Prescott Heat Issues...

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
I agree, intel spends hundreds of millions on thermal research.. they're not idiots
Well, even if they have a problem, they have like two and a half months to enhance the process. Two and a half months is not a short while - even more so if you consider the amount of money flowing into research in the mean time.

<font color=red><b>M</b></font color=red>ephistopheles
 
I haven't seen anyone claim that the 3.2 is 100W.
The max theoretical power draw for the 3.2 Northwood is somewhat over 100W. The 82W or whatever is "thermal design power."

But of course, the 100W max is "theoretical," which means it's rarely ever reached and never stays at that level for long. Using HyperThreading brings typical power consumption rather closer to max though.

<i>I can love my fellow man...but I'm damned if I'll love yours.</i>
 
The PC runs fine. All the heat has to go somewhere though. In my previous place I had the PC in a large room so I never noticeced the heat being blown out, but in my present situation it's noticable. I have heat coming from the CPU, the Videocard, and the monitor mainly of course. It easily heats up my room. There's no way I could tollerate any more heat from the PC.

What I've ended up doing towards the end of the day is opening up the window and blowing a household fan on the monitor towards the window. It keeps the temps in the room down a few degrees. I only wish I could use that fan to blow on me instead of the damn monitor though.

<A HREF="http://forums.btvillarin.com/index.php?act=ST&f=41&t=389&s=1fee5dab901bebe29da7aa1c2658fc6f" target="_new"><font color=red>dhlucke's system</font color=red></A>

<font color=blue>GOD</font color=blue> <font color=red>BLESS</font color=red> <font color=blue>AMERICA</font color=blue>
 
FYI,

it has not been confirmed whether that PC Watch article got it's info based on an old sample, or a recent leaks from Intel. Supposedly, Prescott is said to have a TDP of 103W at 3.6Ghz. This is 15% higher than Intel's original expectations. Of course, as others have mentioned, Prescott's TDP is around 90% of MAX heat output. This implies that not only has HT been improved, but the efficiency of the core overall. Also, many believe that this increased heat output is due to leakage. There is a great debate at Ace's going on about this. The one thing wrong with this argument is that Prescott samples have been running around at 3.2GHz since last November. And, IIRC, Intel demoed Prescott @ 4Ghz at IDF. Even if this leakage problem exists, Intel has probably found a solution. This no doubt ties in with the fact that Intel recently switched to a different low-k dielectric supplier.

- - -
"... In the semiconductor industry, it's good to be paranoid ..." - [Andy Grove]</font color=green>
 
Excellent information you've got there, Dark_Archonis. Indeed, I recall that Prescotts have been circulating for a long while now...

<font color=red><b>M</b></font color=red>ephistopheles
 
I knew Dark wouldn't run away like that. He'd get a kick out of every time THGC has something wrong about Intel...
Pop, he appears. Pop, he disappears.

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>The sexiest website ever, guaranteed XXX!!!</font color=blue></b></A> :wink:
 
I'm still waiting for you to answer me in that thread where you threw a fit.

<A HREF="http://forums.btvillarin.com/index.php?act=ST&f=41&t=389&s=1fee5dab901bebe29da7aa1c2658fc6f" target="_new"><font color=red>dhlucke's system</font color=red></A>

<font color=blue>GOD</font color=blue> <font color=red>BLESS</font color=red> <font color=blue>AMERICA</font color=blue>
 
HAHA! I wanted to direct yer attention to his presence. Was worried you weren't checking this thread out though.

LOL DH, you really do fight till the end to get an answer! :wink: Love that!

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>The sexiest website ever, guaranteed XXX!!!</font color=blue></b></A> :wink:
 
I think we will see all copper Intel heatsinks. Either that or an increase in fan size to 80/92mm and thin fin designs.

People seem to forget that just because it will use a new smaller die size doesnt necessarily mean it will run cooler.
e.g. TbredA/DDR2/Nvidia 5800 ultra.

Smaller die = more heat per unit area
more transistors = more heat
higher Mhz = more heat

And before i get labled a FANBOY. I hope it isnt the case. Though im not holding my breath.

<b>Regards,
Mr no integrity coward.</b>
 
And also, I dont think intel release figues for Maximum thermal output, which is different from nominal max output . So there is variability in there too.

<b>Regards,
Mr no integrity coward.</b>
 
nice sig!

<A HREF="http://forums.btvillarin.com/index.php?act=ST&f=41&t=389&s=1fee5dab901bebe29da7aa1c2658fc6f" target="_new"><font color=red>dhlucke's system</font color=red></A>

<font color=blue>GOD</font color=blue> <font color=red>BLESS</font color=red> <font color=blue>AMERICA</font color=blue>
 
(Not answering to anyone in particular, just clicking on last thread.)

One thing to remember too is that Scotty has more cache. If I remember correctly (someone please point out if I'm wrong) the cache not only takes up a lot of the increase in the transistor count, but also in the increase in thermal output (because cache gates are used tons more often than logic gates and are so tightly packed). And this is the reason why the CPU utilization and TDP is much closer to the theoretical max in Scotty than in Woody. It's not so much because of better processor utilization as just increasing the amount of the most frequently accessed part: cache memory. So if anything that's why Scotty would be so hot.

I mean I could be wrong, but I don't think so. **shrug**

"<i>Let's see what <b>Paragraph 84-B</b> has to say about it.</i>" - Thief from <A HREF="http://www.nuklearpower.com/daily.php?date=030724" target="_new">8-Bit Theater</A>
 
dont the xeons use l3 cache? Isn't L3 cache off die? So it would not effect cpu temps as much as if all the cache was manufactuered right on the cpu core...although it does add some heat....

3 386DX-25's...12 volts...glue some ln2 and a wicked amount of overclocking and you get a willamantee minus 36 pins, 33.75 million transistors and a couple hundred mhz... 😎
 
not answering to anyone in particular, but several people,

dhlucke, In all honesty, I do not remember that argument. If you could dig it up, or refresh my memory, I would gladly repsond.

slvr_phoenix, you're wrong with regards to cache. Cache actually does not output that much heat, which is why Intel added 2MB of L2 cache to Dothan. The reason is that the heat/performance ratio of cache is great. Cache is not used as much as the execution units, and on the P4, the trace cache contributes a large portion of the entire cache thermal output. Cache takes a lot of die space, since so many transistors are needed for cache. Thus, cache dissipates heat relatively easily since it's spread out over a large area. When there is a process shrink, the cache still remains comparatively less dense, then say, the execution (logic) units. As I mentioned before, a large part of Prescott's thermal output is coming from leakage. Alot of people at other forums seem to have agreed that this is indeed the case. And, as that Intel employee over at Anand mentioned, the bigger challenge is cooling the CPU.

I have always stated that a CPU's thermal output is only half of the whole story. The other half is the efficiency of the CPU/HSF to move that heat away from the die efficiently. P4's output more heat than comparable Athlons, yet the P4's usually run at a lower idle temperature. This is a combination of the IHS, efficient HSF, and the fact that the P4 actually idles properly, unlike the Athlon. I forget what the exact problem is, but Athlons run at high temperatrues even when idling, IIRC, due to some idling command that the Athlon does not correctly recognize and execute. This is similar to a HALT or SLEEP state/command.

Eden, I'm shocked as to how little you've changed. You've gone from an overenthusiastic AMD fan, to an overenthusiastic Intel fan. Either way, you stretch the truth no matter which side of the fence you're on.

<i>Originally written by Eden</i>
Prescott will be a new evolutionary CPU both in terms of features/performance AND physical advancements! (process technologies added)
You <i>actually</i> have the guts to call me a fanatical Intel preacher, and yet you make comments like these? I mean, Even I myself would not go so far as to make a comment like that. It's as though you believe Prescott will be the best thing since sliced bread.

<i>Originally written by Eden</i>
While I would usually agree on new process issues, Intel is just NOT THAT clumsy. Yes their early batches could be hot, but it's a new process, which significantly reduces power and heat usually. It also features extra process technologies for leak prevention and such. It just beats me if it reaches 100W like that.
Forgot, but what was the 3.2GHZ Power Consumption? THGC didn't seem to have it on their chart.
Do you fail to ignore the fact the Prescott will be more dense, have a smaller die, AND use a larger percentage of CPU resources than Northwood? Leakage is a big problem at 90nm. Heck, scaling is not the issue here. For Intel, it's leakage. AMD is using SOI for scaling primarly, since otherwise the "cobbled together" K8 core would not be able to scale much past 2ghz. Reduced power consumption is more like a convenient feature for AMD, rather than a necessity at this point. Also, Intel 90nm process really has nothing to prevent leakage. Strained silicon does not prevent leakage; all it does is increase electron mobility resulting in faster transistors, and clock speeds. Rather, I would say that Strained Silicon actually increases leakage somewhat. I'm willing to bet that Intel may introduce it's Terahertz and Tri-gate transistors early.

- - -
"... In the semiconductor industry, it's good to be paranoid ..." - [Andy Grove]</font color=green>


<P ID="edit"><FONT SIZE=-1><EM>Edited by Dark_Archonis on 08/03/03 03:39 AM.</EM></FONT></P>
 
Either way, you stretch the truth no matter which side of the fence you're on.
Spare me how you always judged everyone and thought you were the right person. I could very well right now link you to what you said once in your "departing and sad thread", but I am a fair player and will leave that in the past for now, unless you provoke me to. And for someone who drops by once per millenia to post "correction" facts, you sure do have nerves to judge. I just love talking about you indeed, because it only reminds me of the wussy who couldn't stand fanboys, who preached Intel so much. Of course I like to get on your back. I'd gladly get off soon though, depending on how you are willing to react.

You actually have the guts to call me a fanatical Intel preacher, and yet you make comments like these? I mean, Even I myself would not go so far as to make a comment like that. It's as though you believe Prescott will be the best thing since sliced bread.
One quote of mine does not equal the myriad of quotes you did for Intel my good bud.
Additionally I say this with all due honesty, that Prescott is a great evolution with much better process technologies. Where is it not that much?
It will easily scale to 5GHZ, no doubt, it has even more future-proof features, improved HT and others.
It has new features and HAS new process advancements (the PHYSICAL claim I said). So, why am I being overenthusiastic? I know it has trouble lately, but I'll be damned if one extra stepping won't turn it into a real player.
Please, lest we forget, you were far worse. You were literally traumatized by AMD fanboys.

Do you fail to ignore the fact the Prescott will be more dense, have a smaller die, AND use a larger percentage of CPU resources than Northwood?
Smaller die? Any proof so far? It will after all have an extra 512KB L2, alone cache takes a great amount of space. THEN it has extra Trace Cache space, and possibly extra L1. Furthermore HT is refined, and new instructions are introduced. Frankly if its die is not that of NW or closely by, I'd be surprised.
I realize it will use a higher percentage. However being a 0.09m component at 1.2V, the process technologies should have alleviated most heat problems. I would assume it would churn out 70W at 3.4GHZ, not more. I didn't expect that a higher utilisation of ressources meant ~30W more. Heck, just WHAT ARE those extra resources? I know HT will be refined, but it still has 6 pipes, nothing more nothing less, it's a matter of using them more often. The cache as you explained, is not that heat expensive.

I used to strongly believe in both Intel and AMD. But recently AMD's doings have done nothing to show me they are willing to remain in the desktop game (I could care less about them later on if they do fail at desktop performance and no longer compete there but rather in servers, just like I could care less for Sun's achievements, as server CPUs don't matter to me, Opteron or SPARC or whatever). Opteron is doing great selling to supercomputer labs. I believe it is their lifeline really. Athlon 64, is simply NOT going to save them. No matter how much better it is, it won't. It is not future-proof in any way, both feature-wise and Physically-wise. And you of all, I'd expect to have learned that.
Nonetheless, that doesn't mean I am completely leaving out hope for AMD. Few weeks ago I stated it's their end for sure, after hearing about their losses. But reading about Opteron's recent acquisitions by supercomputer labs, and some possible speed limits broken, I have slightly more faith. Doesn't mean I think AMD will bash Intel out. What I believe in my heart is that AMD is not anymore as good as before, and while I will keep hope, I simply won't look at their products for future-proofness any soon. So that means Prescott or Tejas for me, very likely so. (not to mention the ridiculous 3 sockets introduced in one year for A64)
As a trick question, suppose that the planets aligned on Sept. 23rd, and Athlon 64's performance is monstruous, able to beat Prescott by 10% or more, would you buy the AMD system? In fact, suppose AMD's reign continued for a few more model grades. Would you?
Guess you see why I am more enthusiastic about Intel, all the while without being biased
--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>The sexiest website ever, guaranteed XXX!!!</font color=blue></b></A> :wink:
PS: New sexy users' sites now added! :smile: <P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 08/03/03 04:27 PM.</EM></FONT></P>