Announcing Smithfield; Your Dual Core Intel CPU

13 Comments

View All Comments

• ncage - Tuesday, August 24, 2004 - link

  Ya #12 im thinking on backing my pc up the the nuclear power plant close by :).

• wassup4u2 - Monday, August 9, 2004 - link

  Anemone: The 915/925X were dead ducks at launch. They don't give any performance increase over 865/875, and Intel's bringing out a new chipset in 2005 anyway. Well at least it's supposed to be in 2005. You never know nowadays.
  
  I certainly hope Smithfield isn't based on NetBurst. Single core Prescotts use ~105W, whereas Dothan uses ~35W. A dual core 90nm Netburst chip would just about need water cooling. Somewhere around 200W for a CPU, combined with a fancy new video card, and you'll need a kilowatt PSU. :-)

• KrazyDawg - Saturday, August 7, 2004 - link

  Can we get discounts? ;)

• ThePlagiarmaster - Friday, August 6, 2004 - link

  reread that, should say "BUY the one" not "BUT the one"...meaning buy a plextor (think they work on all).
  
  Also should read haven't sold many 3400+ (and up). Most go for 3200+ or below on A64. Cleared that up, I'm too tired to spellcheck more...ROFL
  
  One more thing, I'm excited about anything dualcore, so bring on the Intel too :) As a reseller I get to play with both anyway for a bit while building somebody's pc..haha. (drooling myself here now...heh)
  
  Plag

• ThePlagiarmaster - Friday, August 6, 2004 - link

  KristopherKubicki,
  I know bandwidth isn't an issue with 1 cpu (as I stated, shows no difference really, as 754's are right on the 939's butts), but surely this will change at least some with 2 cpu's fighting over that bandwidth?
  
  Anemone,
  Where did you read about dead 64's? I have heard none of that and I sell them :) I'd have to think these are newbs pushing things too far with voltage. If you just take what you can get with .05 added or less you should have no problems. Anand's did most of their overlcocks at default voltage. Granted you don't get as much, but is 100mhz more really worth death? I usually tell customers to overclock to whatever possible at default, then go further the day you say its getting "slow". At that point your thinking of buying a cpu already to upgrade. So you wouldn't be depressed if it dies anyway :)
  
  Then again, if you get the boxed version you don't care either. Death just means you send it back to AMD (or your shop if you're under a year still). Not much risk in my book. Even when I'm the one looking at getting them all back...LOL. Not that I want people to kill their chips, and as said I do tell them to just go default voltage until needed. Most of the ones I've sold so far run default at 200mhz more than rated pretty easily (then again, haven't sold many 3400's or fx's).
  
  Why would you need a sata dvdrw? Thats what the Eide's are for (ok, buy Nforce3 it's better anyway). Sata dvdrw's just take away a spot for that next HD. Should be a BIOS issue anyway. If not SATA 2 with Atapi is around the corner. Cheaper boards only come with 2 sata's. There are very few models available for now to begin with. But the one that doesn't say "only works on Intel" *cough* msi *cough*.
  
  I wouldn't stick junk modules in any system. I only sell Corsair, Crucial and Kingston these days. So memory is not an issue. When Corsair or Kingston 512's (Cas2.5's) are going for $90-98 (depending on day of the week it seems) I don't see the point in selling others. Then again most people that have problems have them because they rely on the SPD. Set it up yourself and you can use pretty much any memory. Yes JUNK too, but you wouldn't be buying crap memory to overlock :) SPD's are completely overrated in my book. Almost useless, and it only takes a few seconds to set it up yourself in the bios. The trouble with the world today is we have too many people building pc's that have no business doing it themselves. They end up taking back parts until they get one that works out of the box because they have no idea how to set up a PC. Then they blame the chip or the board and badmouth them. Witness all the big white tags (rma's that work) on boxes at Fry's ;) As far as convincing goes, I only need to know who's faster at the time I buy. A64 is tops now, that should be the same in Dualcore's. Not much different than what we already see with dual Opterons vs. Dual Xeon's. Bandwidth issues on P4's will show up on dualcore p4's even more. I can't believe we have to wait for the dualcore Dothan's. Bummer. Why hot P4's first? Hmm..Dothan is not bandwidth starved quite like a P4, nor even nearly as hot.

• Anemone - Thursday, August 5, 2004 - link

  I'm listening - just not yet convinced I guess Plag. There is a durability issue to consider, and the fact that at the moment at least, the P4's do OC rather well in the right mobo, though admittedly the AMD side of things and the freedom they give is very enticing.
  
  I've read a ton of articles on dead AMD64's though. Push them and you risk killing that lovely memory controller...
  
  I guess durability counts for something in my mind atm, and I'm recalling the years it took for the Nforce boards of old to get things right. And even now, you can't just stick any memory card into an AMD board, you have to check first. The SATA DVD-RW drives not working with Via KT800 boards, the problems are not few, and the worry would include that you might not know you have issues until you come time to upgrade some part, and find that you buy it, get it, and it doesn't work, or works but not well.
  
  But I am reading your comments and they are at least causing me to think on the matter.
  
  :)
  

• KristopherKubicki - Thursday, August 5, 2004 - link

  ThePlagiarmaster: Dual channel really doesnt help as much with A64s because the memory controller is on the CPU; there really is not a bandwidth bottleneck ATM.
  
  Kristopher

• ThePlagiarmaster - Thursday, August 5, 2004 - link

  No idea if it will require a new chipset (these things can change later in the process as they find something doesn't workout). However, what can be said is if you [can] drop them into current boards you're going to be sharing bandwidth that is already fully used by a single chip. P4's need serious bandwidth to work fast.
  
  The Doom3 article here, shows Athlon64's are almost using NOTHING of dual ddr. The 754's are right on their butts the whole way. So, if you (theoretically) [can] drop a dualcore A64 into current 939 boards, you'll have almost the whole 2nd channel dedicated to that chip.
  
  In the end though, I think you'll need a new chipset. Intel loves the switching as they sell chipsets/boards too...not just cpu's. AMD is the opposite, they have to let Mainboard/chipset makers "milk the cow" of current R&D for as long as possible or piss them off. In that respect Intel has a great advantage. They can change sockets/chipsets on a dime and just take more marketshare for their own boards. Remember the 667fsb that croaked and changed to "oops..just kidding 800fsb now guys". Lot of mainboard/chipset makers got screwed there (with 667 designs all done..sheesh), and it gave intel back 50% of both of these markets. With this thinking I'd almost guarantee a chipset change. Or at least some change that will require reworked boards, that won't support the old cpu's. I'd also venture to say most will need a new PSU for Smithfield even if you just bought one last week (on Intel OR AMD's side). TWO of these power mongers on todays PSU? With Intel saying you need 30A on the 12v lead, how many will 2 cpu's take? 45A or 50A? I may be completely off here though. While upping it some makes sense for 2, I'm in no way a PSU/Electical engineer :)

• Anemone - Thursday, August 5, 2004 - link

  Ok then!
  
  I don't think Itanic falls in the same scale of items that are late, but that's just my opinion. You're right, only market and technology leaders introduce more advanced products without having to do so due to competition.
  
  I would tend to say, Intel has hit their goals and releases vastly more often than AMD has. I do agree though both have failed to meet targets.
  
  Smithfield (back on topic) is going to require a new chipset?? Does that indicate that the 915/925 are dead ducks? Who can say?
  
  

• ThePlagiarmaster - Wednesday, August 4, 2004 - link

  Anemone,
  
  Hate to rain on your parade, but AMD has no need to give you anything faster this year. Intel can't catch up and AMD's .13 process will scale to the FX55 chip.
  
  It's standard business practice to NOT forage ahead when your competitors (just Intel in this case) are not forcing you to do it. Lets face it, everybody milks the cow as long as possible to maximize their returns on R&D investments. If they released a .09 chip that overclocked great, how many would buy the .13's that are in the market already? Nobody but the idiots would buy them (unless heavily discounted). That translates into a lot of lost money. With no heat issues on .13, why release your .09's (note engineering samples supposedly overclock quite well..They could put these out if they wanted to). You really should be blaming Intel. When they cancelled their roadmaps and completely shifted direction AMD just laughed and decided to sit on their collective A$$es for another 6 months. What did you expect?
  
  BTW Intel "has this love of putting out press releases on products that miss the mark by a 2 years or more"...LOL (was it 3yrs late? I forget). Just ask how long it took to release that ITANIC! How about the VAPORWARE that is announced then cancelled just months before launch? Just ask where that Tejas went (jayhawk, tulsa?)! Intel (and world+dog) were claiming 4ghz by last december not too long after the P4 northwoods came out. We won't even see 4ghz this year from them (yeah, thats a year+ late too!). Well, maybe another paper launch :) Not surprisingly this used to be AMD's bag. Whoever is BEHIND paper launches while whoever is AHEAD sits on their laurels a bit (same with Nvidia/ATI).
  
  Yes AMD will deliver dual core in 2005 (opterons). To your desktop if Intel can force them to do it. You're forgetting how easy it is for them to do it. It was designed for this already, by Alpha engineers who already knew how to do this. A simple bridge, slap on a core, and bam you have dual core. Slightly more complicated if they have a 2nd cache for the 2nd chip, but still probably not difficult compared to Intel's job. Intel is retrofitting a chip that was NOT designed to do this from the get-go. Much more difficult than a chip that was DESIGNED to be dual core, but they just chose to give you ONE core. Intel also has NO dual core experience in cpu's. Again much more difficult your first time out. Thats TWO very important things working against them that AMD doesn't have to worry about. A chip thats NOT made to be dual core, and engineers learning the ropes. I'm not saying Intel is stupid, they will figure it out. But how long will it take and how much will it cost them (US?) to get them? Intel has this love of releasing products that need to be RECALLED (4 in 4 years or so). Will this be recalled too?
  
  Who cares about FX55 anyway (or any FX when your pockets aren't that deep)? I'm more interested in 90nm CHEAPER chips that overclock to above FX55! Which will be out by years end.