View Full Version : Tbred has disappointing headroom at first glance
OK, you have all seen the initial results. I am very surprised. I think there is a problem here, with the process, the architecture, or perhaps even the testing. What is your take on this?
In all honestly I think AMD has a manufacturing problem. I don't think they're ready for 0.13 micron process yet. Just look at the voltage they need to feed the XP2200+! 1.65V?!? That's 10% more then the T-Bred XP1700-1900+'s! Looks more like they're using an overclockers trick to get the faster CPU's.<BR><BR>The only site which seemed to have a good OC was <a target=new class=ftalternatingbarlinklarge href="http://www.amdmb.com/article-display.php?ArticleID=187">AMDMB's</a> which I have to say is very odd. They do over 2 GHz with air while other sites using Water can't even break 1.9 GHz. <BR><BR>i'll definitely be skipping the initial steppings of the T-Bred and wait for good news before I buy mine.
<i>i'll definitely be skipping the initial steppings of the T-Bred and wait for good news before I buy mine.</i><BR><BR>You and me both.
Wait, what about Barton?<BR><BR>Barton is supposed to be derivative of the Thoroughbred but with 512 L2 cache right?
The Barton will be out l8r this year, yes it will have 512KB L2 cache... It will give the Athlon a nice clock for clock boost! But by that time they'd better have their 0.13 micron manufacturing process all smoothen out since Intel will probably be close to 2.8-3 GHz! <img src="i/expressions/face-icon-small-sad.gif" border="0">
bigchin
06-15-04, 01:25 PM
This is a new process for AMD and they need time to twisk it. You know they relayout the chip and die shrink. They have been twisking for the past couple months and only 1700+ and 1800+ would work well. All others are OC devices. Based upon my understanding, their design would be good to 2800+. It might max out around 2400+ to 2500+ w/o OC. I think the Tbred might max out around 2300+. The new Barton might max out 2500+ with additional twisk.
This will be interesting to see how this works out.<BR> I would have expected to see the voltage drop with the die shrink<BR>Instead as colin pointed out it's higher, higher wattage disapated& smaller surface area =even more heat problems (as if the xp's didn't die quick enough, when the cooling fails)<BR> a few problems on a die shrink are not unexpected, but the shold be fixed <u>before</u> public release! AMD may be risking it all to meet a release date, rather than delaying until its ready (shades of Intel!)
I know AMD is new to the 0.13 micron manufacturing process and AMD did improve their 0.18 manufacturing quite a bit with the T-Bird and Palomino...<BR><BR>Tshirt, I wouldn't say they're problems since they do deliver at rated speed. They're just not good for overclocking... Yet. I'm sure AMD will refine the 0.13 and will produce good overclocking chips.
bigchin
06-15-04, 01:25 PM
The extra power could come from leakage current. Look at P4, after die shrink to 0.13 the leakage current jumps to 18A. In order to fix this power drain issue, AMD need to have good design engineer or product engineer to redesign the chip to reduce leakage current. Intel is in the process of redesign their P4 to fix this problem. AMD is new to 0.13 and probable they don't know how to fix it.
<br><br><< <i>The extra power could come from leakage current. Look at P4, after die shrink to 0.13 the leakage current jumps to 18A. In order to fix this power drain issue, AMD need to have good design engineer or product engineer to redesign the chip to reduce leakage current. Intel is in the process of redesign their P4 to fix this problem. AMD is new to 0.13 and probable they don't know how to fix it. </i> >><br><br> bigchin, thanks for the heads up. <img src="i/expressions/face-icon-small-smile.gif" border="0"><BR><BR>I didn't know that, where did you happen to stumble upon that info?
bigchin
06-15-04, 01:25 PM
It's in the P4 data sheet. You need to read it carefully then you will spot it. I am in the semiconductor sector, we usually know what kind of problem people will know. We have predict that P4 will have at least 10uA per gate-transistor leakage. If you mutiply by 50 mil transistor then you can figure out how much leakage current on the P4. The AMD chip have significant less transistors so their leakage current should be less than Intel part. However, AMD has process issue, the leakage cuurent per transistor is higher than 10uA now. that is the source of problem.
Wow! bigchin thanks for letting me know that!<BR><BR>In your opinion, is it a faulty core design or just immature manufacturing that's causing this rather large leakage?<BR><BR>Sorry, I'm not trying to be a smart ass dude, I know absolutely zero about this... Just trying to learn a bit! <img src="i/expressions/face-icon-small-smile.gif" border="0">
Well you might be interested in doing some reading, the site I moderate for has had a few guys do some test runs with a couple TBreds. <BR>If your interested, check them out here.<BR><a target=new class=ftalternatingbarlinklarge href="http://www.xtremesystems.org/forums/showthread.php?s=&threadid=1261">Tbred</a><BR><a target=new class=ftalternatingbarlinklarge href="http://www.xtremesystems.org/forums/showthread.php?s=&threadid=1264">Unlocking said, TBred</a><BR><a target=new class=ftalternatingbarlinklarge href="http://forum.oc-forums.com/vb/showthread.php?s=&threadid=100600">Another Unlocking</a><BR>The last I dont moderate at, just a senior there.<BR><BR>IFMU
vBulletin® v3.7.2, Copyright ©2000-2010, Jelsoft Enterprises Ltd.