T-Bird in the Irongate boards for, as the KX133 boards offer a lot more features (PC-133 and VCM, 4x AGP, ATA-100, 1.5-2GB memory, etc..)

    The Slot A T-bird is supposed only to be released to OEM's, whereas regular buyers are stuck with Socket A availability only, and therefore must get KT-133 mobos. There has been implication that "Only to OEM's" means only the big guys like Compaq and Gateway (Dell? they still have not sold a non-intel box, unfortunately.) as opposed to your corner parts and slapped together units computer store, or your favorite web protal for computing goodness. If that happens, slot-A tbirds will be harder to come by, and more expensive.

    Personally I am hoping that the Slot A T-bird is more available since I have a board that will use it.

I have the simplest tastes. I am always satisfied with the best. -- Oscar Wilde

Working...